journal of international agricultural trade and … 10... · 2017-04-01 · viju ipe the world bank...

JOURNAL OF INTERNATIONAL

AGRICULTURAL TRADE AND DEVELOPMENT

Volume 10, Number 1

TABLE OF CONTENTS

Effects of Changing Weather Patterns on the Trade of Major Food Crops 1 J. P. Powell, K. Shutes, and A. Tabeau

Changing Technical, Allocative, and Economic Production Efficiency of Small-Scale Farmers in Indonesia: The Case of Shallot Production 31

Sujarwo, Michael R. Reed, and Sayed H. Saghaian

Managing Risk in Cocoa Production: Assessing the Potential of Climate- Smart Crop Insurance in Ghana 53

Justin McKinley, L. Lanier Nalley, Rebecca A. Asare, Bruce L. Dixon, Jennie S. Popp, and Marijke D’Haese

Competitiveness and Comparative Advantage of Important Food and Industrial Crops in Punjab: Application of Policy Analysis Matrix 81

Abdul Salam and Sadia Tufail

Agricultural Trade in South Asia: How Important Are the Trade Barriers? 95 Nitya Nanda

The Trade Embargo and Consolidation of Food Sector in Poland 113 Kazimierz Meredyk

Apple Sector in Poland and Effects of the Russian Embargo 121 A. M. Klepacka

The Russian Embargo: Tribulations of Dairy and Meat Exports from Poland and Other EU Member Countries 135

Marcin Wysokiński and Joanna Baran

New York

Journal of International Agricultural Trade and Development

The Journal of International Agricultural Trade and Development is intended to serve as the primary outlet for research in all areas of international agricultural trade and development. These include, but are not limited to, the following: agricultural trade patterns; commercial policy; international institutions (e.g., WTO, NAFTA, EU) and agricultural trade and development; tariff and non-tariff barriers in agricultural trade; exchange rates; biotechnology and trade; agricultural labor mobility; land reform; agriculture and structural problems of underdevelopment; agriculture, environment, trade, and development interface. The Journal especially encourages the submission of articles that are empirical in nature. The emphasis is on quantitative or analytical work that is relevant as well as intellectually stimulating. Empirical analysis should be based on a theoretical framework and should be capable of replication. Theoretical work submitted to the Journal should be original in its motivation or modeling structure. The editors also welcome papers relating to immediate policy concerns as well as critical surveys of the literature in important fields of agricultural trade and development policy and practice. Submissions of critiques or comments on the Journal’s articles are also welcome.

Journal of International Agricultural Trade and Development is published in 2 issues per year by

Nova Science Publishers, Inc. 400 Oser Avenue, Suite 1600

Hauppauge, New York 11788-3619 USA E-mail: [email protected]

Web: www.novapublishers.com

ISSN: 1556-8520

Subscription Price per Volume

Paper: $505.00 Electronic: $505.00 Paper and Electronic: $707.00

Additional color graphics might be available in the e-version of this journal.

Copyright © 2016 by Nova Science Publishers, Inc. All rights reserved. Printed in the United States of America. No part of this journal may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical, photocopying, recording or otherwise without permission from the Publisher. The Publisher assumes no responsibility for any statements of fact or opinion expressed in the published papers.

Editor-in-Chief and Founding Editor:

Dragan Miljkovic North Dakota State University

Department of Agribusiness and Applied Economics 614A Barry Hall, NDSU Department 7610

Fargo, ND, 58108-6050, U.S.A. E-mail: [email protected]

Editorial Board Members:

Giovanni Anania University of Calabria, Italy

James Rude University of Alberta, Canada

Lilyan Fulginiti University of Nebraska, USA

Bhavani Shankar University of London, UK

Viju Ipe The World Bank

David Skully USDA ERS

William Kerr University of Saskatchewan, Canada

Patrick Westhoff University of Missouri-FAPRI, USA

Jaime Malaga Texas Tech University, USA

Alex Winter-Nelson University of Illinois, USA

William Nganje Arizona State University, USA

Linda Young Montana State University, USA

Allan Rae Massey University, New Zealand

SUBMISSIONS

All manuscripts should be sent as email attachments directly to the Editor-in-Chief, Professor Dragan Miljkovic, at [email protected]

Journal of International Agricultural Trade and Development ISSN: 1556-8520 Volume 10, Number 1 © Nova Science Publishers, Inc.

EFFECTS OF CHANGING WEATHER PATTERNS ON THE TRADE OF MAJOR FOOD CROPS

J. P. Powell1, K. Shutes2, and A. Tabeau1 1LEI, Wageningen University and Research Centre, Wageningen, Netherlands

2Coventry University, Coventry, UK

ABSTRACT

The paper examines the economic effects of expected changes in temperatures and precipitation on the trade of ten major food crops. The relative effects for developing versus developed countries are emphasized. A series of econometric models using panel data and autoregressive integrated moving average models are used to estimate and forecast relationships between yields and weather data, the results of which are used as input into MAGNET, a global computerized general equilibrium modeling framework. Econometric results show that average temperatures have increased across all areas growing major food crops. Results for precipitation are ambiguous, however, there is statistical evidence of two distinct periods, a first in which, on average, precipitation fell, followed by a second in which it increased. Results for crops with statistically significant estimates show that increasing temperatures have negatively affected yields. These results hold for both poor and rich countries, however, the degree to which yields are reduced is crop specific and sensitive to a country’s level of wealth. MAGNET results show that changes in weather are likely to have significant effects on the production, trade, and, in some cases, the consumption of major food crops.

1. INTRODUCTION

1.1. The Problem

The effects of climate change on trade have been identified as an important, neglected, research topic (Josling et al., 2010; Tamiotti et al., 2009). This paper addresses the subject by examining the global consequences for trade of expected changes in temperatures and precipitation. In particular, it asks what the effects will be of changes in those variables on the trade of major food crops. Long-term changes in those important weather variables are expected to affect yields, and thereby the ability and willingness of countries to trade agricultural products and, ultimately, worldwide consumption patterns.

J. P. Powell, K. Shutes, and A. Tabeau 2

It has been argued that the economic effects of changing weather patterns on food consumption will depend as much on its role in trade as it does on crop production directly (Reilly et al., 1994; Sonka, 1991). Trade of agricultural products is particularly important for the well-being of developing countries, and is widely considered to be a major potential contributor to their prospects for growth, food security, and efforts to reduce poverty (Moïsé et al., 2013). Although agricultural trade has fallen as a percentage of the value of total world trade of goods and services, it has been increasing at a faster rate than world agricultural production and remains an important component of the economies of both developing and developed countries (Josling et al., 2010). In particular, FAOSTAT data can be used to show that the share of developing countries in total world exports and imports of agricultural products has been steadily increasing since 1990, and that they have become net importers of agricultural products (FAOSTAT, 2013). For these reasons, it is important that the potential impacts of changes in weather patterns on trade are assessed. Trade, in turn, directly affects domestic prices, quantities produced and, ultimately, consumption.

Arvis et al., 2013, show that the relative costs of trading are higher for developing countries, placing them at a competitive disadvantage when compared to developed countries. There are many reasons cited for this disadvantage, including traditional and non-tariff trade barriers and regional integration agreements. Changing weather patterns represent a potentially significant additional cost for developing countries in that they may be less able than wealthier countries to adapt to changing weather patterns. We know, for instance, that farmers in developed countries are able to adapt to changes in weather patterns by changing when they plant, but also by changing their use of inputs such as applications of fertilizers and pesticides and, in general, their use of inputs requiring energy (Powell, 2015). We hypothesize that developing countries are less able to substitute inputs, particularly high energy inputs, to meet changing climate conditions and are thereby at a competitive disadvantage if weather patterns change. This paper examines whether that hypothesis is supported by testing if there are significant differences between the effects of changing weather patterns on developing versus developed countries. Please note that the paper is concerned with the chronic hazards of changes in weather as opposed to what are known as sudden onset hazards or extreme weather events (Cutter et al., 2009).

1.2. Methodology

In order to determine the effects of precipitation and temperature changes on trade two methodologies are needed, namely, econometric methods are used to forecast changes in yields resulting from changes in weather variables and, thereafter, the forecasts are used within a general equilibrium model to simulate the effects of the altered yields on the exports and imports of food crops. Our approach differs from previous studies by employing a combination of economic tools to address each point of this complex issue.

Pieces of the general topic of the effects of climate change on yields have been previously investigated, (Deschenes, 2006; Licker et al., 2010; Lobell et al., 2011; Schlenker and Roberts, 2009). For instance, statistically significant results on the effects of temperatures and precipitation on yields of major food crops have been reported in the literature, with Lobell and Field providing one of the first articles with results applicable at a global scale (Lobell and Field, 2007). In addition, a series of recent articles using prominent

Effects of Changing Weather Patterns on the Trade of Major Food Crops 3

computerized general equilibrium (CGE) models examined the effects of climate changes on global production and consumption, emphasizing the differences found between model results (Nelson et al., 2013a,b; Valin et al., 2014). Other studies have examined the effects of climate changes at lower levels of aggregation, for instance at regional and country levels of analysis (Erda et al., 2005; Prato et al., 2010). For instance, the literature on the adaptation of farms to climate changes is meant to address farm adaptation to changing weather patterns (Bradshaw et al., 2004; Gebrehiwot and Veen, 2013; Luers et al., 2002; Pandey et al., 2007; van Wijk et al., 2012).

Our approach combines country level data analyzes with a global CGE model. It builds-on and refines previous econometric yield models by providing a more nuanced methodology for examining the effects of changes in weather on yields. In particular, it tests whether country specific heterogeneity is a statistically significant factor in the relationship between yields and weather variables. Estimates are calculated using various panel data models including dynamic models (Baltagi, 2008; Frees, 2004; Hsiao, 2003). Panel models are the preferred method given the global scale of the issue under investigation. They provide insight into general econometric relationships across all countries, and are a means of disentangling the effects of general versus country specific differences. In addition, the wide range of crops considered allows us to test whether there is variation in how crops react to forecasted weather changes. However, panel analyzes provide just one global relationship per crop. Although, country specific forecasts can be retrieved from such a global relationship (Frees and Miller, 2004), a more natural approach given the amount and completeness of the data available is to estimate country specific relationships directly from the data based on variables identified by the general, crop specific panel model. Specifically, when data allows, country specific autoregressive, integrated, moving average (ARIMA) models are used to estimate and forecast the effects of weather variables on yields (Hamilton, 1994).

Those forecasts are then used as input to run a series of simulations using MAGNET, which is a significant extension of the Global Trade Analysis Project (GTAP) model (Hertel, 1997; Hertel et al., 2010). Two general types of simulations are implemented, one in which the calculated effects of weather are included in the model, and the other, a base case, in which the effects of weather are excluded from the model. Results for the separate models are then compared to determine the impacts of changing weather on trade. CGE models have been used extensively to simulate trade related issues (Dixon et al., 1977; Robinson, 1990; Robinson et al., 1990; Rutten et al., 2011). Indeed, as its name implies, analyzing trade was an important motivating factor behind the development of GTAP. By applying simple shocks to yields while retaining default GTAP relationships, MAGNET output will allow us to determine the effects of weather changes on trade at both country and global levels.

Together, the use of econometric techniques and the MAGNET model provide a comprehensive analytical framework which allows the effects of weather events to be linked to a model with a global scope, the result is a unique perspective on the impacts of changing weather patterns on global trade. The proposed approach is admittedly complex in that it involves using descriptive, panel, time-series (ARIMA), and CGE techniques to address the research questions. The first section provides a description of the data and its primary characteristics. Thereafter follows a section describing the panel model and the results thereof. The ARIMA section thereafter describes the procedures used to forecast changes in yields while emphasizing the uncertainty of those forecasts. The results of MAGNET simulations are then presented and, finally, general conclusions are drawn.


2. CHARACTERISTICS OF THE DATA

This section examines the sources of the data used in the analyzes and their important characteristics, in particular, whether it is possible to identify weather trends. Econometric techniques will give an indication of how confident we can be that any observed trends will continue. Yield developments for the ten crops in the study over the last 50 years are then examined. Results in the following sections provide fundamental information on the relationships between variables that eventually need to be addressed in the econometric models.

2.1. Temperatures and Precipitation

Initially, three independent weather variables were used to estimate changes in yields, namely, precipitation, minimum and maximum temperatures (Lobell and Field, 2007). In the data set developed, the globe is divided into 720 * 360 grids according to an area’s latitude and longitude. Temperatures and precipitation are reported for each grid on a monthly basis for each year in the analysis. Precipitation is the monthly average of precipitation in a grid measured in millimeters, while minimum and maximum temperatures are the monthly average low and high temperatures reported in Celsius. Temperature data were obtained from the East Anglia Climate Research Unit (CRU TS) 3.10 (Jones and Harris, 2008). Precipitation data were derived from a revised version of the 3.10 database upon recommendation of East Anglia. Spatially weighted averages of the CRU data were then computed for each crop, with weights defined by the spatial distribution of crop area based on the method by Leff et al. (Leff et al., 2004). The essence of the method used is that the greater the percentage that a grid provides to a country’s total production of a crop, the greater the weight the weather variables in that grid receive in the calculation of a country’s average yearly temperature and precipitation levels. Unfortunately, only one year of data is available to calculate the weights.

We initially followed the methodology as described in Lobell and Field, 2007, which consists of defining global growing seasons for each crop. This approach selects weather data only for the months important to the growth of a crop. However, in contrast to Lobell and Field, whenever the data allowed, we selected growing months per country rather than using a single growing periods for all countries. For instance, say that the months from May to October are important months for the production of wheat in country ’A’, then only weather data from those months was selected and used in the analysis (Sacks et al., 2010). Data on months important for growing certain crops was not always available, in which case a yearly average of the weather variables was used. This use of country specific growing season data seems reasonable given that countries in Northern and Southern hemispheres are likely to grow the same crop in different months. The resulting database contains crop-specific monthly time series data of selected global temperatures and precipitation for the years 1961-2010.

Figure 1a shows monthly average maximum temperatures during growing seasons in grids in which important food crops were grown from 1961 to 2009. An identical plot was made for minimum temperatures, however only the figure for maximum temperatures is included because of the high correlations between the two variables. The average trend line,


the solid smooth gray line in the figure, is positive and statistically significant, as are the trends lines for each crop taken separately.

(a) Maximum Temperatures

b) Precipitation

Figure 1. Average monthly high temperatures and precipitation during growing seasons for ten major food crops.


Average results across all crops show that maximum average temperatures have increased by 0.017 degree C per year in crop growing grids, while average minimum temperatures for all crops have increased by 0.020 degree C per year. The figure emphasizes the extent to which worldwide temperatures are synchronized across grids despite the wide variety of crops grown, the range of growing seasons, and the various environments over which crops are grown.

Global temperatures, whether maximum or minimum, appear to react to the same short-term temperature fluctuations and to be following the same long-term trends. At the individual crop level, cassavas appear to be grown in grids when they are relatively warm, while sugar beets are grown in grids when they have lower temperatures; wheat and barley are grown in grids with nearly the same temperatures. Because temperatures of grids were calculated based on the months in which a crop was grown, it is not possible to directly determine from the figures whether the extent to which the temperature of a grid is determined by its location or by the months in which it was grown. However, the figures do appear to conform to the common knowledge that cassavas, rice and sorghum are grown in warmer regions, sugar beets and rapeseed are grown in cooler climates, and the remaining crops are grown in milder regions or across regions.

The data in this subsection suggests that if underlying data generating processes behind the rise in temperatures persist, as it has throughout the period studied, then we can expect temperatures in the future to rise as well. Note that the intention of this section is not to predict the causes of changes in weather; other than time, no other explanatory variable is included in the model. In the current analysis it is only possible to deduce that if the effects of unidentified underlying variables continue to cause temperatures to increase and nothing of significance changes, then we can be fairly confident that temperatures will increase in the future. These are heavy qualifications, but they need to be explicitly stated in order to understand the limits of the analysis to follow and, at a minimum, they suggest that only short or medium term predictions should be attempted in order to mitigate the issue of potentially changing underlying data generating processes.

In contrast to temperatures, results for precipitation across all crops are, with the exception of sugar beets, insignificant, implying that there is no statistical evidence that precipitation in crop growing grids has significantly changed over the entire period under review. Precipitation in grids growing sugar beets has increased by around 0.075 millimeters per year. However, the trend line in Figure 1b suggests two distinct periods, one from 1961 to around 1987, and the other continuing thereafter until 2010. Running separate regressions for the two periods leads to statistically significant results for crop averages in both periods with an estimate of -1.42 and t-value of -2.2 in the first period, and an estimate of 2.34 with a t-value of 3.0 in the second period. Chow tests confirm the observed structural break, and results from regressions are significant for each period reported.

In general, precipitation in crop growing areas has been increasing since 1988. The figure and tests suggest including a non-linear form of the precipitation variable in the model. They also suggest that, in contrast to temperatures, there has been a large amount of variation in the amount of precipitation falling in different crop growing grids. The ambiguous nature of the effects of precipitation reappear in the econometric results to follow.


2.2. Yields

The FAO data used in the analysis spans a period from 1961 to 2010 and measures hectograms of a crop produced per hectare (FAOSTAT, 2013). The intent of the analysis in this section is to present the general path that yields have followed over the last fifty years in order to identify possible trends which we would expect to reappear in the econometric analyzes. Note that yields are rates in that they measure an amount of output per area and should not be confused with the amount of a crop produced. Yields are a measure of partial productivity in the sense that they inform us about one important aspect of the crop production process, other productivity measures, including labor and capital productivity, are not included in this study. We focus the analysis on yields because we expect that weather will have the greatest effect on that factor and because, frankly, that data are readily available in large quantity.

Figures 2a and 2b show that, on average, crop yields have increased over the period under investigation, however they have in many cases, as has been previously noted, been increasing at a decreasing rate (IFPRI, 2009; Powell and Rutten, 2013; Sims et al., 2008). Figure 2a shows the general tendency of yield increases over the last 50 years. There have been noticeable differences in the growth rates of crops over the period. Yields of cassavas, at 31%, have increased the least over the period, while percentage increases of sorghum (120%), wheat (133%) and maize (180%) have been well over 100%; rapeseed (65%), paddy rice (75%), sugar beet (81%), barley (86%), potatoes (89%) and soybeans (100%) round out the field.

a) Yields of ten crops over 50 years

Figure 2. (Continued).


b) Yields of countries with the highest wheat yields

Figure 2. Yields of ten major crops.

Figure 2b provides a more detailed view of wheat yield developments. Wheat was chosen because of its ubiquity. The figure shows the gradual decrease in the rate of yield increases which has occurred for ten countries with the highest wheat yields in 2009. The thick gray line near the middle of the figure was produced using a weighted regression technique and shows the average path of yield increases through time. Please note that the y-axis in figure 2b does not have a zero origin, this was done in order to direct focus to the marginal changes that have occurred over the period. Although the figure is only for wheat, the observed phenomenon extends to several crops in the study including barley, rapeseed, potatoes and rice, but less so for maize. A reduction in the rate of yield increases is less clear for soybeans and sugar beets, while yield growth appears flat for countries with the highest yields of cassava and sorghum. In any case, results from this section recommend adding a non-linear time trend to the econometric model in order to capture potential effects of observed decreases in yield increases.

3. MODELS

3.1. MAGNET Model

The global economic simulation model used in the analysis is the Modular Applied General Equilibrium Toolbox (MAGNET) model, a CGE modeling framework developed at LEI, an institute which is a part of the Wageningen University and Research Centre in the Netherlands (WageningenUR, 2012). MAGNET has been used extensively to study the


impact of policy changes on international trade, production, and consumption, and it took part in the comparison of CGE models in regards to the effects of climate change on production and consumption as mentioned above (Nelson et al., 2013a; Tabeau and Woltjer, 2010; van Meijl et al., 2006; Verburg et al., 2008). It is an extension and significant reorganization of the GTAP model, a widely used tool for global trade analyzes (Hertel, 1997). The following subsection provides a brief description of the model including the data base, which forms the core of the model, and the method of modeling actor behavior and markets.

The data used in the analysis is based primarily on version 8 of the database collected and processed by GTAP at Purdue University (Hertel, 1997). Version 8 uses 2007 as its base year and contains balanced economic data for 129 regions and 57 economic sectors (GTAP, 2012). For the purpose of the current analysis, the 129 regions and 57 sectors have been aggregated in accordance with the research question. Aggregation allows less important data to be bundled together and the focus to be directed at the specific regions and sectors under investigation. The regional aggregation used consists of all primary crop producing countries and regions. The sectoral aggregation consists of the primary agricultural sectors available in MAGNET, namely, paddy rice, wheat, grains, and vegetable oils. The remaining sectors have been aggregated into a general manufacturing sector and a service sector. The model retains the standard MAGNET specification of five factors of production, namely, skilled and unskilled labor, capital, land, and natural resources.

MAGNET captures the behavior of three types of agents; households, firms and government, for each region in the model. Household behavior is captured via a representative regional household which is assumed to maximize its utility, collect all income that is generated in the economy, and allocate that income over private households, government expenditures, and savings for investment goods. Income is derived from payments by firms to the regional household for the use of endowments of skilled and unskilled labor, land, capital, and natural resources. The regional household also receives income from taxes paid by the private households, firms, and the government expenditures. Firms, profit maximizers, produce commodities by employing the aforementioned endowments and intermediate inputs from other firms based on a constant returns to scale production technology and sells them to private households, the government, and other producers. Domestically produced goods can be either sold on the domestic market or exported. Similarly, intermediate, private household, and government demand for goods can be satisfied by domestic production or by imports.

Demand for and supply of commodities and endowments are traded in markets which are modeled as perfectly competitive and clearing via price adjustments. Because all markets are in equilibrium, firms earn zero economic profits, households are on their budget constraints, and global savings must equal global investments. Since the CGE model can only determine relative prices, the GDP deflator is set as the numéraire of the model against which all other prices are benchmarked. Changes in prices resulting from the model simulations therefore constitute real price changes. For the current study, we are using the model to carry out dynamic analyzes over time, specifically for 2007 (the base year) until 2020. Projections into the future are obtained by allowing the exogenous endowments of capital, land, natural resources, and labor, and the productivity of these factors, most notably yields, to grow according to standard forecast growth paths which are based on readily available economic data.


3.1.1. Data Used to Shock MAGNET A shock in the current context means that crop yield data in MAGNET are changed from

what they would be under the standard assumptions outlined above, to what they are expected to be given forecasted changes in climate. The changes push the model from a starting equilibrium to an equilibrium which accounts for the changes in yields. Essentially, all countries are represented in the MAGNET model either as individual countries or as part of a region. Unfortunately for current purposes, many poorer countries, particularly in Africa, are not directly represented in the GTAP database. This is due primarily to data quality issues. Data for those countries are approximated at Purdue and then aggregated into regions. The aggregation of countries into regions in MAGNET means that our forecasted changes in yields need to be aggregated as well. The procedure used was to weight the ARIMA forecasts (explained below) according to the area that a country uses to grow a particular crop. For instance, if the total area of an aggregated region is 100 units, and the area of a particular country in that region is 20 units, then the forecasted yield change for that country is multiplied by 20/100 and so forth for all countries in the region. Similarly, the number of agricultural sectors represented in MAGNET are aggregates, recall that standard GTAP has fifty-seven sectors to represent the entire global economy. Agricultural sectors, although well represented given their relatively small contribution to the world’s economy, are few. For instance, grains is a sector in GTAP, and therefore MAGNET, which includes maize, but many other grains as well. We do not have data for all grains, consequently, we make the simplifying assumption that the sector grains follows our weighted estimates for maize, barley, and rye–the crops for which we do have data. Weighting the ARIMA forecasts was done according to the area used to grow each of the grain crops in MAGNET countries or regions; for instance, if a region primarily grows wheat, then wheat dominates the calculated shock. Similarly, the category vegetable oils in MAGNET was approximated based on the data we have for soybeans and rapeseed. Wheat and rice are separate sectors in MAGNET so no weighing was required. The resulting percentage change for each crop and each region and each of the five scenarios were used to shock MAGNET.

3.2. Panel Econometric Model

3.2.1. Introduction This section introduces the econometric models used to analyze relationships between

yields and weather. It builds on the results of previous sections by incorporating observed relationships into the models and testing whether those relationships hold statistically. Econometric models are introduced according to their underlying assumptions, with particular attention paid to how the various models account for country heterogeneity. The purpose of this section is to develop a general model, applicable across all countries. The model is, by necessity, rough, and meant to give an overview of the relationship between yields and weather variables on a global scale. The model is then applied to the same data, but partitioned into poor and rich countries to determine whether there are statistical differences between the two groups. Finally, in the section thereafter, the data is further partitioned into country level data and run using the same model with the addition of autoregressive and moving average terms. This partitioning exercise allows us to examine differences between global and country level analyzes.


3.2.2. Model Descriptions Given the high level of correlation between maximum and minimum temperatures as

observed in the figures above and confirmed statistically, minimum temperature was removed as an explanatory variable to avoid the problem of multicollinearity; although multicollinearity does not affect the consistency of estimates, it does adversely affect their efficiency. The choice to remove minimum rather than maximum temperature was somewhat arbitrary and, unsurprisingly given their high correlation, results were similar regardless of which temperature variable was included in the model. Model estimates for wheat, maize, rice, barley and soybeans are reported in Table 1. The first three crops in the table were chosen, out of the ten crops in the study, because they are the world’s top three food crops, while barley and soybeans were included because of the diversity they add in terms of model results. The range of significant outcomes runs the gamut, from cases in which both precipitation and maximum temperatures are statistically significant for a crop model, to cases in which neither variable is significant. Whenever results for the excluded five crops vary greatly from the five included in the table, those results are reported. Finally, this subsection concerns the specification or selection of a general model which is then to be applied to all crops, the results of which are reported in Table 2 in the subsection entitled model estimates.

The number of countries used in the analyzes ranges from 145 for maize to 40 for sugar beets. A balance had to be struck between the number of countries included in a particular crop analysis and the number of years of available data. Although many countries have complete data sets for the entire period from 1961 to 2009, several important developing countries do not. Therefore, the decision was made to reduce the number of periods in the analysis from a potential of 50 to 20 years. This allowed more countries, and therefore more diversity, to be included in a particular crop analysis, and has the practical advantage of providing for data sets that meet the requirement of a large number of observations relative to the number of time periods (Bond, 2002).

yit = xitβ + ziα + εit, where i = 1, ..., N; t = 1, ..., T (1) Equation 1 represents the basic panel framework (Greene, 2007). There are K regressors

in xit. The emphasis in panel models is placed on the second, individual effects term ziα, where zi contains a constant term and individual or group variables which may not be observable. Decisions regarding how ziα should be modeled are reflected in the assumptions made regarding heterogeneity.

The model in the first row of Table 1 is based on Lobell and Field’s model as described in (Lobell and Field, 2007). Their model, identified here as the aggregated model, is included because of its influence and it serves as a point of departure for discussing more standard panel models. Their methodology takes differences of the averages of each of the dependent and independent variables for each year in their analysis (41 years after first-differencing) and performs least squares on the aggregated differences. Their approach is labeled the aggregated model because it averages the hundreds and sometimes thousands of points of country level data available in the data set to one point for each year in the analysis without regard to heterogeneity.

A systematic, if not foolproof, means of interpreting the results for the aggregated model, and all of the other models in the table, is to find those variables in the table with absolute t-


values of around 1.96 or greater and determine how the variable in question affects yields. For instance, using the aggregated model and wheat, we can see that precipitation has an absolute t-value of 2.29 and so can be considered statistically significant. An estimator in a differenced model, a distinction which applies to all models in the table with the exception of the pooled model, shows how a change in, for example, precipitation, affects changes in yields. Consider once again precipitation in the aggregated model for wheat, the significant coefficient of 13.77 tells us that for every 1mm increase in precipitation, we expect wheat yields will increase by 13.77 hectograms per hectare or 1.377 kilograms per hectare. Similarly, the value of -709.05 for soybeans and temperature tells us that for every increase of 0.1 degree C, the change in yields decreases by 709.05 hectograms per hectare or 70.91 kilograms per hectare. A negative estimate does not necessarily imply that yield levels are decreasing, only that yields are, on average, increasing at a decreasing rate (we know from figures above that soybean yields are increasing on average). Note that in the cases of wheat, maize, rice and barley, none of the temperature variables are significant, meaning that no statistical conclusions can be drawn in regards to relationships between yields and temperatures for those crops in the aggregated model. In regards to precipitation, the estimates for wheat and soybeans are significant. The high level of aggregation used in Lobell and Field’s model comes at the cost of reducing the amount of information used to compute the estimates. More concretely, in their model and the two models to follow, no account is taken of individual country characteristics. These models assume, in short, that relations between yields and weather data are homogeneous across all countries.

The second and third models in Table 1 are the first-differenced and pooled (least squares) models. In terms of equation 3, these models, like the aggregated model, force all data to go through one common intercept term, e.g., they do not take into account differences between countries. As in the aggregate model, the first-difference model subtracts, per country, for all variables, the observation in period t from observation in t + 1. Differences are then pooled together across all countries and run together using least squares. The result is a matrix with 19 (20 years minus the first observation) observations per country for each country per crop. This method has the advantage over the aggregated model of using much more of the available data. Comparing estimates from the first-differenced and aggregated models shows that maximum temperature is now significant for wheat, maize and barley. Precipitation is no longer significant for wheat and just significant for barley and soybeans, while the effects of an increase in precipitation are negative for yields. In addition, the very high R-squared terms reported in Lobell and Field and replicated above for many of the crops in our models, fall to nearly zero. By not averaging the data, the correlation between the dependent and independent variables decreases remarkably, implying that the act of aggregating imposes an artificial structure on the data that disappears when the data is left in its disaggregated form.

Whereas aggregated models average all of the data by year, the other two models keep the data in their original, disaggregated, form and thereby retain the available variation in the data. In the pooled model, the 20 years of data for each country are stacked upon one another and run together using least squares. Panel data used in the pooled model was kept in levels form for statistical test purposes discussed below. Temperatures were significant for wheat, maize, rice and barley, while all increases in temperatures had negative effects on yields. Precipitation is significant and negative for wheat, maize, rice and soybeans, but in the case of soybeans an increase in precipitation has a positive effect on yields.

Table 1. Estimates of yield data for 1990-2009

Model Wheat Maize Rice Barley Soybeans Precip. Temp. Precip. Temp. Precip. Temp. Precip. Temp. Precip. Temp.

Aggregated 13.77 −35.74 4.41 −131.39 −0.10 −44.44 2.99 −109.92 −10.55 −709.05 2.29 −0.38 0.61 −0.90 −0.03 −0.55 0.60 −1.45 −2.11 −5.36

First Diff. 0.74 −152.25 0.79 −262.04 0.20 36.30 −1.44 −158.95 −2.58 −143.73 1.16 −9.41 1.16 −9.23 0.41 1.48 −1.92 −8.75 −1.83 −1.38

Pooled −3.34 −142.52 −11.92 −196.87 −2.54 −154.94 −0.60 −140.02 16.48 −28.90 −4.61 −21.53 −14.99 −17.50 −4.80 −16.81 −0.74 −19.20 7.80 −0.67

Within 0.20 −163.31 0.65 −275.47 0.22 46.43 −1.75 −167.35 −2.26 39.11 0.31 −9.71 0.95 −9.28 0.44 1.78 −2.29 −8.63 −1.58 0.32

Between −10.95 −318.75 1.26 66.51 −2.88 574.01 −8.89 −345.38 −24.88 −116.64 −1.87 −2.42 0.18 0.23 −0.54 2.49 −1.62 −3.20 −1.57 −0.11

Random 0.37 −159.76 0.71 −269.68 0.20 37.30 −1.62 −163.64 −2.43 −64.88 0.57 −9.64 1.04 −9.27 0.41 1.51 −2.15 −8.71 −1.71 −0.58

Random Sqr. 2.94 −149.30 5.09 −250.40 −1.70 29.59 2.96 −148.00 −1.14 −61.05 2.12 −8.62 2.92 −8.43 −1.13 1.16 1.77 −7.61 −0.25 −0.54

Dynamic 0.57 −161.25 0.47 −267.92 0.33 24.03 −2.20 −176.70 1.13 −119.42 0.70 −7.18 1.02 −3.94 0.35 0.53 −1.64 −7.53 1.68 −5.10

Note: Estimators and estimates for wheat, maize, rice, barley and soybeans; t-values are reported underneath each estimate. An absolute value t-value of 1.96 or greater is generally considered to be statistically significant. Data for the pooled model are expressed in levels form, all other models use first-differenced data. Year dummies are included in all models, except the between model where it is not appropriate, in order to capture time effects. Tests for first and second-order serial correlation and Sargan tests provide no strong evidence that the dynamic AR(1) model is misspecified. The method by Swamy and Arora was used to calculate the variance components in the random effects model except when year correlations were negative, in which case the Nerlove method was used. Asymptotic assumptions were not used when calculating standard errors because sample sizes do not justify the assumption of an N close to infinity for several of the crops examined. Asymptotic standard assumptions increase the t-value of a coefficient, not the value of the coefficient.


Precipitation becomes significant for several crops and the variables for cassava, for the first time, become significant. Again, in this model, as in the previous two models, no account is made of individual country heterogeneity, thereby effectively circumventing an important advantage of panel data.

And the individual country effects are significant. A standard pooling test was performed using the sum of squared errors from the pooled and the within models. In the case of wheat, an F-ratio of 201 with a degree of freedom in the numerator of 2,237 and a degree of freedom in the denominator of 2,107 has an associated p-value of less than 0.0001, providing strong evidence that country specific parameters should be included in models. Other crops showed similar results; in short, models that do not include country specific characteristics, such as those discussed above, should be avoided because they incorrectly assume that an important explanatory variable, namely country heterogeneity, has no influence on yields when it clearly does.

The first model discussed to actually use the information contained in the panel data format is the within model, a model which is numerically equivalent to the least squared dummy variable model. The within model allows the unobserved country heterogeneity to be correlated with the observed variables. Essentially, country specific intercept terms are included in the model to capture time-invariant, country level effects. The within model is designed to capture and incorporate the individual country variation within the data. In this manner, those effects which are time invariant are removed from the data. For instance, a country’s location does not, under normal circumstances, change significantly through time, the within methodology removes country location from the data and all other static information because subtracting the mean of a static variable from a static variable produces zero. In this manner, the effects of such variables, even those not explicitly in the model, drop out of the analysis and only dynamic effects remain. Results from the within model are similar to those from the first-difference model, a model which also removes time-invariant data. Temperature effects for both models for wheat, maize and barley are nearly identical; while the estimate for soybeans remains insignificant, the estimate for rice is nearly significant and positive. As for precipitation, only the term for barley is significant and similar in magnitude to the first-difference estimate. Whereas the intercept terms in the within model are assumed to be fixed and estimable, in the random effects model they are considered to be random drawings from a given probability distribution. In short, unobserved individual variables are assumed to be uncorrelated with the error term in the model (Baltagi, 2008; Greene, 2007, see, for discussion). The estimator is calculated using generalized least squares, a method which transforms a least squares regression by a weighted average of the residuals of the within and between models. Standard errors for the between model are very small when compared to the within model which means, in short, the variation within a country dominates variation between countries. Finally, in order to capture the effects seen in 1b, a non-linear precipitation variable was added to the random effects model. This model is labeled as the random effects, squared, model. Estimates for the squared term are not included in the table, but were significant yet small for most crops. The random effects squared model is the most inclusive of the non-dynamic models examined in that it uses both within and between variations to estimate coefficients. Results indicate that temperatures for wheat, maize and barley are significant and, in all cases, lead to lower yields for higher temperatures, while increases in precipitation lead to increases in the yields of wheat, maize, and perhaps barley. Estimates for rice and soybeans remain insignificant. A Breusch-Pagan test rejects the null


hypothesis that the intercepts are equal to zero, confirming that the pooled model should not be used. Finally, a Hausman test was run to test the null hypothesis that the coefficients estimated by the efficient random effects estimator are the same as those estimated by the consistent within effects estimator. In the model at hand, there is not significant evidence against the null that the GLS estimates are consistent and so the random effects estimator can be, and is, preferred to the fixed effects estimator.

Table 2. Random effects model estimates of yield data for 1990-2009 for both

poor and rich countries

Poor Countries Rich Countries Crop Temp. Precip. Precip. Sq. Temp. Precip. Precip. Sq.

Barley −61.94 6.60 0.00 −164.81 14.83 −0.01 −1.72 2.70 −2.38 −5.81 4.32 −6.35

Cassava 54.53 11.43 0.00 −247.88 −35.42 0.01 0.47 1.86 −1.95 −0.83 −3.26 2.73

Maize −164.82 3.14 0.00 −276.98 6.97 0.00 −3.97 1.59 −1.46 −5.84 2.12 −1.97

Potatoes 75.54 11.63 0.00 −139.00 −1.06 0.00 0.61 1.77 −1.46 −1.14 −0.11 −0.36

Rapeseeds −33.66 −7.72 0.00 4.86 2.55 0.00 −1.37 −4.13 3.86 0.19 1.22 −1.79

Rice −9.04 −1.43 0.00 64.79 −4.94 0.00 −0.26 −0.85 1.64 1.76 −2.04 1.56

Sorghum −16.24 1.70 0.00 −182.62 −1.50 0.00 −0.96 2.23 −1.54 −2.67 −0.28 0.16

Soybeans −57.99 2.60 0.00 −138.03 12.35 −0.01 −2.23 1.84 −1.72 −3.89 3.61 −3.79

Sugar beets

−2258.91 160.51 −0.06 −899.97 223.41 −0.11 −3.17 3.03 −2.55 −2.40 4.83 −4.40

Wheat −106.36 3.40 0.00 −153.57 10.20 −0.01 -3.85 1.78 -1.4 -6.42 3.64 -4.76

Note: All variables are first-differenced as in Table 1. Year dummies are included in all models. As in Table 1, either the Swamy and Arora or Nerlove methods was used to calculate the variance components of the model.

The last model discussed is the dynamic panel AR(1) estimator. A dynamic estimator

includes a lagged version of the endogenous variable as an explanatory variable. Lagged yields, although not of direct interest in the current study, are needed to calculate consistent estimates of the exogenous variables (Bond, 2002). Note that a squared precipitation term could not be included in the model because of the problem of near collinearity between that term and the time dummies. The time dummies were left in the model for reasons of comparison. Results for temperatures for wheat, maize, barley and now soybeans are all significant and negative. Precipitation is no longer significant for any of the crops in the table.

In general, results indicate that maximum temperatures and precipitation are statistically significant for several crops in several models. Ignoring standard errors for the moment, the results tell a fairly consistent story for models which incorporate individual effects, namely,


increases in temperatures will reduce yields for all crops with the exception of rice. Results for precipitation are mixed, wheat, maize, and barley yields increase, while rice and soybeans decrease. Finally, tests indicate that the random effects model is the preferred choice. An additional recommendation for that model is that its significant estimates fall between those of the levels pooled model and the within model, an indication that the model is well-specified (Bond, 2002). Other models considered and tested include a non-linear time trend as suggested by Figure 2b, and models with interaction terms, but they did not significantly alter the results of the random effects model. Finally, Table 1 illustrates the consistency of the effects of weather on yields; higher temperatures almost certainly have negative effects on the yields of wheat, maize, and barley, and possibly soybeans.

3.2.3. Model Estimates In the previous subsection we derived the general form of the model that will be applied

in the analyzes to follow. This subsection addresses the hypothesis of whether there are statistical differences between how the yields of poor and rich countries have reacted to changing weather patterns.

∆Yieldit = constant + ∆Maximum Temperatureit + ∆Precipitationit + ∆Precipitation2

it + uit i = 1, ..., N; t = 1, ..., T ; where uit = µi + vit, (2) where the µi denotes the unobservable individual specific effects and vit denotes the

remaining disturbance. Table 2 reports regression results for all crops using the random effects model with

maximum temperature, precipitation, and squared precipitation as exogenous terms. The method used to categorize countries was to split the data according to a country’s average real GDP per person for available years in the Penn World Table database (Heston et al., 2011). For both poor and rich countries, F-statistics for barley, maize, sorghum, soybeans, sugar beets and wheat are significant at the 5% or better levels and rice nearly so at that same level. The regressions for potatoes are insignificant for both poor and rich countries, while the remainder are significant for one group or the other. Before turning to the results in the table, we note that equation 2 was used to run regressions for data combining both the poor and rich countries. A dummy variable was used to distinguish between poor and rich countries in the regressions in order to test whether there were significant differences between poor and rich countries. The dummy was positive and significant for maize, potatoes, rice and sugar beets. For those crops, poor and rich countries should be modeled separately in order to account for the higher initial yields of rich countries.

