3-trade agreements in arab countries what are the … · trade agreements in arab countries: what...

Economics World, December 2015, Vol. 3, No. 11-12, 270-282 doi: 10.17265/2328-7144/2015.1112.003

Trade Agreements in Arab Countries: What Are the Effects on

Their Trade? (A Gravity Modeling Approach)

Mahmoud Reda Fath-Allah

League of Arab States, Cairo, Egypt

This paper investigates the effects of proliferating Regional Trade Agreement (RTA) in Arab countries on their

trade performance. The main question that the paper is trying to answer is: Why countries in Arab region are racing

toward signing RTAs with different trading partners. Are there any evidences that the trade performance of the

countries that have membership in many RTAs is better than those of countries that have less number of RTAs? By

focusing on the trade creation and trade diversion effects, a gravity model has been applied using panel data on

variables including Arab countries and their most important trading partners during the period from 1980 to 2010.

The results of the gravity model found that the Arab countries with overlapping membership in a number of trade

agreements have more significant positive effects in the form of having bigger volume of trade than those of the

other countries. The result of the paper supports the argument that countries engaged in many and overlapping trade

agreement would benefit by countervailing their negative effects in some separate agreements with positive overall

gains resulting from being “trade hub” that has an access to different markets with preferential terms.

Keywords: Regional Trade Agreements (RTAs), gravity models, preferential market access, overlapping RTAs,

Arab countries, Pan-Arab Free Trade Area (PAFTA)

Introduction

Arab countries are engaged in many preferential trade agreements with both regional and international

parties. The main reasons for this trend in most cases are to benefit from being a member in the WTO specially

in using article (XXIV) of Regional Trade Agreements (RTAs) and to have preferential access to the markets of

their main trade partners. Proliferation of RTAs in Arab countries reflexes multiple memberships in different

RTAs for many countries and overlapping their commitments in different agreements. This situation raises the

question of “what is the benefit that Arab countries would gain from being members in such overlapped trade

agreements”.

The study will examine the hypothesis test that the country which is member of overlapped RTAs has the

opportunity to trade more than the country that is a member of just one trade agreement.

The paper is organized in five sections: Section II presents an overview of the main trade agreements of

the Arab countries; section III discusses brief literature review; section IV is about model specification; section

V talks about estimation technique; and section VI presents concluding remarks.

Mahmoud Reda Fath-Allah, Ph.D., senior economist at Economic Sector of the League of Arab States, and adjunct lecturer at

Faculty of Economics and Political Science, Cairo University, Cairo, Egypt. Correspondence concerning this article should be addressed to Mahmoud Reda Fath-Allah, League of Arab States, Tahrir

Square, Cairo 11642, Egypt. E-mail: [email protected].

DAVID PUBLISHING

D

A GRAVITY MODELING APPROACH

271

Overview of the Main Trade Agreements in Arab Countries

This section will present an overview of the main trade agreements of the Arab countries. Arab countries

participate in different trade agreements among themselves and with other international parties as shown in

Table 1 and Figure 1. The first is the most important trade arrangement for Arab countries, Pan-Arab Free

Trade Area (PAFTA) which used to be known as the Greater Arab Free Trade Area (GAFTA). This

arrangement has 18 members of Arab countries and has a long history of try and error. The second is the

Euro-Mediterranean Partnership Agreements which seven Arab countries are engaged in. The third is the US

bilateral agreements which four Arab countries have signed.

Table 1

Arab Countries’ Membership in WTO and Selected RTAs Number of RTAs the country has membership

FTA with EFTA

FTA with Singapore

FTA with Turkey

Agadir Agreement

COMESAUS-FTAEU-MEDGCC PAFTAWTO Country

3 X X X X Bahrain

2 X X X Saudi Arabia

2 X X X Qatar

2 X X X UAE

2 X X X Kuwait

3 X X X X Oman

5 X X X X X X Tunisia

1 X X Yemen

3 X X X Lebanon

7 X X X X X X X X Jordan

6 X X X X X X X Egypt

6 X X X X X X X Morocco

4 X X X X Palestine

2 X X Algeria

2 X X Syria

2 X X Sudan

1 X Iraq

1 X X Djibouti

1 X Comoros

0 X Mauritania

2 X X Libya

0 Somalia

6 1 6 4 5 4 7 6 18 12 Arab country membership in each RTA Source: Retrieved from http://www.escwa.un.org/information/pubaction.asp?PubID=1702; http://www.ustr.gov/trade-agreements/ free-trade-agreements; http://www.arableagueonline/org/las/index.jsp; http://rtais.wto.org/UI/PublicMaintainRTAHome.aspx.

First: GAFTA or PAFTA

GAFTA is triggered by the changes in the global economic system. GAFTA is a contractual framework

that is based on legal tools and dispute settlement mechanisms. The legal framework is based on an agreement

of the year 1981 for trade facilitation among the Arab countries. In 1997, 14 Arab nations began talks

concerning the formation of a GAFTA. The end of the grace period for the transition to free movement of


272

goods within GAFTA was moved up from 2007 to 2005. The general secretariat of the League of Arab States

(LAS), mainly the Economic and Social Council, is responsible for the follow-up of the agreement’s

implementation. GAFTA has notified to the WTO as RTA under article XXIV and became known as PAFTA.

Currently, PAFTA has 18 member countries, of which three are less developing countries (LDCs) namely

Palestine, Sudan, and Yemen. The trade of PAFTA members represents 93% of total trade of all Arab countries.

Intra-trade among members’ countries are 6% in exports and 11% in imports. The importance of PAFTA to the

Arab countries represents 99% from the total Arab exports which is 904.5bUS$ in 2010 and represents 93% of

total Arab imports which represents 655.2bUS$. In terms of market size, PAFTA represents a huge market with

population more than 298.2 million which represents about 85% of total population of all Arab countries (Arab

Monetary Fund, 2012).

Figure 1. Overlapping trade agreements in Arab countries.

Second: Euro-Mediterranean Partnership Agreements

The great majority of Arab countries are involved in negotiations with the EU, through the

Euro-Mediterranean or EU-GCC context. Currently, seven Arab countries have signed the agreements and

implemented their commitments. The seven countries are Egypt, Tunisia, Morocco, Jordan, Palestine, Algeria,

and Lebanon. The EU approach is based on the universal model of trade liberalization, which links it to growth,

economic reform, and attraction of foreign direct investments (FDIs). The EU agreements lack agricultural

sector liberalization and free movement of persons. In addition, the EU has linked liberalization of trade flows


273

with “substantial” amount of aid cooperation funds; the EU partnership administers these funds through the

Middle East Development Assistance (MEDA) program. Before the association agreements were signed, most

Mediterranean partner countries accessed the EU market through preferential schemes administered. However,

accession problems are often supply-side and not barriers on the recipient side. Accordingly, these countries

will suffer macro-economic deterioration and their trade balance will be negatively impacted due to the free

trade agreement. This could lead to budgetary and employment vulnerability. EU partnership is a bilateral

agreement. That means, different commitments and aspect coverage, although, in general, a typical EU

partnership agreement has the following aspects (EU Commission, 2012):

market access in goods;

general services;

specific services chapters dealing with financial services and telecommunications;

intellectual property rights;

investment;

government procurement;

competition policy that impacts policy space and country economic structures;

labor and environmental standards and issues.

Third: US Bilateral Trade Agreements

The US is “picking off” individual countries, in pursuit of largely geopolitical and strategic objectives, and

setting its parallel trade agenda in the region. Four Arab countries signed bilateral FTAs with US, namely,

Jordan, Bahrain, Oman, and Morocco. FTAs are a tool for the US to enforce changes in domestic policies and

secure guarantees for US corporations, which are not achieved through the WTO.