Maximum temperatures were significant for both poor and rich countries for barley, maize, sorghum, soybeans and wheat. Only in the case of sugar beets was the maximum temperature significant for poor but not rich countries. And in all cases, both for poor and rich countries, the effects of rising temperatures on yields were negative. Yields of rich countries appear to be more sensitive to temperature changes in the cases of barley, maize, sorghum, soybeans and wheat, although it is not possible to discuss the relative effects for sugar beets because the estimate for rich countries is insignificant. By sensitive we mean, for instance, that a 0.1 degree increase results in a greater decrease in yields for rich countries than the same temperature increase for poorer countries. However, yields in rich countries are generally higher than in poorer countries, in fact, there is a strong positive, non-linear,


relationship between GDP per person and yield that can be observed by plotting yields against real GDP per person (Powell and Rutten, 2013). Yields in poorer countries are somewhere between a half and a third of those of richer countries in our sample for those crops for which estimates were significant. Take the case of wheat, for which the average yields in poorer countries are 20,190 and 33,913 hectograms per hectare (hg/ha) in richer countries. According to Table 2, a 0.1 degree increase in temperature will reduce yields in poor countries by 106.36 hg/ha, while for rich countries that reduction will be 153.57 hg/ha, which converts to a reduction in wheat per ha of around 0.53% for poorer countries and 0.45% for richer countries, the relationship is similar for maize and sugar beets. For sugar beets, poorer countries do far worse than richer countries. However, for soybeans, it is the richer countries that are worse off. These are clearly rough estimates, for instance, standard deviations reported in the table show that there is a great deal of variation within the two broad GDP categories. The results indicate the importance of examining the effects of weather on a per crop basis.

Figure 3. Wheat coefficients for maximum temperatures and precipitation.

For poorer countries, precipitation is significant for rapeseed, soybeans and sugar beets. An increase in precipitation leads to greater yields for those crops. Precipitation is significant for more crops grown in richer countries; in addition to rapeseed, soybeans and sugar beets, estimates for barley, cassava, maize and wheat are also significant for richer countries. Precipitation increases have a positive effect on yields of barley, maize, rapeseed, soybeans, sugar beets and wheat, but a negative effect on cassava yields. For the three crops which can be compared, namely rapeseed, soybeans and sugar beets, poor countries appear to be more susceptible in the case of rapeseed and sugar beets, but less so in the case of soybeans. In other words, a 1 mm increase in precipitation will reduce yields of rapeseed and sugar beets more for poor countries than for richer countries, but will have less of an impact for soybeans.

Conclusions to be drawn from this subsection are that the effects of weather variables on yields are both crop and country specific. Results from table 2 clearly indicate that for those crops with significant outcomes, higher temperatures will reduce yields, while increases in


precipitation will have positive effects for all crops except cassava. In short, the question of whether changes in weather will harm poorer or richer countries is crop dependent.

3.3. ARIMA model

Panel models give a general overview of relationships between yields and weather variables per crop at a global level. However, they produce just one global equation per crop which characterizes the relationship between yields and weather variables, in that sense panel models are rough approximations. In this subsection, country level analyzes are calculated for use in MAGNET. In particular, ARIMA models are used to forecast yields for wheat, rice, barley, maize, rapeseed, sorghum and soybeans. These crops are then aggregated into the following four aggregate agricultural sectors included in MAGNET: wheat, rice, grains, and vegetable oils.

3.3.1. ARIMA Methodology ARIMA models are used to forecast time series variables using information contained in

past realizations of the yields themselves and the weather variables. Estimates of the exogenous variables tell us the additional contribution to yields associated with the weather variables net of information contained in yields from previous periods. The ARIMA models estimated are not complete explanatory models, building that sort of model requires micro-economic, crop level data that is not generally available for a large number of countries.

The basic form of the model is: yt = θ1yt-1 + ... + θpyt-p + εt + α1εt-1 + ... + αqεt-q+ ∆Maximum Temperaturet + ∆Precipitationt + ∆Precipitation2

t. (3) For each country, the model was run across a spectrum of possible, meaningful, lags. The

‘best’ model was then chosen based on the Akaike Information Criterion (AIC). That model was then used to forecast yields ten years into the future using standard time series forecasting techniques. Other methods of determining the best model, such as the BIC and AICc were considered, but those techniques produced similar results. The intuition behind the ARIMA model is simple, the modelling procedure uses linear combinations of past observations and error terms to build a model of any underlying patterns in the data. Those patterns, expressed as a model, are used to predict future observations. The best model is the one that fits the data well with the proviso that given two models that fit the data equally well, the one that uses the fewer number of explanatory variables is preferred. The different criteria used to determine the best model differ only in the penalty they apply to the addition of an explanatory variable. The penalty is applied because, given enough explanatory variables, any econometric model can provide a nearly perfect linear fit of the data. In short, the penalty is applied to avoid the problem of over-fitting.

Figure 3 gives an indication of the range of outcomes for one model. It shows the density distributions of estimates for country level coefficients for wheat. For instance, Figure 3a shows that a large majority of estimates for maximum temperatures for wheat range from -400 to 200. The mean of the data is slightly negative, as expected given the generally negative effects of temperatures on wheat. A value of -100 implies that a 0.1 degree C increase


temperature reduces wheat yields by -10.0 hectograms. The figures show that, although the coefficients for these variables are bundled together near the center of the figure, there is a large amount of diversity in how the yields of countries have reacted to changes in temperatures and precipitation.

The t-values presented in Table 3 for each of the temperature and precipitation variables indicate that only a minority of the estimates are significant. That result is not surprising given that the autoregressive and moving average terms in the model are picking-up a lot of the explanatory variation in the model and that there are many factors that influence yields which are not included in the model. The ARIMA procedure employed selects the best model, given the requirement of parsimony. Although individual explanatory variables may not be significant, the entire model will extract the most amount of information possible given the constraints mentioned. The models and resulting forecasts should be seen as indicators of the direction that yields may take given changes in weather patterns, and not as precise predictions.

Table 3. Significance values of major crops at two levels

Temperature Precip. Precip. Sq. Num. Countries

Crop t-val. > 1.96

t-val. > 1.50

t-val. > 1.96

t-val. > 1.50

t-val. > 1.96

t-val. > 1.50

Barley 12 17 21 26 12 27 58 Maize 19 29 19 38 21 32 132 Rapeseed 3 7 8 8 5 9 26 Rice 18 30 17 28 10 20 101 Sorghum 11 15 9 18 13 19 73 Soybeans 2 7 8 13 9 14 39 Wheat 19 34 28 40 15 27 97

Note: The t-values are reported in absolute values.

3.4. MAGNET Simulations

This section discusses the economic implications of shocks calculated using the ARIMA forecasts. It uses those results to run five simulations, one, the base scenario, in which neither temperatures nor precipitation change, and four scenarios in which yields are shocked assuming changes in temperatures and precipitation are realized. MAGNET simulation results are reported as differences from the base case. Recall that base yields are provided by the FAO. The FAO forecasts do not appear to incorporate climate changes, although the forecast model they use is not well documented. In any case, our results should be interpreted as changes, either increases or decreases, to base yields due to forecasted changes in temperatures and precipitation. MAGNET results indicate how regional economies react to those changes. A strength of the CGE framework is that it demonstrates the importance of relative changes in production. For instance, our results indicate that changes in temperatures are expected to have a range of effects on yields across regions as can be seen in Figure 3. Although, on average, the results are expected to negatively affect yields, regional differences mean that particular regions, those that are relatively good at producing crops, may actually


increase production to meet demand despite lower yields. These economies pull resources away from other sectors to produce more crops for which they have a relative advantage.

Scenario (0,1) shocks yields according to an assumed 0.0mm increase in precipitation and a 0.1 degree C increase in temperature for ten years; while scenario (2,1) assumes that yields are affected by a 2.0 mm yearly increase in precipitation and a 0.1 degree C increase in temperatures. The chosen temperature and precipitation shocks are multiplied by the appropriate coefficient for the ARIMA model for each region and crop category in the MAGNET model. The other two scenarios, (0,2) and (2,2), are calculated using the same methodology. The shock implemented assumes for example, in the case of scenario (2,1), that precipitation increases by 2mm per year for ten years and 0.1 degree C per year for ten years. In short, in ten years precipitation is assumed to have increased by 20 mm and temperatures by 1.0 degree C. This is a big simulated increase, but within the realm of possibility. For instance, the earth’s mean surface temperature has increased in the 20th Century by about 0.8 degree C with about two-thirds of that increase occurring since 1980 National Research Council (2011). However, the point of the MAGNET exercises is to establish direction rather than calculate precise estimates, relatively large shocks help to realize this aim without being so large as to destabilize the model. In addition, the results are reported using an average of the simulations so that the influence of extreme simulations are reduced.

The following two figures present, respectively, the expected changes in exports and imports resulting from changes in weather across four crop categories. As mentioned, each point in a figure is the average yield of the four simulated changes in weather less the base case in which yields are modeled according to standard FAO forecasts for each of the 45 regions in the sample. Together these regions encompass the world. Results should be interpreted as indicating the additional effects of changes in weather on the economy assuming everything else in an economy, including yields, continues to develop as expected. Please note that MAGNET, like most CGE models, is a deterministic model, no stochastic assumptions are made. While the shocks were created using econometric techniques which allow measures of confidence to be associated with forecasts, MAGNET produces unique, non-distributional, outcomes; therefore, terms such as estimates and forecasts do not have their associated technical connotations. The advantage of using MAGNET is that it provides a detailed, global, sectoral, overview of interactions in the economy following a shock, something econometric models are not designed to do.

Figures 4 and 5 show expected changes in exports and imports respectively. More precisely, they present the additional percentage changes in exports and imports expected in 2020 based on the average changes over the weather scenarios. It is important to keep in mind when reading the figures that the values reported are relative to a country’s exports and imports and they are percentage values. For instance, Korea is not a significant exporter of grains, so a nearly 10% increase in exports for that region is only a small amount in absolute terms.

Whereas production changes for the most part were found to be small, changes in trade vary considerably as can be seen by the range of values along the x-axes of the figures. The major exporters of grains are the United States, Argentina, Brazil, Ukraine, South Africa and the European Union. Exports from the United States increase slightly while the region Rest of South America, which includes Argentina, and the region of South Africa, are expected to experience sharp falls in exports of between 4% and 5%. Brazil, on the other hand, is expected to experience a slight increase in exports. In what will become a familiar refrain, the


situation for Europe is mixed, with some countries expected to experience increases in grain exports while others might experience decreases. The major European exporter, the Ukraine (ree), is expected to see a decrease in exports of around 2%, while exports from Russia are expected to increase by around the same percentage.

Of the major palm oil exporters, Indonesia is expected to experience a decrease in exports of vegetable oils of around 5%, while exports from Malaysia (sea), are expected to remain constant. A major soybean exporter, Rest of South America, a region which includes Venezuela, can expect a large decrease in exports, while Brazil is expected to see exports increase by around 5% and the United States by around 2%. The large increases in rice exports for Japan and Korea have to be tempered by the fact that they export very little rice. All of the current major rice exporters, Thailand, Vietnam (both sea), and India, are expected to experience slight decreases in exports.

Note: The values presented in this figure and the next are percentage differences.

Figure 4. Exports.


Of the current major wheat exporters, the United States and Canada are expected to see substantial increases in exports, Europe is expected to experience mixed results with France showing a small decrease and Germany a small increase. Australia is expected to see a slight decrease in exports, while Russia is expected to see a remarkable increase nearing 30%. In 2013 Russia exported around 12 million tons of wheat, so the increase of 30% amounts to around 3.6 million additional tons due to changes in temperatures and precipitation. The Ukraine (ree), which exported around 3.1 million tons in 2013, is expected to experience a substantial decrease in production.

Results reported in the import Figure 5 are dominated by the region Rest of Former Soviet Union (rfsu), which includes countries such as Kazakhstan, Kirghizstan, and Tajikistan. This region shows a remarkable increase in the imports of grains, vegetable oils and wheat. However, these countries import very little of these products so their results should be interpreted with caution. The major importers of maize are Japan, Mexico, Korea, the Western Europe and China. Japan and Mexico are expected to experience little change in amounts imported while Western Europe shows a slight decrease in imports. China is expected to experience a slight decrease in vegetable oil imports, while India is expected to experience a very large, over 20%, increase in vegetable oil imports. Results for Western Europe show a slight increase in imports on average. India is expected to import substantially more rice, 50%, due to changes in weather, however, recall that it is a large net exporter of rice. Of the major importers of rice, the region of Central Africa (caf), which includes Nigeria, is stable, while imports to the region of North Africa (naf) fall, and imports into the region of South Africa increase. Imports into two other major importing regions, Bangladesh and the Middle East, also appear to be stable. Changes for wheat importers are generally small. Imports to Egypt (naf), the world’s largest importer, are expected to fall, as they are expected to for Indonesia. Imports into Brazil are expected to fall by 1% and 2%, while Chinese imports are expected to increase slightly.

Ultimately, policy makers are concerned with consumption, particularly in poorer countries. Changes to consumption were found to be small relative to changes in exports and imports. This is a reasonable outcome given the structure of consumption of agricultural products in most countries. On the whole, countries or regions consume products produced within the country or region, while exports and imports absorb surpluses and deficits. Recall as well, that although the shocks implemented are large in terms of the weather changes simulated, weather effects on yields only explain a small part of the total production of crops.

In terms of consumption, the biggest gainers and losers for grains are people from the regions Rest of former Soviet Union (rfsu), India, Central and South America, Central and North Africa, and some of the newer European countries. Korea, China, Turkey, and the Middle East region experience increases in consumption. Grain consumption in richer countries is not significantly affected by changes in weather. For vegetable oils, Indonesia and India consume less as do many other regions in Asia and the region Rest of Central America. All regions in Africa (naf, caf, saf), consume less oil. The only major increase in oil consumption occurs in Mexico. Consumption of rice in Indonesia, India, and other regions in Asia is also adversely affected, although Africa escapes largely unscathed in terms of rice consumption. Finally, wheat consumption in the region of Central Africa is expected to fall, as it does in the region of the Rest of the Former Soviet Union. The countries that benefit are India, Rest of Central America and the Rest of South America. In general, consumption in rich countries is not expected to be significantly affected by weather changes; for instance,


consumption in Germany, Canada, the United States and France change little. Japan is expected to consume slightly less rice, while Ireland consumes less of everything.

Figure 5. Imports.

CONCLUSION

Conclusions will be drawn from each of the above three sections with an emphasis placed on the contribution of each technique towards answering the research questions and the section’s contribution to the paper’s overall methodology and conclusions.

The first section described the basic characteristics of the data and strong evidence of increasing temperatures in grids growing ten major food crops was found. Temperature fluctuations across crops follow similar short and long-term patterns and the increases were found to be statistically robust. Results for precipitation showed two distinct periods, a first in


which precipitation fell, followed by a second in which it rose. They also suggest that, in contrast to temperatures, there has been a large amount variation in the amount of precipitation falling in different crop growing grids. The results suggest that if the underlying data generating processes behind the observed increases in temperatures persists, as it has throughout the period studied, then we can expect temperatures in the future to rise as well.

The yields subsection found that yields have been increasing but, in many cases, they have been increasing at decreasing rates, results which suggest that a non-linear term should be included in regressions. Combining the weather and yield data stressed the degree to which relationships between those variables vary across the globe. While many countries have experienced increasing yields and temperatures, a sizable minority has experienced increasing yields and decreasing temperatures.

Panel results indicate that temperatures for wheat, maize, and barley are significant and, in all cases, lead to lower yields given higher temperatures, while increases in precipitation lead to increases in the yields of wheat, maize, and perhaps barley. Estimates for rice and soybeans were insignificant. A Breusch-Pagan test rejected the null hypothesis that intercepts are equal to zero, confirming that pooled models should not be used. Finally, a Hausman test was used to test the null hypothesis that the coefficients estimated by the random effects estimator were statically indistinguishable from those estimated by the within effects estimator. There was no evidence against the null hypothesis that the GLS estimates were consistent, therefore a random effects estimator was found to be preferred to a fixed effects estimator. Standard pooling tests were performed using the sum of squared errors from the pooled and within models. In all cases, an F-type test provided strong evidence that country specific parameters should be included in models. In short, we determined that the effects of changes in temperatures and precipitation have been country specific and, accordingly, countries should be taken as the unit of analysis for purposes of forecasting. This outcome was expected given the previously observed diversity of relationships between yields and weather data. In short, models that do not include country specific characteristics should be avoided.

In general, results indicate that maximum temperatures and precipitation are statistically significant for several crops in several models. Other models were considered, but they did not significantly alter the results of the random effects model. Ignoring standard errors, the results tell a consistent story for models with individual effects, namely, increases in temperatures will reduce yields for all crop with the exception of rice. Results for precipitation were mixed, wheat, maize, and barley yields all increase, while rice and soybeans decrease.

ARIMA models were then used to estimate the effects of changes in weather variables on yields for ten year forecasts for all countries with data and all ten crops in the study. The ‘best’ model for each country and each crop in the analysis was chosen according to the AIC criterion; forecasts from the selected models were then used within MAGNET to simulate weather changes. For many countries the weather variables were found to be statistically insignificant. However, the wide spectrum of countries and crops over which results were found to be significant gives some confidence that the basic ARIMA structure, including the weather terms and the AR and MA terms, is identifying important effects of the weather terms on yields.

The MAGNET model was then shocked using the ARIMA forecasts to determine the effects, ten years into the future, of changes in weather on the global production, trade, and


consumption of four major crop groups, namely, grains, vegetable oils, wheat, and rice. Results for exports closely follow changes in production. Wheat exports from the United States and Argentina are expected to increase slightly, but many other countries in the region South America may experience sharp decreases in exports. Exports from Indonesia, a major exporter of vegetable oils, are expected to fall, as will oil exports from South America. However, oil exports from the United States and Brazil may increase following expected increases in production. All of the major rice exporters can expect slight decreases in exports, while the major wheat producers are expected to export more.

The major importers of grains are not expected to be substantially affected by the changes, although Chinese imports fall slightly. China may also experience a decrease in vegetable oil imports. Of the major importers of rice, the region of Central Africa (caf), which includes Nigeria, were found to be stable, while imports to the region of North Africa (naf) fall and imports into the region of South Africa increase. Changes for wheat importers were in general found to be small, while imports to Egypt (naf), the world’s largest importer, may fall, as they might for Indonesia. Brazilian imports are expected to fall by 1% and 2%, while Chinese imports are expected to increase slightly.

In regards to grain consumption, the biggest losers were found to be people from the region Rest of former Soviet Union (rfsu), India, Central and South America, Central and North Africa, and some of the newer European states. Grain consumption in richer countries was not found to be significantly affected by changes in weather. For vegetable oils, Indonesia and India may consume less as will many other regions in Asia and the region Rest of Central America. All regions in Africa (naf, caf, saf), are expected to consume less oil. The only major increase in vegetable oil consumption may occur in Mexico. Consumption of rice in Indonesia, India, and other regions in Asia is also expected to be adversely affected, although Africa escapes largely unscathed in terms of rice consumption. Finally, wheat consumption in the region of Central Africa is expected to fall, as it does in the region of the Rest of the Former Soviet Union. The countries that benefit are India, the regions Rest of Central America and Rest of South America. In general, rich countries are not significantly affected by weather changes.

The entire analysis brought together several techniques, each with its own strengths and weaknesses. Analyzes which relied primarily on figures or simple regressions of variables through time, while fine for investigating general trends of individual variables, can be deceptive when examining interactions between variables. Using panel models it was possible to quantify the importance of country specific characteristics, but that methodology produces just one general model for all countries. The ARIMA methodology was used to forecast yields ten years into the future, but, given the high aggregation level of the analysis and consequent lack of data, the methodology could not be used to analyze the underlying components determining yields. Furthermore, while the MAGNET model has a global scope, the results it produces are heavily dependent on the acceptance of underlying economic assumptions, and its deterministic nature means that it is impossible to attach statistical confidence to results. However, MAGNET provides unique insight into global economic interactions. For instance, while changes in weather are expected to negatively affect production in both rich and poor countries, consumption in rich countries will suffer less from changes in consumption due, in large part, to the relatively small size of that sector in the entire economy. Finally, we asked a very limited set of data to do a lot of work. In our favor, we limited the forecast period to just ten years, a relatively small number of years in


comparison to many CGE experiments, and our shocks were modest. We have attempted to play to the strengths of each method, while being forthright throughout the text concerning their respective limits.

Mindful of the above qualifications, in terms of the main research question, we conclude that changing weather patterns will have significant effects on the trade, and, in some cases, the consumption of major food crops. The effects will be both country and crop specific. As for the ancillary question concerning the relative effects of weather on richer versus poorer countries, results were mixed, but indicate that consumption in richer countries, with few exceptions, will be largely unaffected by changing weather patterns, while consumption in many poorer regions will decrease.

REFERENCES

Anglia Climate Research Unit (CRU). NCAS British Atmospheric Data Centre. Arvis, J., Duval, Y., Shepherd, B., and Utoktham, C. (2013). Trade Costs in the Developing

World: 1995-2010. World Bank. Baltagi, B. H. (2008). Econometric Analysis of Panel Data. John Wiley and Sons, Ltd.

Bond, S. (2002). Dynamic panel data models: A guide to micro data methods and practice. The Institute for Fiscal Studies Department of Economic, UCL cemmap working paper CWPO9/02.

Bradshaw, B., Dolan, H., and Smit, B. (2004). Farm-level adaptation to climatic variability and change: Crop diversification in the Canadian prairies. Climatic Change, 67:119–141.

Cutter, S. L., Emrich, C. T., Webb, J. J., and Morath, D. (2009). Social vulnerability to climate variability hazards: A review of the literature. Staff report no. ages 9049, Hazards and Vulnerability Research Institute, Department of Geography, University of South Carolina, Columbia, SC 29208.

Dixon, P. B., Parmentar, B. R., Ryland, G. J., and Sutton, J. (1977). ORANI: A General Equilibrium Model of the Australian Economy. First progress report of the impact project, vol.2. Canberra, Australian Government Publishing Service.

Erda, L., Wei, X., Hui, J., Yinlong, X., Yue, L., Liping, B., and Luyong, X. (2005). Climate change impacts on crop yield and quality with co2 fertilization in china. Philosophical Transactions of the Royal Society, Biological Sciences, 360(1463):2149–54.

FAOSTAT (2013). United Nations Food and Agricultural Organization (UN-FAO) online statistics database. http://faostat.fao.org/site/291/default.aspx.

Frees, E. W. (2004). Longitudinal and Panel Data: Analysis and Applications in the Social Sciences. Cambridge University Press.

Frees, E. W. and Miller, T. W. (2004). Sales forecasting using longitudinal data models. International Journal of Forecasting, 20:99–114.

Gebrehiwot, T. and Veen, A. v. (2013). Farm level adaptation to climate change: The case of farmer’s in the Ethiopian Highlands. Environmental Management, 52:29–44.

Greene, W. H. (2007). Econometric Analysis. Prentice Hall, sixth edition. Greenstone, M. and Deschenes, O. (2006). The economic impacts of climate change:

Evidence from agricultural profits and random fluctuations in weather. FEEM Working Paper No. 6.


GTAP (2012). GTAP 8 data base description. https://www.gtap.agecon.purdue.edu/databases/ v8/v8_doco.asp. Hamilton, J. D. (1994). Time Series Analysis. Princeton University Press.

Hertel, T. W., editor (1997). Global Trade Analysis: Modeling and Applications. Cambridge University Press.

Hertel, T. W., Martin, W. J., and Leister, A. M. (2010). Potential implications of a special safeguard mechanism in the WTO: the case of wheat. Policy research working paper series 5334, The World Bank.

Heston, A., Summers, R., and Aten, B. (2011). Penn world trade version 7.0. Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania.

Hsiao, C. (2003). Analysis of Panel Data. Cambridge University Press, New York, second edition.

IFPRI (2009). Climate change: Impact on agriculture and costs of adaptation. food policy report. Technical report, IFPRI. Jones, P. and Harris, I. (2008). Cru time series (ts) high resolution gridded datasets. Technical report, University of East.

Josling, T., Anderson, K., Schmitz, A., and Tangermann, S. (2010). Agricultural products: One hundred years of contributions by agricultural economists. American Journal of Agricultural Economics, 92(2):424–446.

Leff, B. N., Ramankutty, and Foley, J. A. (2004). Geographic distribution of major crops across the world. Global Biogeo- chemical cycles, 18.

Licker, R., Johnston, M., Foley, J. A., Barford, C., Kucharik, C. J., Monfreda, C., and Ramankutty, N. (2010). Mind the gap: how do climate and agricultural management explain the ’yield gap’ of cropland around the world? Global Ecology and Biogeography, 19(6):769–782.

Lobell, D. B. and Field, C. B. (2007). Global scale climate–crop yield relationships and the impacts of recent warming. Environmental Research Letters, 2.

Lobell, D. B., Schlenker, W., and Costa-Roberts, J. (2011). Climate trends and global crop production since 1980. Science, 333(6042):616–620.

Luers, A. L., Lobell, D. B., Sklar, L. S., Addams, C. r., and Matson, P. A. (2002). A method for quantifying vulnerability, applied to the agricultural system of the Yaqui Valley, Mexico. Global Environmental Change, 13(4):255–267.

Moïsé, E., Delpeuch, C., Sorescu, S., Bottini, N., and Foch, A. (2013). Estimating the constraints to agricultural trade of developing countries. Trade policy papers no. 142, OECD.

National Research Council (2011). America’s climate choices. Technical report, The National Academic Press.

Nelson, G. C., Valin, H., Sands, R. D., Havlik, P., Derying, D., Elliot, J., Fujimori, S., Hasegawa, T., Heyhoe, E., Kyle, P., von Lampe, M., Lotze-Campen, H., Mason d’Croz, D., van Meijl, H., van der Mensbrugghe, D., Müller, C., Popp, A., Robertson, R., Robinson, S., Schmid, E., Schmitz, C., Tabeau, A., and Willenbockel, D. (2013a). Climate change effects on agriculture: Economic responses to biophysical shocks. PNAS. Epub ahead of printing.

Nelson, G. C., van der Mensbrugghe, D., Blanc, E., Calvin, K., Hasegawa, T., Havlik, P., Kyle, P., Lotze-Campen, H., von Lampe, M., Mason d’Croz, D., van Meijl, H., Müller, C., Reilly, J., Robertson, R., Sands, R. D., Schmitz, C., Tabeau, A., Takahashi, K., and


Valin, H. (2013b). Agriculture and climate change in global scenarios: why don’t the models agree. Agricultural Economics, 45(1):85–101.

Nerlove, M. (1971). Further evidence on the estimation of dynamic economic relations from a time-series of cross-sections. Econometrica, 39:359–382.

Pandey, S., Bhandari, H., and Hardy, B. (2007). Economic costs of drought and rice farmers’ coping mechanisms: a cross- country comparative analysis. Technical report, IRRI (International Rice Research Institute.

Powell, J. and Rutten, M. (2013). Convergence of European wheat yields. Renewable and Sustainable Energy Reviews, 28:53–70.

Powell, J. P. (2015). Productivity implications of extreme precipitation events: the case of Dutch wheat farmers. Resubmitted to Climatic Change, 2015.

Prato, T., Zeyuan, Q., Pederson, G., Fagre, D., Bengston, L. E., and Williams, J. R. (2010). Potential economic benefits of adapting agricultural production systems to future climate change. Environmental Management, 45(1):577–89.

Reilly, J., Hohmann, N., and Kane, S. (1994). Climate change and agricultural trade: Who benefits, who loses? Global Environmental Change, 4(1):24–36.

Robinson, S. (1990). Analysing agricultural trade with single-country computable general equilibrium models. In Goldin, I. and Knudsen, O., editors, Agricultural Trade Liberization. Paris:OECD.

Robinson, S., Kilkenny, M., and Hanson, K. (1990). The USDA/ERS computable general equilibrium (CGE) model of the United States. Staff report no. ages 9049, Economic Research Service, U.S. Department of Agriculture.

Rutten, M. M., Chant, L. J., and Meijerink, G. W. (2011). Sit down at the ballgame: How trade barriers make the world less food secure. Available at SSRN: “http://ssrn.com/ abstract=1769745” or “http://dx.doi.org/10.2139/ssrn.1769745”.

Sacks, W. J., Derying, D., Foley, J. A., and Ramankutty, N. (2010). Crop planting dates: an analysis of global patterns. Global Ecology and Biogeography, 19:607–620.

Schlenker, W. and Roberts, M. J. (2009). Nonlinear temperature effects indicate severe damages to U.S. crop yields under climate change. Proceedings of the National Academy of Sciences of the United States of America (PNAS), 106(37):15594– 15598.

Sims, R., Taylor, M., and Bioenergy, I. (2008). Biofuels technologies: An overview of current industry and R &D activities. Sonka, S. (1991). Methodological guidelines for assessing the socio-economic impacts of climate change on agriculture. In Climate Change: Evaluating the Socio-Economic Impacts, pages 21–45. OECD, Paris.

Swamy, P. and Arora, S. (1972). The exact finite sample properties of the estimators of coefficients in the error components regression models. Econometrica, 40:261–275.

Tabeau, A. and Woltjer, G. (2010). Modelling the agricultural employment development within the CGE framework: the consequences for policy responses. In Trade for Sustainable and Inclusive Growth and Development, Bangkok, Thailand. Thirteenth Annual Conference on Global Economic Analysis.

Tamiotti, L., Teh, R., Kulaçglu, V., Olhoff, A., Simmons, B., and Abaza, H. (2009). Trade and climate change. A report by the united nations environment programme and the world trade organization, World Trade Organization.

Valin, H., Sands, R. D., van der Mensbrugghe, D., Nelson, G. C., Ahammad, H., Blanc, E., Bodirsky, B., Fujimori, S., Hasegawa, T., Havlik, P., Heyhoe, E., and Kyle, K.


(2014). The future of food demand: Understanding differences in global economic models. Agricultural Economics, 41(1):51–67.

van Meijl, H., van Rheenen, T., Tabeau, A., and Eickhout, B. (2006). The impact of different policy environments on land use in Europe. Agriculture, Ecosystems and Environment, 114(1):21–38.

van Wijk, M. T., Rufino, M. C., Enahoro, D., Parsons, D., Silvestri, S., Valdivia, R. O., and Herrero, M. (2012). A review on farm household modelling with a focus on climate change adaptation and mitigation. Working Paper 20, CGIAR Research Program on Climate Change(CCAFS), Copenhagen, Denmark.

Verburg, P., Eickhout, B., and van Meijl, H. (2008). A multi-scale, multi-model approach for analyzing the future dynamics of European land use. Annals of Regional Science, 42(1):57–77.

WageningenUR (2012). WageningenUR site. http://www.wageningenur.nl/en/About-Wa geningen-UR.htm.


CHANGING TECHNICAL, ALLOCATIVE, AND ECONOMIC PRODUCTION EFFICIENCY OF SMALL-SCALE FARMERS IN INDONESIA:

THE CASE OF SHALLOT PRODUCTION

Sujarwo, Michael R. Reed, and Sayed H. Saghaian Department of Agricultural Economics

University of Kentucky, Lexington, KY, USA

ABSTRACT

This study investigates the efficiency of Indonesian small-scale shallot farmers. It uses an efficiency measurement approach that handles situations when there is little variation in input and output prices. The changing production efficiency of small-scale farmers is measured during two periods: 2005 and 2012. The empirical results confirm that there is a 2.8% improvement in technology between 2005 and 2012. However, the increase in technical efficiency (from 0.84 in 2005 to 0.85 in 2012) was not statistically significant. Allocative efficiency and economic efficiency significantly decreased in 2012 from 0.76 and 0.64 in 2005 to 0.68 and 0.58 in 2012, respectively.

Keywords: stochastic frontier production function, technical efficiency, allocative efficiency, economic efficiency

JEL Codes: Q12

INTRODUCTION

This study measures production efficiency levels (technical, allocative, and economic efficiency) of shallot production. Shallots are a vegetable that have a short growing period and are attractive to small-scale farmers because of their high potential profit and low-level of required technology. Shallot production takes less than three months (60-80 days) from planting to harvest. Indonesian farmers produce shallots during three production periods with planting in February/March, May/June, and August/September. The major harvest season is July/August each year for planting time in May/June. This is the peak season for shallot production because of adequate water availability and relatively optimum weather, which greatly improve shallot yields. According to Statistics Indonesia (2013), shallot yields in East-Java province Indonesia were 9.98 tons/hectare for 2012.

Sujarwo, Michael R. Reed, and Sayed H. Saghaian 32

Shallots, however, are vulnerable to water shortages, weather changes, plant pests and diseases, and their yield is also responsive to fertilizer application. Shallot farmers also face high price volatility for output and inputs caused by local supply and demand conditions, exchange rates (pesticides and fertilizers are imported), and distribution problems. These factor changes (technical and economic factors) could reduce the production efficiency of the farmers because their observations and accumulated knowledge are less relevant in this new environment. The effect of these changes in technical and economic factors on efficiency levels is an interesting research question, which is examined in this paper.

Estimating a proper production function is essential to this analysis. The production function is estimated using three steps (Koopmans 1951, Afriat 1951). The first step estimates the production function for each year using OLS; then, we test whether negative skewness exists in the model (Meeusen and Broeck 1977, Aigner et al. 1977, Schmidt and Lovell 1979, Førsund et al. 1980, Jondrow, et al. 1982). The results show that we have enough proof that there is negative skewness, which indicates the existence of technical inefficiency in the model. Accordingly, we must depart from OLS to an MLE estimator for the frontier production function (Aigner et al. 1977). The existence of negative skewness possibly affects the expected value of error term (no longer zero); therefore, the intercept is biased upward (Satchachai and Schmidt 2010).

The second step is to evaluate the properties of the frontier production function, especially related to concavity with respect to inputs (Fare and Lovell 1978, Just and Pope 1978, Coelli et al. 2005). Multicollinearity has affected coefficient estimates greatly in the past studies and caused incorrect signs for input coefficients (Farrar and Glauber 1967, Grapentine 1997). Principal Component Analysis (PCA) is applied in the third step for the MLE frontier production function to deal with harmful multicollinearity. PCA is a very powerful approach to overcome multicollinearity problems. As a result, we are able to estimate a proper production function which is concave in inputs. After fitting a proper production function with well-behaved input coefficients, the cost frontier function is derived based on the invariance property of MLE.

This study relies on estimating a proper frontier production function which satisfies the diminishing marginal productivity of the inputs to determine the efficiency levels in the shallot production. Previous approaches focused on estimating the frontier cost function (Schmidt and Lovell 1979, Førsund, et al. 1980, Kopp and Diewert 1982, Bravo-Ureta and Pinheiro 1997). This study provides an alternative approach generating efficiency levels (technical, allocative, and economic efficiency) for cross-sectional data, which have limited price variations.

Finally, technical efficiency (TE), allocative efficiency (AE), and economic efficiency (EE) are used to examine changes in efficiency levels for Indonesian shallot farmers in 2005 and 2012. The results are used to examine Schultz’s hypothesis of “poor but efficient” small-scale farmers in traditional agricultural settings.

Changing Technical, Allocative, and Economic Production Efficiency … 33

CONCEPTUALIZATION

Stochastic Frontier Production Function (SFPF) and Technical Inefficiency (TE)

TE is measured based on Aigner et al. (1977) by decomposing disturbance terms into two components. The first of these is a normal random error (v), and the second is technical inefficiency (u). For as input coefficients and as inputs, the stochastic frontier can be specified as equation (1). Equation (2) and equation (3) are disturbance term ( ) and probability distribution function of , respectively.

= ( ; ) exp( ) = 1,2, … , (1)

= − ~ (0, ); ~ (0, ); (2) ( ) = 2 ∗ 1 − ∗( ) − ∞ +∞

(3) The best practical production is ∗ when there is no technical inefficiency ( = 0); then,

efficient production will be ∗ = ( ; )exp ( ). This function is called the stochastic frontier production function (SFPF). Additionally, ∗(. ) and ∗(. ) are PDF and CDF, respectively, of the normal distribution function. If = 0 there is no technical inefficiency and all deviation from the production frontier is due to uncontrollable factors (v).

Coelli et al. (2005) described the output-oriented measure of technical efficiency of firm-i ( ) as:

= ∗ = ( ; ) exp( − )( ; ) exp( ) = exp(− )

(4) A decomposition of is required to obtain . Jondrow et al. (1982) came up with the

idea of conditional distribution of inefficiency, ( | ), which contains all the information about at given . Therefore, ( | ) = .

Jondrow et al. (1982) defined technical inefficiency as the value of at given observed output .

= ∗| = ∗ + ∗ ∗ ∗ F ∗ ∗

(5)


where (. )and (. ) are normal CDF and PDF, respectively, ∗ = −(ln − ) / and ∗ = / . Equation (5) holds when the inefficiency component ( ) is a half normal distribution. Battese and Coelli (1988) derived the predictor of technical efficiency (TE) as follows:

= exp (− )| = F ∗∗ − ∗ /F ∗∗ ∗2 − ∗

(6) Battese and Corra (1997) specified a log-likelihood function to develop the estimator

(MLE) for the parameters as follows: ln ( /| , , ) = − 2 2 + ln 1 − Φ 1 − − 12

(7) The MLE are consistent and asymptotically efficient (Aigner et al., 1977). This study uses FRONTIER 4.1 package software developed by Coelli to obtain

parameter estimates and technical efficiency (TE). Several hypothesis tests are conducted. They are (1) to test individual input coefficients ( s); and (2) to test the existence of TE using the Likelihood Ratio (LR). The LR, calculated as = −2 ln ( ) − ( ) . ( ), is the restricted log-likelihood assuming no technical inefficiency, and ( ) is the unrestricted log-likelihood. Coelli et al. (2005) explain that the LR test uses the Kodde and Palm critical values (Kodde and Palm, 1986), which is 2.71 or 1.642 for 5% or 10% level, respectively, and df=1 (the number of restrictions).

Koop and Smith (1980) found that the level of productive efficiency is more dependent on the frontier estimator used than on the functional form of the production frontier. Having that information, the econometric model preferred in this study is a double-log production function. The reasons are that the functional form has special properties, such as a homogenous function and input coefficients as elasticities. Moreover, coefficient inputs of this function also represent the factor shares assuming the farmers are at the efficient level of production.

The double-log frontier production function is specified in equation (8). There are 17 explanatory variables, which consist of 8 production inputs, 1 dummy variable for year (D2012 = 1, D2005 = 0), and 8 interaction variables, which are the interaction of dummy and input variables.

= ( + + ( 1 ∗ ) + 1 + ) = 1, … , 76 (8)

where is shallot production, is land used for shallot production, is seed used in shallot production, and is labor use. Those three inputs are considered as primary inputs in the


shallot production because they are required for production to take place. Supported inputs are represented by fertilizers, pesticides, and fungicides. Fertilizer consists of 3 variables which are nitrogen use (N), phosphate use (P), potassium use (K); is insecticide use and is fungicide use. D1 is dummy time identifying 2012 and = − is the error term.

Maximum Likelihood Estimator (MLE) and Principal Component Analysis (PCA)

Principal component analysis (PCA) is an effective approach to deal with multicollinearity. This approach eliminates the correlation among independent variables through the transformation of the variables ( ) into new orthogonal variables ( ). First, the vector inputs (logarithmic form for this case), , are transformed into standardized data, ,

(mean=0 and variance=1) to avoid the effect of different units among the variables. = − = 1,2,3, … , =

(9) Secondly, eigenvalues and eigenvectors are calculated to generate principal components

scores. These eigenvalues and eigenvectors satisfy equation − = 0, where I is the square identity matrix, is the eigenvalue for principal component, and are eigenvectors of the principal component. The first principal component ( ) has the highest eigenvalue, which explains the most variation in the original data. The second principal component has the highest eigenvalue from the remaining variation. This process is continued until all of the variation in the original data set is captured by the principal components. This is shown in equation (10):

= (10) Instead of using equation (8) for the SFPF, equation (11) is used in the existence of

harmful multicollinearity: = + + − (11) The original input elasticities/coefficients are obtained by employing the invariance

property of the MLE and transforming the coefficients on principal components back to the . If there is as a MLE for , then for as some function, we have ( ) as a MLE for ( ), which possesses a one-to-one relationship (Zehna, 1966):


= + + − (12) for = − ; =

The variance of the original coefficient, and , is

( ) = = ( )

(13) ( ) = ( ) + ( ) ( )

(14) There are three steps for obtaining the efficiency measures using PCA-MLE. The first is

to choose how many principal components (P) will be used in the stochastic frontier model. Fekedulegn et al. (2002) suggest that the principal components explain at least 85% of the original data variation. The second step is to transform coefficients of P variables back to the input coefficients. The third step uses the frontier production function to construct the frontier cost function.

Factor Share and Input Elasticity

This study uses a double-log frontier production function. Considering the profit maximization assumption and first order condition (FOC), we find that the value of marginal product ( = . ; Py is output price) is equal to its input price ( ). Then,

= .. = … = ℎ

(16) Assuming that producers maximize profit and that the production function is in the

double-log form, the input coefficient should be equal to the factor share. This study tests statistically whether the input coefficient is equal to the factor share using a t-test.


Deriving Economic Efficiency (EE) and Allocative Efficiency (AE)

Significant differences between the coefficients and the factor shares indicate inefficiency (EE and AE) in shallot production. Economic efficiency (EE) is derived from the frontier cost minimization (dual approach). Instead of direct estimation of the frontier cost function, the frontier cost function in this study is generated from the frontier production function. This step exploits the invariance principle of MLE for frontier production function coefficients.

Suppose a production function incorporating technical inefficiency (a simplification of equation (8)) is

= … . . Where ~ (0, ) and ~ ℎ (17) Using the dual, we have a constrained cost minimization problem as: = . + − . .

(18) Getting the first order condition (FOC) by taking the first derivative of the Lagrangian

function ( ) with respect to Xi and Xj for j ≠ i: // = = ∗∗ = (19)

where and , ∈ m for = 1,2,3, … , 17 . Equation (19) assumes that there is no inefficiency in input allocations. If cost minimization does not hold due to inefficiency (Schmidt and Lovell 1979); then,

= ⇒ =

(20) Substituting into the production function, one obtains the demand for Xj at a given

production level:

= ( )

(21)


Since = ; then:

= ( )

(22) The cost function given technical and allocative inefficiency is specified as follow: = +

(23)

= ∑ ( ) +

If the firm is perfectly efficient in its operation (u=0 and =0), then:

∗ = ∑ ( )

(24) Based on equation (15) and equation (16), one can measure economic efficiency as: = ∗ = (∑ ) +

(25) Technical inefficiency (u) is obtained from the frontier production function estimation

and i is obtained from − . If technical efficiency (TE) and economic

efficiency (AE) are known allocative efficiency is the ratio of EE and TE (Farrel 1957).

DATA DESCRIPTION

This study uses cross sectional data collected from survey in Ngadiboyo Village, East Java Province, Indonesia, in 2005 and 2012. The sample farmers were selected using simple random sampling of over 650 shallot farmers. The data collected in 2005 and in 2012 were


purposively selected from the same period of production (May-August). The samples in dataset 2005 and dataset 2012 are independent and drawn from the same target population.

The formula specified by Parel et al. (1973) was used to determine the sample size. The variables employed in the formula are the target population, which is 650 shallot farmers; the farmers estimated homogeneity level, which is 90% because most have less than one hectare of land and they apply the same production techniques; the significance level, which is 5%; and the maximum sampling error, which is 10%. The representative sample size based on this formula is 32.82 farmers. This study has 35 observations in 2005 and 35 observations in 2012. Therefore, the samples are representative of the target population for each year.

Data collected from the individual farmers were output price, input prices, and physical amount of output and inputs employed in shallot production. Unfortunately, there is little variation in several input prices because the farmers purchase from the same suppliers. The wage of hired labor for instance was identical among farmers within the same year. The other inputs, such as seed and land rent are quite homogenous as well. Therefore, prices of primary inputs (land, seed, and labor) use averaged prices for each grower within a year.

Table 1. Summary Statistics for Dataset 2005 and 2012

Inputs Unit Average Standard deviation

Minimum Maximum

Dataset 2005

Output kg 2037.03 915.67 624.00 5000.00

Land m2 1706.67 723.81 706.67 3333.33

Productivity ton/hectare 11.94 1.93 7.80 15.00

Seed kg 153.80 64.94 47.00 300.00

Labor man days 126.32 50.51 41.49 246.60

Nitrogen kg 31.07 25.39 7.90 144.72

Phosphate kg 22.83 13.93 4.32 56.72

Potassium kg 15.18 8.94 3.17 39.20

Insecticide gram 676.65 697.82 30.72 2482.67

Fungicide ml 474.20 316.81 68.85 1249.17

Dataset 2012

Output kg 3954.29 1770.40 1600.00 7500.00

Land m2 2421.43 1145.28 1000.00 5500.00

Productivity ton/hectare 16.62 2.36 10.91 21.43

Seed kg 254.71 119.72 100.00 500.00

Labor man days 109.11 41.38 62.50 249.50

Nitrogen kg 58.92 25.76 15.70 135.00

Phosphate kg 39.14 17.52 8.00 70.00

Potassium kg 21.34 13.33 7.80 54.00

Insecticide gram 3179.21 2815.11 339.46 13802.46

Fungicide ml 563.54 338.96 116.00 1415.22


Table 2. Input Prices and Prices Changes (Rupiah)

Price Unit 2005 2012 Price Increase (%) Land m2 350.00 1,071.00 206.00 Seed kg 5,000.00 13,000.00 160.00 Labor labor day 10,000.00 40,000.00 300.00 Nitrogen kg 3,443.00 8,297.00 140.98 Phosphate kg 3,965.00 10,957.00 176.34 Potassium kg 4,225.00 14,352.00 239.69 Insecticide gr 150.00 850.00 466.67 Fungicide ml 387.00 725.00 87.34 Output rp/kg 2200.00 3350.00 52.27

Note: One US dollar is approximately 10,000 Indonesian rupiahs. Shallot farmers employ fertilizers and pesticides with many different formulations and

brand names. Fertilizers were transformed into macro-nutrient equivalents (N, P, and K). Insecticides and fungicides are transformed based on the assumption that price reflects the formulation’s strength. For instance, an insecticide selling for double the price of the base formulation was assumed to be double the strength in calculating the grams of insecticides used. The same was true for fungicide. Table 1 provides the descriptive statistics of the data for both dataset 2005 and 2012 including physical unit of inputs, output, and the yield. Table 2 presents prices for inputs and output from the sample.

The average shallot yield was 11.94 tons/hectare in 2005 and 16.62 tons/hectare in 2012. The standard deviation of the yield in 2012 was 2.36; higher than in 2005, 1.93. The shallot yields in the study area are much higher than the average for East-Java Province, which was 9.98 tons/hectare (Statistics Indonesia 2013).

The average farm size for shallots was 0.24 hectare in 2012, an increase of 41.9% compared to 2005 (only 0.17 hectare). Among other inputs used for shallot production, almost all (seed, fertilizers, and pesticides) increased along with increasing farm size (including seed even though the seed price increased by 160% (Table 2)). However, labor use decreased in 2012 by 13.5%, which was likely caused by the 300% increase in wages. Nitrogen had the highest percentage increase among fertilizers (89.6%). Nitrogen fertilizer boosts the vegetative phase of the shallot, but that attracts insects and fungi, such as Spodoptera litura, Spodoptera exiqua, and Alternaria porri.

EMPIRICAL RESULTS

Estimation of the Frontier Production Function in 2005 and 2012

Using cross-sectional data and a limited number of observations exacerbate the problem of estimating the production function and calculating efficiency measures. The result of an OLS estimation of the production function using the shallot product data confirms this assertion (Table 3). The variance inflation factors (VIFs) indicate serious multicollinearity since land, seed, labor, and all dummy variables (slope and intercept) have VIFs more than


10. The other symptom is a high coefficient of determination (R2) but only 3 significant coefficients at the 5% level.

Table 3. Regression of Cobb-Douglas Production Function Using OLS

Variable Parameter Estimate

Standard t-value

Variance Inflation Error

Intercept 2.38855 ** 1.0751 2.2200 0.0000 Land (m2) 0.30474 0.3502 0.8700 101.1123 Seed (kg) 0.31037 ** 0.1230 2.5200 16.6121 Labor (man days) 0.02107 0.3339 0.0600 62.4457 Nitrogen (kg) 0.12200 0.0613 1.9900 6.5828 Phosphate (kg) 0.08865 0.0654 1.3600 7.9551 Potassium (kg) 0.00250 0.0595 0.0400 5.7758 Insecticide (gram) 0.04274 0.0247 1.7300 4.2439 Fungicide (ml) 0.05818 0.0447 1.3000 3.5944 D1* Land 0.02616 0.3963 0.0700 8993.0619 D1* Seed 0.08698 0.2067 0.4200 1229.2075 D1* Labor 0.48249 0.3747 1.2900 2927.7894 D1* Nitrogen -0.24601 ** 0.1118 -2.2000 195.3544 D1* Phosphate -0.14951 0.0924 -1.6200 107.9166 D1* Insecticide 0.04450 0.0855 0.5200 63.4440 D1* Insecticide -0.04295 0.0438 -0.9800 113.3435 D1* Fungicide -0.09785 0.0715 -1.3700 189.4161 D1 -0.41678 1.1821 -0.3500 1343.8914 R2 = 0.9631 R2

adj = 0.9517 F-test = 84.49 RMSE = 0.12558 1) Natural logarithmic form, *** Significant at the 0.01 level, ** Significant at the 0.05 level.

An evaluation of the disturbance terms shows that the mean, standard deviation, and

skewness are 0.000, 0.117, and -0.298, respectively. Several normality tests are conducted using SAS 9.1 statistical package software (Table 4) yet none of the tests are found to be significantly different from zero so normality is not rejected. However, there is a long left tail in the histogram (Figure 1), indicating negative skewness or technical inefficiency in the production function. So, a production frontier incorporating technical inefficiency as proposed by Schmidt and Lovell (1979) and Jondrow, et al (1982) is pursued. Then, the LR test is used to check for technical inefficiency in the model.

Table 4. Descriptive Statistics and Normality Tests for Disturbance Terms

Descriptive statistics

Size No Normality test Notation Statistic Probability

Mean 0.0000 1 Shapiro-Wilk W 0.9844 0.537 Std. Deviation 0.1171 2 Kolmogorov-Smirnov D 0.0674 0.150 Skewness -0.2982 3 Cramer-von Mises W-Sq 0.0657 0.250 Kurtosis 0.2858 4 Anderson-Darling A-Sq 0.3966 0.250


Figure 1. Histogram of OLS Residuals.

Table 5. Cobb-Douglas Frontier Production Function Using MLE

Variable Parameter Estimate

Standard t-value Error

Intercept 0.1439 0.4848 0.2969 Land (m2) 1.0427 *** 0.1929 5.4063 Seed (kg) 0.0180 0.0940 0.1919 Labor (man days) -0.2078 -0.1855 1.1202 Nitrogen (kg) 0.1036 ** 0.0429 2.4161 Phosphate (kg) 0.0026 0.0362 0.0724 Potassium (kg) 0.0180 0.0503 0.3588 Insecticide (gram) 0.0048 0.0181 0.2679 Fungicide (ml) 0.0605 0.0351 1.7238 D1* Land -0.9243 *** -0.2603 3.5511 D1* Seed 0.3157 ** 0.1508 2.0934 D1* Labor 0.8692 *** 0.2174 3.9979 D1* Nitrogen -0.1283 0.0803 -1.5985 D1* Phosphate -0.0611 0.0601 -1.0157 D1* Insecticide 0.0489 0.0663 0.7376 D1* Insecticide 0.0116 0.0343 0.3390 D1* Fungicide -0.0243 0.0554 -0.4383 D1 2.1375 *** 0.6438 3.3198 Sigma-squared 0.0328 *** 0.0027 12.2614 Gamma 0.999995 *** 0.00001 97267.0250 LR test of the one-sided error = 21.6274

Notes: 1) Natural logarithmic form, *** Significant at the 0.01 level, ** Significant at the 0.05 level.

Curve: Normal(Mu=2E-16 Sigma=0.1171)

Percent

0

5

10

15

20

25

30

35

resd2

-0.40 -0.32 -0.24 -0.16 -0.08 0.00 0.08 0.16 0.24


The SFPF is estimated using MLE (as proposed by Aigner et al. 1977). The results are presented in Table 5. The Likelihood Ratio test for one-sided skewness is 21.63, which is greater than critical value of 2.71 at the 5% level (Kodde and Palm), therefore, it is concluded that the SFPF is preferred to the OLS estimation because of technical inefficiency. Yet the SPPF MLE seems to have a problem with multicollinearity as well. Among 18 coefficient estimates, only five are statistically significant at the 5% level. Moreover, the coefficient for labor is negative in 2005 and the coefficients for nitrogen and phosphate are negative in 2012. Violating the concavity condition for inputs prohibits further efficiency calculations so PCA is applied to transform the highly correlated dataset into an orthogonal transformed dataset.

Table 6 shows the eigenvalues and the proportion by which the variation of the original dataset is explained by the PCA. This study employs four principal components, which explain 93.3% of the variation in the original dataset. The results of SFPF using PCA-MLE confirm that the first two principal components, P1 and P2, are significant at the 1% level, while P3 is significant at the 5% level (Table 7). Even though P4 is not significant at the 5% level, P4 is a critical component because including P4 affects the significance of the LR test, which is used to test the existence of technical inefficiency. As a result, the LR test verifies that there is skewness in the residuals and technical inefficiency exists.

The final step of SFPF estimation is shown in Table 8. Among 18 coefficients, there are two coefficients (for potassium and insecticide), which are not statistically significant at the 5% level but both are significant at the 10% level. More importantly, the model is concave in X, so all coefficients satisfy the law of diminishing marginal productivity, making the PCA-MLE preferable to MLE. A dummy variable is applied to differentiate between data 2005 and 2012. The dummy variable is used as a slope shifter, allowing input productivity to differ between the two production periods, and as an intercept shifter, allowing neutral technological progress (Ismail, 1981). The coefficient for the dummy intercept is 0.028 and is significantly different from zero at the 1% level, which means technological progress increased production by 2.8% between 2005 and 2012. This technological progress could come from better timing decisions for input applications and other experience-related production practices.

Table 6. Eigenvalues and Eigenvector

Principal Component

Eigen value

Propor-tion

Cumu-lative

Principal Component

Eigen-value

Propor-tion

Cumu-lative

1 11.14 0.66 0.66 10 0.02 0.00 1.00

2 3.67 0.22 0.87 11 0.01 0.00 1.00

3 0.57 0.03 0.90 12 0.01 0.00 1.00

4 0.49 0.03 0.93 13 0.01 0.00 1.00

5 0.46 0.03 0.96 14 0.00 0.00 1.00

6 0.27 0.02 0.98 15 0.00 0.00 1.00

7 0.21 0.01 0.99 16 0.00 0.00 1.00

8 0.10 0.01 0.99 17 0.00 0.00 1.00

9 0.05 0.00 1.00


Table 7. Frontier Production Function Using PCA-MLE

Variable Parameter Standard t Value Intercept 8.0312 *** 0.0289 278.0290 Prin1 0.1360 *** 0.0048 28.5454 Prin2 0.1493 *** 0.0081 18.5323 Prin3 0.0476 ** 0.0191 2.4987 Prin4 -0.0390 0.0228 -1.7122 Sigma-squared 0.0517 *** 0.0134 3.8438 Gamma 0.9156 *** 0.0740 12.3741 LR test of the one-sided error = 3.4988

Notes:1) Using three principal components explain 93.32% of total variance the original data All of the slope dummy variables are significant at 1% level and have positive signs,

which show that there are increased input productivities in 2012 compared to 2005. Nitrogen fertilizer and insecticide are the inputs with the highest productivity increase. Those input elasticities increase 1.13% and 1.28%, respectively.

Table 8. Cobb-Douglas Frontier Production Function Using PCA-MLE

Variable Parameter

Estimate Standard t-value Error

Intercept 2.8783 *** 0.15979 18.0130 Land (m2) 0.1946 *** 0.01040 18.7092 Seed (kg) 0.1573 *** 0.00805 19.5345 Labor (man days) 0.2250 *** 0.01338 16.8229 Nitrogen (kg) 0.1257 *** 0.01686 7.4526 Phosphate (kg) 0.0617 *** 0.01393 4.4281 Potassium (kg) 0.0421 0.02173 1.9360 Insecticide (gram) 0.0145 0.01104 1.3132 Fungicide (ml) 0.1297 *** 0.02052 6.3188 D1* Land 0.0050 *** 0.00054 9.1774 D1* Seed 0.0076 *** 0.00072 10.5403 D1* Labor 0.0085 *** 0.00087 9.7567 D1* Nitrogen 0.0113 *** 0.00129 8.7462 D1* Phosphate 0.0098 *** 0.00144 6.8173 D1* Insecticide 0.0128 *** 0.00161 7.9133 D1* Insecticide 0.0040 *** 0.00054 7.3599 D1* Fungicide 0.0074 *** 0.00084 8.8573 D1 0.0280 *** 0.00416 6.7343 Sigma-squared 0.0517 *** 0.01345 3.84380 Gamma 0.9156 *** 0.07399 12.37406

Notes: 1) Natural logarithmic form 2) Using three principal components explain 93.32% of total variance the original data *** Significant at the 0.01 level. ** Significant at the 0.05 level.


Primary inputs are those absolutely necessary for agricultural product. They are land, seed, and labor. This study finds that those inputs have the highest elasticities, which are 0.195, 0.157, and 0.225 in 2005 and 0.199, 0.165, and 0.234 in 2012, for land, seed and labor, respectively. Among the primary inputs, labor has the highest elasticity. Labor has various roles in the production process starting from pre-planting and ending with harvesting. Additionally, land is obviously another essential factor and has the second highest elasticity. The total elasticity of the primary inputs is 0.598, which means that increasing primary inputs by 10% holding others constant (ceteris paribus) potentially increase shallot production by 5.98%.

Fertilizers and pesticides are considered supported inputs. Nitrogen and fungicide have the greatest elasticity among supported inputs, which are 0.137 for both inputs. The best time for shallot planting is from the middle of May to the beginning of June, which is the production period observed in this study. The months following those planting times have optimum sunshine and water. This combination of a favorable environment and optimum nitrogen fertilizer use can boost production significantly. Unfortunately, fungi and worms also attack the crop during this production period. The growth of Alternaria porri is supported by the humid environment in the field (micro-climate), making the production function very elastic with respect to fungicide application.

Input Elasticity and Factor Share

Table 9 shows the input elasticities and factor shares for all inputs in 2005 and 2012. Assuming that the farmers are able to obtain the efficient level of the shallot production, the input elasticities and factor shares are equal for each input within a year. The t-test is used to test the equality of input elasticity and factor share (the null hypothesis).

The results in Table 9 show that only insecticide use in 2005 was allocated efficiently by shallot farmers. Moreover, all of the other inputs were either underused or overused. In 2005, land, seed, nitrogen, phosphate, potassium, and fungicide were underused and labor was overused. In 2012, seed, labor, and insecticide were overused and the other inputs were underused.

Table 9. Input Elasticities and Factor Shares

Input 2005 2012 Elasticity Factor share t-value Sig Elasticity Factor share t-value Sig

Land (m2) 0.195 0.113 -33.629 *** 0.200 0.054 -69.846 *** Seed (kg) 0.157 0.146 -2.160 ** 0.165 0.209 8.001 *** Labor (man days) 0.225 0.241 2.921 *** 0.234 0.291 6.927 *** Nitrogen (kg) 0.126 0.019 -79.277 *** 0.137 0.032 -81.033 *** Phosphate (kg) 0.062 0.016 -46.132 *** 0.072 0.028 -31.340 *** Potassium (kg) 0.042 0.012 -36.935 *** 0.055 0.019 -27.099 *** Insecticide (gram) 0.015 0.018 1.441 0.018 0.179 7.011 *** Fungicide (ml) 0.130 0.033 -35.610 *** 0.137 0.026 -60.320 ***

Note:*** Significant at the 0.01 level, ** Significant at the 0.05 level.


Technical, Allocative, and Economic Efficiency in 2005 and 2012

The predicted yield using MLE and PCA-MLE are depicted in Figure 2. There are 70 observations, 35 observations for each year. The first 35 observations in the figure are for 2005. The range of the shallot yield was smaller in 2005 than in 2012. The predicted yield of MLE SFPF and PCA-MLE SFPF are quite close and near the actual observations in most instances. On average, the absolute percentage error from MLE SFPF is 20.0% with a standard deviation of 20.75 and the absolute percentage error from PCA-MLE SFPF is 20.4% with a standard deviation of 17.07. However, the PCA-MLE SFPF estimation results are preferred to the MLE SFPF results because PCA-MLE SFPF satisfies the concavity condition of X and has more reasonable input elasticities.

Figure 2. Actual Output and Predicted Output Based on SF MLE and SF PCA-MLE.

Figure 3. Technical Efficiency, Allocative Efficiency, and Economic Efficiency from the PCA-MLE Stochastic Frontier Production Function for 2005.


The next step is to examine the efficiency level of shallot production. Figures 3 and 4 depict the individual farmer efficiency levels of TE, AE, and EE for 2005 and 2012. Table 10 summarizes those efficiency measures for 2005 and 2012. On average, TE was 0.84 in 2005 and 0.85 in 2012, but this small increase in average TE is not statistically significant at the 5% level. In contrast, both allocative efficiency and economic efficiency for 2005 and 2012 are statistically different at the 1% level. The t-test for AE is 4.51 and that for EE is 3.16.

Figure 4. Technical Efficiency, Allocative Efficiency, and Economic Efficiency from the PCA-MLE Stochastic Frontier Production Function in 2012.

Table 10. Descriptive Statistics for Efficiency Measures

Descriptive statistics

2005 2012 TE AE EE TE AE EE

Average 0.84 0.76 0.64 0.85 0.68 0.58 Standard deviation 0.10 0.06 0.09 0.09 0.09 0.08 Coefficient variation 11.55 7.31 13.33 10.73 13.38 13.95 Minimum 0.63 0.66 0.50 0.61 0.44 0.42 Maximum 0.97 0.90 0.84 0.97 0.86 0.75 Median 0.86 0.77 0.64 0.88 0.69 0.58

Notes: T-test of TE = -0.5601 T-test of AE = 4.5137 T-test of EE = 3.162753.

Table 11. Efficiency Level of Farmers Based on Efficiency Measures from PCA-MLE

Efficiency level 2005 2012 TE AE EE TE AE EE

<0.60 0.00 0.00 37.14 0.00 17.14 57.14 0.60 - < 0.70 8.57 17.14 42.86 5.71 37.14 37.14 0.70 - < 0.80 28.57 57.14 14.29 22.86 34.29 5.71 0.80 - < 0.90 22.86 22.86 5.71 37.14 11.43 0.00 0.90 - < 0.95 31.43 2.86 0.00 22.86 0.00 0.00 > =0.95 8.57 0.00 0.00 11.43 0.00 0.00


Table 11 shows that 48% of the farmers had technical efficiency levels between 0.80 and 0.95 in 2005 and another 33% had technical efficiency levels between 0.60 and 0.80. In 2012, the technical efficiency level was considerably different. About 60% of farmers had technical efficiency levels between 0.8 and 0.95, higher than for 2005; while 28% had technical efficiency levels between 0.60 and 0.80.

Table 12. Farming Size and Average Efficiency Measures

Farm size (hectare)

2005 2012 TE AE EE TE AE EE

<0.10 0.8126 0.7524 0.6138 - - - 0.10 - < 0.15 0.8343 0.7578 0.6266 0.8429 0.6557 0.5508 0.15 - < 0.20 0.8629 0.7538 0.6341 0.8348 0.6866 0.5713 0.20 - < 0.25 0.8804 0.7555 0.6400 - - - 0.25 - < 0.30 0.8572 0.7584 0.6399 - - - 0.30 - < 0.35 0.7799 0.7631 0.6400 0.8825 0.6815 0.5787 0.35 - < 0.40 - - - - - - 0.40 - < 0.45 - - - - - - 0.45 - < 0.50 - - - 0.8820 0.6805 0.5784 > =0.50 - - - 0.7493 0.7170 0.5373

The level of allocative efficiency is lower than technical efficiency for almost every

instance. On average, AE was 0.76 in 2005 and 0.68 in 2012. The decreasing allocative efficiency level of 2012 compared to 2005 suggests that farmers failed to respond appropriately to changes in input prices. Almost three-quarters of the farmers (74.3%) had AE between 0.60 and 0.80 in 2005 compared with 71.4% in 2012. However, 17.1% of the farmers had AE levels below 0.60 in 2012. No farmer had AE lower than 0.60 in 2005. Another significant difference between 2005 and 2012 is the percentage of farmers that reached AE greater than 0.80 -- 25.7% in 2005 versus 11.4% in 2012. The decreasing allocative efficiency is in line with reduced economic efficiency of shallot farmers. On average, economic efficiency was 0.64 in 2005 and 0.58 in 2012. The percentage of farmers with economic efficiency less than 0.60 increased from 37.1% in 2005 to 57.1% in 2012.

Table 12 shows how the efficiency level varies with respect to farm size for each year. In 2005, farms from 0.20 to 0.25 hectare had the highest technical efficiency level (0.88). The farms that operate with less than 0.1 hectare and greater than 0.30 hectare (lower and upper tail of land distribution) have the lowest technical efficiency levels (0.81 and 0.78, respectively). In 2012, the two highest technical efficiency levels were for 0.30 - 0.35 hectare (TE = 0.88) and 0.45-0.50 hectare (TE = 0.88). The pattern of technical efficiency by farm size for both years is similar with lower and upper tails having lower TE.

The allocative and economic efficiency measures have a different relationship relative to farming size. Allocative efficiency in 2005 was around 0.75 for almost all farm size categories. This pattern for AE did not hold for 2012, where AE was lower for the smallest farm size (0.66) and higher for the largest farm size (0.72). Economic efficiency in 2005 increased (in general) as farm size increased; going from 0.61 for the small size category to 0.64 for the largest. The pattern of EE for 2012 was more like the TE pattern for 2012 with the smallest and largest size farms having lower efficiency relative to the middle sizes.


Economic efficiency is consistently lower than TE or AE, suggesting that farmers have a more difficult time adjusting to changes in prices rather than changes in technical factors (such as weather).

CONCLUSION

This study investigated the efficiency of Indonesian shallot farmers using cross sectional data from 2005 and 2012. Standard approaches for the analysis with a frontier cost function could not be used because of limited variation in input prices. Instead a stochastic frontier production function is estimated using principal components analysis and maximum likelihood estimation which resulted in input elasticities with the necessary concave properties. Efficiency measures were then calculated from these well-behaved production functions.

The results do not support Schultz’s hypothesis that the small-scale farming in traditional agricultural setting is efficient. There were inefficient farmers in both years of the study. Furthermore, increasing farming size in 2012 reduced farmer efficiency. It is not possible to say if this finding represents a trend over time for efficiency, but it is something that the Indonesia authorities should watch. Further analysis for other years and other crops would definitely pinpoint the problems.

It seems that accumulation of knowledge from experience is not strong enough to maintain or increase efficiency level when the producer is facing a volatile technical and economic environment. The limited managerial skill, especially capability to adapt changing prices, is a problem for small scale agricultural producers. This finding suggests that the local research and extension stations have an important role in educating and assisting farmers in adjusting to their changing environment. A local research and education presence is needed because of the specific issues associated with small-scale shallot production. Shallot farmers need assistance concerning which technology can be applied so they can change their input proportions in an efficient way.

REFERENCES

Afriat, Sidney N.,1972. Efficiency Estimation of Production Function. International Economic Review 13.3: 568-98.

Aigner, D., Lovell, C. A. A., and Schmidt, P., 1977. Formulation and Estimation of Stochastic Frontier Production Function Models. Journal of Econometrics, 6(1), 21-37.

Aigner, D., C. Lovell, and P. Schmidt, 1977. Formulation and Estimation of Stochastic Frontier Production Function Models. Journal of Econometrics, 6(1): 21-37.

Battese, George E., and Tim J. Coelli, 1988. Prediction of Firm-Level Technical Efficiencies with a Generalized Frontier Production Function and Panel Data. Journal of Econometrics 38.3: 387-399.

Battese, George E., and Greg S. Corra, 1977. Estimation of a Production Frontier Model: with Application to the Pastoral Zone of Eastern Australia. Australian Journal of Agricultural and Resource Economics 21.3: 169-179.


Bravo-Ureta, Boris E., and Antonio E. Pinheiro, 1977. Technical, Economic, and Allocative Efficiency in Peasant Farming: Evidence from the Dominican Republic. The Developing Economies 35.1: 48-67.

Coelli, T. J., Rao, D. S. P., O’Donnell, C. J., and Battese, G. E., 2005. An Introduction to Efficiency and Productivity Analysis. Springer.

Färe, R., and Knox Lovell, C. A., 1978. Measuring the Technical Efficiency of Production. Journal of Economic Theory, 19(1), 150-162.

Farrar, D. E., and Glauber, R. R., 1967. Multicollinearity in Regression Analysis: the Problem Revisited. The Review of Economic and Statistics, 92-107.

Farrell, M., 1957. The Measurement of Productive Efficiency. Journal of the Royal Statistical Society. Series A (General), 120(3): 253-290.

Fekedulegn, B., J. Colbert, R. Hicks, and S. Michael, 2002. Coping with Multicollinearity: an Example on Application of Principal Components Regression in Dendroecology. United States Department of Agriculture – Forest Service.

Førsund, F. R., Lovell, C. A., and Schmidt, P. 1980. A Survey of Frontier Production Functions and of Their Relationship to Efficiency Measurement. Journal of Econometrics, 13(1), 5-25.

Grapentine, Terry. 1997. Managing Multicollinearity. Marketing Research 9.3: 10-21. Jondrow, J., L. Knox, I. Materov, and P. Schmidt, 1982. On the Estimation of Technical

Inefficiency in the Stochastic Frontier Production Function Model. Journal of Econometrics, 19(2): 233-238.

Just, R. E., and Pope, R. D., 1978. Stochastic Specification of Production Functions and Economic Implications. Journal of Econometrics, 7(1), 67-86.

Kodde, D. A., and Palm, F. С., 1986. Notes and Comments Wald Criteria for Jointly Testing Equality and Inequality. Econometrica, 54(5), 1243-1248.

Koopmans, T. C., 1951. Analysis of Production as an Efficient Combination of Activities. New York: Wiley Press.

Kopp, R. J., and Diewert, W. E., 1982. The Decomposition of Frontier Cost Function Deviations into Measures of Technical and Allocative Efficiency. Journal of Econometrics, 19(2), 319-331.

Kumbhakar, S. C., and Tsionas, E. G. (2006). Estimation of Stochastic Frontier Production Functions with Input-oriented Technical Efficiency. Journal of Econometrics, 133(1), 71-96.

Meeusen, W., and Van den Broeck, J. (1977). Efficiency Estimation from Cobb-Douglas production functions with composed error. International Economic Review, 435-444.

Parel, Cristina P., G. C. Caldito, and P. L. Ferrer. 1973. Sampling Design and Procedures. New York: Agricultural Development Council.

Pa Ismail, 1981. A Note on the Interpretation of The Intercept of Cobb-Douglas Type of Productions Through Dimensional Analysis, Akademika, No. 19. p22- p27

Kennedy, Peter E., 2005. Oh no! I got the Wrong Sign! What Should I Do? The Journal of Economic Education 36.1: 77-92.

Marine and Fisheries Department, 2005. Prototype Climate and Weather Information for the Salt Ponds. Research Center of Marine and non-Biological Resources. Indonesia.

Satchachai, P., and Schmidt, P. (2010). Estimates of Technical Inefficiency in Stochastic Frontier Models with Panel Data: Generalized Panel Jackknife Estimation. Journal of Productivity Analysis, 34(2), 83-97.


Schmidt, P. andC. Lovell, 1979. Estimating Technical and Allocative Inefficiency Relative to Stochastic Production and Cost Frontiers. Journal of Econometrics, 9(3), 343-366.

Statistic Indonesian, 2013. Vegetables Harvested Area in Indonesia. Retrieved from http://www.bps.go.id/.

Zehna, Peter W., 1966. Invariance of Maximum Likelihood Estimators. The Annals of Mathematical Statistics 37.3: 744-744.


MANAGING RISK IN COCOA PRODUCTION:

ASSESSING THE POTENTIAL OF CLIMATE-SMART

CROP INSURANCE IN GHANA

Justin McKinley1, L. Lanier Nalley2,, Rebecca A. Asare3, Bruce L. Dixon4, Jennie S. Popp4, and Marijke D’Haese5

1Graduate Student, University of Arkansas, Department of Agricultural Economics and Agribusiness, Fayetteville, USA;

Graduate Student, Ghent University, Department of Agricultural Economics, Ghent, Belgium; and Consultant, International Rice Research Institute,

Social Sciences Division, Los Baños, Philippines 2Associate Professor, University of Arkansas, Department of Agricultural Economics

and Agribusiness and Division of Agriculture Fayetteville, USA 3Director of Programs and Research, Nature Conservation Research Centre, Accra,

Ghana and Africa Coordinator, Forest Trends, Washington D.C., USA 4 Professor, University of Arkansas, Department of Agricultural Economics

and Agribusiness and Division of Agriculture, Fayetteville, USA 5Associate Professor, Ghent University, Department of Agricultural Economics,

Ghent, Belgium

ABSTRACT

Cocoa production in Ghana increases deforestation as poor landholders encroach into forests to expand production. The management practices proposed in Climate-Smart Cocoa (CSC) can increase livelihoods while concurrently abating deforestation. This study investigates how yield and yield variation are affected by the adoption of CSC practices in order to determine the feasibility of a CSC crop insurance program in Ghana. To answer this question, we divided producers into two groups: those following CSC practices and those not following; then we used a multiple regression model to estimate yield and used the parameters of the model to simulate yields for CSC and non-CSC producers in 19 districts. Our results show that producers who followed the CSC recommended practices had higher estimated yields by 19–25%, were 5–25% less likely to have a yield loss large enough to receive an insurance payment, and the total expenses associated with indemnity payments in an insurance program were 20% less for CSC.

Corresponding author: L. Lanier Nalley, 217 Agriculture Building, Fayetteville, Arkansas 72701, USA.

Email: [email protected]

Justin McKinley, L. Lanier Nalley, Rebecca A. Asare et al. 54

Keywords: climate-smart agriculture, cocoa, crop insurance, deforestation, risk JEL: G22, Q10, Q23

LIST OF ABBREVIATIONS

AIC Akaike Information Criterion APH Actual Production History CAT 50% Catastrophic Coverage CLP Cocoa Livelihoods Program CRIG Cocoa Research Institute of Ghana CSA Climate-Smart Agriculture CSC Climate-Smart Cocoa CSCWG Climate-Smart Cocoa Working Group FCPF Forest Carbon Partnership Facility GAIP Ghana Agricultural Insurance Program HFZ High Forest Zone IIPACC Innovative Insurance Products for the Adaptation to Climate Change IP Input Promoter MPCI Multiple Peril Crop Insurance MRV Measurement, Reporting, and Verification NCRC Nature Conservation Research Centre REDD+ Reducing Emissions from Deforestation and Forest Degradation RSD Relative Standard Deviation SRID Ghana Statistics, Research, and Information Department WCF World Cocoa Foundation WII Weather-based Index Insurance

1. INTRODUCTION

In 2013, agriculture accounted for 21.5% of Ghana’s total GDP and 56% of the labor force, including many smallholder producers who contribute up to 80% of agricultural production in Ghana (CIA 2014). In Ghana, approximately 44.7% of households depend on agriculture for their livelihood (CIA 2015) and of those households, cocoa-producing households rely on cocoa (Theobroma Cacao) for 70-100% of their household income (Ntiamoah and Afrane 2008). Moreover, cocoa is Ghana’s largest agricultural export and has the largest share of agricultural GDP (CIA 2014) with total Ghanaian cocoa exports exceeding more than $2.2 billion USD in 2011 (FAO 2014a).

Surprisingly, Ghana has one of the lowest cocoa yields per hectare in the world, less than 400 kg ha-1 (Dormon et al. 2004; Ruf and Schroth 2004; Stutley 2010) despite being the second largest cocoa exporter in 2011. Neighboring Côte d’Ivoire has yields twice that amount at 800 kg ha-1 (Dormon et al. 2004). Over the years, low yields may be the result of disease and pests, failure to adopt research recommendations, e.g., high-yielding hybrid varieties, proper pesticide usage, shade management, weed control, a low producer price, or an inadequate extension system (Dormon et al. 2004). Consequently, the failure to adopt

Managing Risk in Cocoa Production 55

recommended procedures, e.g., inorganic fertilizer use, could be the consequence of constrained, physical access to fertilizer from limited supply and underdeveloped distribution systems, as well as limited credit to purchase fertilizer. Even if credit were available, many producers would likely forego the opportunity to borrow because of prohibitively high interest rates and stringent collateral standards for loans, which is the case in most low-income countries. In Ghana, rural finance loans have interest rates ranging from 25-40% per year, depending on the loan provider (Steel and Andah 2003). In addition, smallholders have historically been risk averse because they are vulnerable to adverse shocks (e.g., crop failure) that can sink their household deeper into poverty (World Bank 2014; Todaro and Smith 2012; IFAD 2011a). By avoiding these adverse shocks, producers pass up opportunities, i.e., increased yields and incomes, that could improve their livelihoods (IFAD 2011a; Dercon and Christiaensen 2011; Cole, Giné, and Vickery 2013; Karlan et al. 2012; Dercon, Gunning, and Zeitlin 2011; Morduch 1995).

Historically, cocoa expansion into undisturbed forests has increased cocoa production in Ghana and has been largely driven by: (1) a customary system of land tenure throughout West Africa commonly known as droit d’hache, or, “right of the axe” (R. A. Asare 2010), (2) the avoidance of pests and disease in current, cocoa-producing areas (Berry 1992), and (3) the utilization of forest rent: which is the nutrient stock that has built up on the forest floor through time, rather than replant cocoa trees on existing farms (Ruf and Schroth 2004; Ruf and Zadi 1998; Ruf, Schroth, and Doffangui 2014).

Furthermore, forest degradation and deforestation in Ghana could be reduced through the implementation of Climate-Smart Agriculture (CSA). The FAO (2013) defines climate-smart agriculture as, “agriculture that sustainably increases productivity, resilience (adaptation), reduces/removes greenhouse gases (mitigation), and enhances achievement of national food security and development goals.” In Ghana, CSA practices have been adapted to cocoa production to form what is called Climate-Smart Cocoa (CSC), which has been endorsed by the Ghanaian Government and promoted by the CSC Working Group, established in 2011 (NCRC 2012). CSC is a holistic approach to cocoa production with four fundamental elements: (1) access to extension services and financial credit, (2) access to crop insurance to decrease farming risk, (3) land use planning, and (4) cocoa producer data management, coupled with REDD+ measurement, reporting, and verification (MRV). Additionally, there are environmental concerns that deforestation may be accelerated if the focus is only on raising farm yields, as producers see the opportunity for increased revenue from investing their increased income into new farms (R. A. Asare 2014). As a solution to this problem, cocoa producers could be incentivized not to expand or encroach into forest areas by linking their access to extension, trainings, and other farming resources, including crop insurance, to their demonstrated adoption of CSC practices and reductions in rates of deforestation or forest degradation in the landscape. Crop insurance can specifically help to reduce deforestation in two ways. First, it reduces risk in cocoa farming by providing protection against weather-related cocoa losses and in effect, removing the need for producers to exploit forests to reduce their risk through land diversification. Second, access to crop insurance would require that producers follow CSC practices that prohibit such expansion into forests.