Literature Review

The debate of the effects of trade agreements on the trade of a specific country manly lays on one of the

two arguments (Urata & Okabe, 2007). The first one is that the countries which have membership in multiple

trade agreements are likely to face positive effects on the trade. This argument, supported by the

hub-and-spokes (HAS) concept, which is prevalently used in the transportation literature and first introduced to

international trade as a “two-sided triangle” by Wonnacott (1975), is a useful framework for unraveling this

noodle bowl of FTAs. The HAS is unique to FTAs, because there is no restriction on the number of FTAs that a

region can sign. As a result, the region acts like a “hub”, linking up several free trade areas and trading on

preferential terms with every “spoke” partner. Chong and Hur (2007) have highlighted the importance of

hub-and-spokes as a trading system in a world of overlapping free trade agreements. Their study indicates that

small and open economies like Singapore prefer hub status to a free trade zone involving the same country

group. They are not likely to stop at one agreement, once they embark on the FTA path. Although FTAs

especially those with deep integration can be attractive, significant changes in industrial composition (due to

specialization in production) can lead to temporary spells of frictional unemployment.

The second argument is that the countries which have membership in multiple overlapping RTAs are

likely to have negative effects on the trade flow. This argument is supported by the idea of “transaction cost of

trade”, as it will be higher if the number of trade agreement for respective country gets higher. The explanation

is that the membership of a country in different RTAs where each of RTA has its own rules of origins,


274

certificate of origins, customs procedures, trade provisions, list of eligible goods, and standards requirements

will complicate trade system and needs more managerial effort. All of those will lead to increasing transaction

cost of trade in terms of time of release for shipment and number of documents required. So, this paper needs to

imperially investigate which one of those arguments works in the case of Arab countries.

To do so, this paper estimates whether or not an FTA has had a statistically significant effect on trade

flows. If the respective coefficient of FTA is statistically significant with a positive sign, that would lead to

concluding that the FTA has a positive effect on trade flows with a magnitude relation to the size of the

coefficients. This, however, is an inference about the FTA’s effect on total trade flows and not due to trade

creation, trade diversion, or both. To estimate these effects separately, another binary variable would need to be

included. With this extended specification, the binary variable for observations where both the importing and

exporting countries are members of the FTA would capture trade creation, while the second binary variable for

observations where one of the trading partners is not a party to the FTA would capture trade diversion.

The gravity model is an econometric method of estimating trade flows. This model has been used to

analyze the impact of not only FTAs, but also the effects of General Agreement on Tariffs and Trade-WTO

membership, currency unions, migration flows, foreign direct investment, and even disasters (Freund &

McLaren, 1999). The main benefit of the gravity model in evaluating an FTA is that it can control the effects of

as many other trade determinants besides the FTA as necessary and can therefore isolate the effects of the FTA

on trade. The basic gravity model of trade, which is analogous to Newton’s law of universal gravitation in

physics, relates the imports of country i from country j (Mij) positively to the gross domestic product (GDP) of

the importing country (Yi) and the GDP of the exporting country (Yj), but negatively to the geographical

distance between the importing and exporting countries (Dij) (Anderson, 1979):

ij i j ijM GYY D (1)

where G is a constant.

Expressed in logarithmic form and attaching a random error term (uij), the basic gravity equation becomes:

1 2 3ln ln ln ln

jij i ij ijM G b Y b Y b D u (2)

where b is coefficient. Given the hypothesized relationships contained in the gravity model, b1 and b

2 are

expected to be positive, while b3 is expected to be negative. In the gravity equation, geographical distance

between the importing and exporting countries is actually a proxy for trade costs, which impede bilateral trade.

Other variables that capture trade costs (e.g., adjacency, common language, colonial links, common currency,

or whether the importing or exporting countries are islands or landlocked) may be added to the equation

(Plummer, Cheong, & Hamanaka, 2010).

Empirical Model

Extended gravity model in double-log form as in the work of Lee, Innwon, and Kawanho (2005) will be

used:

0 1 2 3 4

5 6 1 2 3

GDP GDPln(Trade ) ln(GDP GDP ) ln( ) ln Dist ln(Area Area )

PDP PDP

Border Language RTA Insiders RTA Outsiders RTA Overlap

i j

ijt i j t t ij i j

i j

ij ij ijt ijt ijt ijt

(3)


275

where Tradeijt is real bilateral trade between i and j; GDPiGDPj are real GDP for countries i and j respectively;

PDPiPDPj are population of countries i and j; Distij is the distance between countries i and j; AreaiAreaj are land

mass of the country i and j respectively; Borderij is binary variable which is in unity if i and j share a land

border; Languageij is binary variable which is in unity if i and j have a common language; RTA Insidersijt is

dummy variable which is in unity if i and j belong to the same RTA and take zero otherwise; RTA Outsidersijt

is the dummy variable which is in unity if country i has existing trade agreement with a country other than j;

RTA Overlapijt is the dummy variable which is in unity if country i has existing trade agreement with at least

one country other than j in addition to the case which RTA Insidersijt and/or RTA Outsidersijt dummies are in

unity; εijt is the independent and identically distributed (IID) random variable with zero mean and constant

standard deviation.

The data came from Rose (2004) which covered 175 countries from 1948 to 1999 and the dataset has been

expanded by adding more observations using UN-COMTRADE (United Nations Statistics Department, 2012),

WITS (World Integrated Trade Solution, 2012), WTO (WTO, 2011; 2012), World Bank’s world development

indicators databases. Table 2 shows list of countries included in data sample.

Table 2

List of Countries Included in Sample

Arab countries EU-27 EFTA Other countries

1 Algeria 1 Austria 1 Iceland 1 China

2 Bahrain 2 Belgium 2 Liechtenstein 2 India

3 Comoros 3 Bulgaria 3 Norway 3 Japan

4 Djibouti 4 Cyprus 4 Switzerland 4 Australia

5 Egypt 5 Check Republic 5 USA

6 Iraq 6 Denmark 6 Canada

7 Jordan 7 Estonia 7 Brazil

8 Kuwait 8 Finland 8 Mexico

9 Lebanon 9 France 9 Turkey

10 Libya 10 Germany 10 Singapore

11 Mauritania 11 Greece 11 Iran

12 Morocco 12 Hungary 12 South Korea

13 Oman 13 Ireland 13 Russia

14 Qatar 14 Italy

15 Saudi Arabia 15 Latvia

16 Somalia 16 Lithuania

17 Sudan 17 Luxembourg

18 Syria 18 Malta

19 Tunisia 19 Netherlands

20 UAE 20 Poland

21 Yemen 21 Portugal

22 Romania

23 Slovakia

24 Slovenia

25 Span

26 Sweden

27 UK


276

Three groups of explanatory variables have been used as follows:

(1) Variables affect demand and supply of importing and exporting partners: Those variables include

economic mass of paired countries, such as GDP and population. The expected signs of respective variables are

positive (β1 and β2 > 0) (Shepherd, 2012);

(2) Variables represent natural factors affecting trade between respective partners: Those variables include

distance, land area, border, and language. The expected signs of respective variables are: β3 < 0 and β4, β5, and

β6 > 0;

(3) Variables capture trade policy effects: Those variables include RTA Insiders, RTA Outsiders, and RTA

Overlap. The dummy RTA Insiders measures the degree of trade-creation effects of the RTA among members,

while the dummy RTA Outsiders captures the degree of trade-diverting effects between members and

nonmembers, compared to the “normal” bilateral trade flows. RTA Overlap captures just the additional trade

creation taking place between an overlapped country and a member country not overlapped together. Table 3

shows possible combinations of dummy variables and respective implications on trade policy.

Table 3

Possible Combinations of Dummy Variables and Respective Implication on Trade Policy RTAIn RTAOut RTAOverlp

Respective meaning of the dummy variable combination Trade agreement exists between i and j

Country i has existing trade agreement with a country different than j

Country i has at least one trade agreement with a country other than j

0 0 0 Arab country has no existing trade agreement with its respective trade partner or any other country. Though, no overlap exists in this case.