Funding for the subsidized crop insurance could come through contributions from the private sector as well as through payments for emission reductions, which would result from Ghana’s implementation of at Reducing Emissions from Deforestation and Forest Degradation (REDD+), a United Nation’s collaborative initiative in forested, low-income


countries aimed REDD+ (R. A. Asare 2014). Specifically, Ghana’s Emission Reductions Program for the Cocoa Forest Mosaic Landscape articulates five major pillars to help Ghana reduce its emissions from deforestation and forest degradation, as a result of cocoa farm expansion, by: (1) supporting yield increases through the adoption of CSC practices; 2) enabling institutional collaboration across government agencies and with the private sector; (3) implementation of land use planning at the local level; (4) development of a cocoa sector data management platform and REDD+ MRV system; and (5) reducing farmers’ risks (FCPF 2014) By using a holistic approach that includes crop insurance, CSC practices can help to abate deforestation while concurrently raising cocoa yields for Ghanaian cocoa producers. Yield risk could also be reduced among CSC producers if proper training is provided to deal with pests and diseases on the farm while concurrently realizing increases in average yield.

Figure 1. Map of cocoa-producing regions of Ghana in green with survey locations.


A major hurdle in the proposed CSC holistic approach is the absence of crop insurance products in Ghana for cocoa producers. Traditional crop insurance programs, like those offered in the United States, are expensive to operate and depend heavily on government subsidies (Barnett and Coble 1999; USDA 2011). Subsidies are problematic in Ghana where a large portion of the population is still engaged in agriculture. This problem is amplified in low-income countries due to lack of quality farm data, the small size of the average farm, and governments’ inability to consistently maintain the high costs of subsidies (Linnerooth-Bayer, Mechler, and Hochrainer-Stigler 2011).

For this study, we estimated fair-market premium levels in 19 districts of Ghana for two groups of cocoa producers: (1) those that follow CSC practices and have undergone input-use training, use inorganic fertilizer, and practice shade management (defined in the data collection as 12-16 trees taller than 12m ha-1), and (2) those that do not use CSC practices but do use inorganic fertilizer. Our analysis involved estimating a regression model that explains yield variability using household-level data on yield, input use, weather, and farm characteristics, obtained through the Cocoa Livelihoods Program (CLP) of the World Cocoa Foundation (WCF) (WCF 2014a). These data are unique because they provide a large spatial coverage (1,200 households in 108 villages, 19 districts, and five regions) for a location and a crop for which large data sets are scarce (Figure 1).

The objective of this study is to investigate how mean yield and yield risk behave under CSC conditions. Accordingly, this should give insight into the feasibility of providing crop insurance as an incentive for producers to follow CSC practices and reduce forest degradation and deforestation. To determine how the yield functions under CSC conditions, we: (1) estimated yield differences among producers who follow CSC and non-CSC practices and (2) estimated the impact of CSC practices on yield variability using percent chance of indemnity payments to producers. If producers who follow CSC practices have higher yields and/or less yield variability, then CSC may be a viable option, when coupled with crop insurance, to protect producers from yield risk, while also reducing forest degradation and deforestation through the incentivization of cocoa production.

2. LITERATURE REVIEW

2.1. Cocoa in Ghana: Seasons, Trends, and Deforestation

There are generally two harvest periods for cocoa in Ghana. The main harvest is from early September to late February; and the light harvest is from mid-May to mid-July.1 The main harvest corresponds with southern Ghana’s rainy season in April because cocoa pods require approximately five months to mature for harvest (Glendinning 1972). Cocoa precipitation requirements are between 1,500 mm and 2,000 mm per year (ICCO 2013). If annual precipitation is below 1,250 mm, the trees will lose more water through evaporation than they receive from precipitation (Wood and Lass 1985). Annual precipitation above 2,500 mm greatly increases the incidence of fungal diseases such as black pod (Phytophthora megakarya (Wood and Lass 1985). Additionally, cocoa requires minimum temperatures of

1 Personal communication Dr. S.T. Ampofo, Cocoa Abrabopa Association.


18o-21oC and maximum temperatures of 30o-32oC (ICCO 2013), which are present in the moist, semi-deciduous forests of southern Ghana.

Another driving factor in deforestation in Ghana is the prevalence of cocoa diseases. The first, substantial expansion of cocoa cultivation and forest degradation occurred in Ghana in the 1930s and ‘40s when cocoa farms in today’s Eastern Region were afflicted with an endemic outbreak of the cocoa swollen shoot virus (CSSV) (badnavirus) (Berry 1992). CSSV is spread by mealy bugs (Planococcus citri) (Wood and Lass 1985) and is capable of killing a cocoa tree within a few months of initial infection (Wilson 1999). There is no cure for CSSV, it can only be managed by destroying the trees that are infected with the virus (Wood and Lass 1985). Those whose crops were infected with CSSV, looked to new, un-diseased land that was mostly virgin rainforest, in order to produce cocoa. As such, producers at that time abandoned their failing farms and migrated westward into the moist, semi-deciduous forests of the Ashanti and Brong-Ahafo regions (Berry 1992). Recent statistics from the Government of Ghana, as cited in Forrest Carbon Partnership Fund (FCPF) (2014) show that in areas of the High Forest Zone (HFZ), the conversion of intact forests to other land uses (predominantly cocoa production) has accelerated from 2.8% per annum in the period of 1986-2000, to 6.1% per annum in the period of 2000-2011.

2.2. Shade in Cocoa Production

Shade plays an important role in cocoa production through numerous interactions with the environment (i.e., solar radiation, air movement, temperature, relative humidity, and soil moisture) (Wood and Lass 1985). A study conducted in Ghana by Murray (1954) found that cocoa yields increased on cocoa farms that did not use nitrogen fertilizer when light levels increased up to a light level of 50%. When light levels exceeded 50%, cocoa yields began to decline without nitrogen application (Murray 1954). The same study found that when nitrogen fertilizer was applied, reduction in shade led to increased cocoa yields until light levels reached 75%, at which point yields plateaued and then declined slightly as light levels approached 100% full-sun exposure (Murray 1954). Previous studies also found that shade reduction led to higher yields with some varieties (Ahenkorah, Akorifi, and Adri 1974; Gockowski and Sonwa 2011; Gockowski and Sonwa 2008). The increased yield from decreased shade is the result of cocoa trees producing more leaves in direct sunlight. The additional leaves stimulate flowering, leading to more cocoa pods and ultimately to more cocoa beans (Asomanin, Kwakwa, and Hutcheon 1971; Boyer 1974; Cunningham, Smith, and Hurd 1961).

There is no definitive consensus in the literature that shade reduces yield. One example from Campbell (1984) is the reduction of pest and disease damage, which shade helps with and can actually increase yield. Ahenkorah et al. (1974) also cites decreased weed growth as a yield enhancer in shaded production systems. More recently, a forthcoming publication covering a four-year period in Ghana also found shade to have a positive and significant effect on cocoa yields (R. Asare et al. n.d.) The relationship between shade and cocoa yields is complex and is still not completely understood as there are many dynamic factors at play. Nonetheless, the Cocoa Research Institute of Ghana (CRIG) recommends that cocoa should be grown under approximately 30-40% canopy cover, noting that this constitutes approximately 16-18 mature, large shade trees per hectare. Growing cocoa under shade has


immediate implications such as: reductions in forest degradation leading to eventual deforestation, and even increases in carbon stocks. In the future, shade may play an even more important role in mitigating the effects of climate change by buffering temperature changes as well as improving hydrological cycling (Kenneth and Baba 2011).

2.3. Climate-Smart Cocoa

As a consequence of this rapid deforestation, the concept of climate-smart cocoa (CSC) was initiated. CSC is a local adaptation of CSA in Ghana. The FAO (2013) defined CSA as, “agriculture that sustainably increases productivity, resilience (adaptation), reduces/removes greenhouse gases (mitigation), and enhances achievement of national food security and development goals.” There is mounting interest in CSA because of the realized economic and environmental benefits. In Ghana, interest from stakeholders in the cocoa industry in adopting a CSA approach for the cocoa sector and production system, led to the establishment of the Climate-Smart Cocoa Working Group (CSCWG) in 2011. The initiation of the CSCWG came from Forest Trends and the Nature Conservation Research Centre (NCRC) in partnership with over seventeen national and international entities, including: the private sector, government institutions, civil society, and research organizations.

The main purpose of the working group was to better understand the threats to Ghana’s cocoa sector and to define strategies to reduce the illegal entry of cocoa farms into forest reserves (R. A. Asare, 2014). The CSCWG reported that cocoa production in Ghana was on an unsustainable pathway for three primary reasons: (1) predicted changes in temperature and rainfall patterns, due to climate change, threaten the productive capacity of the trees, (2) the cocoa sector’s primary emphasis on intensification without thought to how production increases could promote deforestation (and ultimately undermine the condition of the forests, which provide critical ecosystem services to the cocoa production landscape), and (3) a total lack of land-use planning on the landscape level (NCRC 2012). The results of the report stressed a need to improve cocoa producers’ livelihoods by increasing yields and accessing mitigation and adaptation knowledge. Additionally, the report stressed the importance of improving livelihoods with efforts to decrease greenhouse gas emissions and reduce the amount of cocoa farms in forests and on lands with medium carbon stocks. Finally, the report stressed the importance of landscape level land-use planning as a means to obtain increased income for cocoa producers and promote biodiversity and ecological resilience across cocoa-producing landscapes, and not just on farms (NCRC 2012).

The CSCWG (2014) identified key gaps that need to be addressed or improved for CSC to reach the “desired future state”: (1) increasing yields: the need to define and describe climate-smart best practices in cocoa, including a focus on extension, inputs, and appropriate soils that will lead to substantial yield increases and income increases for farmers; (2) reducing risk in cocoa farming: the need to develop yield insurance products for cocoa in the face of climate change and expand access to credit sources to reduce farmer risk that results from climate change; (3) landscape planning: the need to harness a process or mechanism to implement landscape planning on the community level in order to curb expansion into forest reserves, target the most appropriate cocoa soils, retire over-age, high biomass cocoa farms, and grow forest in the landscape; and (4) managing data and MRV: the need to construct a


platform and system to manage and link data at multiple scales related to climate-smart cocoa, so that mitigation impacts can be measured and monitored over time.

The first element, increasing yields, is currently being practiced and is available in Ghana through various programs like those offered by the World Cocoa Foundation (WCF) and the Cocoa Abrabopa Association (R. A. Asare, 2014; WCF, 2014). These projects primarily increase yields through access to extension services and access to credit for the purchase of agricultural inputs, once the producer has completed the necessary training programs. The primary distinction between these projects and the proposed CSC program is that the other existing projects and initiatives focus exclusively on raising yields and incomes, without addressing the landscape issues of potential deforestation and forest degradation that often occurs when farmers invest their increased income into new cocoa plantings (R. A. Asare, 2014). In addition to increasing producers’ yields and incomes, a crucial element of CSC is mitigating forest degradation and deforestation through land-use planning of the landscape, while also monitoring to ensure the sustainability of cocoa in West Africa.

2.4. Crop Insurance

Crop insurance was identified as a key element to the success of CSC in Ghana because of its ability to help producers mitigate risk and as incentive for producers to follow CSC production practices, which protect against yield variability related to weather events. Furthermore, the CSC practices include recommended management practices, e.g., shade management, and give producers access to trainings and credit so that they can properly apply agrochemical inputs (inorganic fertilizer, pesticides, fungicide). In addition to adopting these practices, CSC producers would be required to refrain from expanding into undisturbed forests or other areas designated by a localized land-use planning process. If this scheme is applied across a landscape, sizable reductions in deforestation could happen through proper planning of land use (R. A. Asare 2014). The final element of the CSC approach is the accurate measurement of impact at the livelihood and landscape level through the management of cocoa data and forest monitoring. To illustrate, accurate monitoring of the landscape using an MRV system could enable CSC producers to access carbon (REDD+) or emission reduction (ER-Program) payments.2 Hence, this revenue could be used to partially subsidize a crop insurance program for eligible CSC producers.

There are two primary types of crop insurance that could be utilized in CSC: crop yield insurance and crop revenue insurance (Barnett and Coble 1999). In the case of crop yield insurance, indemnities (money paid from the insurer to the insured in the event of a loss) are paid to producers when their yield falls below their insured level, and is based on the producer’s actual production history (APH)3 (USDA 2011). Alternatively, crop revenue insurance provides revenue protection to producers, in addition to physical losses, by guaranteeing commodity prices, which protects producers from volatile markets (Barnett and Coble 1999; USDA 2011). For the CSC practices program, the intention is to insure yield rather than revenue.

2 For more information see: FCPF (2014) Emission Reductions Program Idea Note (ER-PIN). Accra and UN-

REDD. (2011). The UN-REDD Program Strategy. Geneva. 3 Actual production history is based on historical on-farm yields.


Conventional crop yield insurance has two primary policy types: Multiple-Peril Crop Insurance (MPCI) and Crop Index Insurance. MPCI is used to cover against a wide range of perils such as drought, flooding, disease, pests, hail damage, and fire damage (Barnett and Coble 1999). Policy holders choose and pay for the perils they want protection against. MPCI insures against perils that can be difficult to measure (Roberts 2005). For instance, black pod

(Phytophthora megakarya) is a fungal infection that affects cocoa, but the severity of the infection and the associated yield losses are not easily determined without an on-farm visit. Unfortunately, the need for an adjuster to make an on-farm visit to inspect losses, makes MPCI expensive to offer, especially in low-income countries where the agricultural system is dominated by smallholders (IFAD 2011b). MPCI programs also have higher operational costs because of additional claims of moral hazard and adverse selection. MPCI programs typically operate at a loss and depend on government subsidies to function (Stutley 2010; Roberts 2005).

Recently, index insurance, such as Yield Index Insurance (YII) and particularly Weather-based Index Insurance (WII) have become possible alternatives to MPCI in low-income countries. A WII advantage over MPCI means that the high transaction costs associated with indemnity-based systems such as MPCI are avoided (Linnerooth-Bayer et al. 2011). In WII programs, a proxy measure like precipitation is used to determine at what point, i.e., too much or too little precipitation, should an indemnity be paid (IFAD, 2011b; Roberts, 2005; Stutley, 2010). By using a proxy variable, transaction costs are reduced because there is no need for on-farm assessments to quantify losses (Stutley 2010). Operational costs are also reduced in WII (IFAD, 2011b) because adverse selection and moral hazard are eliminated since indemnities are only paid when the proxy variable falls below a trigger point (Roberts 2005). WII packages only cover a limited number of perils, which does not always satisfy the risk-management needs of the insured (IFAD, 2011b). In addition, technical capacity development and expertise in agro-meteorology still need to be improved as this is a relatively new product (IFAD, 2011b). Data used in agro-meteorology can be obtained through remote sensing. However, even though the insurance industry is very information-intensive (Triplett and Bosworth 2006), the use of remote sensing data is still limited in the industry (de Leeuw et al. 2014).

At present, Ghana has very limited crop insurance. In December of 2009, a project called Innovative Insurance Products for the Adaptation to Climate Change (IIPACC) was established with funds from the German Federal Ministry for the Environment, Nature Conservation, and Nuclear Safety (Stutley 2010). Funds from the IIPACC were used to create the Ghana Agricultural Insurance Programme (GAIP). The first insurance product from GAIP became available in 2011. This was a WII product, which used on-the-ground weather stations and insured 3,000 smallholder maize (Zea mays) farmers from three different regions in Ghana (GAIP 2013). By 2012, the GAIP’s coverage had expanded to cover maize, soya (Glycine max) and sorghum (Sorghum bicolor), and was available in seven regions of Ghana (Gille 2013). The GAIP has struggled to create an insurance product that has affordable premiums and adequate risk coverage for the insured, which is partly due to its inability to create affordable channels to distribute low-cost insurance and to actively engage the Ghanaian government in the insurance program to contribute to affordable premiums, comprehensive coverage, and up-scaling of coverage (Gille 2013).

Currently, there is no crop insurance program available for cocoa in Ghana due to the many challenges involved in insuring a perennial crop like cocoa. First, reliable farm-level


production data are largely unavailable and dependable insurance requires accurate data. The Ghana Statistics, Research, and Information Department (SRID) has not maintained any time-series production and yield databases for plantation tree crops, which includes cocoa (Stutley 2010). Second, the Ghana Cocoa Board and the CRIG are responsible for nearly all aspects of research and development for the cocoa sector in Ghana, yet collecting and maintaining information about cocoa producers, including yield data have not been priorities for these organizations. Finally, the Cocoa Board does not have accurate data on the total number of cocoa producers in Ghana, nor does it have accurate data on the total area under cocoa production, much less individual producer yield data.4

Furthermore, perennial crops such as cocoa are difficult to insure because they have different life-cycle stages of production. Vilsack (2009) describes the cocoa production cycle in four stages: (1) establishment: zero yield, (2) development: exponential yield growth, (3) maintenance: relatively constant yields, and (4) decline: yield reduction. This cycle poses a challenge for cocoa insurance because the amount of cocoa harvested varies, given the age of the tree. This phenomenon makes it difficult to accurately predict yield without knowledge or at least estimates of the ages of cocoa trees for a given owner.

A final challenge in writing a WII policy for cocoa is accurately defining the ‘trigger points’ at which indemnities are paid to the insured. Cocoa production requires a specific range of precipitation and temperature; it prefers precipitation between 1,500 mm and 2,000 mm per annum (ICCO 2013). Annual precipitation below 1,250 mm is unfavorable because high tropical temperatures cause higher evaporation from the tree than the precipitation that is received (Wood and Lass 1985). Conversely, annual precipitation above 2,500 mm increases the prevalence of fungal diseases such as black pod (Wood and Lass 1985). Cocoa also requires minimum temperatures of 18o–21oC, and maximum temperatures of 30o –32oC to be productive (ICCO 2013). The complexity of such environmental needs makes it difficult to model cocoa yields. However, even though these complexities provide many difficulties for crop modeling, the challenges would be surmountable in the future with the acquisition of reliable data.

Regarding all of the aforementioned challenges in establishing a cocoa crop insurance program that is focused on the framework of CSC, WII appears to be the most suitable because of its reduced transaction costs and the absence of adverse selection and moral hazard. Particularly under an MPCI program, moral hazard could reduce the effectiveness of the training programs because CSC producers would have an incentive not to treat disease or pest outbreaks. With WII, however, CSC producers would have an incentive to use the training they received to treat the disease and pest outbreaks because WII only pays indemnities on weather variability, not for damage from disease and pests. Further, the reduced transaction costs of WII would make the program more affordable to producers. Moreover, WII covers regions rather than individuals, making it more compatible for landscape planning associated with CSC, the elimination of moral hazard, and coverage for CSC producers against climate variability.

4 Personal communication Mr. E.T. Quartey, Director of Research, Monitoring and Evaluation, Cocobod.


3. METHODOLOGY

3.1. Data and Model Specifications

Primary household data were provided by the World Cocoa Foundation’s Cocoa Livelihoods Program (CLP) (WCF 2014a). Twelve hundred farms were sampled representing five regions, 19 districts, and 109 villages in Ghana (Table 1). These data were collected by WCF to assess the impact of their CLP. Furthermore, a vital element of the CLP was the three farmer training programs offered: (1) training on procedures in crop cultivation, which included some agrochemical training, (2) training on agribusiness management, and (3) a capstone course, called input promoter (IP), that taught the participants proper levels and timing of usage of agrochemical inputs to maximize the effectiveness and efficiency of agrochemical use. Subsequently, the principles of IP were also an integral part of CSC in this scenario. Moreover, participants who successfully completed all three training courses were granted access to credit for farm inputs. In addition to data on the completion of the training courses, other WCF data used here include: a binary variable for fertilizer use (use vs. non-use), gender, a binary variable for shade management (12-16 trees taller than 12m ha-1 vs. not), farm size in hectares, and total cocoa yield in kilograms per hectare. These data were collected in four different surveys from February 2011 to August 2012 and are inclusive of the two, main cocoa seasons in Ghana.

Table 1. Distribution of Survey Observations by Region and District

Location n Ashanti (region toal) 194 Adansi South 44 Ahafo Ano South 36 Atwima Nwabiagya 61 Bosome Freho 53 Brong-Ahafo (region total) 124 Asunafo North 40 Asunafo South 42 Asutifi 42 Central (region total) 115 Asin North 60 Upper Denkyira West 55 Eastern (region total) 267 Akyemansa 54 Birim North 45 Birim South 168 Western (region total) 500 Aowin Suaman 55 Bia 52 Bibiani Awiaso Bekwia 49 Juaboso 138 Sefwi Akontombra 44 Sefwi Wiawso 112 Wassa Amenfi West 50 All Regions 1,200


Daily weather data were obtained from Awhere Incorporated’s online weather platform (Awhere Incorporated 2015). Weather data were available at a five arc-minute resolution, approximately 9km2, and were collected through a combination of orbiting weather satellites and on-the-ground meteorological stations. Weather data were merged with WCF data at the village level based on village GPS coordinates. This provided unique spatial and temporal data on precipitation for villages, unless multiple villages were contained within the same 9km2 grid cell.

A traditional approach to valuing crop insurance is not feasible when using actual production history because farm-level data for cocoa in West Africa are difficult to obtain and time series producer data are nearly non-existent. For this reason, cocoa yields were simulated from a regression model based on a limited number of observed yields in order to obtain yield distributions at the district level, and to assess various thresholds for estimating a fair-market value for crop insurance. Previously, Norton et al. (2014) assessed the effectiveness of the CLP program using a cost-benefit analysis to compare the profits and yields of farmers who had participated in the CLP program to those who had not. The present study repurposed the data in order to assess the risk that farmers who follow CSC practices face. This was done by using available data on the completion of the IP training because it was considered to be equivalent to the training in which CSC producers would participate. Also, the variables for fertilizer use (with training) and the use of a shaded production system (12-16 trees taller than 12m ha-1) were obtained from the CLP data. Furthermore, rainfall data was included in this model to introduce climate risk factors.

3.2. Regression Analysis

An ordinary least squares regression was specified to predict how cocoa yields vary as a function of different demographic, locational, weather, and agronomic factors. This regression has a semi-log, functional form assuming that the effect of precipitation and land size both have diminishing returns on yield. The regression model was specified as:

= + + + ∗ + ∗ + +ℎ + + + , (1) where Yi is observed annual yield (kg ha-1) for the ith farmer; Disti is a vector of binary

variables representing individual districts (locations) for the ith farmer with the district Wassa Amenfi West included in the intercept; CSCi is a binary variable to capture input-use efficiency constructed by taking a value of one for farmers who have completed the WCF’s IP training program for the ith farmer; CSC*Ferti is a binary variable taking a value of one for farmers who have received training in agrochemical input-use and who also use inorganic fertilizer; nCSC*Ferti is a binary variable taking a value of one for a farmer who does not receive training, but uses inorganic fertilizer; Gndri is a binary variable taking a value of one if the ith household head is male; CSCshadei is a binary variable taking a value of one if the ith farmer uses shade management; lnSizei is the natural logarithm of the ith farm’s size in hectares; and lnPrecipi is the natural logarithm of the ith farm’s average precipitation (mm per day) during the pod maturation period of the main crop such that:


= ∑ 153 , (2)

where Precipi is average daily precipitation in mm for the ith farmer in each location and time period; Pdi is daily precipitation in millimeters for the ith farmer; and 153 is the duration in days of the pod maturation period for the main crop from June 1 to October 31. Flowering for cocoa trees in Ghana occurs approximately from early January to late May, proceeded by a period of pod maturation, which occurs approximately from early June to late October.5 Jozwik (2013) reports that 73% of farmers in his study cited precipitation during this period as a major driver of yield. The model used in this study differs from Norton et al. (2014) in that it uses a different functional form and includes precipitation as an explanatory variable. A complete list of dependent and independent variables with definitions and responses is shown in table 2.

Table 2. Definition of Regression Model Variables

Variable Definition Units Yieldi Amount of cocoa beans produced kg ha-1

Disti District where respondents farm is located district name CSCi Completed input-use training 1=yes, 0=no CSC*Ferti Completed input-use training and used fertilizer 1=yes, 0=no nCSC*Ferti Did not complete input-use training but used fertilizer 1=yes, 0=no Gndri Gender of household head 1=male, 0=female CSCshadei Practiced shade management (12-16 mature trees ha-1) 1=yes, 0=no lnSizei Natural logarithm of respondents reported farm size hectares lnPrecipi Natural logarithm of precipitation in pod maturation period mm day-1

3.3. YIELD SIMULATIONS

Using the estimated yield regression model (equation 1), simulations were conducted based on specific farm characteristics for two groups of cocoa producers: (1) those who follow CSC practices: receive input-use training, use inorganic fertilizer, and practice shade management (12-16 trees taller than 12m ha-1), and (2) those who do not use CSC practices: no input-use training, no shade management, but do use inorganic fertilizer. By allowing both groups to use inorganic fertilizer, the impact that input-use training had on the effectiveness of fertilizer use can be more clearly computed. Furthermore, simulations were conducted at 1,000 iterations each, for CSC and non-CSC groups from 19 different districts.

In these iterations, three sources of uncertainty were modeled, which are as follows: uncertainty in precipitation, beta coefficients, and the error term. Potential precipitation distributions for each district were examined by using the software package @RISK (Palisade Corporation 2014) to fit observed data into various distributions and rank them based on Akaike Information Criterion (AIC). The lognormal distribution was determined to be the most appropriate distribution for precipitation in all districts, based on the results of AIC. The distributions of all of the estimated regression coefficients in equation (1) as well as the error 5 Personal communication Dr. S.T. Ampofo, Cocoa Abrabopa Association.


term, are assumed to be distributed normally, consistent with conventional regression theory, and are the empirical evidence of the distribution of the residuals. The three sources of uncertainty in this model are:

Uncertainty in precipitation, distributed as: ~ , (3)

The random draw of precipitation for an individual farmer, i, in location λ has a log

normal distribution with a mean of observed precipitation mean, and variance of observed precipitation variance, .

Uncertainty in the beta coefficients (due to sampling error) where b is the full vector of simulated regression coefficients distributed as:

b~N( ,Cov( )), (4) where is estimated betas in equation (1) and Cov( ) is the corresponding covariance

matrix. Uncertainty in the error term is distributed such that: ~ (0, ), (5) where the error, εi, is normally distributed with mean of zero and standard deviation of . The simulated yields required random draws for all three, independent sources of

uncertainty. Simulated yields were used to estimate the mean and variance of yield by production system (CSC and non-CSC) and by location. The yield estimates resulted from multiplying the simulated values of the independent variables by the draws of the estimated regression coefficients from equation (1). The variables CSC, CSC*Fert, nCSC*Fert, Gndr, and CSCshade are binary and take a value of zero or one. For yield simulations, the CSC producers were given a value of one for CSC, CSC*Fert, Gndr, and CSCshade. Non-CSC producers were given a value of one for nCSC*Fert and Gndr. Farm size, although not binary, was held constant at the average reported farm size in each district, in order to capture the difference between CSC and non-CSC producers. Furthermore, negative yield observations in the simulations were discarded before computing the yield means and variances, rather than being truncated at zero because: (1) there are no instances of zero yields in the observed data set, (2) truncating at zero results in lower simulated yields than discarding those observations, and (3) the simulated mean yields with this adjustment are less than observed mean yields. Approximately 13% of simulated observations were removed, with 9% of the CSC simulations and 17% of the non-CSC simulations.

3.4. Indemnity Payments

The cost of an index crop insurance program can be measured through the probability of receiving an indemnity payment. An indemnity payment is made when the on-farm yield is less than some percentage of the expected yield – in the following example district averages


are used as the expected yield. Additionally, the trigger yield at which an indemnity payment is disbursed is set at 50% of the expected yield. This level of insurance coverage is often referred to as catastrophic (CAT) coverage. The probability of receiving an indemnity payment is defined as:

% ℎ = ∑ ∑ (6)

The percent chance of indemnity was estimated for all locations, j, which was calculated

as the proportion of observations that fell below the trigger yield for an indemnity payment to all observations in location j (i.e., insurance pool). Using simulated yields, a test was performed to determine if the proportion of producers receiving an indemnity payment was significantly different between the CSC and non-CSC groups. Because simulated observations falling below zero were removed from the final simulated sample, CSC and non-CSC groups have different sample sizes. As such, there are two resulting proportions: pcscj and pnon-cscj for CSC and non-CSC producers in location j, respectively. Differences of the proportions were tested using a z statistic such that:

= , (7)

where pcscj and pnon-cscj are the proportions of indemnity payments made for CSC and non-

CSC producers, respectively, in location j and Pallj where there is a pooled proportion for both groups such that:

= , (8)

where Xcscj and Xnon-cscj are the number of instances in which an indemnity payment was

made to CSC and non-CSC producers, respectively, in location j and ncscj and nnon-cscj are the total number of observations in the samples (i.e., insurance pool) for CSC and non-CSC producers, respectively, in location j.

3.5. Relative Standard Deviation

Another way to compare differences in yield risk for CSC and non-CSC producers is by estimating the coefficient of variation (Scheel 1978). Risk can be considered as the dispersion of the simulated yield ha-1 measured as the standard deviation of the simulated results. However, because the mean yields of the two groups, CSC and non-CSC, were not equal, the results should be normalized to have a fair comparison of risk. This normalization was accomplished with the coefficient of variation expressed as:

= , (9)


where the coefficient of variation, (CVλ) is equal to the ratio of the standard deviation (σλ)—yield per hectare, to the mean (μλ)—yield per hectare—for the λth location. Relative standard deviation (RSD) is the absolute value of the coefficient of variation, multiplied by 100, to be expressed as a percentage.

Table 3. Regression Results for Yield (kg/ha)

Variable Coefficient Standard Error Constant‡ –277.37*** 74.94 Adansi South 145.49*** 42.90 Ahafo Ano South 33.74 44.55 Akyemansa 175.51*** 45.50 Aowin Suaman 86.99** 40.13 Asin North 104.82** 40.94 Asunafo North 97.53** 43.12 Asunafo South 104.91** 42.87 Asutifi 96.36** 43.25 Atwima Nwabiagya 83.43* 46.07 Bia 147.79*** 39.94 Bibiani Awiaso Bekwia 176.88*** 41.74 Birim North 156.04*** 44.03 Birim South 220.32*** 39.72 Bosome Freho 80.46* 41.76 Juaboso 53.07 35.15 Sefwi Akontombra 132.14*** 41.69 Sefwi Wiawso 172.32*** 36.62 Upper Denkyira West 137.16*** 40.14 CSC 15.71 35.20 CSC*Fert 156.48*** 35.61 nCSC*Fert 118.07*** 13.78 Gndr 35.47*** 12.23 lnSize –63.85*** 9.91 lnPrecip 367.94*** 50.48 CSCshade 42.51** 20.58

Note: '***', '**', and '*' are significant at the P < 0.01, 0.05, and 0.10 levels, respectively. '‡' Wassa Amenfi West in included in the constant. R2 = 0.24 RMSE = 201.4 n = 1,200.

4. RESULTS

4.1. Regression Results

Results of the estimated regression model specified in equation 1 are presented in Table 3. The fixed-effect coefficients for location (Dist) were found to be significant at the one (P < 0.01), five (P < 0.05), and ten (P < 0.10) percent levels in nine, five, and two districts,


respectively with Wassa Amenfi West as the reference district. Precipitation was also found to be significant (P < 0.01) in the model. These results confirm the importance of precipitation during the pod maturation period of June through October. Precipitation during this period, although highly significant, has diminishing returns. Following Yang (2012), a one percent change in precipitation with a coefficient of 367.94, results in a yield increase of approximately 3.67 kg ha-1.6 As precipitation increases, the benefit of each additional mm decreases.

Regression results from Table 3 also indicate that larger farms have lower yields. For each one percent increase in farm size, yields decreased by approximately 0.63 kg ha-1.7 Similar results were also found by Norton et al. (2014). One probable explanation for smaller farms having higher yields per hectare is that producers can manage smaller farms more efficiently. As farms become larger, it becomes more difficult to properly maintain the cocoa trees, and losses could result from pests or disease. Furthermore, the variable CSC alone was not found to be significant. It is likely that the benefit of the input-use training program was captured by the variable (CSC*Fert), which represents producers who underwent input-use training and also used inorganic fertilizer. Model results show that the use of fertilizer combined with training (CSC*Fert) was significant (P<0.01). More notably however, is that the coefficient for CSC*Fert (156.48 kg ha-1) was higher than that of a producer who used fertilizer without training (nCSC*Fert) at 118.07 kg ha-1. The difference between the coefficient estimates of CSC*Fert and nCSC*Fert is significant (P<0.01), suggesting that training improves the effectiveness of fertilizer.8 Second, shaded production (CSCshade) was found to be significant (P<0.05) with a coefficient of 42.51 kg ha-1. The positive coefficient of the shade management variable is contrary to other, previous literature on the topic (Ahenkorah, Akorifi, and Adri 1974; Gockowski and Sonwa 2011; Gockowski and Sonwa 2008; Murray 1954). However, there are numerous reasons why shade management would result in higher cocoa yields including: the reduction of pest and disease damage (Campbell 1984) and decreased weed growth in shaded production systems (Ahenkorah, Akorifi, and Adri 1974).

4.2. Yield Simulation Results

Average simulated yield results were investigated for the two different groups: (1) those who followed CSC practices and (2) those who did not use CSC practices. As reported in Table 4, CSC producers were found to have significantly higher average yields (P<0.01) in all 19 surveyed districts. Specifically, the largest difference in yield between CSC and non-CSC producers was in Juaboso, where CSC producers yielded an estimated 77.31 kg ha-1 more cocoa than non-CSC producers, creating a disparity of 24.7%. Furthermore, the region with the lowest yield difference was Atwima Nwabiagya, where CSC producers yielded 52.26 kg ha-1 more than non-CSC producers, creating a disparity of 19.2%.

6 367.94 ∗ ≈ 3.67.

7 −63.85 ∗ ≈ −0.63. 8 Because fertilizer use is a binary variable, we are forced to assume that CSC and non-CSC producers use equal

amounts of fertilizer per hectare.


Table 4. Simulated Yield (kg/ha) Comparison between CSC and Non-CSC

Non Climate-Smart Climate-Smart Difference‡

Adansi South 275.17 341.84 66.67

(167.99) (189.53)

Ahafo Ano South 286.29 356.31 70.01

(171.49) (187.85)

Akyemansa 248.97 310.78 61.81

(162.03) (185.13)

Aowin Suaman 310.07 384.97 74.90

(178.09) (197.49)

Asin North 271.17 335.37 64.20

(171.46) (188.78)

Asunafo North 313.12 389.65 76.53

(182.43) (193.78)

Asunafo South 296.07 371.52 75.45

(174.44) (192.16)

Asutifi 297.51 366.88 69.37

(171.69) (186.44)

Atwima Nwabiagya 271.57 323.83 52.26

(188.89) (207.85)

Bia 310.81 385.85 75.04

(182.63) (191.37)

Bibiani Awiaso Bekwa 298.40 367.08 68.68

(177.14) (187.49)

Birim North 243.16 304.59 61.43

(165.90) (181.79)

Birim South 227.40 282.70 55.30

(158.25) (179.03)

Bosome Freho 239.31 292.64 53.33

(154.84) (177.95)

Juaboso 312.74 390.05 77.31

(184.22) (192.51)

Sefwi Akontombra 303.05 373.66 70.61

(183.05) (196.59)

Sefwi Wiawso 300.57 369.47 68.91

(177.75) (196.41)

Upper Denkyira West 267.42 333.95 66.54

(167.43) (188.23)

Wass Afenfi West

298.01 367.25 69.24

(176.14) (193.89)

Note: Figures in parentheses are standard errors. ‘‡’All differences are significant at P<0.01.


Table 5. Percent Chance of Indemnity Payment at the District Level

Non-CSC CSC Difference Adansi South 23.6% 21.0% 2.6%* Ahafo Ano South 23.0% 18.9% 4.1%** Akyemansa 25.3% 23.3% 2.0% Aowin Suaman 21.7% 18.5% 3.1%** Asin North 25.0% 21.8% 3.1%* Asunafo North 22.2% 17.0% 5.2%*** Asunafo South 23.5% 17.7% 5.8%*** Asutifi 20.3% 17.4% 2.9%* Atwima Nwabiagya 28.5% 25.1% 3.4%* Bia 22.9% 18.0% 4.9%*** Bibiani Awiaso Bekwa 22.4% 17.9% 4.5%*** Birim North 25.9% 23.5% 2.4% Birim South 28.8% 24.9% 3.9%** Bosome Freho 24.5% 23.4% 1.1% Juaboso 22.9% 15.9% 6.9%*** Sefwi Akontombra 22.9% 19.6% 3.3%** Sefwi Wiawso 22.2% 19.0% 3.2%** Upper Denkyira West 24.2% 20.8% 3.4%** Wass Afenfi West 21.7% 20.4% 1.3%

Note: '***', '**', and '*' are significant at the P < 0.01, 0.05, and 0.10 levels, respectively.

4.3. Indemnity Payments Results

Table 5 shows the results for the percent chance of an indemnity payment for CSC and non-CSC producers in each district. The results reveal that non-CSC producers had the highest probability of receiving an indemnity payment in all districts. This result is likely driven by the higher average yield of producers being more than 67 kg ha-1, with only slightly larger standard errors of 16.7 kg ha-1. Similar to differences in yield, the largest differences in percent chance of indemnity is found in Juaboso, and this disparity is significant (P < 0.01). The lower probability of receiving an indemnity payment for CSC producers is an encouraging result for the future of a CSC crop insurance program. Fewer indemnity payments equates to a more affordable insurance program.

4.4. Relative Standard Deviation Results

Normalized standard deviations or relative standard deviations (RSD) were computed to more fairly compare the yield variance among the two groups of producers. This is a necessary adjustment because the average yields per group are significantly different (P < 0.01). Additionally, lower estimated RSD percentages equate to lower yield risk. Table 6 shows the results of RSD percentages for CSC and non-CSC producers; and in every district, CSC producers have lower RSD than non-CSC producers. Accordingly, the primary catalyst for these results comes from the higher average yields of the CSC producers, with only


slightly higher standard deviations. Specifically, the difference in RSD between non-CSC and CSC ranges from 4% in Bosome Freho to 10% in Juaboso with an average difference in RSD of 7%. These results indicate that CSC practices reduce relative yield risk in cocoa production, although the greatest driving factor of the low RSD values for CSC producers is higher yields obtained through CSC practices.

Table 6. Relative Standard Deviation for CSC and Non-CSC

Non-CSC CSC Difference Adansi South 61.0% 55.4% 5.6% Ahafo Ano South 59.9% 52.7% 7.2% Akyemansa 65.1% 59.6% 5.5% Aowin Suaman 57.4% 51.3% 6.1% Asin North 63.2% 56.3% 6.9% Asunafo North 58.3% 49.7% 8.5% Asunafo South 58.9% 51.7% 7.2% Asutifi 57.7% 50.8% 6.9% Atwima Nwabiagya 69.6% 64.2% 5.4% Bia 58.8% 49.6% 9.2% Bibiani Awiaso Bekwa 59.4% 51.1% 8.3% Birim North 68.2% 59.7% 8.5% Birim South 69.6% 63.3% 6.3% Bosome Freho 64.7% 60.8% 3.9% Juaboso 58.9% 49.4% 9.5% Sefwi Akontombra 60.4% 52.6% 7.8% Sefwi Wiawso 59.1% 53.2% 6.0% Upper Denkyira West 62.6% 56.4% 6.2% Wass Afenfi West 59.1% 52.8% 6.3%

The largest cause of reduction in yield risk in Table 6 appears to be the completion of

input-use training coupled with access to credit, even though one of the main causes of reduction in yield risk is the increase in the average yield of CSC producers, with only slightly higher yield variance as seen in Table 4. The increase in yield seems to be less pronounced with the use of agrochemicals such as: herbicides, pesticides, and fungicides and more pronounced in the use of inorganic fertilizer. Consequently, the increase in the average yield of CSC producers is mostly driven by use of fertilizer combined with training, and to a lesser extent, by the use of shade in production. The result of higher yields in shaded production systems is however still ambiguous because the variable for shade is binary and based on the producer’s decision to use shade rather than the actual percent canopy cover.