1 0 0 Arab country has existing trade agreement with only its respective trade partner. There are no agreements with any other countries. Though, no overlap exists in this case.

0 1 0 Arab country has only one existing trade agreement with a country otherthan its respective trade partner. Though, no overlap exists in this case.

0 1 1 Arab country has trade agreements with more than one country other than respective partner. Though, overlap is existing in this case.

1 1 1 Arab country has trade agreement with its respective trade partner and with other countries. Though, overlap is existing in this case.

Estimation Technique

The data set has a feature of panel structure consisting of 32,886 annual observations clustered by 1,134

country pair groups from 1980 to 2010. The number of observations varies per year as the data set is

Un-Balanced Panel Data. Summary statistics for the whole data used in the estimation are presented in Table 4

and diagnostic scatterplots appear in Figure 2.

The following steps in estimation process have been applied (Baltagi, 2005):

(1) estimate pooled data using OLS;

(2) estimate fixed effects model using least square dummy variable (LSDV);

(3) testing pooled model vs. fixed effects model;

(4) estimate random effects model;

(5) testing fixed effect vs. random effects models.

The followings are illustration of those steps.


277

Table 4

Summary Statistics

Variable Mean Std. Dev. Min Max Observations

ltrade overall 11.73751 2.781319 0.2307641 18.80282 N = 16,109

between 3.054619 1.705306 17.69675 n = 1,040

within 0.9082277 1.068073 17.27022 T-bar = 15.4894

lrgdp overall 49.20263 2.290296 40.99001 56.01982 N = 16,109

between 2.431357 41.08295 54.9786 n = 1,040

within 0.4314612 47.07972 51.13304 T-bar = 15.4894

lrgdppc overall 17.08778 1.400499 11.71127 20.80505 N = 16,109

between 1.430628 12.25521 20.16171 n = 1,040

within 0.3437182 14.75803 19.08219 T-bar = 15.4894

ldist overall 7.779295 0.742171 4.016798 9.226302 N = 16,109

between 0.687129 4.016798 9.226302 n = 1,040

within 0 7.779295 7.779295 T-bar = 15.4894

lareap overall 24.47382 3.238323 12.18546 31.33648 N = 16,109

between 3.284881 12.18546 31.33648 n = 1,040

within 0 24.47382 24.47382 T-bar = 15.4894

border overall 0.0277485 0.1642564 0 1 N = 16,109

between 0.1590971 0 1 n = 1,040

within 0 0.0277485 0.0277485 T-bar = 15.4894

comlan overall 0.1623316 0.3687662 0 1 N = 16,109

between 0.3664252 0 1 n = 1,040

within 0 0.1623316 0.1623316 T-bar = 15.4894

rtain overall 0.066857 0.249782 0 1 N = 16,109

between 0.1595534 0 1 n = 1,040

within 0.1609761 -0.883143 1.031143 T-bar = 15.4894

rtaout overall 0.4225588 0.4939818 0 1 N = 16,109

between 0.4200256 0 1 n = 1,040

within 0.2673273 -0.5404041 1.386845 T-bar = 15.4894

rtaove~p overall 0.0733131 0.2606578 0 1 N = 16,109

between 0.0743976 0 0.3103448 n = 1,040

within 0.2452565 -0.2370318 1.034852 T-bar = 15.4894

Source: Stata output.

Estimating Pooled Data Using OLS

Pooled model is a restricted form of the panel data assuming no difference across all units or time. That

means equation (1) is a single equation for all country pairs (i), in this case (Baltagi, 2005):

1 2

0 = 0 = ... = 0 = 0i

(4)

Table 5 shows estimation results of the four forms of the gravity model of equation (1).

Model P1 fits the data well, explaining a major part of the variation in bilateral trade flows. The signs of

the coefficients meet the economic expectations except those (lareap and comlang) of which both have negative

signs. Lee et al. (2005) encountered the same results and found that the variable of the land area could lead to

the same results of the distance variable (Mayer & Zignago, 2011) in a country with large land area. Model P2

fits data slightly better than P1 but with a negative signs for RTAin and RTAout which lead to the result that

trade agreements of the Arab countries have trade diversion effect. In model P3, the introduction of the variable


278

(RTA Overlap) has significant positive effect which means that Arab countries with overlapping trade

agreement encountered better trade flows than other normal Arab countries. Finally, model P4 slightly less fits

data with significant parameters except those of RTA Overlap.

Figure 2. Diagnostic scatterplots.

Table 5

OLS Estimates for Pooled Data

(P1)

Standard gravity model

(P2) (P1) + trade policy variables without overlap variable

(P3) (P2) with overlap variable

(P4) (P3) after omitted non-significant variables

ltrade Estimated parameter

t-statistics Estimated parameter



t-statistics

lrgdp 1.04994 117.76 1.071241 117.75 1.0671 116.06 0.955379 146.23

lrgdppc 0.22367 18.33 0.257506 20.53 0.264093 20.76

ldist -0.86244 -41.5 -0.86135 -40.21 -0.86399 -40.31 -0.97142 -46.28

lareap -0.16469 -25.89 -0.17941 -27.72 -0.17819 -27.49

border 1.091818 12.42 1.091883 12.45 1.096459 12.5

comlang -0.11477 -2.83 0.010194 0.24 0.01324 0.32

rtain -0.30262 -5.28 -0.37143 -6.04 0.200394 3.13

rtaout -0.30944 -10.28 -0.34344 -10.71 0.295374 9.68

rtaoverlp 0.178938 3.07 -0.10918 -1.77

_cons -33.0161 -100.9 -34.1601 -99.62 -34.0728 -99.05 -27.8429 -87.98

Adjusted R2 0.6417 0.6444 0.6446 0.588

RSS 44630.54 44,288.32 44,262.35 51,318.93

F test 4809.05 3649.76 3,246.98 4,599.32

Notes. P1 estimates the core gravity variables regardless any trade policy variables; P2 estimates the same variables in P1 plus trade policy variables except that of overlap; P3 estimates P2 model plus overlap variable; and P4 estimates P3 model after omitting variables with non-significant parameters.


279

Estimating Fixed Effects Model Using LSDV

The problem in estimating pooled data model is that this is a restricted form of specification and if this

assumption does not meet the real situation, the OLS estimates yield biased and inconsistent estimates of the

regression parameters. Because this is an omission variable bias, due to the fact that OLS deletes the individual

dummies when in fact they are relevant. To overcome this problem, starting by estimating fixed effects model

and then testing wither pooled or fixed effect are the best specification for the data. Using LSDV which has two

sided estimations, one is called within units’ estimation which assumes the focus on the units (country pair)

though, variables, such as area, language, and border do not change over time. The second side is called among

units, in this case, the focus is on the cross-sections and its variations across time (Retrieved from

http://www.wto.org/english/res_e/publications_e/practical_guide12_e.htm). Table 6 shows fixed effect

estimates of the gravity model.

Table 6

Fixed Effects Estimates Using Within and Between Units Variations

(F1)

FE—Within units (F2)

FE—Between units ltrade Estimated parameter t-statistics Estimated parameter t-statistics

lrgdp 0.307045 11.44 1.102518 31.32

lrgdppc -0.10957 -3.51 0.210389 3.99

ldist -0.98845 -11.36

lareap -0.18665 -7.71

border 1.133116 3.26

comlang 0.127101 0.8

rtain 0.022933 0.44 -0.99911 -2.77

rtaout -0.23432 -7.14 -0.72786 -4.83

rtaoverlp 0.202156 5.35 4.329017 5.42

_cons -1.41487 -1.42 -33.9836 -25.9

R2 0.0193 0.7259

F test 59.22 303.02

Model F1 has low fits data, even though the parameters are significant and lead to concluding that

overlapping trade agreements have small positive effects on a trade flow of a country. Model F2 fits data better

than model F1 and also F2 has significant parameters except for the variable common language which has

negative sign anticipated as the majority of trade for Arab countries existing with non-Arab speaking countries.