4.5. Costs and Feasibility of CSC Crop Insurance

Although the data show that the use of CSC practices is likely to reduce costs of an insurance program through fewer indemnity payments, the question still remains: what will be the cost of insuring a hectare and who will bear this cost? It is difficult to estimate the cost


of such a program for several reasons: (1) as stated earlier, there is no exact measure of cocoa hectares harvested in Ghana; (2) farm gate prices are variable because the producer price, offered by COCOBOD is variable; (3) there is currently no stated, expected area that is to be covered by insurance from the CSCWG; and (4) the transaction and operational costs of an insurance program are unknown in Ghana. For the sake of estimating the cost of an insurance program in Ghana, the following actions were taken to overcome the aforementioned difficulties: (1) the total area harvested is allowed to be variable using maximum, minimum, and average values of FAO (2014b) total harvested estimates for the past five years, (2) the farm gate price is allowed to be variable with two parameters: first being the percent share of international cocoa price that COCOBOD gives to producers (between 60-80% recently) and the international cocoa price itself from the previous five years, (3) the total coverage area is assumed to be the entirety of the harvested area in Ghana because specific hectares for the program are unknown, and (4) the transaction and operation costs are omitted to be investigated in a later study. From these assumptions we overestimate some costs by assuming all cocoa area is covered but underestimate the costs in other areas as we assume insurance is offered at a fair market value with no transactions costs.

The results of the total estimated indemnity costs of crop insurance programs for cocoa in Ghana are shown in Table 7. The results indicate that a crop insurance program would be nearly 20% cheaper to implement if all insured producers followed CSC practices. This result holds in all three estimates: minimum, maximum, and average scenarios. The estimated cost of operating the program varies from a minimum of 6.2 million USD to a maximum of 12.7 million USD with an average estimate of 10.8 million USD, annually. These results assume that each hectare of cocoa under cultivation in Ghana is insured and all producers follow CSC practices. Considering that cocoa production in Ghana was valued at 913.2 million USD in 2012 (FAO 2014b), the average estimated cost of a CSC crop insurance program represents just over one percent of the total value of coca production in Ghana. However, this estimate does not include the operational costs of implementing such a program, so the cost of the program would be higher. In addition to operational costs, climate and precipitation variability is expected to increase in the future, and will likely raise the frequency and size of indemnity payments to producers, making the program even more expensive.

While the objective of this research was to explore if adoption of CSC would increase yields and decrease yield variability the obvious follow up question is, who is going to pay for it as most small scale producers are living hand-to-mouth? One logical funding source are private sector cocoa companies who have been using certification standards (Fair Trade, Rainforest Alliance, etc.) for years to both help the environment and small scale producers via a price premiums and access to farming resources. Private companies such as Mars, Nestle, Cadbury, Hersheys, etc. all have vested interests in the success of Ghanaian producers to manage pest and diseases to maintain supply chains. Coupling CSC trainings with crop insurance is one feasible option to better manage these threats. Beyond maintaining supply chains, private companies are also increasingly more concerned with corporate responsibility and sustainability. Again, CSC offers a potential in this area. Payment for CSC via private companies could come in the form of higher prices paid for cocoa to the marketing board which in turn would purchase blanket insurance for each producer in Ghana. Another option would be to charge cocoa consumers more per unit in an effort to appeal to the enhancement of a public good.


Table 7. Estimated Cost of Cocoa Crop Insurance Program in Ghana

Non-CSC CSC Difference Cocoa Area Harvested (ha) Minimum (a) 1,600,000 1,600,000 0.00 Maximum (b) 1,822,500 1,822,500 0.00 Average (c) 1,644,660 1,644,660 0.00 Average Indemnity Cost ($USD) ‡ Minimum (d) $19.45 $19.06 $0.39 Maximum (e) $35.12 $34.41 $0.71 Average (f) $33.16 $32.49 $0.67 Average Indemnity Probability (%) Average (g) 23.77% 20.23% 3.54% Total Cost ($USD) Minimum (a*d*g) $ 7,396,798.18 $ 6,167,996.11 $ 1,228,802.07 Maximum (b*e*g) $ 15,212,609.90 $ 12,685,396.63 $ 2,527,213.27 Average (c*f*g) $ 12,963,665.90 $ 10,810,061.18 $ 2,153,604.72

‡: calculated using average kg below CAT coverage times producer price as percent of f.o.b. cocoa price. Another potential funding source is through independent consumer labeling, supported by

some international NGOs or companies working to maintain biodiversity and ensure sustainable livelihoods through sustainable land-use practices. If all elements of CSC were enacted, including MRV, then labeling would be feasible because the land-use changes could be quantified. The premium paid by consumers for the “climate-smart cocoa” label could be used to subsidize the program. Furthermore, proper MRV could quantify reduction in deforestation and could potentially increase reforestation, both of which provide funding opportunities. On this same note, the implementation of REDD+ in Ghana could be instrumental in generating performance-based revenue from the World Bank’s Carbon Fund through implementation of Ghana’s Emission Reductions Program for the Cocoa Forest Mosaic Landscape, for which the Forestry Commission and Ghana’s Cocoa Board are dual proponents.

Finally, the fact that Ghana still has a marketing board for cocoa, COCOBOD, could be advantageous in implementing CSC insurance and the distributions of indemnities in the event of a loss. COCOBOD is deeply integrated into various aspects of cocoa production in Ghana. As such, they could be an integral part of the program by directly collecting revenues on carbon payments to distribute as indemnities in the future. Beyond managing the crop insurance program, COCOBOD could also be an important funding source by taking current subsidies, such as fertilizer and fungicide input support, and refocusing them in support of a CSC program for crop insurance.

CONCLUSION

This study investigates how average cocoa yield and yield risk are affected by adopting CSC practices. Results show that CSC producer yields averaged nearly 25% more than non-


CSC producers. Moreover, CSC producers were nearly 7% less likely to receive an indemnity payment than non-CSC producers in some districts, and had a relative standard deviation that was 6.91% less than non-CSC producers on average. Finally, our work analyzes the financial feasibility of providing crop insurance for cocoa producers who follow climate-smart agricultural practices in Ghana. Future research should investigate the incentive aspects of such a program.

The practices embodied in CSC appear to reduce relative yield risk as well as increase average yields. Such an effect is largely driven by proper training and access to agro-chemical inputs, particularly inorganic fertilizer. Furthermore, our results lead to the conjecture that increased average yields and less yield risk for CSC producers could have beneficial effects on minimizing deforestation in the future given that participation in a CSC program would require that producers no longer enter protected forests for cocoa production. If CSC producers were to receive proper training and access to fertilizer and agro-chemicals, then they could increase production without encroaching further into forests.

Because CSC is interested in working at the landscape level, WII seems to be well-suited since it also works on the landscape level rather than by individual producer. Besides working at the landscape level, WII insurance is also a more suitable option because the cost of operating such a program is less than a traditional MPCI. The yield risk reductions can primarily be attributed to access to training, inputs, and credit. Increased yields and less yield variation could prevail in the future by providing producers with the means to better manage non-weather related risks (e.g., pests, disease, etc.), coupled with an insurance package such as WII, which only covers weather-related risks.

The sample size of 1,200 households was satisfactory but the scope of the data were limited. For example, many variables were collected as binary variables (yes or no) such as: fertilizer use and the use of shade management. The predictive power of the model could be improved by using continuous observations for these variables as well as including more years. Although the scope of the data was limited for this study, the results were promising for the feasibility of CSC crop insurance. If an insurance program were instituted in the future, the predictive power of models would surely improve as data are collected from participants in the insurance program.

The analysis suggests that CSC can in fact have positive impacts on both the producers and the environment in Ghana. However, providing evidence that the program would work is only half of the battle; funding it is likely the more difficult obstacle. Unfortunately, the environmental services provided by forests are a public good; and as such, it is difficult to motivate private individuals to pay for these services. On the other hand, it is possible to acquire the funds to partially subsidize crop insurance through a combination of funding from private sector cocoa and chocolate companies who have an interest in the sustainability of supply, emission reductions payments via Ghana’s Cocoa Forest REDD+ Program, and possible subsidies from COCOBOD. The CSC insurance program for cocoa should be implemented in Ghana to increase yields and reduce yield variance for producers. While the program appears to be economically feasible if adopted, future research should investigate if the CSC program is successful at minimizing deforestation and increasing reforestation.


ACKNOWLEDGMENTS

The authors would like to extend their deepest gratitude to the World Cocoa Foundation for providing household data as well as technical support during this research, the U.K. Department for International Development (DFID) for providing in-country financial support to portions of the work, Natalie Shew for providing technical edits on the manuscript, Diana Danforth for programming assistance, the two journal referees and the editor for taking the time to read the manuscript, and especially the farmers who participated in this study.

REFERENCES

Ahenkorah, Yaw, G S Akorifi, and A K Adri. 1974. “The End of the First Cocoa Shade and Manurial Experiment at the Cocoa Reserach Institute of Ghana.” Journal of Horticultural Science 49: 43–51.

Asare, Rebecca Ashley. 2010. “Implications of the Legal and Policy Framework for Tree and Forest Carbon in Ghana: Redd Opportuniities Scoping Exercise.” Washington DC. http://www.forest-trends.org/publication_details.php?publicationID=245276.

———. 2014. “Understanding and Defining Climate-Smart Cocoa : Extension, Inputs, Yields, and Farming Practices.” Accra.

Asare, Richard, Rebecca Ashley Asare, Anders Ræbild, and Gilbert J. Anim-Kwapong. n.d. “On-Farm Cocoa Yields Increase with Emergent Shade Cover in Two Regions in Ghana.”

Asomanin, E J A, R S Kwakwa, and W V Hutcheon. 1971. “Physiological Studies on an Amazon Shade and Fertilizer Trial at the Cocoa Research Institute, Ghana.” Ghana Journal of Agricultural Science 4: 47–64.

Awhere Incorporated. 2015. “The Power of Agricultural Intelligence.” http://www.awhere .com.

Barnett, Barry J, and Keith H Coble. 1999. “Understanding Crop Insurance Principles : A Primer for Farm Leaders.” Miississippi State.

Berry, Sara. 1992. “Hegemony on a Shoestring: Indirect Rule and Access to Agricultural Land.” Africa: Journal of the International African Institute 62 (3): 327–55.

Boyer, J. 1974. “Etude Ecophysiologique Du Developpement de Cacaoyers Cultives Au Cameroun.” Café, Cacao, Thé 18: 3–30.

Campbell, C A M. 1984. “The Influence of Overhead Shade and Fertilizers on the Homoptera of Mature Upper-Amazon Cocoa Trees in Ghana.” Bulletin of Entomological Research 74: 163–74. doi:C. A. M. Campbell (1984). The influence of overhead shade and fertilizers on the Homoptera of mature Upper-Amazon cocoa trees in Ghana. Bulletin of Entomological Research, 74, pp 163-174. doi:10.1017/S0007485300010038.

CIA. 2014. “The World Fact Book.” https://www.cia.gov/library/publications/the-world-factbook/geos/gh.html.

———. 2015. “Ghana.” The World Factbook: https://www.cia.gov/library/publications/the-world-factbook/geos/gh.html.

Cole, Shawn, Xavier Giné, and James Vickery. 2013. “How Does Risk Management Influence Production Decisions ? Evidence from a Field Experiment.” 13-080. Boston.


Cunningham, R K, R W Smith, and R G Hurd. 1961. “A Cocoa Shade and Manurial Experiment at the West African Cocoa Research Institute, Ghana II. Second and Third Years.” The Journal of Horticultural Science & Biotechnology 36 (2): 116–25.

De Leeuw, Jan, Anton Vrieling, Apurba Shee, Clement Atzberger, Kiros Hadgu, Chandrashekhar Biradar, Humphrey Keah, and Calum Turvey. 2014. “The Potential and Uptake of Remote Sensing in Insurance: A Review.” Remote Sensing 6 (11): 10888–912. doi:10.3390/rs61110888.

Dercon, Stefan, and Luc Christiaensen. 2011. “Consumption Risk, Technology Adoption and Poverty Traps: Evidence from Ethiopia.” Journal of Development Economics 96 (2): 159–73.

Dercon, Stefan, Jan Willem Gunning, and Andrew Zeitlin. 2011. “The Demand for Insurance under Limited Credibility : Evidence from Kenya.”

Dormon, E. N. A., A. Van Huis, C. Leeuwis, D. Obeng-Ofori, and O. Sakyi-Dawson. 2004. “Causes of Low Productivity of Cocoa in Ghana: Farmers’ Perspectives and Insights from Research and the Socio-Political Establishment.” NJAS - Wageningen Journal of Life Sciences 52 (3-4). Koninklijke Landbouwkundige Vereniging (KLV): 237–59. doi:10.1016/S1573-5214(04)80016-2.

FAO. 2013. “Climate-Smart Agriculture Sourcebook.” Rome, Italy. ———. 2014a. “FAOSTAT.” http://faostat.fao.org/desktopdefault.aspx?pageid=342&lang

=en&country=81. ———. 2014b. “FAOSTAT.” Rome, Italy: Food and Agriculture Organization of the United

Nations. http://faostat.fao.org/site/567/default.aspx#ancor. FCPF. 2014. “Emission Reductions Program Idea Note ( ER-PIN ).” Accra. GAIP. 2013. “Ghana Agricultural Insurance Programme.” Accra. http://www.gaip-

info.com/wp-content/uploads/gaip-brochure-screen.pdf. Gille, Susan. 2013. “Challenges and Lessons Learned in Implementing a Sustainable

Agricultural Insurance System.” Bonn: Gesellschaft für Internationale Zusammenarbeit. Glendinning, D R. 1972. “Natural Pollination of Cocoa.” New Phytologist, no. 71: 719–29. Gockowski, James, and Denis Sonwa. 2008. “Biodiversity and Smallholder Cocoa Production

Systems in West Africa.” 6. STCP Working Paper Series. Accra. ———. 2011. “Cocoa Intensification Scenarios and Their Predicted Impact on CO2

Emissions, Biodiversity Conservation, and Rural Livelihoods in the Guinea Rain Forest of West Africa.” Environmental Management 48 (2): 307–21. doi:10.1007/s00267-010-9602-3.

ICCO. 2013. “Growing Cocoa.” http://www.icco.org/about-cocoa/growing-cocoa.html. IFAD. 2011a. “Rural Poverty Report 2011.” Rome, Italy. ———. 2011b. “Weather Index-Based Insurance in Agricultural Development.” Rome, Italy. Jozwik, Jan. 2013. “Research on Perils in Cocoa Production and Agri-Cultural Index

Insurance Options for GAIP.” Bonn. Karlan, Dean, Robert Osei, Isaac Osei-Akoto, and Christopher Udry. 2012. “Agricultural

Decisions after Relaxing Credit and Risk Constraints.” 18463. Cambridge, MA. Kenneth, Ofori-boateng, and Insah Baba. 2011. “An Empirical Analisis of the Impact of

Climate Change on Cocoa Production in.” Journal of Sustainable Development in Africa 13 (8): 24–50.


Linnerooth-Bayer, Joanne, Reinhard Mechler, and Stefan Hochrainer-Stigler. 2011. “Insurance against Losses from Natural Disasters in Developing Countries.” Journal of Integrated Disaster Risk Management 1 (1): 59–81. doi:10.5595/idrim.2011.0013.

Morduch, Jonathan. 1995. “Income Smoothing and Consumption Smoothing.” The Journal of Economic Perspectives 9 (3): 103–14.

Murray, D B. 1954. “A Shade and Fertiliser Experiment with Cacao.” In Annual Report Cacao Research Imerial College Tropical Agriculture1, 102:25. St. Augustine. http://www.ncbi.nlm.nih.gov/pubmed/19943419.

NCRC. 2012. “Climate-Smart Cocoa in Ghana Achievements and a Way Forward.” Accra. Norton, Mike, L. Lanier Nalley, Bruce L Dixon, and Jennie S Popp. 2014. “Cost-Benefit

Analysis of Farmer Training Schools: The Case of Ghanaian Cocoa.” Journal of International Agricultural Trade and Development 9 (1).

Ntiamoah, Augustine, and George Afrane. 2008. “Environmental Impacts of Cocoa Production and Processing in Ghana: Life Cycle Assessment Approach.” Journal of Cleaner Production 16 (16): 1735–40. doi:10.1016/j.jclepro.2007.11.004.

Palisade Corporation. 2014. “@RISK.” Ithica: Palisade Corporation. Roberts, Richard AJ. 2005. Insurance of Crops in Developing Countries. Rome: Food and

Agriculture Organization of the United Nations. Ruf, François Olivier, and Götz Schroth. 2004. “Chocolate Forests and Monocultures : A

Historical Review of Cocoa Growing and Its Conflicting Role in Tropical Deforestation and Forest Conservation.” In Agroforestry and Biodiversity Conservation in Tropical Landscapes, 06:107–34. Island Press.

Ruf, François Olivier, and H Zadi. 1998. “Cocoa: From Deforestation to Reforestation.” Workshop on Shade Grown Cocoa. Washington DC.

Ruf, François Olivier, Götz Schroth, and Kone Doffangui. 2014. “Climate Change, Cocoa Migrations and Deforestation in West Africa: What Does the Past Tell Us about the Future?” Sustainability Science 10 (1): 101–11. doi:10.1007/s11625-014-0282-4.

Scheel, William C. 1978. “Comparisons of Riskiness as Measured by the Coefficient of Variation.” The Journal of Risk and Insurance 45 (1): 148. doi:10.2307/251816.

Steel, William F., and David O. Andah. 2003. “Rural and Micro Finance Regulation in Ghana : Implications for Development and Performance of the Industry.” 49. World Bank - Africa Region Working Paper Series. Africa Region Working Paper Series. Washington DC. http://www.worldbank.org/afr/wps/wp49.pdf.

Stutley, Charles J. 2010. “Innovative Insurance Products for the Adaptation to Climate Change Project Ghana (IIPACC).” Bonn.

Todaro, Michael P., and Stephen C. Smith Smith. 2012. Economic Development. 11th ed. Boston: Addison-Wesley.

Triplett, Jack E, and Barry P Bosworth. 2006. “‘Baumol’s Disease’ Has Been Cured: IT and Multifactor Productivity in U.S. Services Industries.” In The New Economy and Beyond, Past, Present and Future, edited by D W Jansen, 1995:34–71. Cheltenham, UK: Edward Elgar Publishing.

USDA. 2011. “Risk Management Glossary.” Washington DC. http://www.rma.usda.gov/pubs /2011/rma_glossary.pdf.

Vilsack, Thomas J. 2009. “Report to Congress : Perennial Crops , Pecans and the Federal Crop Insurance Program Stop 0801.” Washington DC.


WCF. 2014a. “Cocoa Livelihoods Program Dataset.” Accra, Ghana: World Cocoa Foundation.

———. 2014b. “WCF Cocoa Livelihoods Program.” http://worldcocoafoundation.org/wcf-cocoa-livelihoods-program/.

Wilson, K C. 1999. Coffee, Cocoa and Tea. Cambridge, MA: CABI Publishing. Wood, G A R, and R A Lass. 1985. Cocoa. 4th ed. Ames, IA: Blackwell Science. World Bank. 2014. “Risk and Opportunity.” Washington DC. doi:10.1596/978-0-8213-9903–3. Yang, Jing. 2012. “Interpreting Coefficients in Regression with Log-Transformed Variables.”

83. StatNews.


COMPETITIVENESS AND COMPARATIVE ADVANTAGE OF IMPORTANT FOOD AND INDUSTRIAL

CROPS IN PUNJAB: APPLICATION OF POLICY ANALYSIS MATRIX

Abdul Salam1 and Sadia Tufail2 1Professor of Economics,

Federal Urdu University of Arts, Science and Technology, Islamabad 2M. Phil. Student,

Federal Urdu University of Arts, Science and Technology, Islamabad

ABSTRACT

The study was designed to ascertain the competitiveness and comparative advantage of wheat, rice (food crops), cotton and sugarcane (industrial crops), in the Punjab. Analysis of crop budgets for wheat, rice, cotton and sugarcane, 2010 to 2012 crops, has generally indicated their profitability and competitiveness. The degree and extent of competitiveness, however, were sensitive to output prices. The results of social profitability analysis have, inter alia, indicated comparative advantage of Punjab in production of both these groups of food and industrial crops. The domestic resource cost coefficients for all these crops, though varying from year to year, were consistently less than one, confirming their comparative advantage. The nominal protection coefficients for wheat and sugarcane were consistently less than one, indicating implicit taxation of their domestic producers. Estimates of effective protection coefficients for these crops also confirmed their implicit taxation. Estimation and analyses of protection coefficients suggest continued taxation of the food and industrial crops in the Punjab and significant resource transfers from their producers, undermining incentives to domestic production.

Keywords: competitiveness, comparative advantage, protection, implicit taxation, incentives, distortions, resource transfers, revenue, policy analysis matrix

JEL classification codes: Q17, Q18

1. INTRODUCTION

Agriculture in Pakistan has experienced significant transformation and progress since independence in 1947. At the time of independence there was not much use of purchased

Abdul Salam and Sadia Tufail 82

inputs in farm production and bulk of the production was directed for domestic use except for industrial and export crops. In the 1960s many favorable developments like the introduction of high yielding seeds of wheat and rice, increasing use of fertilizers, tractor, tube well irrigation and other irrigation related projects helped in expanding farm output. These trends continued in the subsequent period as well. The successive governments adopted a number of initiatives such as development of rural infrastructure, pricing policy aimed at assuring a reasonable return to farmers by stabilizing output prices and subsidizing prices of important inputs. The government policy interventions impacting on prices of farm inputs and outputs have been periodically reviewed in the light of domestic and international developments, pressures and availability of resources. These policy revisions and shifts, influencing the structure of production incentives of various farm enterprises have been the subject of several studies: (Appleyard 1987, Hamid, Ijaz and Anjum 1990, Orden et al. 2006, Dorosh and Salam, 2007, Salam 2009 and 2010).

Wheat and rice are the staple food crops. Their production also provides the raw material for wheat flour and rice mills. Cotton farming provides raw material for the textile industry, the largest industry in Pakistan. Cotton seed, a byproduct from cotton production, is an important source of edible oil and extensively used by oil mills in the country. Sugarcane is the principal source of raw material for sugar mills, the 2nd largest industry, manufacturing an important constituent of the food basket in the country. The production, marketing and processing of these important food and industrial crops are in the private sector and important sources of employment, both on and off farm. Total contribution of these four crops is estimated at 6 percent in the GDP, 29 percent of value added by agriculture sector and 91 percent of the value added by the major crops in the country (Government of Pakistan 2012). These crops are important not only for food security, industrial growth and employment generation but also from the perspective of trade and balance of payments, since rice and cotton made ups are amongst the major exports. A good harvest of wheat and sugarcane crops can make a difference between imports and exports of wheat and sugar. In view of the foregoing, it is important to examine and analyze not only their financial viability from the producers’ view point but also their comparative advantage from the country’s perspective. The subject is of critical importance and interest for policy planners and those interested in the development of agriculture. Punjab, with more than 55 percent of the total population (Government of Pakistan 2012) and 57 percent of the total cultivated area is a dominant player in the agriculture sector of Pakistan. Wheat, rice, cotton and sugarcane are the most important crops in the province, contributing 75, 68, 70 and 68 percent, respectively, of their total production in the country (Government of Pakistan 2013) It is in this context and background that the present study has been designed to examine the competitiveness and comparative advantage of wheat, rice, cotton and sugarcane farming in the Punjab- the major producer of these commodities in Pakistan, since results of the earlier studies, especially those conducted in 1980s (Appleyard 1987), and 1990s (Hamid, Ijaz and Anjum 1990), may have become dated.

Rest of the paper is organized under following sections. Objectives of the study are set out in section 2. Section 3 on methodological framework and data explains the rationale and construction of policy analysis matrix and its data requirements. Different measures and indicators for estimating competitiveness, comparative advantage and distortions to incentives in production of various farm enterprises are also explained in section 3. Findings of empirical analysis regarding private and social profitability and distortions to incentives in

Competitiveness and Comparative Advantage of Important Food … 83

production of important crops are detailed in section 4. The salient features of empirical results and policy implications thereof are summarized in section 5.

2. OBJECTIVES OF THE STUDY

The principal objectives of this study are: To examine financial viability and competitiveness of cultivation of wheat, long

grain aromatic basmati rice, cotton and sugarcane crops in the Punjab, for 2009-10, 2010-11 and 2011-12 crop years.

To analyze the comparative advantage of Punjab in production of wheat, basmati rice, cotton and sugarcane cultivation in Punjab – Pakistan and examine their comparative advantage; 2009-10 to 2011-12 crops.

To estimate distortions to incentives in the domestic production of wheat, rice, cotton and sugarcane in the Punjab.

3. METHODOLOGICAL FRAMEWORK AND DATA

3.1. Methodological Framework

The Policy Analysis Matrix (PAM), developed by Monke and Pearson, provides a conceptual and analytical framework for estimating the financial viability and comparative advantage of various farm enterprises. PAM as a tool of analysis provides a complete coverage of policy impacts on costs and revenues. The format of PAM is based on two accounting identities: one defining profitability as the difference between revenues and costs and the other measuring the effects of divergences (distorting policies and market failure) as the difference between the observed parameters and parameters that would exist if the divergences were removed. By filling in the entries in PAM for a given enterprise, the analyst can estimate the extent of transfers imposed on the enterprise by a set of policies as well as its inherent economic efficiency (Monke and Pearson 1989). Analysis of PAM data begins with estimation of existing private and social profits, revenues and costs. Different measures of private and social profitability and distortions to incentives in the production of a given enterprise, as used in the analysis here, are set out below.

Private profit. It is the excess of gross revenue over total costs entailed in

production of a given crop / enterprise; costs and revenues, both estimated at market prices. It is a measure of the competitiveness of a given enterprise. In terms of symbols in table 1:

Private profit = D = A – (B+C) (1)


Output – input ratio. It is the ratio between gross revenue and total costs, and indicative of the overall index of relative profitability of various enterprises. In terms of symbols used in table 1 :

Output- input ratio = A/(B+C) (2) Profitability ratio. It is the cost of domestic factors per unit of value added at private

prices and used to compare relative profitability of various crops. In terms of symbols in table 1:

Profitability ratio= C/(A–B) (3) Social profit. This is an indicator of the efficiency or comparative advantage of a

given enterprise. It is estimated by subtracting the total costs of a given enterprise from its gross revenues, both calculated at social prices. A value of social profit >0 indicates social profitability of the enterprise. In terms of symbols employed in Table 1:

Social profit, H = (E– F– G) (4) Domestic resource cost coefficient (DRC). It is the ratio between cost of domestic

factors and value added at social prices. It represents the cost of domestic factors to earn / save a unit of foreign exchange from the crop under consideration. In terms of symbols used in table 1:

DRC = G/(E – F) (5) The value of DRC > 1 is indicative of comparative disadvantage and a DRC < 1 is a case

of comparative advantage. Nominal protection coefficient (NPC) is defined as the ratio between gross revenue

at private prices and gross revenue at social prices. In terms of symbols used in table 1:

NPC = A/E (6) When the value of NPC > 1, it indicates protection and encouragement to domestic

production and when its value < 1 it shows implicit taxation of domestic producers. Effective protection coefficient (EPC) is the ratio between the value added at

private prices and value added at social prices. In terms of symbols used in table 1: EPC = (A – B)/(E – F) (7)


Table 1. Policy Analysis Matrix

Items Revenue Costs Profit Tradable Inputs Domestic Factors

Private Prices A B C D Social Prices E F G H Divergence I J K L

Source: Monke and Pearson (1989).

When value of EPC > 1, net result of interventions in input and output markets is

protection to the domestic producers and when EPC < 1, net result of various interventions is discouragement and implicit taxation of domestic production.

The formulae for various coefficients, as given above, are based on the discussions in Scandizzo and Bruce (1980) and Monke and Pearson (1989).

3.2. Data Requirements of Policy Analysis Matrix

The construction of PAM for various crops and farm enterprises requires data relating to respective crops’/enterprise budgets: data on the use level of various inputs, prices thereof and crop output and its market prices. Empirical data used in this study were drawn from the crop budgets reported in the policy reports of the Agriculture Policy Institute (API). These were supplemented, where needed, from the data and analysis reported in Sadia Tufail’s (2014) M. Phil. Thesis. This thesis has also used crop budgets presented in the policy reports of the API, successor to the Agricultural Prices Commission, Government of Pakistan. A few points on the API’s data and crop budgets are in order.

The API periodically organizes crop specific farm surveys in major producing areas of the crops to collect farm level inputs-output data. From these input - output data crop budgets are prepared for various farm groups. The survey respondents at the time of data analysis, inter alia, are classified into: “progressive”, “average” and “traditional” farmers. The criteria used in this classification relate to the adoption of technology, use level of modern inputs, cultural practices, etc., on the farms (Salam, Siddiqui and Zaidi 2002). The API in its policy analysis use crop budgets of “average category” of farmers and report these budgets in its annual policy reports on various crops. These “average” farms fall in between the “progressive” and “traditional” farms in terms of their adoption of technology, use level of modern inputs and cultural practices. Based on these crop budgets and prices of farm inputs and outputs policy analysis matrices were constructed for the commodities included in this study and are given at Annexes 1 to 3.

Farm inputs, as used in raising various crops, for constructing PAM are divided into tradable and non- tradable while prices of inputs and outputs are differentiated into private and social prices and explained below.

Tradable inputs include all those inputs which were either purchased or can be traded

in the international market. In our analysis these are seed, chemical fertilizers, pesticides, services of farm machinery i.e., tractor, thresher and tube well etc.

Non-tradable inputs include land, labor, and farm yard manure and canal water.


Private prices are the market prices faced by the farmers in their purchases of inputs and selling their outputs.

Social prices of inputs and outputs reflect the opportunity cost for the resources used in their domestic production. The social or economic prices of various farm inputs and output as used in our analysis were estimated from their corresponding international prices, adjusted for import / export costs entailed in the process.

4. EMPIRICAL RESULTS FROM POLICY ANALYSIS MATRIX

From the data presented in the PAMs, Annexes 1-3, various indicators of competitiveness, comparative advantage and distortions to incentives in production of different crops, as detailed in section 3.1 above, were estimated. These indicators and co-efficients are discussed below.

4.1. Private Profitability and Competitiveness

The private or actual market prices reflect the underlying economic costs and valuation and the effects of all policies and market failures (Monke and Pearson 1989). The calculations of private profitability of various enterprises, as provided in the relevant column of PAM in Table 2, show their competitiveness or otherwise; positive values indicating competitiveness while negative values implying non competitiveness of the enterprise under consideration. As all the entries in the profitability column in table 2, except that for wheat in 2011-12 crop year, were positive i.e., revenues exceeding all the costs incurred in the use of tradable and non- tradable factors, producers received above normal returns. However, in case of 2011-12 wheat crop gross revenues fell short of the expenses involved in its production, leading to negative entry in profits’ column. As farmers received suboptimal rate of returns in wheat during 2011-12 crop year, this was likely to discourage its future cultivation, if the things continued in this fashion.

In view of the varying use levels of farm inputs and differences in duration of growing periods of various crop absolute values of profits may not be appropriate for inter crop comparisons. Accordingly, we estimated the ratio of the domestic factors’ costs and the value added at private prices for various commodities. The ratio between the costs of domestic factors and value added at private prices shows how much a given enterprise can afford to pay domestic factors and still remain competitive (Monke and Pearson 1989). Assuming cost minimization behavior of farmers, lower the ratio between cost of domestic factors and the value added, higher the profitability ranking of that crop/ enterprise. These rankings are subject to change over time, depending on the changes in the prices of outputs and inputs and or on account of technological developments resulting in changes of the underlying input - output relationship. In addition, we calculated the ratios between gross revenues and gross costs i.e., output –input ratio for various crops to estimate the overall rate of return to the costs incurred in the process. These ratios are also presented in Table 2.

During 2009-10 crop year sugarcane enjoyed the highest profitability ranking in terms of domestic factors cost per unit of value added, followed by cotton and a tie between basmti rice and wheat for third place. The analysis based on the ratio between gross revenue and


gross costs also confirmed these rankings, also indicated very high rate of return to investment in sugarcane cultivation. During 2010-11 crop year profitability rankings of crops changed, primarily on account of changes experienced in output prices. There was a dramatic improvement in the ranking of cotton, rising to first position, followed by sugarcane with not much improvement in comparative positions of rice and wheat. The output - input ratios for both cotton and sugarcane, being greater than two, were indicative of very high returns to investments in their farming. The situation in 2011-12 crop year witnessed a significant improvement in profitability ratio of basmati rice, resulting in number one position for this crop, followed by sugarcane and cotton in that order. Wheat having sub optimal rate of returns had negative profitability as gross revenues at market prices were not sufficient to recover all the costs entailed in its production.

Table 2. Private Profitability and Competitiveness of Important Crops: 2010 to 2012

Crops Private profit Profitability ratio Output - input ratio

Rs. / acre C /( A-B) A/(B+C) 2009-10 Crop

Wheat 3,702 0.77 1.16 Basmati 3,557 0.77 1.16 Cotton 7,413 0.65 1.29 Sugarcane 52,515 0.30 2.39

2010-11 Crop Wheat 4,061 0.77 1.17 Basmati 5,035 0.72 1.20 Cotton 30,537 0.35 2.02 Sugarcane 53,938 0.36 2.04

2011-12 Crop Wheat (1,072) 1.07 0.97 Basmati 12,824 0.58 1.39 Cotton 2,634 0.90 1.06 Sugarcane 22,711 0.64 1.33

Notes: Private profit= Gross revenue at private prices – Total costs at market prices. Values in parenthesis are in negative. Private cost ratio indicates the cost of domestic factors per unit of value added at market prices. Output-input ratio is the ratio between gross revenue and total costs of all inputs, valued at market prices. The output–input ratio for sugarcane though had declined as compared to previous two

crop years but still suggested substantial rate of return as was also true for cotton. These results of profitability and competitive analysis suggest relative rankings of various crops with regard to their profitability were quite sensitive to changes in output prices.

4.2. Social Profitability and Comparative Advantage

Social profits, equaling social revenue less social costs, measure efficiency or comparative advantage of a production system or farm enterprise (Monke and Pearson 1989).


Data relating to social profits and domestic resource cost coefficients of crops, under study, are presented in table 3. As all the entries in the column for social profits, for both food grains (wheat and basmati) and industrial crops of cotton and sugarcane, for the three crop years included in this study, were positive, cultivation of all these crops was economical from the country’s perspective. In other words, Punjab has had a comparative advantage in their cultivation during the crop years under reference. Prima facie, sugarcane production seemed much more attractive as it had the highest, though varying, values for social profitability. In view of the varying intensity of factor inputs used and varying periods of crop duration we turn to DRCs to rank the relative position of various crops in terms of their comparative advantage and economic efficiency. The DRC measures the country’s international comparative advantage in production and foreign exchange generating capacity of specific production activities (FAO 1991).

Table 3. Social Profitability and Comparative Advantage: 2010 to 2012

Crops Social profit DRC

Rs/acre 2009-10 Crop

Wheat 6,890 0.65 Basmati rice 8,229 0.66 Cotton 5,144 0.65 Sugarcane 73,836 0.24

2010-11 Crop Wheat 16,489 0.45 Basmati rice 2,514 0.88 Cotton 24,774 0.40 Sugarcane 81,332 0.27

2011-12 Crop Wheat 11,250 0.60 Basmati rice 5,551 0.81 Cotton 7,285 0.77 Sugarcane 33,408 0.55

Note: Social profit= Revenue – (costs of tradable and non- tradable factors), all valued at social prices. Appleyard (1987) noted since DRC coefficient shows the domestic resource costs

incurred per unit of foreign exchange earned or saved, if the value of DRC is greater than one then comparative disadvantage exists and when DRC was less than 1 it implied comparative advantage.

2009-10 crop: The DRCs estimated, for all the crops, for 2009-10 crop year were less than one, (<1), thus Punjab province enjoyed comparative advantage in production of wheat, basmati rice, cotton and sugarcane. In terms of inter crop comparison, sugarcane with the lowest value of DRC, 0.24, had the greatest comparative advantage, followed by wheat and cotton, and rice with not much difference in their relative positions of economic efficiency in resource use.

2010-11 crop: Sugarcane, with its DRC estimated at 0.27, lowest of all the crops included in analysis, again enjoyed the highest ranking in terms of comparative advantage. It


was followed by cotton and wheat as basmati rice, an important export, had the lowest ranking for comparative advantage.

2011-12 crop: For the 2011-12 crop year again sugarcane had the lowest value of DRC, 0.55, and the highest ranking in terms of comparative advantage followed by wheat, cotton and basmati in that order. The results of comparative analysis, as discussed above, are sensitive to the changes in international commodity prices. Nevertheless, comparative advantage of Punjab in production of wheat, basmati rice, cotton and sugarcane crops seems well established. These results of comparative advantage of Punjab in the context of crops mentioned above are in line with those published by Appleyard based on crop budgets of the 1980s (Appleyard 1987).

The comparative advantage of both sugarcane and wheat as estimated at import parity prices is limited to the extent of import substitution. The comparative advantage of Punjab - Pakistan in exporting wheat and sugarcane would be predicated on significant improvements in their efficiency of production, processing and marketing of these crops.

4.3. Distortions to Incentives

To examine and analyze incentives to domestic production and distortions thereof, nominal and effective protection coefficients, from the data provided in PAM (Annexes 1-3), were estimated for each crop by crop year. These coefficients are presented Table 4.

2009-10 crop year: As per estimates of the nominal protection coefficients, presented in table 4, only the coefficient for seed cotton was greater than one, (>1), in this season. Thus, cotton farmers in domestic market received higher prices than corresponding world prices. Domestic seed cotton prices on the average were 11 percent higher than its export parity prices, as worked back from export prices of lint. Estimate of effective protection coefficient, reflecting the net result of various interventions in input as well as product markets, also endorsed protection and incentive to domestic cotton producers. However, farmers producing wheat, sugarcane and basmati rice in 2009-10 crop year were implicitly taxed, as the output prices received by them were less than their corresponding opportunity costs, estimated from the world prices. Implicit taxation of these crops worked out to 11 percent for wheat, 20 percent for sugarcane and 26 percent for basmati (paddy). These findings of implicit taxation of wheat and basmati confirm continued taxation and resource transfers from the domestic producers as also found in the earlier studies by Appleyard (1987), Hamid, Ijaz and Nasim (1990) Dorosh and Salam (2009) and Salam (2009 and 2010). The calculations of EPCs also supported the contention of implicit taxation of wheat, sugarcane and basmati, which in fact increased somewhat on account of various taxes imposed, by the government, on farm inputs.

2010-2011 Crop: Estimates of NPCs and EPCs obtained from the PAM for this crop season, are quite similar to the ones discussed above for the 2009-10 crop year. Therefore, inferences drawn from these coefficients about incentives, and protection to cotton crop and implicit taxation of domestic production of wheat, basmati, and sugarcane crops during 2010-11 were similar, with marginal differences, to those discussed and explained earlier with respect to 2009-10 crop year.

2011-12 Crop Year: The values of NPCs and EPCs for 2011-12 crop season, as reported in table 4, for wheat and sugarcane are less than one as was also the case for 2009-10 and 2010-11crop years. Thus, producers of both these crops continued to face implicit taxation


and resource transfers like the previous two crop years. Magnitude of net taxation of wheat as suggested by EPC in 2011-12, however, increased to 45 percent from 42 percent in 2010-11 while that of sugarcane declined to 16 percent as compared to 26 percent in the previous crop year. The situation for cotton with NPC and EPC estimates of 0.65 and 0.86, respectively, in 2011-12 changed to one of implicit taxation from that of protection, noted in the previous two crop seasons when its NPCs and EPCs hovered around 1.11 and 1.13 in 2009-10 and 2010-2011 crop years, respectively. For basmati (paddy) also position of incentives in 2011-12 experienced a qualitative change; from one of implicit taxation of 14 to 36 percent in the previous two crop seasons to that of marginal protection of 1-3 percent in 2011-12.

CONCLUDING REMARKS AND POLICY IMPLICATIONS

Analysis of crop budgets for wheat, rice -basmati, cotton and sugarcane, 2010 to 2012 crops has generally indicated their profitability and competitiveness. Farmers producing these crops were able to recover their total costs and earn some profit. The competitiveness of various crops however remains sensitive to output prices as reflected by varying amounts of profit in general and negative profit for wheat during 2011-12 crop year in particular. The results of social profitability analysis have, inter alia, indicated comparative advantage of Punjab in production of wheat, basmati rice, cotton and sugarcane crops. The DRCs for all these crops, though varying from year to year, were consistently less than one, confirming comparative advantage of Punjab in production of the above mentioned crops. The DRC estimates for sugarcane, though showing considerable inter year fluctuations nevertheless have also confirmed comparative advantage of Punjab in its production, notwithstanding its high water requirements. Our analysis of comparative advantage of wheat, basmati rice, cotton and sugarcane in Punjab has also endorsed the results obtained by Appleyard (1987) from his analysis based on crop budgets related to the 1980s.

Distortions in incentives to domestic production of food and industrial crops, listed above, were analyzed by estimating their NPCs and EPCs. The results have revealed no protection to wheat under importing scenario. The prices received by wheat farmers were less than its import parity price, indicating implicit taxation of domestic producers. The EPC estimates also confirmed implicit taxation of wheat. The NPCs for basmati paddy, for 2009-10 and 2010-11 crops, being less than one, indicated its implicit taxation. However, the situation changed somewhat during 2011-12 crop year, as the paddy prices in domestic market marginally exceeded its export parity prices. The comparison of domestic prices of seed cotton, with corresponding world prices, worked back from export prices of lint during the study period, has indicated some protection to cotton. EPC analysis has also confirmed protection to cotton. This may have been on account of expanding domestic textile industry and thriving exports of cotton made ups. The NPC’s for sugarcane, 2010 to 2012 crops, have indicated no protection to sugarcane. The prices of sugarcane received by farmers in domestic market were considerably less than import parity price of sugarcane, worked back from international prices of white sugar, suggesting implicit taxation of domestic producers. EPC estimates have also endorsed the implicit taxation of sugarcane growers. The results are consistent but sensitive to shifts in international prices of outputs as well as inputs.

The analysis of the NPCs and EPCs as discussed above suggest continued taxation of the important food and industrial crops in the Punjab and significant resource transfers,


undermining incentives to domestic producers. In the earlier studies on the subject: Appleyard (1987), Hamid, Nabi and Nasim (1990), (Government of Pakistan 1988), Dorosh and Salam (2009) and Salam (2009 and 2010) had also reported significant implicit taxation of important crops in the country. In spite of many economic reforms and policy initiatives leading to reduced role of public sector and increasing role of private sector in farm output and input markets implicit taxation of domestic production of important food and industrial crops continues, as also noted elsewhere (Salam 2012), depriving the sector of investment funds via resource hemorrhages. Implicit taxation and resource transfer on the scale, 11 to 45 percent, as observed in this study has adverse impacts not only on farm investments but also on rural poverty and employment as a substantial proportion of the production of these crops is contributed by small and marginal farmers. As per the data from 2010 Agricultural Census, farms operating less than 12.5 acres accounted for 60 percent of wheat area, 57 percent of rice, 56 percent of cotton and 51 percent of the sugarcane acreage in the country (Government of Pakistan 2012). In view of the importance of agriculture in the development of economy and rural poverty alleviation resource transfers from agriculture through implicit taxation needs to be arrested.

Table 4. Nominal and Effective Protection Coefficients : 2010 to 2012

Crop Nominal protection coefficient Effective protection coefficient 2009-10 Crop

Wheat 0.90 0.84 Basmati 0.74 0.65 Cotton 1.12 1.11 Sugarcane 0.81 0.81

2010-11 Crop Wheat 0.70 0.58 Basmati 0.91 0.86 Cotton 1.13 1.14 Sugarcane 0.80 0.75

2011-12 Crop Wheat 0.74 0.56 Basmati 1.01 1.03 Cotton 0.95 0.87 Sugarcane 0.90 0.85

Note: Export parity prices were used for basmati (paddy) and seed cotton while import parity prices used for wheat and sugarcane to represent their respective social prices.

ACKNOWLEDGMENTS

The crop budgets forming the basis of analysis reported in this paper were adapted from the policy analysis reports of the Agriculture Policy Institute. The authors would like to express their gratitude to the officers of the Institute for providing the policy analysis reports, sharing unpublished data on prices and sparing time to discuss many issues that arose during the course of this study. The authors wish to acknowledge the comments and suggestions of two anonymous referees and the guidance by the Editor of this journal which were helpful in


improving the manuscript. The authors alone, however, are responsible for the views expressed in this paper.

ANNEX 1. POLICY ANALYSIS MATRIX (PAM): 2009-10 CROP YEAR

Crop Gross Revenue Tradable

inputs cost Non-tradable inputs cost

Profit

Wheat Rs/acre Private prices 26,554 10,182 12,670 3,702 Social price 29,551 9,957 12,704 6,890 Transfers (2,997) 225 (34) (3,188)

Basmati (paddy) Private prices 25,150 9,392 12,201 3,557 Social prices 33,774 9,374 16,171 8,229 Transfers (8,624) (18) (3,970) (4,672)

Cotton Private prices 32,590 11,121 14,056 7,413 Social prices 29,173 9,866 14,163 5,144 Transfers 3,417 1,255 (107) 2,269

Sugarcane Private prices 90,345 15,617 22,213 52,515 Social price 111,815 14,894 23,085 73,836 Transfers (21,470) 723 (872) (21,321)

Note: Export parity prices were used for valuing basmati (paddy) and seed cotton while import parity prices used for wheat and sugarcane. Values in parentheses are in negative.

Sources: (API 2010, 2010a, 2010b and 2010c) and (Tufail 2014).




Profit

Wheat Rs/acre Private prices 28,150 10,657 13,432 4,061 Social prices 40,354 10,401 13,464 16,489 Transfers (12,204) 256 (32) (12,428)

Basmati (paddy) Private prices 30,290 12,295 12,960 5,035 Social prices 33,260 12,385 18,361 2,514 Transfers (2,970) (90) (5,401) 2,521

Cotton Private prices 60,449 13,179 16,733 30,537 Social prices 53,275 11,673 16,828 24,774 Transfers 7,174 1,506 (95) 5,763

Sugarcane Private prices 105,869 22,150 29,781 53,938 Social prices 132,989 21,068 30,589 81,332 Transfers (27,120) 1,082 (808) (27,394)

Note: Export parity prices were used for valuing basmati (paddy) and seed cotton crops while import parity prices were used for wheat and sugarcane. Values in parentheses are in negative.

Sources: (API 2011, 2011a, 2011b and 2011c) and (Tufail 2014).





Profit

Wheat Rs./acre

Private price 31,000 15,424 16,648 (1,072)

Social price 41,962 14,053 16,659 11,250

Transfers (10,962) 1,371 (11) (12,322)

Basmati (paddy)

Private price 45,884 15,635 17,425 12,824

Social price 45,224 15,756 23,917 5,551

Transfers 660 (121) (6,492) 7,273

Seed cotton

Private price 43,625 16,136 24,855 2,634

Social price 45,750 14,134 24,331 7,285

Transfers (2,125) 2,002 524 (4,651)

Sugarcane

Private price 91,179 28,300 40,168 22,711

Social price 100,784 26,458 40,918 33,408

Transfers (9,605) 1,842 (750) (10,697) Note: Export parity prices were used for valuing basmati paddy and seed cotton while import parity prices

were used for valuing wheat and sugarcane. Values in parentheses are in negative. Sources: ( API 2011, 2011a, 2011b and 2011c) and (Tufail 2014).

REFERENCES

Agriculture Policy Institute. 2010. Rice Paddy Policy Analysis for 2010-11 Crop. Islamabad. Agriculture Policy Institute. 2010a. Wheat Policy Analysis for 2010-11 Crop. Islamabad. Agriculture Policy Institute. 2010b. Cotton Policy Analysis for 2010-11 Crop. Islamabad. Agriculture Policy Institute. 2010c. Sugarcane Policy Analysis for 2010-11 Crop. Islamabad. Agriculture Policy Institute. 2011. Sugarcane Policy Analysis for 2011-12 Crop. Islamabad. Agriculture Policy Institute. 2011a. Wheat Policy Analysis for 2011-12 Crop. Islamabad. Agriculture Policy Institute. 2011b. Rice Paddy Policy Analysis for 2011-12 Crop. Islamabad. Agriculture Policy Institute. 2011c. Cotton Policy Analysis for 2011-12 Crop. Islamabad. Appleyard, Dennis .1987. Comparative advantage of agricultural production systems and its

policy implications in Pakistan. Rome: FAO Economic and Social Development Paper 68.

Food and Agriculture Organization .1991. Economic analysis of agricultural policies: a basic training manual with special reference to price analysis. Rome: Training Materials for Agricultural Planning 30.

Government of Pakistan. 1988. Report of the National Commission on Agriculture. Islamabad.

Government of Pakistan. 2012. Pakistan Economic Survey (Statistical Supplement) 2011-12. Islamabad.

Government of Pakistan. 2012a. Agricultural Census 2010. Pakistan Report. Lahore.


Government of Pakistan. 2013. Agricultural Statistics of Pakistan 2011-2012. Islamabad. Hamid ,Naved, Ijaz Nabi and Anjum Nasim .1990. Trade, Exchange Rate and agricultural

Pricing Policies in Pakistan (The Political Economy of Agricultural Pricing Policy). Washington: world Bank.

Monke, Eric A. and Scott R. Pearson .1989. The Policy Analysis Matrix for Agricultural Development. Ithaca and London: Cornell University Press.

Orden, David . etal. 2006. “Impact of Global Cotton and Wheat Prices on Rural Poverty in Pakistan”. The Pakistan Development Review. Vol.45: 4 part II. Winter 2006. Pp 601-17

Salam. Abdul, M.B. Siddiqui and Waseem Raza Zaidi. 2002. Economics of Wheat Production: Results from Field Surveys in Punjab and Sindh. Agricultural Prices Commission. Islamabad.

Salam, Abdul . 2009. “Distortions in Incentives to Production of Major Crops in Pakistan: 1991-2008”. Journal of International Agricultural Trade and Development. Volume 5. Issue No 2. 2009. Pp 185-208.

Salam, Abdul. 2010. “Distortions in Prices of Food Grains in Pakistan: 1996-2006.” Pakistan Journal of Applied Economics. Vol xx Nos 1 & 2, 2010. pp 13-28.

Salam, Abdul. 2012. “Domestic Terms of Trade and resource Transfers from Agriculture: A case Study of Pakistan.” Journal of International Agricultural Trade and Development. Volume 8, Issue 2. pp.

Scandizzo, Pasquale L. and Colin Bruce .1980. Methodolgies for Measuring Agricultural Price Intervention Effects, World bank Staff Working Paper No. 394. Washington D.C.

Tufail, Sadia .2014. Comparative Advantage of Major Crops in Punjab- Pakistan: An Application of Policy Analysis Matrix. M. Phil. Thesis submitted to Federal Urdu University of Arts, Science and Technology, Islamabad.


AGRICULTURAL TRADE IN SOUTH ASIA: HOW IMPORTANT ARE THE TRADE BARRIERS?

Nitya Nanda The Energy and Resources Institute (TERI),

New Delhi, India

ABSTRACT

Greater intra-regional trade in agricultural goods by reducing tariff and non-tariff barriers is often considered to be a good way of promoting rural development and improving food security in South Asia. Openness to trade of agricultural sector varies across South Asian countries, yet agricultural trade plays a crucial role as, in many countries, its share in exports is reasonably high, apart from the fact that most of them are net food-importing countries. Analysing different trade-related indictors and using a modified gravity model, the paper concludes, while factors like lack of complementarities, lack of diversification of export baskets and lack of trade facilitation are important barriers to trade, supply constraints appear to be the most important barrier, though there is scope for improvement in reducing tariff and non-tariff barriers as well.

Keywords: Agricultural trade, South Asia, regional trade, gravity modelling, trade barriers, trade complementarity, trade diversity

JEL codes: C51, F15, Q17, R11

1. INTRODUCTION

Agriculture plays a unique role in South Asian economies. Though share of agriculture in GDP has come down over the decades, it continues to employ majority of the population in all South Asian economies. In 2010, while agriculture contributed only about 18 per cent to GDP, it provided employment to 51 per cent of the population in South Asia. Agriculture is thus extremely important for providing livelihood and food security in these countries.

An earlier version of the paper was presented at the Regional Consultation on “Trade, Climate Change and Food

Security in South Asia”, Kathmandu, 20–21 December 2012, organized by South Asia Watch on Trade, Economics and Environment (SAWTEE). The author wishes to thank SAWTEE and the participants in the regional consultation. The author solely takes the responsibility of any errors and omissions and the views expressed.

Table 1. Agriculture and South Asian Economies 2010

AFG BGD BHU IND MDV NPL PAK SLK SAS World

Agriculture value added , % of GDP 29.92 18.59 23.18 17.74 3.14 36.53 21.18 12.79 18.28 2.81

Share of agriculture in employment (2005)

48.10 43.60 55.80

43 30.70 53.53 35.02

Crop Prodn Index (2004-6 = 100) 125.03 131.01 91.99 119.07 84.17 113.02 100.14 122.42

Cereal yield (kg per hectare) 1908.1 4143.5 2177.2 2536.6 2000 2294.5 2591.9 3974.3 2690.6 3563.5

Food exports (% of merchandise exports) 40.05 6.22 7.17 8.26 96.15 19.08 16.79 26.89 11.69 8.21

Food imports (% of merchandise imports) 13.69 13.90 11.47 3.95 22.35 13.56 13.08 15.35 6.86 7.43

Source: World Development Indicators (WDI). Note: AFG = Afghanistan; BGD = Bangladesh; BHU = Bhutan; IND = India; MDV = Maldives; NPL = Nepal; PAK = Pakistan; SLK = Sri Lanka;

SAS = South Asia. These abbreviations have been used in later tables and figures as well.

Agricultural Trade in South Asia 97

It is also quite obvious that the people dependent on agriculture are relatively poorer compared to other sectors in these countries. Being a primary sector, it also has a close economic link with other sectors of the economies especially when external economic linkages of these countries are relatively weak.

Share of agriculture in employment is quite high in most countries of the region. In some countries, the share of food exports in total merchandise exports is also very high (Table 1). This shows that agricultural trade plays an important role in the livelihood security of a large number of people in South Asia. Except India, food imports as a share of total merchandise imports are also quite high in all South Asian countries. Hence agricultural trade plays a significant role in ensuring food security in all these countries. In some countries, along with high share of food imports in total imports, the share of food exports in total exports is also high indicating that their agricultural trade also determine their ability to import food. Afghanistan, Maldives, Nepal, Pakistan and Sri Lanka fall in this category. Trade policy of a country, however, can often have conflicting impacts on livelihood and food security. South Asian countries thus need to maintain a fine balance between these two policy objectives.

Source: WITS Database and WDI.

Figure 1. Trade Intensity of South Asian Agriculture (Per Cent).

Trade intensity, measured as the share of agricultural exports in value added in agriculture, has been going up in South Asia as a whole (Figure 1). However, the picture is not quite similar in all countries. Trade intensity of agriculture is the highest in Sri Lanka. But it has been declining over time. What actually drives the rising trade intensity of agriculture is that of India which has almost doubled between 2003 and 2011. Even Pakistan has shown similar trend. Trade intensities of Afghanistan, Bhutan and Nepal have shown wide fluctuations and the overall trend shows almost stagnating levels over the years. Another noteworthy feature of agricultural trade in South Asia is that all countries, except India and Sri Lanka, have maintained deficits in agricultural trade (Figure 2). Even Sri Lanka became a net-importing country in agricultural goods recently.

0

10

20

30

40

50

2003 2004 2005 2006 2007 2008 2009 2010 2011

AFG BGD BTN IND MDV

NPL PAK LKA SAS

Nitya Nanda 98

Source: WITS Database.

Figure 2. Agricultural Trade Balance in South Asian Countries (‘000US$).

South Asian countries have signed the Agreement on South Asian Free Trade Area. Moreover, some countries in the region have also signed bilateral trade agreements with deeper commitments. India’s agreements with Bhutan, Nepal, Sri Lanka and Afghanistan and Pakistan’s agreements with Sri Lanka and Afghanistan are worth notable in this context. Agricultural trade among South Asian countries have shown rising trend in recent years which could be due to pruning of sensitive lists as well as India’s offer of duty free access to imports coming from the LDCs. However, there are concerns about existing tariff and non-tariff barriers as well as lack of adequate infrastructure. Lack of trade complementarities is also often cited as a factor in regional trade (Mamoon et al. 2011).

It is also believed that the South Asian trade negotiations have yielded relatively fewer opportunities for agricultural trade compared to non-agricultural trade as agriculture is a politically sensitive issue in most countries in the region (Samaratunga et al. 2007). Some studies, however, argued for a cautious approach as liberalization of trade in agricultural goods can have both costs and benefits (Ghimire and Adhikari 2001; Razzaque and Laurent 2006). India’s unilateral offer on duty-free access to LDCs of course might be a game changer, particularly for the LDCs in South Asia as four out of eight countries in the region are LDCs.

The next section of the paper analyses the trends in trade in agricultural goods in South Asia. The analysis is based on the 24 commodity groups as followed in the commodity classification of agricultural goods in UN COMTRADE database (Annexure 1). The third


section looks at the diversity of their trade baskets and also deals with trade complementarities of agricultural goods between different pairs of countries in the region. The fourth section discusses the tariff and non-tariff barriers prevailing in the region. The fifth section provides an econometric analysis to understand the determinants of agricultural trade in South Asia. While the sixth section analyses the supply side constraints to agricultural trade in different countries in the region, the seventh and the last section, concludes the paper.

Table 2. Agricultural Trade Flows in South Asia (Average of 2008-2010 in ‘000$)

Exports Imports ↓ AFG BGD BTN IND MDV NPL PAK SLK RSAS AFG 86802 80060 125 166987 BGD 76 1528 24384 86 1507 10660 416 38657 BTN 9141 92561 201 101903 IND 46772 1090890 9698 28152 204568 752082 320338 2452500 MDV 74 0 124 198 NPL 51921 414 128823 3 365 1693 183219 PAK 745158 166852 94932 4624 7 89269 1100841 SLK 144 4825 203778 25252 232 46500 280730 RSAS 792150 1323628 11640 631353 58116 206514 889668 411965 4325036

Source: WITS (World Integrated Trade Solution) Database, World Bank (wits.worldbank.org). Note: Where there are some data gaps here and there, for Bangladesh, data were not available after

2007. In all such cases mirror data were used as far as possible.

2. TRENDS IN AGRICULTURAL TRADE IN SOUTH ASIA

During the year 2008-09, the average annual trade in agricultural goods within South Asia was in excess of US$4bn and India accounted for about 57% of it. While India trades with all other countries of South Asia that is not true for many other countries. This is probably due to the fact that most South Asian countries are not geographically connected to each other. While India shares borders with all countries except Afghanistan and Maldives, other countries share borders only with India, except Pakistan and Afghanistan who also share a border with each other. This is also reflected in regional orientation of trade flows of agricultural goods (Table 2).

Share of regional trade in agricultural exports is quite low for most countries in South Asia except the landlocked countries like Afghanistan, Nepal, Bhutan and the small island nation of Maldives. Share of regional trade in agricultural exports has shown a stagnant or declining trend over 2002-2011, except in cases of Bangladesh and Maldives (Figure 3). However, complete data for these two countries were not available for the period considered here.

Similarly, the share of regional trade in agricultural imports also shows a declining or stagnating trend in most countries of the region except Nepal and India who showed an upward trend in recent years (Figure 4). Bhutan, of course, has a high share as it imports almost its entire requirement of agricultural goods from the region, notably from India and, to some extent, from Bangladesh. Nepal also has relatively high share of its imports coming

Nitya Nanda 100

from the region, followed by Maldives and Sri Lanka. Hence, as like exports, larger economies of the region import relatively smaller share of their agricultural goods import requirements from the region while, the smaller, landlocked and island nations import relatively larger share of their agricultural goods import requirements from the region. However, it is worth noting that although intra-regional trade in agricultural goods appears to be low overall, it is higher than the share of intra-regional trade in total trade, i.e., South Asian countries trade more intensively intra-regionally in agricultural products compared to trade in other products, which is natural, in view of the nature of agricultural products (perishability, bulkiness etc.).


Figure 3. Share of Regional Trade in Agricultural Exports (Percent).


Figure 4. Share of Regional Trade in Agricultural Imports (Percent).


3. DIVERSITY AND COMPLEMENTARITY OF TRADE BASKET

It is often believed that trade between countries is driven by the comparative advantages and differences in technology, economies of scale or preferences, natural resources, climatic conditions and, in some circumstances, by strategic trade policies. For agricultural trade, agro-climatic conditions of countries play a significant role. Hence, an important factor that very often determines the trade intensity or trade performance of a country or region is the sectoral diversity or concentration of export basket. In case of analysing intra-regional trade flows, both export and import concentration or diversity can give crucial indications. The sectoral Hirschmann-Herfindahl Index (HHI) is a measure of the sectoral or product concentration of a country’s exports or imports. It tells us the degree to which a region or country’s exports or imports are dispersed across different economic activities. High concentration levels of exports are sometimes interpreted as an indication of vulnerability to economic changes in a small number of product markets. This can also be used to measure the import diversification of a country.

The sectoral HHI is defined as the square root of the sum of the squared shares of exports of each industry in total exports or imports for the region or country under study. It takes a value between 0 and 1. Higher values indicate that exports or imports are concentrated in fewer sectors. Alternatively, lower values indicate that exports or imports are diversified across sectors or products. Mathematically, the HHI for exports of a country can be denoted as:

HHIX = √Σi(Σdxisd/ΣdXsd)2 (1a) where s is the country of interest, d is the set of all countries in the world, i is the sectors

of interest, x is the commodity export flow and X is the total export flow. Each of the bracketed terms is the share of good i in the exports of country s.

Similarly, the HHI for imports of a country can be denoted as: HHIM = √Σi(Σdmisd/ΣdMsd)2 (1b)

where s is the country of interest, d is the set of all countries in the world, i is the sectors of interest, m is the commodity and M is the total import.

Concentration of export basket is fairly high in all countries except India (Figure 5). The HHI is higher than 0.8 for Maldives and higher than 0.5 in Bangladesh. However, import concentration is much lower in most countries in the region indicating that though they mostly export few goods, they import a wide range of agricultural products. Interestingly, while India has the lowest export concentration in the region, it has the highest import concentration. Since more than half of agricultural trade of South Asia is accounted for by India, its trade pattern is important for the whole region.

As can be seen from Figure 6, five out of eight countries in the region are dependent on single commodity group where one such group accounts for more than fifty percent of agricultural exports. In fact, in four of them, it is even more than sixty percent. The single commodity and the associated countries are: Fish etc. for Bangladesh and Nepal, cereals for

Nitya Nanda 102

Pakistan, tea and spices for Sri Lanka and fruits and nuts for Afghanistan. These are generally not among the top import items in South Asian countries except cereals which is among the major import items in Bangladesh, Bhutan and Sri Lanka. Apart from India, the two other countries that have fair diversification of agricultural exports are Nepal and Bhutan, both are landlocked and mountainous.

Source: UN COMTRADE Database.

Figure 5. Concentration in Agricultural Trade Baskets 2010.


Figure 6. Export Diversification of Agricultural Products 2010.



Figure 7. Import Diversification of Agricultural Products 2010.

When it comes to imports, the picture is just the opposite (Figure 7). Since most countries have concentrated export basket, it is quite likely that they will have concentrated production pattern as well. This means that they will have to import almost everything. It is also interesting, though not surprising, India has the least diversified import basket. About half of its imports consist of fats and oils. While other countries do not show such high level of concentration in one product group, in three other countries this product group has high share and the largest import item. Even in other countries, this product category is among the largest import item if not the largest one. This, along with the fact that fats and oils do not occur as a major export item in any of the south Asian countries, implies that there would be relatively little scope for intra-regional trade in south Asia.

Linked to diversity of trade basked is the degree of trade complementarities. Actual trade between two countries is also determined by the degree of trade complementarities that exist between them. As has been seen above, import basket of a country is generally diverse. Hence, a country is likely to export more to another country if its export basket is also diversified. The complementarity index measures the degree to which the export pattern of one country matches the import pattern of another. It is defined as the sum of the absolute value of the difference between the import category shares and the export shares of the countries under study, divided by two. The index is converted to percentage form and hence the values range between 0 and 100. Mathematically, the index can be denoted as:

TCIsd = [1-{Σi|(Σwmiwd/ΣwMwd- Σwxisw/ΣwXsw) |}/2]X100 (2)

Nitya Nanda 104

where d is the importing country of interest, s is the exporting country of interest, w is the set of all countries in the world, i is the set of industries, x is the commodity export flow, X is the total export flow, m the commodity import flow, and M the total import flow. In words, we take the sum of the absolute value of the difference between the sectoral import shares of one country and the sectoral export shares of the other. Dividing by 2 coverts this to a number between 0 and 1, with zero indicating all shares matched and 1 indicating none did. Subtracting from one reverses the sign, and multiplying by 100 puts the measure in percentage terms.

Export complementarities as well as import complementarities are generally low among South Asian countries (Table 3). For Afghanistan, the highest export complementarities are with India and Sri Lanka, just at 16. For Bangladesh, the highest export complementarities are with Maldives and Nepal at 22 and 19 respectively. Bhutan has relatively better export complementarities with India and Nepal at 34 and 30 respectively. At 36, Nepal’s export complementarity is the highest with Pakistan followed by 35 with Sri Lanka. Pakistan has the highest complementarity with Bangladesh at 42, followed by 39 with Maldives. Sri Lanka’s export complementarities are relatively lower, the highest being 25 with Maldives, followed by Nepal and Pakistan at 22 and 21 respectively.

Since India has the most diversified export basket, it also has relatively better export complementarities with all countries in the region, the lowest being 24 with Afghanistan and the highest is 56 with Nepal. At 53 and 51, it also has good export complementarities with Bhutan and Maldives. However, India’s import complementarities are much lower, the highest being 34 with Bhutan, followed by Nepal and Pakistan at 33 and 25 respectively. It is interesting to note that, except Maldives, India’s import complementarity is the lowest with Sri Lanka, and yet Sri Lanka’s share in Indian imports from South Asia is the highest.

Table 3. Trade Complementarities between South Asian Countries 2010

Export Imports

↓ AFG BGD BTN IND MDV NPL PAK SLK SA-IND

AFG 8 16 4 16 BGD 9 16 15 22 19 15 16 BTN 16 34 30 IND 24 45 53 51 56 41 44 52 MDV 2 2 NPL 26 32 33 36 35 PAK 14 42 25 39 36 32 SLK 16 12 14 25 22 21 SA-IND 25

Source: UN COMTRADE: International Trade Statistics Database (comtrade.un.org).

4. TARIFF AND NON-TARIFF BARRIERS IN THE REGION

As has been seen in the previous sections, there are some natural disadvantages that adversely affect the intra-regional trade in South Asia. These are concentration or lack of


diversity in export baskets of agricultural goods, lack of trade complementarities etc. However, there are some barriers or lack of facilitation that are often policy induced. Of course, the two types of factors often interact with each other in determining the trade flows among countries. The following paragraphs discuss some of the barriers that affect agricultural trade in South Asia.

Import tariff is one of the oldest instruments that have been used to protect domestic production of goods as well as to raise revenue. Import tariffs have generally come down across the globe, yet tariffs on agricultural goods remain much higher compared to other goods, particularly in the developed world. Considering that, import duties on agricultural goods in South Asian countries are relatively lower. Among the South Asian countries, the average effective tariff on agricultural goods is the highest in Bhutan followed by India and Sri Lanka (Table 4). It may be noted that Bhutan is not yet a member of the WTO and hence it has no global commitment to bind its tariffs.

The other two high tariff countries, India and Sri Lanka have one interesting similarity. They are the only two South Asian countries that maintained trade surplus in agricultural goods over the years, though in recent years, Sri Lanka has also turned to be a deficit country. It appears that countries with high trade deficit in agricultural goods tend to impose lower import duty. May be they cannot afford to impose high import duty due to food security concerns. Bhutan is, of course, an exception in this regard as it has high trade deficit in agricultural goods and at the same time imposes high import duty. Among the South Asian countries, Nepal faces the highest average effective tariffs in Bhutan and alternatively Bhutan also faces the highest average effective tariffs in Nepal. The two countries also have similar trade baskets.

Table 4. Average Effective Tariffs in South Asia - 2009

Exporting Country/Region Importing Country AFG BGD BTN IND MDV NPL PAK SLK

WLD 7.26 18.52 49.34 34.6 17.07 14.56 19.1 21.8 AFG 19.5 34.69 11.54 15 BGD 4.38 36.67 37.36 23.93 19.29 12.22 24.78 BTN 24.05 1.07 22.5 15.88 IND 5.57 14.64 44.48 15.51 11.29 9.16 19.6 MDV 65 31.43 25.15 NPL 11.34 46 39.17 25 8.67 13.66 PAK 6.61 17.76 27.29 14.13 8.91 11.02 SLK 7.13 18.27 9.13 15.48 18.33 15.93

Note: The tariff figures are simple average of effective tariffs. In some cases tariff figures for 2009 were not available, where figures for earlier year available were considered.

Source: WITS Database. While India imposes the highest average effective import duty in the region, its duty is

lowest on agricultural goods coming from Bhutan which is also the lowest on agricultural goods imposed by any country in the region. Interestingly, this is despite the fact that Bhutan’s duty is the second highest for its regional partners only after Nepal. The tariff data for latest years were not available and hence the data provided in Table 4 might not have

Nitya Nanda 106

captured the impacts of India’s policy of offering duty free access to imports coming from LDCs and the pruning of sensitive lists in all countries. There of course continue to be imposed several para-tariff measures that also affect agricultural trade in the region.

NTMs (non-tariff measures) affecting trade in agricultural goods relate to standards, testing and certification procedures. For example, in India, there are bio-security and SPS (sanitary and phyto-sanitary) requirements. Import of nearly all agricultural goods, including livestock and food products require some kind of SPS certificate and import permit. Getting such certificates are often time and resource consuming (Raihan 2012). Bangladesh continued to ban imports of poultry products from India despite India having regained avian influenza-free status and many other countries lifting the ban. Till recently, Pakistan used to import from India on the basis of a positive list. Incidentally, it has not yet given India the MFN status, which now is expected to be forthcoming (Mukherjee 2012). It is also noteworthy that South Asian countries also use export restrictions to deal with seasonal shortages of goods such as rice, onion, fish etc.

5. DETERMINANTS OF AGRICULTURAL TRADE

To better understand the factors determining trade flows of agricultural goods among the South Asian countries, an econometric model is also used here. A gravity model type econometric specification has been considered. The gravity model of trade is similar to other gravity models and it predicts bilateral trade flows based on the economic sizes of (often using GDP) and the distance between two units. The model was first used by Tinbergen in 1962. The basic theoretical model for trade between two countries (i and j) takes the form of:

Tij = K(MiMj/Dij) (3) where Tij is the trade flow between countries i and j (export from country i to j), Mi and

Mj are the economic masses of country i and j respectively, Dij is the distance between countries i and j and K is a constant.

In general a gravity model uses GDP of the trading partners as the economic masses and hence is not able to distinguish between the directions of trade flows (Tij and Tji) between the countries. Ordinary gravity model is also unable to explain trade in a particular commodity or a sector. The specification considered here uses different economic variables for the economic masses of the exporting and importing countries and hence is able to distinguish between Tij and Tji. The basic model considered here can be expressed by the following function:

Tij = f(AVAi, GDPj, AEDij, TCIij, D) (4) where Tij is the trade flow between countries i and j (export from country i to j), AVAi is

value added in agriculture in country i, GDPj is gross domestic product in country j, AEDij is the average effective import duty that the country i faces in country j, TCIij is the trade complemetarity index for the exports of country i in country j, and D is a dummy variable which takes the value 1 if the countires i and j share border with each other, and 0 otherwise.


5.1. Justification and Data

For Tij, trade flows of agricultural goods are used. Data were collected from WITS. Since agricultural trade is subject to high fluctuations, average of trade flows of thre years, 2008, 2009 and 2010 is considered. In some cases trade flows of all the three years were not available and hence average of two years or even trade figure of a single year was considered. The data used in this analysis have been presented in Table 2.

Since we are concerned with agricultural trade only, the appropriate economic mass for the exporting country should be the agricultural value added as the size of value added in other sectors is unlikely to influence exports of agricultural goods from a country. In other words, the size of the agriculture sector is the appropriate economic size for the gravity model in the current context. The data for agricultural value added were obtained from the World Development Indicators.

Even though agricultural value added is considered as the economic mass of the exporting countries, the economic mass in importing country would still be the GDP. Since agricultural goods are used by the entire population as foods and also used as raw materials in industry, the GDP can be considered. Once again, the data for the variable were obtained from World Development Indicators.

Average effective import tariff data were obtained from WITS database, where data for TCI were estimated using trade flow data of agricultural goods obtained from the UN COMTRADE database. Data used for AED are presented in Table 4, whereas data for TCI are presented in Table 3.

In a gravity model, it is customary to use distance between two countries as one of the variables. Often a dummy for neighbouring countries is also used. However, measuring the distance has always been a difficult issue. In the South Asian context, the sizes of the countries are quite skewed and even their position being quite peculiar where the largest country, India is at the centre and other countries being largely in the periphery, this becomes even more difficult. Moreover, for many of the smaller countries, trade occurs largely with the neighbouring regions of India, and hence, measuring distance poses even greater challenge. Therefore, only a dummy for neighbouring countries is used here.

The basic model considered here is: LnTij = LnAVAi + LnGDPj + D (5)

where LnTij is the natural logarithm of Tij, LnAVAi is the natural logarithm of AVAi, and LnGDPj is the natural logarithm of GDPj.

However, as indicated earlier, the basic model was extended to include TCI and AED and the following two versions were estimated:

LnTij = LnAVAi + LnGDPj + LnTCIij + AEDij + D (6a) LnTij = LnAVAi + LnGDPj + LnTCIij + LnAEDij + D (6b) Some other forms were also used where variables like population in exporting country,

and population in importing country were also included. The following two specifications were tried:

Nitya Nanda 108

LnTij = LnAVAi + LnGDPj + LnTCIij + AEDij + LnPOPi +Ln POPj + D (7a) LnTij = LnAVAi + LnGDPj + LnTCIij + AEDij + LnPOPi + D (7b)

where LnPOPi is the logarithmic value of the population of exporting country and LnPOPj is the logarithmic value of the population of the importing country.

Table 5. Regression Results of Gravity Models

Eq 6 Eq 7a Eq 7b Eq 8a Eq 8b Intercept -22.22* -19.80* -20.21* -37.40* -31.33* LnAVAi 0.71* 0.49* 0.51* 2.56** 2.26*** LnGDPj 0.56* 0.57* 0.57* 1.07** 0.60* LnTCIij 0.84*** 0.86*** 0.34 0.46 AEDij -0.02 -0.01 -0.01 LnAEDij -0.16 LnPOPi -1.88*** -1.62*** LnPOPj -0.42 CTEi CTTij Dummy 2.53* 2.64* 2.61* 2.15 2.32* Multiple R 0.75 0.77 0.77 0.79 0.79 R Square 0.57 0.59 0.59 0.63 0.62 Adj R Square 0.53 0.53 0.53 0.54 0.55 Std Error 2.19 2.18 2.18 2.16 2.15 Observations 39 39 39 39 39

Note: * indicates significant at 95% level of confidence, ** indicates significant at 90% level of confidence and *** indicates significant at 80% level of confidence respectively.

5.2. Regression Results

Estimated results of different regression equations are presented in Table 5. In the basic model, all the variables included are highly significant. All the variables also show expected signs, i.e., agricultural value added in exporting country, GDP of importing country as well as the neighbourhood dummy, all influence trade positively. When trade complementarity index and average effective duty are included, trade complementarity shows significance but only at the level of 80%, but average effective duty turns out to be insignificant. Significance of the original variables remains almost similar.

When populations of both exporting and importing countries are also included in the equation, population of exporting country appears to be significant only at 80% level of confidence, while population of importing country turns out to be insignificant. But the status of other variables remains almost similar, except the trade complementarity index which turns out to be insignificant. However, when only the population of exporting country is included, the significance of agricultural value added comes down a bit.

Table 6. Indicators of Supply Side Constraints in Agriculture 2010

AFG BGD BTN IND MDV NPL PAK SLK SAS WLD

Agriculture value added per worker (constant 2000 US$)

480.14 465.32 479.01 2671.38 242.05 962.62 906.88 510.35 1064.38

Arable land (hectares per person) 0.23 0.05 0.11 0.13 0.01 0.08 0.12 0.06 0.12 0.20

Average precipitation in depth (mm per year) 327.00 2666.00 2200.00 1083.00 1972.00 1500.00 494.00 1712.00

Cereal yield (kg per hectare) 2045.20 4140.80 2159.10 2571.90 2041.70 2373.90 2789.70 3663.60 2728.87 3567.94

Droughts, floods, extreme temperatures (% of population, average 1990-2009)

1.06 4.58 0.01 4.36 0.03 0.70 1.06 2.16

Nitya Nanda 110

It might not be a good idea to include them together, as South Asian countries being at similar stage of development, population has a link with GDP and agricultural value added. Nevertheless, the negative sign of population in exporting country might be an important indication that availability of exportable surplus can be a factor in agricultural trade. Results pertaining to the variables considered in the basic models are quite stable in different specifications indicating that the basic model is quite robust.

6. SUPPLY SIDE CONSTRAINTS

Regression results in all the specifications used here show that agricultural value added in exporting countries is a significant factor in determining intra-regional trade in agricultural goods in South Asia. Hence, this section tries to have a better understanding of the constraints (Table 6). The most important supply side constraint in agriculture is the availability of land. Except Afghanistan, in all South Asian countries, availability of arable land is much lower than the global average. Among the other countries, India has the highest availability of per capita arable land. Bangladesh, Maldives, Sri Lanka and Nepal have very low per capita availability of arable land. Though Afghanistan has the highest per capita availability of land, its agriculture is seriously affected by shortage of water. Similarly, though Pakistan’s availability of arable land is slightly below the Indian level, it faces serious water shortage.

Together with the availability of water and occurrence of extreme weather conditions, the availability of land ensures that the supply capacity of agricultural goods is weak in most South Asian countries. Such weak supply capacity might have partly been responsible for the fact that most countries in the region are net importers of agricultural goods. Productivity of agriculture is another factor that might have caused constraint in trade in agricultural goods and, in some countries, there is substantial scope for improving agricultural productivity.

Another supply side constraint could be the availability of infrastructure. Lack of proper roads can mean that farm lands are not well connected with the ports. In countries like Nepal, traders in Kathmandu might find it difficult to access farms in hinterland in Nepal and may find it easier to import agricultural goods from India than to source from within the country, contributing to making Nepal a net importer of agricultural goods.

CONCLUSION

Though share of agriculture has been declining in South Asian countries, it continues to provide employment to a large section of the population, and hence also plays an important role in providing them with livelihood and food security. Agricultural goods have a relatively high share in exports in most countries of the region, and at the same time, most of them are net food importing countries. Hence trade in agricultural goods has a complex role in livelihood and food security in South Asian countries. To that extent the issue of trade in agricultural goods needs to approached with caution.

There are several barriers that are hindering the growth in intra-regional trade in agricultural goods in South Asia. However, it should be recognized that share of intra-regional trade in agricultural goods in South Asia is much higher than the share of intra-


regional trade in overall trade. The lack of trade complementarities between South Asian nations as well as lack of diversification of export baskets might have been important barriers.

The econometric analysis shows that tariff barriers are unlikely to have acted as a major hindrance in enhancing intra-regional agricultural trade in South Asia. It may be noted that Sri Lanka increased its trade in agricultural goods with India upon signing a bilateral free trade agreement with India. However, this case may not be generalized, as Sri Lanka, apart from India, has been the only South Asian country that has maintained surplus in trade in agricultural goods, while all other countries have consistently maintained deficit in trade in agricultural goods.

The analysis also shows that the lack of supply capacity might have been the most important barrier in agricultural trade in South Asia. It is difficult to quantify non-tariff barriers and hence difficult to include them in an econometric model. However, given that tariff barriers have not played any decisive role, it is unlikely that non-tariff barriers might have been significant obstacles to trade in agricultural goods within South Asia. It is also not clear if, intra-regional trade will play any major role in addressing food security concerns, though it can make a contribution.