Testing Pooled Model vs. Fixed Effects Model

This is a simple chow test with the restricted residual sums of squares (RRSS) being that of OLS on the

pooled model and the unrestricted residual sums of squares (URSS) being that of the LSDV regression. If N is

large, one can perform the “within” transformation and use that residual sum of squares as the URSS. In this

case (Baltagi, 2005), it is as follows:

0 1, 1

RRSS URSS / ( 1)F ~ F

URSS / ( ) N N T K

NNT N K

(5)

Applying this test on model F1 leads to the result: F (1,039; 15,064) = 45.09. According to this test, the

specification of fixed effect model better fits the data.


280

Estimating Random Effects Model

Random effects models assume that the individual-specific effect is a random variable that is uncorrelated

with the explanatory variable. This setup allows to estimate coefficients for time-invariant variables, such as

language, distance, border, and land area which fixed effect models do not allow to do so.

Table 7 shows RE estimate of the gravity model (Baltagi, 2005). The results indicate low value of overall

coefficient of determination with higher respective value for between estimates. That would imply the

importance of variations across units (cross sections). The explanation of that results meets the reality of the

agreements of Arab countries which are not having the same effects across all Arab countries, because each of

them has different aspects and scope in terms of inclusive sectors and goods.

Table 7

Random Effects Estimates of the Gravity Model1

ltrade Estimated parameters z-test

lrgdp 0.636634 29.48

lrgdppc -0.12116 -4.61

ldist -0.93785 -10.86

lareap -0.01063 -0.53

border 0.896302 2.53

comlang -0.51451 -3.33

rtain -0.01447 -0.28

rtaout -0.2629 -8.26

rtaoverlp 0.038094 1

_cons -10.3674 -11.19

R2

within 0.0168

between 0.6241

overall 0.5192

sigma_u 1.557954

sigma_e 0.930076

rho 0.73725

Wald chi2 1,884.28

Testing Fixed Effect vs. Random Effects Models

Now, after having fixed effect and random effects results, both results are consistent estimators, while only

random effects is efficient estimator, because random effects model has more degrees of freedom than those of

fixed effect model. So, this is a decision point to select the true specification model between random effect and

fixed effect. If the true model specification is random effects, it means that the estimator is a constant and

efficient. If this is not the case, it has to take fixed effect estimator, as it is consistent estimator even it is not

efficient. In order to take a decision in this regard, Hausman specification test has been used which takes the

form as the following (Baltagi, 2005):

12

'ˆ ˆ ˆ ˆ ˆ ˆVar( ) Var( ) ~FB RB FB RB FB RB kx

(6)

1 Coefficients estimated in random effects models have normal distribution (Gaussian); that is why using z-test instead of t-test in this case.


281

where ˆFB is the estimated parameters of fixed effect model; ˆ

RB is the estimated parameters of RE model;

ˆVar( )FB is the variance matrix of fixed effect model; ˆVar( )RB is the variance matrix of RE model; 2kx is

the critical value of qui-square statistics with k degrees of freedoms; and k is the number of explanatory variables.

The resulting value of the test is 1,099.74 which will lead to rejecting the null hypothesis. Consequently,

fixed effects model is the right model to fit the data.

Conclusions

This paper applied gravity model approach on trade flows of Arab countries in order to have empirical

evidence on the effects of proliferating RTA in Arab region on their trade flows and to contribute to the debate

of the two mentioned arguments of whither membership in multiple trade agreements which would gain

positive or negative results. The results of this study indicate that the effects of the overlapping agreements are

positive on the trade flows of the respective countries. The respective parameter of trade agreement without

overlap resulted in negative effects as the agreements itself may have trade diversion effects. Countries engaged

in many overlapping trade agreement would benefit by countervailing their negative effects in some separate

agreements with positive overall gains resulting from being “trade hub” that has an access to different markets

with preferential terms.