Since trade in agricultural goods is significantly linked to livelihood and food security concerns, reducing tariff and non-tariff barriers may not be seen as the panacea. Such sentiments have been expressed by other studies as well. Hence the major focus in South Asian countries should be on reducing the supply constraints. Such measures are also unlikely to affect the livelihood and food security of people in South Asian countries adversely, though there is a strong likelihood that they will make a positive contribution.

Annexe 1. Classification of Agricultural Commodity

01 Live animal 02 Meat and edible meat offal 03 Fish & crustacean, mollusc & other aquatic invertebrate 04 Dairy prod; birds' eggs; natural honey; edible prod nes 05 Products of animal origin, nes or included. 06 Live tree & other plant; bulb, root; cut flowers etc. 07 Edible vegetables and certain roots and tubers. 08 Edible fruit and nuts; peel of citrus fruit or melons. 09 Coffee, tea, matï and spices. 10 Cereals 11 Milling products, malt, starches, inulin, wheat gluten 12 Oil seed, oleagic fruits, grain, seed, fruit, etc., nes 13 Lac; gums, resins & other vegetable saps & extracts. 14 Vegetable plaiting materials; vegetable products nes 15 Animal/veg fats & oils & their cleavage products; etc. 16 Prep of meat, fish or crustaceans, molluscs etc. 17 Sugars and sugar confectionery. 18 Cocoa and cocoa preparations. 19 Prep.of cereal, flour, starch/milk; pastrycooks' prod 20 Prep of vegetable, fruit, nuts or other parts of plants 21 Miscellaneous edible preparations. 22 Beverages, spirits and vinegar. 23 Residues & waste from the food indust; prepr ani fodder 24 Tobacco and manufactured tobacco substitutes

Nitya Nanda 112

REFERENCES

Ghimire, H., and R. Adhikari. 2001. Agricultural Liberalisation and its Impact on South Asia, SAWTEE Discussion Paper, SAWTEE, Kathmandu and CUTS, Jaipur.

Mamoon, D., S. Paracha, H. Mughal, and A. Ayesha. 2011. Pakistan's trade competitiveness & complementarities in South Asia. Munich Personal RePEc Archive (MPRA).

Mukherjee, I. N. 2012. Non-tariff barriers facing South Asia: case of India, Trade Insight, 8(3): 30-31.

Raihan, S. 2012. Non-tariff barriers facing South Asia: case of Bangladesh, Trade Insight, 8(3): 28-29.

Razzaque, M. A., and E. Laurent. 2006. Agriculture and Rice Trade Liberalisation: Potential Implications for South Asian Countries, Commonwealth Trade Hot Topics, Issue No 8.

Samaratunga, P., K. Karunagoda, and M. Thibbotuwawa. 2007. Mapping and Analysis of the South Asian Agricultural Trade Liberlization Efforts. In Agricultural Trade: Planting the Seeds of Regional Liberalization in Asia, ed. ESCAP. A Study by the Asia Pacific Research and Training Network on Trade, Studies in Trade and Investment 60, UN-ESCAP, Bangkok.

Tinbergen, J., 1962. “Shaping the world economy: Suggestions for an international economic policy.” New York: The Twentieth Century Fund.


THE TRADE EMBARGO AND CONSOLIDATION OF FOOD SECTOR IN POLAND

Kazimierz Meredyk University of Bialystok, Management and Economics Faculty,

Political Economy Chair, Bialystok, Poland

ABSTRACT

This study discusses the issue of the stability of the food sector in Poland during the period 2001-2030 in the context of economic sanctions imposed by Russia in 2013. Food exports from Poland to Russia have fluctuated prior to the introduction of sanctions and Polish food exporters have dealt with numerous regulations. Some of the recently observed trends, including deflation, might have been exacerbated, but not because of sanctions. The GDP growth rate may have declined by 0.3-0.5% in 2014 and the food sector lost about $340 mln as a result of sanctions. Paradoxically, the Russian trade embargo will accelerate the modernization of Poland’s food sector and enhance competitiveness.

Keywords: economic expansion, trade embargo, administrative intervention, market mechanism, effectiveness

JEL: E65, F13

INTRODUCTION

This study discusses the issue of the stability of the food sector in Poland during the period 2001-2030. Production volume may incidentally and locally rise, but over a longer period of time, growth remains insignificant. Countries at the upper middle and high levels of development face a problem in this sector of inefficiency, potential production surplus, and the need for acceleration of structural changes.

Structural and market weaknesses of the primary sector, including mainly its agricultural and raw materials, lower the long-term dynamic of the entire national economy. Therefore the problem is of a strategic nature and its solution is both an economic and political necessity, and in this case, the market mechanism may be used only on a limited scale (see section 2).

In the short term, counteracting perturbations in the market resulting from the trade embargo mutually imposed by the European Union (EU) and Russia (see section 3) requires diversification of ready markets and improvement in the organization of the food trade. At the

Kazimierz Meredyk 114

same time, it is believed that a prerequisite for the increased stability of the food sector in Poland is not significant growth in production volume, but better competitiveness: improved quality of products and a significant decrease in production costs. Such changes lower the sensitivity of the food economy to market perturbations, such as trade embargos, and maintain high export growth.

Furthermore, large portions the potential of the primary sector will be released and transferred to sectors that are economically more efficient. Such deep restructuring of the economy will, at the same time, encourage growth of the entire national economy, enhancing employment and raising average income (see section 4).

ECONOMIC GROWTH OF THE POLISH ECONOMY DURING THE PERIOD 1995-2020

For the past twenty years, there has been a long-term expansion trend in the Polish economy. The average annual growth rate of GDP during the period 1995-2014 was approximately 4% (Rocznik Statystyczny, 2014, p.701), above the average in Europe. The cumulative growth index in Poland amounted to 123.8% and for the period 2007-2014 was the highest among European Union economies (Ministerstwo Finansow, 2015). Forecasts for the next decade (2015-2025) are for continued growth (see Figure 1).

Long-term trend Smoothed line of trend based on real data. Source: Instytut Badań Strukturalnych, 2006, p. 17; Meredyk, 2012; Biuletyn Statystyczny, 2014a,

p. 58; Rocznik Statystyczny Rzeczypospolitej Polskiej 2014, 2014, p. 701; Ministerstwo Finansów, 2015.

Figure 1. GDP growth rate in constant prices in Poland during the period 1993-2020, in %.

Simultaneously with the growth of the national economy, a relatively fast growth rate in the food sector together with its raw material and agricultural base has occurred. During the period 1995-2013, the final output index in agriculture, in constant prices, amounted to 135.8% and the market output index was 152.4 % (Rocznik Statystyczny, 2014, p. 476). Hence, the average annual growth rate was 1.8% and 2.4%, respectively. It should also be kept in mind that the share of the primary sector (agriculture, forestry, fishery, and mining)

0

1

2

3

4

5

6

7

8

1990 1995 2000 2005 2010 2015 2020 2025 2030

The Trade Embargo and Consolidation of Food Sector in Poland 115

amounted to approximately 6% of the GDP in 2005 and remained stable at about 5 % in 2013 (Rocznik Statystyczny, 2014, p. 719).

The share of the primary sector’s employment fell from 19% to 16% of the total number of employed in the analyzed period and remains 2-3 times higher as compared to the leading EU countries. In Poland, the decrease in employment was negligible and amounted to only 7% during the period 2006-13 (Rocznik Statystyczny, 2014, p. 718). Nevertheless, it should be stressed that it was generally a function of demographic changes in rural areas and not a characteristic of the entire economy.

Table 1. Dynamics of Exports of Food in the SITCa System, in Constant Prices, in PLN

- SITC : Standard International Trade Classification.

The previously noted growth in exports has been unprecedented in the history of the

Polish economy (Table 2). The average annual pace of growth in exports during the last two decades reached the level of 10% accompanied by a rise in modernization and an improvement in efficiency. This is a significant achievement since avoiding the pitfalls of becoming raw material suppliers in the international market is a prerequisite for stable market expansion.

Moreover, recent years brought a sharp decrease in the deficit in international trade from about PLN60 billion in 2011 to about PLN10 billion in 2014 (Biuletyn Statystyczny, 2015, p. 52). The public sector debt, in relation to GDP, levelled off at around 50% in that period (Rocznik Statystyczny, 2014, p. 727).

A synthetic measure of unused production capacity of the Polish economy (second largest in the entire EU) is the cumulative labor efficiency index of 116.2% for the period 2008-2014 . Additionally, minimal real wages increased in that period by 50% (Sytuacja, 2015) and public finance sector servicing costs fell to the level of 1.8% in relation to GDP (Ministerstwo Finansow, 2015).

The above description of the Polish economy in the last 20 years confirms a fairly common opinion that the economy is more stable and efficient than ever before in modern history. The imposed sanctions affected some sectors, but the economy as a whole continues to expand.

Indices Average growth rate (%)

2013 (2000 = 100) - 457.3 12.3

2013 (2005 = 100) - 205.6 9.4

2013 (2010 = 100) - 135.1 10.6


SHORT TERM PERTURBATIONS IN THE MARKET

Short-term analyses obviously reveal adverse phenomena and trends blurring the general view. (However, in the market economy especially open to the world, such situation appears quite normal.)

For example, in the second quarter of 2014 (Figure 2), deflation appeared – a market phenomenon unprecedented in recent history of the Polish economy. In all likelihood it will last until mid-2015. Projections show that the inflation index for the entire year of 2014 dropped to the level of 99.3%. One of the most important reasons for the observed trend was the Russian embargo on food imports from EU-member countries.

Table 2. Indices of the Foreign Trade Dynamics during the Period 1995-2013, 1995=100

Year Index 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

195.0 211.2 250.7 296.3 327.7 380.5 416.3 444.6 409.0 463.0 500.5 517.5 551.1 590.0

Source: Rocznik Statystyczny Handlu Zagranicznego, 2014, p. 52; Biuletyn Statystyczny, 2014a, p. 191; own forecasts. However, it should be remembered that a significant element of the expansion of the

Polish economy during the period 1995-2014, in addition to opening to international markets and an increase in efficiency, was the growth of food exports (Table 1) (Rocznik Statystyczny, 2014, p. 578 and Biuletyn Statystyczny, 2014b, p. 189). Paradoxically, in spite of the technological inadequacy of the food sector in Poland, the economy appeared to be competitive in terms of product quality. Even in 2014, despite the embargo, exports rose by more than 3%, exceeding the average rate achieved during the period 2000-2013.

In recent years, the relatively fast growth in food exports to Russia fluctuated due to the instability of trade conditions and continued political tensions. The share of exports to Russia amounts to 4.4% and is much lower than in the 1990s (Kublik, 2014, p. 18). The 2014 Russian embargo reduced Polish food exports to Russia but at the same time, increased sales in other markets, which confirms the importance of political relations and administrative regulations to international trade.

In the first half of 2014, exports of Polish food to Russia reached the value of $4.84 billion (Kublik, 2014, p. 18) and by the end of 2014 did not exceed $9 billion. What emerged


from the current statistical data was that the total value of food exports in 2014 was expected to be 3-4% higher than in 2013 (Biuletyn Statystyczny, 2014b, p. 189).

However, it should be emphasized that the Russian trade embargo is only one factor in Poland’s economy. The growth and structure of world trade in 2014 were also negatively affected by such factors as political and economic sanctions imposed on the Russian Federation by the United States, the EU, and other developed countries as well as by a dramatic fall in prices of crude oil on the world market.

Source: Rocznik Statystyczny Rzeczypospolitej Polskiej 2014, 2014, p. 457; Roczne wskaźniki

makroekonomiczne, 2014; own forecasts.

Figure 2. Price trend of consumer goods and services during the period 2000-2015, in %.

It has been well established that governments and industries generate instruments which have an effect on the economy both in the long and short term. However, politicians and managers often do not recognize the possible full effects of their actions. The Russian embargo on food imports from the EU, including Poland, seems to be such a case.

In Poland, the embargo resulted in market tensions, but much suggests that these issues will cease within 1 to 2 years since the impact of those adjustments results from high flexibility of trade links. However, considering the inertia of the system, the deflation will probably have ceased by the middle of 2015. Russia will likely face a recession that may obviously lower growth of bilateral trade in general.

It should be emphasized that total exports amount to about 40% of annual GDP in Poland (Rocznik Statystyczny, 2014, p. 563) of which only about 5% are destined for the Russian market (Rocznik Statystyczny, 2014, pp 567, 699) ($10 billion). Assuming that in 2015, GDP will reach the level of $500 billion (Rocznik Statystyczny, 2014, pp 567, 699) (at the forecasted growth rate of 3.4%, see Table 2) and the share of food in exports to Russia will not exceed 5% ($500 million) in absolute terms, the rate of economic growth may decrease by about 0.3-0.5%, i.e., to the level of 3% in Poland. Hence chances are that market perturbations and economic losses generated by the embargo will be temporary and of local importance.


MEDIUM AND LONG TERM: CONSOLIDATION OF FOOD SECTOR

Radical administrative interference may be effective in the realization of short-term political and propaganda goals. However, as an instrument of economic creation it is usually not effective enough, and it is definitely ineffective in raising the effectiveness. Thus, the trade embargo may slow down the expansion of Poland’s food sector, but it will not stop it.

Considering the experience of the previous quarter of the last century, it is plausible the trade embargo will strengthen structural changes in the Polish farm and food sector that in the long run will streamline this sector and the entire economy. Undoubtedly, a more difficult market will accelerate processes of concentration not only of raw material links within the whole primary sector, but also growth and modernization in processing and distribution. It will also speed up development in the sphere of marketing, promotion, and advertising. Finally, it will increase the pace of development of domestic funding for the entire sector.

Therefore, it should be stressed that although several thousand firms sent their goods to Russia in recent years, the number is relatively small and very flexible. Therefore, due to trade sanctions, food exports would fall by $340 million over the entire year of 2014 (Gazeta Wyborcza, 2014). Last but not least, those exports were mainly less processed goods.

Following the sanctions, adjustment processes on the food market in Poland and the entire EU assumed various forms. Almost immediately, institutional shock absorbers were generated based on a joint agricultural policy of the EU. Thus, a system of compensation for sectors and producers of fruit and vegetables was initiated. The authorities made every effort to diversify agricultural production and development of food processing as well as to promote Polish food on new markets: ''The Ministry of Economy is looking for new markets for our food in Asia and in the Balkan countries. Many markets are already open today'' (Rocznik Statystyczny Handlu Zagranicznego, 2014).

An increased supply of food on the domestic market due to the export-limiting trade embargo caused a fall in prices. As a result, there was an increase in domestic food consumption at the expense of imported food. In 2014, exports increased by about 3% (Biuletyn Statystyczny, 2015, p. 189), and food retail sales decreased (in constant prices) by about 5%, even though the increase amounted to a mere 0.3% in 2013 (Biuletyn Statystyczny, 2015, p. 177).

The experience of Poland’s food sector, which until recently supplied mainly the domestic market, confirms the common economic principle that relatively high pressure on prices and costs usually accelerates modernization and system consolidation. The acceleration is all the more present because the dominant characteristic of a modern development mechanism is the highly dynamic creation of new functions, forms, and methods of management rather than reconstruction of the existing system. Thus, a modern economy is more resistant to external disruptions of the system such as the trade embargo in question.

CONCLUSION

Expansion is a natural market phenomenon. However, the key to long-term effectiveness and competitiveness is efficiency. Poland’s food sector, as well as the entire economy, has to increase its competitiveness as fast as possible by fostering advancement in technology and organization as well as offering competitive products. With fewer barriers in world trade, the


market mechanism is more efficient, the economy is more creative, and technological advancement moves faster. After all, a market mechanism is more effective than administrative instruments in trimming the inefficiency tail in the economy. On a global scale it is possible to market any product providing that the ratio between its value and price is reasonable.

Paradoxically, the Russian trade embargo will accelerate the modernization of the Poland’s food sector. The development mechanism which enforces effective adaptation to the market based on new demand requires both a detailed analysis of the effectiveness of investment, trends, and forms of investing as well as the presence of strong market incentives. Such incentives have been generated by the Russian trade embargo itself.

REFERENCES

Biuletyn Statystyczny. 2014a, No. 10 (684), GUS, Warszawa. Biuletyn Statystyczny. 2014b, No. 12 (686), GUS, Warszawa. Biuletyn Statystyczny. 2015, No. 1 (687), GUS, Warszawa 2015. Gazeta Wyborcza. 2014. Polska żywność ruszy w świat?, No. 260(8291). Instytut Badań Strukturalnych 2006. Źródła i perspektywy wzrostu produktywności w Polsce,

Warszawa, p. 17. Kublik, A., W. Matusiak, P. Maciejewicz. 2014. Rosja po rublowym szoku. Gazeta

Wyborcza” 2014, no. 293(8324). Meredyk, K. 2012. Tendencje rozwojowe gospodarki polskiej, w: Między ekonomią a

historią. Studia ofiarowane Prof. Czesławowi Noniewiczowi z okazji 75. urodzin. R. Dziemianowicz, A. Kargol-Wasiluk, J.Wilkina,M. Zalesko (eds). Wydawnictwo Uniwersytetu w Białymstoku, Białystok, pp. 75-94. ISBN 978-83-7431-324-7.

Roczne wskaźniki makroekonomiczne. 2014 GUS, Warszawa. Available online at http://stat.gov.pl.

Rocznik Statystyczny Handlu Zagranicznego 2014, GUS, Warszawa 2014, p. 52. Rocznik Statystyczny Rzeczypospolitej Polskiej 2011, GUS, Warszawa 2011, p. 675. Rocznik Statystyczny Rzeczypospolitej Polskiej 2014, GUS, Warszawa 2014. Ministerstwo Finansow. 2015. Sytuacja makroekonomiczna w Polsce. Departament Polityki

Makroekonomicznej, February 27, 2015. Available online at www.mf.gov.pl. Accessed March 3, 2015.


APPLE SECTOR IN POLAND AND EFFECTS OF THE RUSSIAN EMBARGO

A. M. Klepacka Faculty of Production Engineering, Wydział Inżynierii Produkcji, Warsaw Univeristy of Life Sciences (SGGW), Warszawa, Poland

ABSTRACT

This paper examines the aftermath of a series of embargoes targeting EU food exports and especially fresh produce exports from Poland to Russia. Poland has been exporting large volumes of fresh apples to Russia for decades and was its primary supplier in 2013. The embargo resulted in substantial losses for Polish apple and fruit growers that were only covered in part by the EU compensation program. Promotion of domestic apple consumption might have reversed the observed decline in per capita apple consumption, but the majority of surplus apples were processed into juice concentrate and cider. Energetic efforts on developing new export markets have been undertaken leading to some success. However, additional markets need to be developed, while growers have to select varieties preferred by consumers in less traditional markets. Some growers may be forced out of apple production if unable to manage costs. The embargo led to accelerated food price inflation and higher fresh apple prices in Russia, at least in the short term.

Keywords: re-export, apple variety, exchange rate, food price inflation, policy response JEL: F14, F51

INTRODUCTION

Apples are the main fruit produced in Poland. For decades, fresh apples have been exported from Poland to the Soviet Union (later, Russia). The apple sector has undergone a major restructuring and traditional orchards have been replaced by high density orchards. The change in production also involved planting new varieties. The investment in tree replacement led to a rapid expansion of apple production in recent decades. In 2004, Poland produced 2.522 million tons of apples, 2.626 million tons in 2009, and 3.084 million tons in 2013 (Trajer, 2015).

As apple production expands, the annual per capita apple consumption has been decreasing in Poland. In 2006, the average consumer ate 1.97 kg of apples per year (CSO, 2007), 1.76 kg in 2010 (CSO, 2011), and 1.57 kg in 2013 (CSO, 2014). The decreasing apple

A. M. Klepacka 122

consumption varies across household types (Klepacka et al., 2014), and tends to be more important for low income households (Florkowski, 2012). Declining consumption and increasing production encouraged growers to seek markets outside Poland.

Fresh apples have become the primary fruit exported from Poland. Between the crop years 2001/02 and 2010/11-2012/13, the share of exports in total apple production increased from 14% to 27% (Nosecka, 2015). During the same period, the share of apples in all horticulture exports increases from 7% to 17% and the share of apples in fruit exports rises from 47% to 71% (Nosecka, 2015). The majority of Polish apples have been exported to the well-established markets in the former Soviet Union republics, i.e., Russia, Ukraine, and Belarus. Other destinations for Polish apples include countries in Scandinavia and Bosnia & Hercegovina. The importance of the Russian market was indisputable in 2013.

This paper provides insights into the effects of the embargo introduced on August 7, 2014 by the Russian Federation on imports of fruit and vegetables from the European Union (EU) country-members with an emphasis on Poland’s apple exports. The Russian Federation introduced a temporary ban on August 1, 2014, retaliating for the EU sanctions imposed after the annexation of Crimea (a part of Ukraine) in March 2014. That temporary ban was followed by the announcement of the embargo on August 7, 2014, with the effective date of August 8, 2014. The primary motive of the embargo was political, but the result has wide trade and economic implications for Poland, other EU member countries, Russia, and a number of other countries. An additional embargo on re-exporting of Polish produce was introduced in October 2015 reflecting the wide efforts to limit access to the Russian market.

THE IMPORTANCE OF APPLE EXPORTS

The transition to market economy forced Polish apple growers to focus on cutting production costs and improving fruit quality. Efforts in the early 1990s included the adoption of cultural practices limiting the use of pesticides (Florkowski et al., 1996). Growers experienced competition from imported tropical fruit on an unprecedented scale following the opening of borders to free international trade. The resentment of competition was due to the substantial decrease in revenues. Under the system of central planning, with its controlled prices and restricted varieties of fresh fruit, apple sales generated relatively high revenues and a good standard of living for producers in an environment of permanent shortages of almost all consumer goods. But Polish apple growers have adapted to new economic conditions. Apple production has been expanding in recent decades in Poland stimulated by the replacement of old orchards by high-density, irrigated orchards.

The investment in new technology and planting of new varieties has resulted in increased production (Figure 1), although the area of orchards also increased in recent decades (Pizlo, 2011). The calculated trend using data from the most recent period shows that production has been increasing rather rapidly. In 2014, the apple crop was the highest on record. Figure 1 also shows the volume of apple exports. Exports had been increasing during the period under consideration and the rate of growth of exports closely corresponded to the production growth. But in 2014 the production and exports volumes diverted; the year of the record crop witnessed a decline in exports.

Apple Sector in Poland and Effects of the Russian Embargo 123

Source: Based on Trajer, 2015.

Note: Figures for 2014 are preliminary.

Figure 1. Trends in volume of harvested apples and apple exports, Poland, 2004-2014.

The cause of decline in fresh apple exports was the embargo on importation of Polish apples to Russia. Russia had become the primary export market for Polish apples in the years leading up to the embargo. Two other East European countries, Belarus and Ukraine, were also importing large volumes of apples from Poland. The importance of these three markets is captured by the increasing share in apple exports. In the period 2010/11-2012/13, the share of apple exports to Russia grew from 44% to 54%, while that of Ukraine and Belarus increased from 15% to 26% (Nosecka, 2015).

POLICY RESPONSE

The embargo on food imports declared by the Russian Federation immediately disrupted the shipments of fresh fruit and vegetables from the EU. Growers and shippers faced sudden losses in the middle of the harvesting season. Poland suffered the largest losses in apple exports among the EU countries due to the embargo (Makosz, 2015) as her estimated share of apple exports to Russia was about 80-85% in the preceding two years.

EU Compensation Scheme

Fruit and vegetable growers across the EU were affected by the trade embargo introduced by the Russian Federation. The temporary restrictions on fruit and vegetable imports from Poland were announced on August 1, 2014. It was followed by the announcement of an embargo on food imports from the EU, the United States, Australia, Canada, and Norway. It

A. M. Klepacka 124

became effective on August 8, 2014 and was to remain in place for 12 months (it has been extended and remains in place). Besides fruits and vegetables, the embargo banned the importation of fish, milk and dairy products, and meat and meat products. Soon, another embargo followed. Announced on October 6, 2014, it targeted only the re-export of Polish plant products to counter the shipments of, primarily, apples through other countries, such as Belarus and Serbia. Still another embargo was introduced on October 21, 2014 preventing importation of beef offal and other beef by-products from the EU. The last one did not apply to fruits and vegetables, but reflects the attitudes of Russian policy decision-makers.

In response to the embargo, on August 18, 2014 the EU released 125 million euros to compensate the withdrawal from the market of perishable fresh fruits and vegetables (AgroNews, 2014a; 2014b). Of that amount, 82 million euros was to support the apple and pear sectors, while 43 million euros was for other produce. An additional 33 million euros was allocated for the support of peach and nectarine growers (AgroNews, 2014a). Compensation applications soon exceeded the allocated amount. More than 85% of applications for compensation originated in Poland. The European Commission (EC) questioned the amount of losses on some applications and suspended the program in the middle of September 2014. The delays and the total amount of compensation intended for growers of all produce in all 28 EU countries led to tensions. In October 2014, the EC implemented another compensation program to supplement the earlier scheme in the amount of 165 million euros.

Overall, the EC allocated the compensation funds for 12 affected countries. The program encompassed up to 181,000 tons of apples, 96,000 tons of citrus fruit, 44,300 tons of vegetables, and 78,800 tons of other fruit (kiwi, plums, table grapes). The program excluded compensation for growers of cauliflower, cabbage, broccoli, button mushrooms, and berries. The EC allocated compensation funds for Poland to cover losses by withdrawing from the market 18,750 tons of apples (an estimated 0.005% of the crop and 3.13% of 2013-14 exports to Russia), 3,000 tons of pears, and other fruits and vegetables (Wsparcie z Unii dla Rolnikow, 2015). The highest compensation was received by those growers who withdrew their fruit from the market and allocated it for free distribution. The importance of the program is reflected in the speed with which the government agency in charge of its implementation received applications – the agency closed the period of applying for compensation after 24 hours because the claimed value of applications exhausted the allocated amount. The proposed funding was clearly insufficient to compensate for losses due to the embargo not only for apple growers, but for all growers of fruits and vegetables.

Expectations for full compensation were unrealistic. The majority of perennial fruit growers in Poland and elsewhere in the EU suffered substantial economic losses in the 2014/15 crop year. The perennial nature of apple trees and the implied asset fixity forced many growers to re-evaluate production practices (to lower production costs) as well as apple varieties planted in new orchards (see below) or used to replace old trees, as orchards have been converted from low- to high-density. Although current supply is determined by existing orchards, the future supply of apples and other fruits and vegetables will be affected by the Russian embargo.


Growers’ Reaction

The harvest of early maturing apple varieties started in August 2014, but the peak harvest period is typically in the second half of September and October in Poland. However, apple traders and processors factored the effects of the embargo in their purchase plans and procurement prices were so low that some growers considered harvesting only some varieties or none at all. The domestic apple supply conditions changed overnight, while the domestic consumption level was anticipated to remain at the level experienced in previous years.

Faced with the market uncertainty threatening the economic existence of many apple farms, growers organized a series of demonstrations in early September 2014. They targeted major institutions, e.g., the Parliament, the Ministry of Finance, and the Ministry of Agriculture and Rural Development, and partially blocked traffic in Warsaw (AgroNews, 2014b) and other parts of the country. Demonstrations were accompanied by petitions to the government demanding assistance in alleviating the sudden and painful economic consequences of the Russian embargo. To alleviate the supply glut, a major effort was undertaken by the Ministry of Agriculture and Rural Development and other agencies to promote Polish apples in domestic and foreign markets.

PRICE RESPONSE

Disappearance

The record Polish apple crop in 2013 coincided with a very good total crop of 12 million tons for all 28 EU countries, 2 million tons above the 10-year average (Makosz, 2015). The surplus crop in combination with the embargo led to difficulties in apple disposition. In the fall of 2014, the expected disappearance of apple crop in Poland included 2.1 million tons processed into juice concentrate and cider, 0.7 million tons for the domestic fresh consumption, 0.55 million tons of high quality for export market, and 0.05 million tons for school programs and charity. The balance of about 0.3 million tons was the balance that would require compensation, subsidies, or other form of assistance. The losses resulting from the embargo in the apple sector were estimated at 1.3 billion Polish zloty (PLN). Another 200 million PLN could be losses resulting from unsold inventory. Additional losses were expected by transportation firms and input suppliers.

Prices

In September 2014, retail apple prices hovered around 1.5 PLN per kg, or about 40% below the level of a comparable period in 2013. A common opinion among growers was that such a price level could not sustain production. Procurement prices for processing apples ranged from 0.10 to 0.20 PLN per kg (about $0.03-$0.07) and were 0.55 PLN lower than a year earlier. Fresh apple procurement prices increased to 0.40-0.60 PLN per kg as the export to East European countries started later in the fall 2014. But since March 2015, the price increase was noticeable as apples were exported in different directions. Additionally, United

A. M. Klepacka 126

States-based buyers purchased enough apple juice concentrate to lift procurement prices for processing apples to 0.50 PLN per kg.

Source: Multiple issues of Rynek owoców i warzyw świeżych, 2013, 2014, 2015.

Note: The series is not continued and consists of three periods. The new period is preceded by either a drop or an increase in price.

Figure 2. Lower range of apple prices for Gala variety at Bronisze wholesale market between Oct 1, 2013 and Nov 26, 2015, per kg.

Figure 2 provides an illustration of changes in fresh apple prices using the example of “Gala,” second to “Golden Delicious” in volume produced in the EU and well-liked by EU consumers. All prices were the recorded at the Bronisze wholesale market, located near Warsaw, Poland. The prices through May 2014 were quite strong and in the months leading to May 2014 resemble the typical seasonal pattern. The opening prices in October 2014 were much lower, about 50% below the 2013 level. The prices recovered somewhat and in October 2015, “Gala” prices opened at 1.33 PLN per kg and increased in November 2015 to 1.5 PLN per kg.

The embargo did not affect all apple growing regions in Poland to the same extent. Some Polish growers specialized in varieties preferred by consumers in Eastern Europe like “Idared” (consumers in Western Europe prefer “Gala”). This variety is more commonly grown in eastern and south-eastern regions of Lubelszczyzna and Sandomierz. The variety’s share in apple exports to the Eastern European market was about one third prior to the embargo. Its popularity among Polish consumers was limited. From a post-embargo viewpoint, the variety is only suitable as a processing apple which means much lower prices for growers. The World Apple Grower Association estimated the 2015 “Idared” crop at 1.11 million tons (FAMMU/FAPA, 2015), much of it originating in Poland. With the imposition of the embargo, growers replacing trees in their orchards are less likely to plant “Idared,” slowly changing the varietal composition of the Polish apple crop over time.


Search for New Markets

The rapidly accumulating surplus supply of harvested apples in 2014 happened after years of declining per capita apple consumption. The fruit purchasing behavior of Polish consumers may reverse in response to unprecedented low prices. Media began promoting apple consumption, and these promotions were taking a more organized character. A few days after the embargo was announced, free apples were offered in the streets of two cities, Suwalki and Bialystok, to promote Polish fruit and the regional economic food processing zone (onet.pl, 2014). A Polish railroad company initiated on September 9, 2014 (as the apple growers’ demonstrations were most visible) a program of serving free apples to its passengers traveling in all high-speed trains under the program of serving healthy snacks by its catering service “Wars.” The volume of free apples, 40 tons, was inconsequential to market prices, but the program was very visible given the substantial number of passengers (and their average or above average purchasing power). However, the major user of domestic apples was apple juice concentrate producers who received a substantial volume of the variety “Idared.” “Idared,” unless it could be sold to consumers in Eastern Europe could not outcompete other varieties on the fresh apple market, and even if offered at low prices could not find enough buyers.

One particular effort to stimulate the domestic disappearance, and consistent with the intent of the EU compensation program, is the program of serving fresh fruit and vegetables in Polish schools. In the year the embargo was introduced, 1.34 million school children from a little over 11,000 schools participated in the program. The program expanded in the current crop year by another 100 schools. Children are served fresh fruits, vegetables, and fruit and vegetable juices at least twice a week. The program costs 22.5 million euros and 88% of it is funded by the EU. The goal of the program is to develop healthy eating habits and in each school it is supplemented by other activities including growing vegetables in school gardens, tours of orchards, and culinary workshops for children.

The energetic search for new markets by a sector which habitually shipped apples with well-established quality attributes to the same destinations for decades predictably resulted in a trial-and error process. For example, the Polish-Saudi Business Council representative commented on the lack of interest among Polish apple growers regarding the embargo effects (Puls Biznesu, 2015). The trial shipments sent to the UAE in anticipation of large imports were rejected by the importing country inspection and later, growers doubled their prices. Export to Nigeria did not materialize despite a concluded contract.

However, the majority of efforts in finding new markets were successful. In November of 2014, Polish apples reached retail outlets in Canada, but this effort was a result of the processes already started in 2012. On October 23, 2014, Canadian authorities finalized their evaluation of the Polish apple sector and permitted the importation of fresh apples. This has been a substantial development given the importance of the domestic apple sector in Canada, the proximity of the suppliers from the United States, and a declining consumption of apples in the country (Carew et al., 2006). The competitively priced Polish apples found a niche on the Canadian market and the exports will likely continue given the formal approval, although the volume exported may fluctuate. A much easier effort was in shipping apples to Singapore (with an initial shipment of 120 tons in November 2014) because the phytosanitary certification required is the importer’s responsibility.

A. M. Klepacka 128

But the search for new markets was quite broad and led by the Ministry of Agriculture and Rural Development. The short apple crop in Turkey opened another opportunity as did the drought in China’s apple growing regions. China is viewed as a potential major market for Polish apples in the future. Other places that became targets of exporters of Polish apples included India, Japan, Kazakhstan, Azerbaijan, northern Africa, Chile, and the Middle East. Within the region, however, Polish apple exporters countered some opposition from growers in other countries (like the Czech Republic) promoting their own domestic production.

The 2015 crop was expected to establish a new volume record, but the summer drought lowered the crop about 5% below the previous year’s crop. Apple exports started early, during the harvest, unlike in the past. The primary exported variety was “Gala” shipped to other EU countries and high quality fetched high prices.

Polish Zloty and Ruble Exchange Rates

The exchange rates most important for Polish apple growers are those with the euro and the American dollar because of their common use in export contracts. Since January 2014, the Polish zloty strengthened a bit against the $. The value of the PLN eroded starting in July 2014. The decrease in the value of Polish currency between July 4, 2014 and January 2, 2015 the zloty was 14.8% against the dollar. The PLN also weakened against the euro, but considerably less.

The weakening Polish currency had multiple immediate effects. Prices of Polish apples were more attractive to importers paying in dollars. But, the weaker PLN decreased export revenues for growers who chose to ship apples abroad. To protect their revenues, as apple import contracts were negotiated, some growers attempted to increase their prices (see earlier discussion about the search for markets) which, in turn, made the imports less price competitive. The weakening of PLN against the euro was less and the value of compensations expressed in euros changed little. Additional exports of apples to EU countries have become relatively more attractive to Polish growers.

The value of PLN strengthened, however, against Russia. Not only did the embargo restrict apple exports, but the deteriorating ruble value increased the relative price of Polish and other imported apples. The lower procurement price of apples in Poland offset the strengthening exchange rate, at least in the short term.

The value of the ruble remained relatively unchanged until early July 2014. But, between July and December 2014, the ruble fell precipitously against the dollar and the euro. For example, on July 2, 2014 it cost 34.46 rubles to buy a dollar, but on January 2, 2015 the price of a dollar increased to 58.75 rubles. The value of the ruble reflected the price of oil on the world market, and oil and gas revenue is of critical importance for the Russian economy and national budget. Russian consumers were the primary loser in the process. The embargo, the deteriorating exchange rate, and the economic slowdown (which started in 2013) resulted in weakening domestic demand. The decision to retain the trade embargo for the foreseeable future and the continuing softness in the oil and energy markets make chances for domestic demand recovery in Russia slim.


Price Inflation

The sudden increase in the supply of apples and other fresh produce dampened price increases. Figure 3 shows the monthly changes in the CPI in Poland from the year preceding the embargo until October 2015. The slowdown in price inflation was already observed in July, i.e., the month prior to the declaration of the trade embargo. The deflation deepened in subsequent months and has been present in 2015. Because the deflation was already reported in other EU countries earlier in 2014, the embargo was not the only contributing factor. Nevertheless, food prices together with decreasing energy prices contributed to the deflation.

Source: GUS, 2015.

Note: Monthly values for 2015 available through October.

Figure 3. Monthly index of consumer prices, Poland, 2013-2015.

In contrast, Russia experienced accelerated inflation since July 2014. Initially the price increases were slow, but the food price inflation exceeded 16% in December 2014. Retail apple prices ranged from $1 to $1.5 per kg in Russia from December 2014 to January 2015. High food prices are concerning because the average Russian household spends one half of its income on food. The inflation rate stabilized until the fall of 2015. However, the recently introduced embargo on food imports (November 2015), including fresh vegetables and fruit, from Turkey suggests that Russia chooses to use selective trade embargos regardless of its effects on the domestic consumer.

A. M. Klepacka 130

CONCLUSION

It appears that the effects of the embargo led to short term disturbances in the Polish fruit market, product innovation, and the search for new markets, while in Russia there appears to be, at least in the short run, a gap between demand and supply causing an accelerated inflation limiting the consumption of fresh apples. In the medium term, Polish apple growers are forced to aggressively search for new markets, whereas Russian consumers may face limited or more expensive deliveries from alternative suppliers. Polish fruit growers have shown agility in their adjustment to changing economic conditions since the transition to a market economy in 1989 and are likely to improve efficiency to remain competitive. Russian consumers face price increases eroding their purchasing power and potentially weakening the domestic demand and harming economic growth.

The embargo appears to have stimulated the movement towards establishing apple varieties that could lead to certification of geographic origin under the EU law. For example, in the region of Zulawy (AgroNews, 2015) scientists assisted in an inventory of apple (and other fruit tree) varieties in abandoned and old orchards. The goal is to eventually establish local varieties that could distinguish regional production, although this region has not been known as an important apple producer. This regional effort reflects the movement towards planting new apple varieties suitable to Poland’s climatic conditions in order to win new consumers in both domestic and foreign markets.

The transition that the Polish apple sector has been undergoing is not without its economic effects. The loss of revenues will force some growers to re-consider whether to continue apple production or replace apple orchards with other crops. In the short term, all growers will seek ways to lower production costs, while seeking and establishing niches in new markets overseas.

REFERENCES

AgroNews. 2015. Dzialacze z Zulaw chca ocalic lokalne historyczne drzewa owocowe. Available online at http://www.agronews.com.pl/pl/4,7,19349,dzialacze_z_zulaw_chca_ ocalic_lokalne_historyczne_drzewa_owocowe.html. Accessed December 2, 2015.

AgroNews. 2014a. Siekierski: KE moze odrzucić większość wniosków (Rosyjskie “niet”). Available online at www.agronews.com.pl Accessed September 12, 2014.

AgroNews. 2014b. Sadownicy zaprotestowali w Warszawie (Rosyjskie “niet”). Available online at www.agronews.com.pl Accessed November 4, 2014.

Carew, R., W. J. Florkowski, and E.G. Smith. 2006. Apple Industry Performance, Intellectual Property Rights and Innovation: A Canada-U.S. Comparison, International Journal of Fruit Science, 6(1):93-116.

Central Statistical Office. 2007. Households budget surveys 2006 Available online at http://stat.gov.pl/obszary-tematyczne/warunki-zycia/dochody-wydatki-i-warunki-zycia-ludnosci/budzety-gospodarstw-domowych-w-2006-r,9,1.html. Accessed Jan. 16 2015.

Central Statistical Office. 2011. Households budget surveys 2010 Available online at http://stat.gov.pl/cps/rde/xbcr/gus/LC_household_budget_surveys_in_2010(1).pdf. Accessed January 16 2015.


Central Statistical Office. 2014. Households budget surveys 2013. Available online at http://stat.gov.pl/obszary-tematyczne/warunki-zycia/dochody-wydatki-i-warunki-zycia-ludnosci/budzety-gospodarstw-domowych-w-2013-r-,9,8.html. Accessed Jan. 16 2015.

FAMMU/FAPA. 2015. WAPA. Prognozy dla tegorocznych zbiorów jabłek i gruszek. Available online at www. agronews.com.pl. Accessed August 25, 2015.

Florkowski, W. J. 2012. Evidence of household consumption changes associated with the global economic slowdown. Poznan University of Economics Review, 12(3):42-79.

Florkowski, W.J., E. Majewski, and M. Bagel. 1996. Opportunities in Production and Marketing of Integrated Fruit Production (IFP) Certified Apples in Poland. Journal of Tree Fruit Production, 1(2):15-31.

GUS, 2015. Available online at http://stat.gov.pl/obszary-tematyczne/ceny-handel/wskazniki-cen/wskazniki-cen-towarow-i-uslug-konsumpcyjnych-pot-inflacja-/miesieczne-wskazniki -cen-towarow-i-uslug-konsumpcyjnych-od-1989-roku/. Accessed Novemebr 30, 2015.

Klepacka, A.M., Meng, T., Florkowski, W.J. 2014. Apple or oranges? Recent household fruit consumption in Poland. Annals SERiA, XVI(6):220-229.

Makosz, E. 2015. Towarzystwo Rozwoju Sadów Karłowych, Lublin. Available online at http://www.trsk.pl/aktualnosci.html Accessed January 19, 2015.

Nosecka, B. 2015. Konferencja Polskie Jabłka. IERiGŻ-PIB. Available online at http://tsw. targi.pl/blog/2014/06/05/polski-eksport-jablek/. Accessed January 21, 2015.

Onet.pl. 2014. Przedstawiciele SSE rozdają jabłka Available online at www.agronews.com.pl Accessed August 11, 2014.

Pizlo, W. Economic situation of the Polish fruit growers in the period 1999-2009. Acta Scientiarum Polonorum – Oeconomia, 10(3):123-134.

Puls Biznesu. 2015. Polskie jabłka nie chcą podbijać świata. Available online at www. agronews.com.pl Accessed January 13, 2015.

Rynek Owoców i Warzyw Świeżych - notowania za okres: 07.01.2013 - 17.01.2013. 2013. Available online at http://www.minrol.gov.pl/Rynki-rolne/Zintegrowany-System-Rolni czej-Informacji-Rynkowej/Biuletyny-Informacyjne/Rynek-owocow-i-warzyw-swiezych/ Rynek-Owocow-I-Warzyw-Swiezych-notowania-za-okres-07.01.2013-17.01.2013. Accessed January 18, 2015.

Rynek Owoców i Warzyw Świeżych - notowania za okres: 11.02.2013 - 21.02.2013. 2013. Available online at http://www.minrol.gov.pl/Rynki-rolne/Zintegrowany-System-Rolni czej-Informacji-Rynkowej/Biuletyny-Informacyjne/Rynek-owocow-i-warzyw-swiezych/ Rynek-Owocow-I-Warzyw-Swiezych-notowania-za-okres-11.02.2013-21.02.2013. Accessed January 18, 2015.

Rynek Owoców i Warzyw Świeżych - notowania za okres: 09.09.2013 - 19.09.2013. 2013. Availbale online at http://www.minrol.gov.pl/Rynki-rolne/Zintegrowany-System-Rolni czej-Informacji-Rynkowej/Biuletyny-Informacyjne/Rynek-owocow-i-warzyw-swiezych/ Rynek-Owocow-I-Warzyw-Swiezych-notowania-za-okres-09.09.2013-19.09.2013. Accessed January 18, 2015.

Rynek Owoców i Warzyw Świeżych - notowania za okres: 21.10.2013 - 31.10.2013. 2013. Availbale online at http://www.minrol.gov.pl/Rynki-rolne/Zintegrowany-System-Rol niczej-Informacji-Rynkowej/Biuletyny-Informacyjne/Rynek-owocow-i-warzyw-swiez ych/Rynek-Owocow-I-Warzyw-Swiezych-notowania-za-okres-21.10.2013-31.10.2013. Accessed January 18, 2015.

A. M. Klepacka 132

Rynek Owoców i Warzyw Świeżych - notowania za okres: 04.11.2013 - 14.11.2013. 2013. Availbale online at http://www.minrol.gov.pl/Rynki-rolne/Zintegrowany-System-Rolni czej-Informacji-Rynkowej/Biuletyny-Informacyjne/Rynek-owocow-i-warzyw-swiezych/ Rynek-Owocow-I-Warzyw-Swiezych-notowania-za-okres-04.11.2013-14.11.2013. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 13.01.2014 – 23.01.2014 r. 2014. No 3/, January 23, 2014. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2003_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 17.02.2014 – 27.02.2014 r. 2014. No 8, February 27, 2014. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2008_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 24.02.2014 – 06.03.2014 r. 2014. No 9, March 6, 2014. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2009_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 31.03.2014 – 10.04.2014 r. 2014. No 14, April 10, 2014. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2014_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 19.05.2014 –29.05.2014 r. 2014. No 21, May 29, 2014. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2021_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 06.10.2014 –16.10.2014 r. 2014. No 41, October 16, 2014. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2041_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 24.11.2014 –4.12.2014 r. 2014. No 48/, December 4, 2014. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2048_14.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 01.12.2014 –11.12.2014 r. 2014. No 49, December 11, 2014. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2049_14.pdf Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. Notowania W Dniach 22.12.2014 –05.01.2015 r. 2015. No 52, January 9, 2015. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2001_15.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach 5.01.2014 –15.01.2015 r. No 2, 1 January 15, 2015. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2002_15.pdf. Accessed January 18, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach 19.01.2014 –22.01.2015 r. Nr 3/2015 22 stycznia 2015r. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2003_15.pdf Accessed November 27, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach: 13.07.2015 - 23.07.2015 r. Nr 29/2015 23 lipca 2015r. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2029_15.pdf Accessed November 27, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach: 12 - 22.10.2015 r. Nr 42/2015 22 października 2015r. Available on line at http://www.agronews.com.pl/files/ File/ceny_rynkowe/ow_wa%2042_15.pdf Accessed November 27, 2015.


Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach: 19 - 29.10.2015 r. Nr 43/2015 29 października 2015r. Available on line at http://www.agronews.com.pl/files/ File/ceny_rynkowe/ow_wa%2043_15.pdf Accessed November 27, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W DniACH: 26.10 – 05.11.2015 r. Nr 44/2015 5 listopada 2015r. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2044_15.pdf Accessed November 27, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach: 02 - 12.11.2015 r. Nr 45/2015 12 listopada 2015r. Available on line at http://www.agronews.com.pl/files/File/ ceny_rynkowe/ow_wa%2045_15.pdf Accessed November 27, 2015.

Rynek owoców i warzyw świeżych. 2015. Notowania W Dniach: 16 - 26.11.2015 Nr 47/2015 26 listopada 2015r. Available on line at http://www.agronews.com.pl/files/File/ceny _rynkowe/ow_wa%2047_15.pdf Accessed November 27, 2015.

Trajer M. 2015. Handel zagraniczny na rynku owoców i warzyw, in: Biuletyn informacyjny 3/2015, Rynek owoców i warzyw, Produkcja, Konsumpcja, Handel Zagraniczny, ARR, pp 16-19.


THE RUSSIAN EMBARGO: TRIBULATIONS OF DAIRY AND MEAT EXPORTS FROM POLAND

AND OTHER EU MEMBER COUNTRIES

Marcin Wysokiński and Joanna Baran Warsaw University of Life Sciences (SGGW), Poland

ABSTRACT

The paper describes the increasing exports of food and agricultural products from the EU, especially Poland, and the aftermath of the ban on pork imports followed by a trade embargo declared by the Russian Federation in August 2014. The effects of the latter are illustrated, with particular focus on milk and dairy and meat products. While pork imports to Russia have already been banned because of the African swine fever (ASF) in January 2014, the food trade embargo contributed to a decline in hog and pork prices in the EU and a likely decrease in hog herds. In Poland and several other countries, the embargo also severely impacted the dairy industry in the wake of a global decline of milk and dairy product prices. Within the EU, efforts were undertaken to offset the sudden supply increase by supporting inventory stocking, but the measure had a limited effect. Although the embargo affected a small section of the EU agricultural production and exports, the dairy and hog sectors were strongly affected in some countries and consequently the economy of regions heavily dependent on agriculture was impacted as well. However, the poultry sector reached record production and export levels in the year after the embargo was implemented. The embargo compelled producers to the search for new markets within and outside the EU to satisfy consumers, an in the longer run it may cause a major restructuring of the Polish hog and dairy sector.

Keywords: trade embargo, dairy products, meat exports, re-export, trade policy JEL: F14, Q19, F52

INTRODUCTION

Trade stimulates economic development and contributes to national wealth (Pawlak and Poczta 2011). The rate of growth of the value of international trade exceeds the world’s GDP growth rate. Such trade also has a rising share in Poland’s GDP (Szajner, 2012). The accession of Poland to the European Union (EU) in May 2004 drastically changed the trade

Corresponding author’s e-mail: [email protected]

Marcin Wysokiński and Joanna Baran 136

policy and altered the country’s competitive position with regard to both other EU members and non-EU countries. Poland enjoyed a relative competitive advantage and the value of its exports increased fivefold by 2014 as compared to the pre-accession level in the international food and agricultural trade (Szajner, 2012). Trade in food and agricultural commodities is, to an increasing degree, tied to global developments, as reflected in its share in the total value of Polish foreign trade. The share increased from 8.8% in 2004 to 13.2% in 2014 (IERiGŻ, 2005; 2015). Similarly, the share of exports in total sales of agricultural commodities and food products increased to 35% in 2014 (ARR, 2015).

The susceptibility of international trade relations to political influences is clearly illustrated by developments in international political and economic relations between Russia and the world in 2014. The annexation of Crimea and the following military conflict between Russia and Ukraine instantly worsened international relations. In response to these events, the United States, EU, Australia, Norway, Japan, Canada and other countries introduced economic sanctions against Russia. Russia responded by declaring a trade embargo on food imports from these countries on August 7, 2014. The embargo took effect on the next day (August 8, 2015) and covered meat, dairy products, fruits, vegetables, and processed foods, among others (see Appendix). EU food exports, with the exception of poultry and fish and seafood exports, were among the most affected (Table 1).

Table 1. Shares of Selected Foods in Russia’s Total Food Imports Affected by the 2014

Ban Prior to the Ban in 2013, in Percent

Country Products

Beef Pork Poultry Fish& seafood

Milk & milk products

Vegetables Fruits

Australia 4.1 0.0 0.0 0.0 0.9 0.0 0.1

Canada 0.0 11.1 0.0 4.1 0.0 0.1 0.0

EU 4.6 58.9 10.6 7.5 37.4 31.9 23.5 Norway 0.0 0.0 0.0 39.0 0.1 0.0 0.0

USA 0.0 0.9 37.7 2.6 0.0 0.3 3.6

Total 8.7 70.9 48.3 53.2 38.4 32.3 27.3 Source: UN Comtrade.

This paper tracks agricultural and food trade between the Russian Federation and the

European Union (EU) and examines the scope of the Russian embargo on food and agricultural products exported from the EU. The focus is on EU member states bordering Russia, especially Poland, because of the substantial exports of food products and identifiable short-term effects of sanctions. An overview of the trade pattern prior to the embargo imposed by Russia in August 2014 is followed by a description of the immediately observed reaction of EU and Russian markets to the sudden disruption of trade flows. Illustrations of short-term effects are drawn from macro- and microeconomics. The description is based on information and data from the European Commission, the Eurostat database, and data from the Ministry of Finance of the Republic of Poland, the Institute of Agricultural Economics and Food Economy, and other public sources related to trade in food and agricultural products.

The Russian Embargo 137

GROWTH OF POLISH FOOD EXPORTS

Following integration with the EU, the development of the Polish agriculture and food industry has increasingly been export-oriented. The index of export-oriented production in food industry and agriculture, i.e., the ratio of export revenues to the sector’s total sales value, increased from 22% in 2008 to 35% in 2014 (IERiGŻ, 2009; 2015. IERiGŻ PIB). The rapid increase of the value of commodity and food exports from 7.1 billion euro in 2005 to 21.9 billion euro in 2014 (Rocznik Statystyczny Handlu Zagranicznego, 2008; 2015) has been matched by the fast growth of the food processing industry. Sales of processed foods measured in constant prices increased by 42% in the period from 2005 to 2014 (Rocznik Statystyczny Przemysłu, 2015). The growth of animal production exceeded the growth of domestic demand. Indeed, domestic sales of food and beverages decreased by 1.5% in constant terms during the period 2005-2014 (Rocznik Statystyczny RP 2014, 2015), forcing producers and processors to seek buyers outside the country.

Geographically, EU countries are the primary destination for Polish food and agricultural products. Their share was 79.6% in 2014 (Ministerstwo Rolnictwa i Rozwoju Wsi RP, 2015). The factor driving export expansion is the unrestricted access to markets of EU countries, but also the ability of Polish producers to meet the sanitary and veterinary EU standards. Last but not least, Polish food products are competitively priced. Polish food and agricultural exports have not been limited to EU states. Russia has been a major trade partner of Poland and other EU countries for years.

FOOD SECURITY DOCTRINE

Large food imports have been noted and perceived as a threat to Russia because of the low level of food self-sufficiency (Ishchukova and Smutka, 2015). Consequently, in 2010 the Russian Federation issued a decree concerning food security. The new policy defines certain minimum shares of domestic production in the total domestic supply of basic agricultural commodities. The major objective is to reach a minimum share of domestic supplies on the level of 95% for grains, 80% for sugar, 80% for vegetable oils, 85% for meat and meat products, 90% for milk and dairy products, 80% for fish and processed fish, 95% for potatoes and 85% for food-grade sodium (salt). These objectives are expected to be reached by 2020 (Lewandowska, 2010).

The decree of the Russian Federation concerning food security has become a tool to control food imports. Even before the 2010 decree, the nature of Russian economic policy included restrictions imposed on trade with Poland from 2005 to 2008. Specifically, a sudden decision to invalidate phytosanitary certificates issued in Poland led to a decrease in exports of plant products, especially fruits and vegetables. Later, in June and July 2011, Russia introduced an embargo on exports of vegetables from Poland. The embargo was introduced during the peak production season. Finally, since January 2014, Russia terminated pork and pork product imports from EU because of the African swine fever (ASF) incident in Poland (Ministerstwo Rolnictwa i Rozwoju Wsi RP, 2015).

The list of embargoed products corresponds to food security objectives aiming at replacing imports with domestic production defined in the food security doctrine of the Russian Federation. Application of similar domestic measures is not otherwise permitted


under the WTO policy, although even prior to sanctions the Russian trade policy was highly protectionist. The sanctions support Russian farmers and food processors by restricting imports. However, imports essential for the development of Russia’s domestic food production, such as fish stocks or breeding animals, have not been included in the list of embargoed items.

Table 2. Value and Shares of the 2013 EU Agricultural and Food Exports to Russia

Covered by the Embargo Implemented in August 2014

Country The embargoed value of agricultural food products in million euro

Share of embargoed products by country in the value of all EU embargoed good products, in percent

Total agricultural and food exports to Russia in million Euros

Share in embargoed products in country’s exports to Russia, in percent

UE-28 5064 100 11864 42.68

Lithuania 922 18.21 1374 67.10

Poland 840 16.59 1267 66.30

Germany 598 11.81 1649 36.26

Holland 503 9.93 1551 32.43

Denmark 341 6.73 627 54.39

Spain 326 6.44 572 56.99

Belgium 281 5.55 558 50.36

Finland 273 5.39 464 58.84

France 229 4.52 756 30.29

Italy 163 3.22 705 23.12

Greece 114 2.25 158 72.15

Austria 103 2.03 247 41.70

Hungary 77 1.52 266 28.95

Ireland 70 1.38 216 32.41

Latvia 67 1.32 628 10.67

Estonia 60 1.18 228 26.32

Great Britain 20 0.39 148 13.51

Portugal 13 0.26 48 27.08

Sweden 13 0.26 107 12.15

Cyprus 12 0.24 12 100.00

Czech Republic

11 0.22 96 11.46

Slovenia 10 0.20 34 29.41

Bulgaria 8 0.16 56 14.29

Croatia 7 0.14 16 43.75

Slovakia 6 0.12 32 18.75

Luxembourg 5 0.10 8 62.50

Romania 1 0.02 42 2.38 Source: Based on Eurostat data.


EFFECTS OF THE EMBARGO ON FOOD EXPORTS FROM EU TO RUSSIA

Until August 2014, Russia was one of the leading food importers in the world. Trade with EU states accounted for 43% of total Russian imports. For example, according to FAO, Russia’s share in the global dairy product trade amounted to five percent in 2014, second only to China. Russia was the second most important destination of EU food and agricultural exports, behind the United States. Food and agricultural commodities accounted for ten percent of the total EU exports to Russia, but represented between one and three percent of the total EU agricultural production. In the year preceding the embargo, total agricultural exports from EU to Russia amounted to 11 billion euro and the Russian sanctions affected exports valued at 5.1 billion euro, or 43% of total EU exports to that country (Table 2).

Among all the countries and areas affected by Russia’s embargo on food imports, 73% of imports covered by the trade embargo originated from the EU. Within the EU, the effects of sanctions were unevenly distributed (Figure 1). The importance of the Russian market varied with regard to various products and regions. In terms of value, nearly one half of EU agricultural and food exports covered by the embargo originated from Lithuania, Poland, and Germany. In the case of Lithuania and Poland, products banned from import represented almost 70% of the total food exports to Russia, and their value for either country exceeded one billion euro. In the case of Finland, Spain, Denmark, and Belgium the banned goods accounted for more than 50% of their total exports to Russia. The country most affected, however, was Cyprus, in whose case all exported agricultural products have been banned, although they were valued at only 12 million euro per year. Such exports accounted for more than 5% of agricultural production of Cyprus.

For individual EU countries affected by the embargo, the importance of individual markets differed. From Poland’s perspective, exports to Belarus are of great relevance. In the years 2014/15, Belarus was the largest recipient of Polish apples with shipments in excess of 220 thousand tons. Given the absorption capacity of the Belarusian market, it is probable that at least part of that volume was re-exported to the Russian Federation. It is not unlikely that the re-export was taking place with the tacit approval of the Russian authorities, although Russia issued another embargo on re-export of produce from Poland in October 2014. The restrictions on import of selected products would have negatively affected the Russian economy.

SHORT-TERM EFFECTS OF THE EMBARGO IN RUSSIA

Growth and exchange rate. Factors limiting food imports to Russia included not only trade restrictions, but also the slowing growth of GDP since the middle of 2013. It is likely that exports from the EU would have declined somewhat even without the embargo. The cause was the suspension of trade in hogs and pork products from the EU, introduced in late January 2014 following an occurrence of the African swine fever (ASF). IMF data show that the GDP growth rate in constant prices was 0.6% in 2014. The IMF forecast for 2015 is a 3.8% contraction of Russia’s GDP.


An additional element affecting EU trade with Russia was the weakening of the ruble against euro that took place mostly after the embargo has been declared and coincided with a decline in oil prices. Between July 2014 and June 2015, the average exchange rate of ruble to euro was 0.017 euro for one ruble, a 23% slide year to year according to the ECB. In the same period, the ruble lost 33% against the US dollar. A slowdown in food exports to Russia was observed since the beginning of 2014 and the embargo certainly contributed.

Source: Based on Eurostat data.

Figure 1. Share in food products exports to Russian Federation from EU countries in 2013, in percent.

Given the low level of self-sufficiency and limited import possibilities from countries not covered by the trade embargo, a total ban on supplies from the EU led to a substantial price increase on the domestic Russian food market. This increase in turn accelerated inflation. According to the European Central Bank (ECB), inflation exceeded 15% in the first quarter of 2015, the highest increase in the last 10 years, deepening the economic slowdown.

An increase in retail cheese prices was noted within the first month of the embargo taking effect (tygodnik Poradnik rolniczy, 2014). Experts anticipated the share of food expenditures in the average family budget to increase from 35% to 40-42% of total by the end of 2014. Rising butter prices were expected to shift the demand to margarine. As a result, milk processors would decrease the volume procured, causing domestic milk prices to drop and threatening efforts to expand the cattle population. Indeed, the goals of the adopted food security doctrine might have been thwarted or met at a much lower volume level (although the percentage share might have been achieved). Increased spending on food would limit expenditure on non-food products.

BEFORE AND AFTER: EXPORTS OF MEAT AND DAIRY TO RUSSIA

The embargo imposed on EU and the loss of Russian market sharpened the competition among EU food suppliers, both within and outside EU markets. Such acute competition was observed, among others, on the milk and dairy product markets.

Dairy products. Poland is a major milk producer in the EU, next to Germany, France, Italy and the United Kingdom (FAO, 2015). Although the EU milk production growth was


restricted by the milk quota, milk production in Poland has been increasing since the country joined the EU in 2004. Moreover, after years of managing dairy production through the allocation of milk quotas among EU countries, the scheme was terminated on April 1, 2015. In anticipation of abolishing the quota system, dairy producers in several EU countries, most notably in Poland, Germany, and the Netherlands, have been expanding production already in 2013 (Sobczynski et al., 2015). An increase in the EU production in 2014 coincided with the embargo-induced export restrictions and contributed to the decline of prices. The combined effect of increased supply and decreased exports led to a decrease in the value of exports than was relatively larger than the decrease in volume.

Dairy product exports declined by 8%, the third largest decline among all food categories. The primary casualty of Russia’s embargo was hard cheese exports. One third of the EU cheese exports, a volume of 257,000 tons, were destined for Russia (TVP Parlament, 2015). Among the EU countries, the share of Finland, Lithuania, Latvia and Estonia cheese exports to Russia accounted for 90% of their total cheese exports. Cheese exports from Poland, the Netherlands and Germany to Russia represented 43%, 42% and 38% of their total cheese exports, respectively.

The geographical proximity created an advantage, while Russia absorbed dairy product surpluses. The products were wholesome and safe, but differed in their sensory qualities from those preferred by consumers in Western European countries. Lithuanian cheeses have been well known to Russian consumers since Soviet days and the proximity of large urban markets in northwestern Russia encouraged their imports. Finnish producers have an even easier access to Sankt Petersburg than do Lithuanian cheese makers. Dutch dairy farms have the highest milk yields in the world (Dairy Report, 2011), a long cheese-making tradition and quality products, but their domestic demand is limited. Cheese exports have been important for decades and Russia absorbed a large portion of the Dutch cheese output.

Besides the embargo, the decline of dairy product exports was affected by substantially constrained shipments of fat-free and whole powdered milk to China. Increasing production in China and the high level of inventories worldwide, including Chinese holdings, led to a decreasing tendency of milk prices to spiral down both globally and on the EU markets since the early 2014. The embargo only accelerated the process. According to the European Commission, the average price of defatted powdered milk in the 28 EU member states fell by 16% between January and July 2014. Between August and December 2014, however, the decrease was larger and amounted to 39%. In the case of butter, the decrease in prices between January and July 2014 was 17%. An even larger decline of 23% took place between August and December 2014.

Dairy product compensation program. The European Commission faced some desperate EU dairy farmers who already experienced declining milk prices in 2014 resulting from overstocking on powdered milk inventory. Already in early September 2014, the Commission proposed to compensate dairy product producers with amounts ranging from 10 million to 25 million euro depending on the manufacturers’ interests (TVP Parlament, 2014). The program was considerably more complex because the embargo was introduced when the milk quota system and the associated infrastructure were being phased out (April 1, 2015 was the termination date for the milk quota system in the EU). Only those exporting cheese to Russia could expect payments, which in turn could have an impact on milk buying-in prices. Specifically, the compensation was to cover private costs of storing selected types of cheese, butter and defatted powdered milk for the period from three to seven months. The rate at


which the compensation was calculated was 15.57 euro plus 0.40 euro per ton/day (TVP Parlament, 2015). For example, storing 10,000 tons for a month could lead to receiving 275,700 euro in compensation. Additionally, in 2014, the European Commission proposed to extend the annual intervention buying-in until the end of the calendar year rather than the end of September.

Unofficial information suggested that in the first week of accepting compensation applications, 87% of them originated from Italy, and payments for storing 74,000 tons of cheese were claimed (Polskie Radio, 2014). Italy shipped a relatively small volume of cheese to Russia. For example applications from Poland in the same period concerned 60 tons of cheese, a small portion of Polish cheese exports to Russia. Indeed, 75% of cheese exports to Russia originated from Finland, Latvia, Lithuania, and Estonia.

An example of bypassing the embargo is an investment made by a Finnish dairy processing company in Russia. Disregarding the embargo, the company opened a milk processing plant near Sankt Petersburg that produces milk, cream, yogurt and kefir (Suwalki24.pl, 2015). Interestingly, the milk for the plant is imported from countries excluded from the embargo, for example Belarus and New Zealand.

Milk and dairy product prices in Poland. The process of global accumulation of dairy product inventories was already underway when Russia imposed the trade embargo, obscuring the examination of the embargo’s effects on milk prices in Poland. The typical seasonal pattern is that dairy product prices tend to increase in the second half of the year. In 2014, milk prices in Germany and France, which are among the leading milk producing countries in EU, dropped rapidly. The decline was noticeable, but less rapid, in Poland-bordering Slovakia and the Czech Republic. In Poland the drop was relatively small because local prices were already among the lowest in the EU. Farmer organizations appealed to the Polish government to stabilize the milk market (Suwalki24.pl, 2015) as prices failed to recover by the fall of 2015. In contrast with the slow response of the Polish government, the Lithuanian government aggressively promoted its dairy products and increased exports, including exports to new markets in Asia.

Cheese price changes are illustrated using changes of two types of cheeses, Gouda-like and Ementaler-like. After a substantial increase throughout 2013, the trend reversed and showed a steady decline in 2014, with a noticeable drop in September, i.e., the month following the imposition of trade embargo. More importantly, the prices did not recover in 2015.

In Poland, butter prices showed a decline since January 2014, reaching 13.60 PLN per kg in August (from nearly 17.00 PLN per kg in January) and dropping to 12.60 PLN per kg in September. The decline continued and, through much of 2015, the price was below 12 PLN per kg, not returning to the level observed in late 2014 until October 2015. However, it was the prices of powdered milk that fared worst. In decline through 2014 with a pronounced drop in September, they slumped even more in January 2015, and remained low.

Milk and dairy prices had an immediate impact on the prices of cows and heifers. First there was the rapid slide of cow prices since September 2014, although they ascended in 2013. In 2015, cow prices were below the level reported in 2014 using month-to-month comparisons. Heifer prices were relatively steady in 2014. Their decline started in March-April 2015 and prevailed until November 2015. The combined effect of price changes for outputs and inputs indicates that the sector is still undergoing major restructuring processes (Wysokinski et al., 2015), unless the prices recover soon. Unfortunately, livestock farmers


are, next to the unemployed, the most likely victims of suicide among all social groups in Poland (Suwalki 24. pl, 2015).

Pork and pork products. Polish hog farmers suffered losses since January 2014 after a dead wild boar, found in woods, tested positively for the African swine fever (ASF). To protect the Polish hog sector, several communes in the county of Łosice were punt under quarantine. Hog herds in the area were preventively destroyed and hog production temporarily suspended. Russia promptly banned all pork imports from the EU. Farmers from the quarantined area, situated in one of the least developed areas of EU, were highly affected. The county is not far from the Poland-Belarus border, i.e., the eastern fringe of the EU.

After the 2014 August embargo, farmers in Łosice county blockaded major routes. The protests led to the Minister of Agriculture and Rural Development visiting the village of Platerów on 26 October 2014 (Podlasie 24, 2014). Hog farmers demanded government assistance, including aggressive hunting of wild boars, viewed as the source of the ASF. Indeed, the ban on hog production combined with the August trade embargo caused substantial losses in the local economy and affected the standard of living of farming households.

Despite the suspension of trade caused by the ASF, the EU livestock and meat exports increased by 12% and 4%, respectively, in 2014. The increase was a result of expanding demand in countries other than former Soviet republics, China among others. The increased demand was, at least in part, a response to pork prices, which were declining after the ban on EU pork exports to Russia. Already in March 2014, the European Commission implemented a support program to subsidize the keeping of private pork inventories, but prices remained low (Malkowski and Zawadzka, 2015). For the rest of 2014 and in 2015, the pork market continued to feel the pressure of the Russian embargo. For example, in the second quarter of 2015, pork prices were 15% lower than in the corresponding period of 2014. The prices remained below the 2014 level until November 2015, and then rapidly declined on a weekly basis. Although feed grain prices declined less than hog prices, farm profits were substantially reduced. Rather as in the case of dairy cow herds, hogs numbers could reach the lowest levels recorded in the past 40 years (Malkowski and Zawadzka, 2015). Because the hog and pork price decline affects all EU countries, while the EU pork consumption seems to lag, increasing exports are urgently needed, or else a major supply adjustment will take place. In Poland, the restructuring of the hog sector would affect primarily family farms, many of which are located in the eastern part of the country that is among the least developed EU regions.

Poultry. In the case of poultry, there were changes in trade flow directions that offset the effects of Russian embargo. As a result of restricting poultry imports to EU from Brazil, the latter has emerged as a supplier for the Russian market. Poultry production in Poland does not seem to be affected by the Russian trade embargo. The production expanded in 2014 and the pace of expansion accelerated in 2015. Based on data from poultry farms employing at least 50 permanent workers, the poultry production level posted a record high in October 2015 (Dybowski, 2015). Low feed grain prices had minimal effect on profitability; a slight decrease in case of broilers, but an increase in case of turkeys.

In the first nine months of 2015, poultry exports increased by 14.6% as compared to 2014 (Dybowski, 2015). Exports were probably encouraged by diverging prices, which decreased slightly in Poland while increasing by 1.2 % in EU. The domestic consumption of poultry meat increased, but that of processed poultry products, such as sausage, declined as


consumers responded to price changes. The increasing demand for poultry meat in the wake of unprecedented low pork prices may signal stronger underlying changes in consumer preferences. Poultry producers seem to gain from the embargo, while the woes of hog farmers continue.

CONCLUSION

The effects of Russia’s embargo may have been limited overall. The value of EU agricultural and food exports increased by 4% in the years 2014-15. The sanctions, valued at 5.1 billion euro, were cushioned by increased exports to the United States and China in the amount of almost 4 billion euro. Additional shipments to the United States and China offset 75% of the loss caused by closing the Russian Federation market. Despite the decline in exports, Russia still ranks third among the EU agricultural and food trade destinations. Moreover, the EU also increased its exports of grain and grain products by 13% (the highest rate of growth), because this category was not included in the embargo imposed by Russia. This trend can be sustained if the level of imports to Russia will depend on the domestic demand for meat and dairy products. Indeed, given the reaction of the dairy sector in Poland in the second half of 2014 and 2015, decreased demand for dairy products induced a decrease in prices of major inputs, such as heifers and cows. In Russia, the decreased demand for dairy products poses a real threat to meeting the goals of domestic self-sufficiency level in dairy products despite breeding livestock being excluded from the embargo. Gains in domestic cow herds might thus be reversed.

Because milk and dairy products are important sources of nutrients, decreased dairy product consumption may also affect the health of the population. Combined with lower consumption of fruits and vegetables, also included in the trade embargo, the diet becomes not only more expensive overall, but less healthy. Gains made in this area in recent years are likely to be reversed, with losses to consumer welfare and economic growth that are difficult to estimate.

In spite of decisive restrictions on trade with Russia in 2014, Poland managed to increase food exports by 5%. The rate of growth was, however, smaller than in previous years. The primary cause of food exports growth was increased exports to, among others, France, Belgium, and Italy; the increase amounted to 23%, 19%, and 8% respectively. Although the Russian embargo did not reverse the growth of the total food exports, it had a substantial negative effect on some sectors in certain countries, including Poland. There are some noticeable regional differences within the countries most affected by the meat and dairy export ban. Because the least developed (and lowest income) regions are primarily agricultural, the trade embargo had a particularly severe impact on incomes of rural communities.

The real winner may be Polish consumers. The food price CPI has increased by 0.2% so far in 2015. With a continued increase of the country’s GDP and incomes, the share of food expenditure may decrease further. The share decreased from about 30% in 1995 to just under 18% in 2010 (Hanne and Roux, 2012). In the longer term, consumers will have access to enhanced quality products as farmers and processors are likely to respond to price decline by shifting focus to improvements in that area.


APPENDIX

Imports of the following foodstuffs are prohibited per the customs classification of the Russian Federation (Ishchukova and Smutka, 2015):

1) Fresh and frozen meat (beef, pork, poultry meat and its by-products). Mutton and

goat meats were not included; 2) Meat, salted, dried, smoked or pickled in brine, including mutton, lamb and goat

meats; 3) Fish and fish “by-products,” shellfish, mollusks and other water invertebrates (live,

fresh, refrigerated, frozen, salt-cured, dried, smoked or pickled in brine, fish meal), including boiled or steamed shellfish in shells, as well as heat treated fish, shellfish, mollusks and other water invertebrates which were subsequently smoked;

4) Milk and dairy product; 5) Vegetables, edible roots and tubers (fresh, chilled, frozen, canned, dried); 6) Fruits and nuts (fresh, chilled, frozen, canned, dried); 7) Sausage and like products made of meat, offal, or blood, including those made from

mutton, lamb or goat meat; 8) Food products and supplements containing certain amounts of malt extract, milk or

vegetable fat as specified in the Customs Classifications, including certain dietary supplements and cooking ingredients;

9) Food products, not containing milk fats, sucrose, isoglucose, glucose and starch, or containing less than 1.5% milk fat, 5% sucrose or isoglucose, 5% glucose or starch.

REFERENCES

Dairy Report 2011. 2011. International Far Comparison Network. ISSN 1610-434X. OFCN AG, Kiel, Germany.

Dybowski, G. 2015. Rynek miesa drobiowego i jaj. Rynek Rolny, IERGiZ, December, Warszawa, pp 55-60.

European Central Bank. 2015. Available online athttps://www.ecb.europa.eu/stats/exchange/ eurofxref/html/index.en.html. Accessed September 11, 2015.

EUROSTAT Data, http://ec.europa.eu/eurostat/data/database. Accessed August 18, 2015. Hanne, H., N. Roux. 2012. Changes in food expenditure and how prices have impacted on

French and European households’ consumption since 1959. DGCCRF – ECO, No 4, May, 16 p.

IERiGŻ.2005. Analizy Rynkowe Handel zagraniczny produktami rolno-spożywczymi, stan i perspektywy. December, No 22, 56 p.

IERiGŻ.2008. Analizy Rynkowe Handel zagraniczny produktami rolno-spożywczymi, stan i perspektywy. December, No 28, 64 p.

IERiGŻ.2009. Analizy Rynkowe Handel zagraniczny produktami rolno-spożywczymi, stan i perspektywy. April, No 29, 64 p.

IERiGŻ.2015. Analizy Rynkowe Handel zagraniczny produktami rolno-spożywczymi, stan i perspektywy. September, Nr 42, 80 p.


Ishchukova, N.,L. Smutka. 2015. The impact of Russian import embargo 2014 on the structure of international trade in agricultural products. Typescript.

International Monetary Fund 2015, http://www.imf.org/external/pubs/ft/weo/2015/02/weo data/weoselgr.aspx. Accessed October 28, 2015.

Lewandowska, I. 2011. Polityka rolna – przyklad Rosji. Available online at https://www. minrol.gov.pl/content/.../24584/.../15_Lewandowska_K.pdf. Accessed October 15, 2015.

Malkowski, J., D. Zawadzka. 2015. Rynek wieprzowiny. Rynek Rolny, IERGiZ, December, Warszawa, pp 51-54.

Ministerswo Rolnictwa i Rozwoju Wsi RP, 2015. Available online athttps://bip.minrol. gov.pl/Informacje-Branzowe/Handel-zagraniczny-artykulami-rolnospozywczymi/ Polski-Handel-Zagraniczny-Artykulami-Rolno-Spozywczymi-w-2014-r: Accessed November 12, 2015.

Mroczek R., 2009. Ocena wpływu Wspólnej Polityki Rolnej na podstawowe rynki rolne w Polsce, In: Wpływ instrumentów polityki handlowej UE na handel zagraniczny produktami rolno-spożywczymi, Chapter 7, R. Mroczek, editor. Publikacja Programu Wieloletniego 2005-2009 – Ekonomiczne i społeczne uwarunkowania rozwoju polskiej gospodarki żywnościowej po wstąpieniu Polski do UE), N. 155, IERiGŻ, Warszawa

Pawlak K., and W. Poczta. 2011. Międzynarodowy handel rolny. PWE, Warszawa. Podlasie24. 2014. Spotkanie z Ministrem Rolnictwa w Platerowie. Available online at

http://www.podlasie24.pl/losice/region/spotkanie-z-ministrem-rolnictwa-w-platerowie-14a67.html. Accessed November 30, 2015.

Polskie Radio. 2014. Unia wstrzymuje rekompensaty producentow sera. Wszystko przez Wlochow. Available online at http://www.polskieradio.pl/42/273/Art kul/1241186,Unia-wstrzymuje-rekompensaty-dla... Accessed November 30, 2015.

Rocznik Statystyczny Handlu Zagranicznego. 2007. 2008. GUS. Available online athttp://stat.gov.pl/obszary-tematyczne/roczniki-statystyczne/roczniki-statystyczne/roc znik-statystyczny-handlu-zagranicznego-2015,9,9.html#archive. Accessed September 17, 2015.

Rocznik Statystyczny Handlu Zagranicznego. 2015, GUS. Available online athttp://stat.gov.pl/obszary-tematyczne/roczniki-statystyczne/roczniki-statystyczne/roc znik-statystyczny-handlu-zagranicznego-2015,9,9.html. Accessed September 17, 2015.

Rocznik Statystyczny Przemysłu. 2014. 2015, GUS. Available online at http://stat.gov .pl/obszary-tematyczne/roczniki-statystyczne/roczniki-statystyczne/roc znik-statystyczny-przemyslu-2014,5,8.html. Accessed September 19, 2015.

Rocznik Statystyczny RP 2014, GUS. Available online athttp://stat.gov.pl/obszary-tematyczne/roczniki-statystyczne/roczniki-statystyczne/rocznik-statystyczny-rzeczypospolitej-polskiej-2014,2,9.html. Accessed September 2, 2015.

Rosyjska Gazeta. 2014, Available online athttp://www.rg.ru/2014/08/08/postanovlenie-dok.html].Accessed January 12, 2015.

Sobczyński, T., A.M. Klepacka, C. Revoredo-Giha, W.J. Florkowski. 2015. Dairy farm cost efficiency in leading milk-producing regions in Poland. J. Dairy Sci. 98:8294-8307, doi:10.3168/jds.2014-9030.

Soczewska I., Ginter A., 2013. Handel międzynarodowy artykułami rolno-spożywczymi w Polsce w latach 2007-2011. Journal of Agribusiness and Rural Development, 2(28):225-234.


Suwalki24.pl. 2015. Rolnicy samobojcy. Available online at http://www.suwalki24.pl/article /3,rolnicy-samobojcy. Accessed December 3, 2015.

Szajner P.2012. Polski handel zagraniczny produktami rolno-spożywczymi. Biuletyn Informacyjny ARR 2012, pp 1-15. Available online at http://www.arr.gov.pl/data/400/ biuletyn_informacyjny_arr_4_2012.pdf.Accessed July 15, 2015).

TVP Parlament. 2014. Walka ze skutkami rosyjskiego embarga: KE rozwaza rekompensaty dla producentow. Available online http://www.tvpparlament.pl/aktualnosci/walka-ze-skutkami-rosyjskiego-embarga-ke-roz. Accessed November 17, 2015.

tygodnik Poradnik rolniczy. 2014. Czy polski eksport rolniy musza ratowac bialoruskie ostrygi? Available online at http://www.tygodnik-rolniczy.pl/articles/aktualnosci_/czy-polski-eksport-rolny-musza-rat. Accessed November 14, 2015.

UN Comtrade 2015. Available online at http://comtrade.un.org/data/. Accessed October 10, 2015.

Wysokinski, M., J. Baran., W. J. Florkowski. 2015. Concentration of milk production in Poland. In: Economic Science for Rural Development, J. Kusis, ed., Conference proceedings of the International Scientific conference for Rural Development, 37:93-104.

journal of international agricultural trade and … 10... · 2017-04-01 · viju ipe the world bank...

Documents