References Anderson, J. E. (1979). A theoretical foundation for the gravity equation. American Economic Review, 69(1), 106-116. Arab Monetary Fund. (2012). The joint Arab economic report 2012. Retrieved from

http://www.amf.org.ae/content/joint-arab-economic-reportwww.amf.org.ae Baltagi, B. H. (2005). Econometric analysis of panel data (3rd ed.). New York: John Wiley & Sons. Chong, S., & Hur, J. (2007). Overlapping free trade agreements of Singapore-USA-Japan: A computational analysis. Retrieved

from http://www.fas.nus.edu.sg/ecs/pub/wp-scape/0711.pdf EU Commission. (2012). Euro-Mediterranean partnership (EUROMED). Retrieved from

http://ec.europa.eu/economy_finance/economic_governance/sgp/index_en.htm Freund, C. L., & McLaren, J. (1999). On the dynamics of trade diversion: Evidence from four trade blocs. Retrieved from

http://korea.ssrn.com/delivery.php?ID=380124069122110094090012068121072081027089022064027023064106068098085004104004001098067089097102064028088042031073089113000026080020094100031070066088089005101067030103091092126100074123124&EXT=pdf

League of Arab States. (2012). Greater Arab free trade area. Retrieved from http://www.arableagueonline/org/las/index.jsp Lee, J. W., Innwon, P., & Kawanho, S. (2005). Proliferating regional trade agreements: Why and whither? Retrieved from

http://ssrn.com/abstract=604101 Mayer, T., & Zignago, S. (2011). Notes on CEPII’s distances measures (GeoDist). Retrieved from

http://www.cepii.fr/anglaisgraph/workpap/summaries/wp_resume.asp?annee=2011&ref=25&NoDoc=3877 Plummer, M. G., Cheong, D., & Hamanaka, S. (2010). Methodology for impact assessment of free trade agreements.

Mandaluyong City: Asian Development Bank (ADB). Rose, A. K. (2004). Do we really know that the WTO increases trade? American Economic Review, 94, 98-114. Shepherd, B. (2012). The gravity model of international trade: A user guide. Retrieved from

http://www.unescap.org/tid/publication/tipub2645.pdf UN and WTO. (2012). A practical guide to trade policy analysis. Retrieved from

http://www.wto.org/english/res_e/publications_e/practical_guide12_e.htm UN-ESCWA. (2007). Survey of economic and social developments in the Arab region. Retrieved from

http://www.escwa.un.org/information/pubaction.asp?PubID=1702 United Nations Statistics Department. (2012). Commodity trade statistics database (COMTRADE). Retrieved from

http://unstats.un.org/unsd/cr/registry/regcst.asp?Cl=14


282

Urata, S., & Okabe, M. (2007). The impacts of free trade agreements on trade flows: An application of the gravity model approach (RIETI discussion paper series 07-E-052, The Research Institute of Economy, Trade and Industry). Retrieved from http://www.rieti.go.jp/en/

US-Trade Representative. (2012). Trade agreements. Retrieved from http://www.ustr.gov/trade-agreements/free-trade-agreements Wonnacott, R. J. (1975). Canada’s future in a world of trade blocks: A proposal. Canada Public Policy, 1(1), 118-130. World Integrated Trade Solution [WITS]. (2012). Metadata. Retrieved from

http://wits.worldbank.org/CountryProfile/Metadata/Country/All World Trade Organization [WTO]. (2011). World tariff profiles 2011. Retrieved from

http://www.wto.org/english/res_e/booksp_e/tariff_profiles11_e.pdf World Trade Organization [WTO]. (2012). Regional trade agreement information system. Retrieved from

http://rtais.wto.org/UI/PublicMaintainRTAHome.aspx

3-trade agreements in arab countries what are the … · trade agreements in arab countries: what...

Documents