economic and social analysis of projects: coast '...
TRANSCRIPT
Economic and Social Analysis of Projects:A Case Study of lvory Coast '
SWP253
World Bank Staff Working F,aper No. 253
May 1977
This paper is prepared for staff use. The views expressed are those of the rauthor and rnot necessarily those of the Warld Bank. j
y: Johannes F. Linn s D
i Regiorial Economics Division L_J \iPUB ent Economic Department j HG8.5 lent Policy Staff
3881 .5 EN o 1.W573W67 1Vno.253
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
This paper is prepared forstaff use. The views arethose of the author and notnecessarily those of the Bank.
THE WORLD BANK
Bank Staff Working Paper No. 253
ECONOMIC AND SOCIAL ANALYSIS OF PROJECTS:A CASE STUDY OF IVORY COAST
May 1977
The purpose of this paper is to apply to a concrete country case theapproach of economic anaLysis of projects proposed by Lyn Squire and Herman G.van der Tak in their boolc Economic Analysis of Projects (Baltimore: JohnsHopkins University Press, 1975). After a brief introduction to the principlesof economic and social analysis of projects the present study estimates shadowprice parameters for the Ivory Coast which permit the evaluation of projectsfrom an efficiency as weLl as from a "social" point of view, where the formertakes no account of the distributive or growth implication of project selection,while the latter does so eyplicitly by introducing distributive weights forprivate costs and benefits. The parameter values are based on the author'sbest interpretation of the available information base at the time of writing(Spring 1974), but they 3hould not be interpreted as necessarily reflectingcurrent Bank views.
The author is grateful to Mr. B. A. de Vries who initiated andencouraged this study, and to Messrs. L. Squire, A. Ray, and H. van der Tak,from whose extensive comments on the methodological content and estimationprocedures he has extensively benefitted. Access to data and informationconcerning the Ivory Coast, as well as useful discussions on the empiricalcontent of the study were provided by numerous economists in the Bank, includingMessrs. B. den Tuinder, R. Glaeser, L. Goreux, and G. Pursell.
Prepared by:Johannes F. LinnUrban and Regional Economics DivisionDevelopment Economics DepartmentDevelopment Policy Staff
IVORY COAST
CURRENCY EQUIVALENTS
Currency Unit: CFA Franc (CFAF)
A fixed parity existsbe:tween the CFA and the French francs:
FF 1 = CFAF 50
The CFA franc floats against the dollar. Between February 12, 1973 and theend of November 1973, the rate has fluctuated as follows:
US$ 1 = CFAF 205-230
Throughout this report the following rates have been used for the conversionof CFA francs into ÙS!dollars and vice versa:
1968 and earlier years: US$ 1 = CFAF 2471969: US$ 1 = CFAF 2561970: US$ i = CFAF 2781971: US$ 1 = CFAF 2721972: US$ 1 = CFAF 2561973: US$ 1 = CFAF 230
WEIGHT AND MEASURES
1 metric Ton (t) = 2,205 pounds1 Kilogram (kg) = 2.2 pounds.1 Kilometer (km) = 0.62 mile1 Meter (m) = 3.28 feet
TABLE OF CONTENTS
Page No.
I. INTRODCTION..1
II. ECONOMIC AND SOCIAL ANALYSIS OF PROJECTS .2
Adjustment for Transfer Payments. Application of the Shadow Exchange Rate.3Application of the Shadow Wage Rate .3Efficiency Pricing: The Squire/van der Tak Approach .4Social Pricing: An Experimental Approach ....................... 6Summary of Estimation Results for Ivory Coast ................... 10
III. EFFICIENCY PRICING PARAMETERS IN IVORY COAST .................... 15
Conversion Factors .............................................. 15The Marginal Productivity of Capital, q ......................... 22
IV. SOCIAL PRICING PARAMETERS IN IVORY COAST ......................... 31
Value Judgements: The CRI, n and p ............................. 31The Consumption Distribution Weight, d .......... ................ 34The Summary Distribution Measure ................................ 34The Value of Public Income, v .; ................................. 37The Accounting Rate of Interest, ARI ............................... 44
V. THE SHADOW WAGE RATE ............................................ 46
Ivorian Labor (Excluding Urban lJnskilled) ....................... 46Urban Unskilled Labor ........................................... 47Non-Ivorian Rural African Labor ................................. 50Non-African Expatriate Labor .................................... 54
VI. IMPLICATIONS FOR PROJECT SELECTION .............................. 55
The Grand Bereby Rubber Estate Project, 1973 .................... 55The Third Highway Project, 1972 ................................. 60The Cocoa Project, 1970 ......................................... 64
LIST.OF TABLES
Page No.
Table 1: Summary of Country Parameters ............................ i l
Table 2: Shadow Wage Rates .12
Table 3: Results of Project Appraisals ............................ 13
Table 4: Imports and Exports of Consumption Goods .17
Table 5: Computation of Marginal Propensities to Spend on
Importables .18
Table 6: Computation of Marginal Propensities to Consume
Exportables .19
Table 7: Data for Computation of Standard Conversion Factor 20
Table 8: Valuation of Inputs for Construction Sector .23
Table 9: Output/Capital Ratio, 1965-1972 .24
Table 10: Alternative Estimates of Output/Capital Ratio .24
Table 11: Employment/Capital Ratio, 1965-1972 .25
Table 12: Informal Sector Productivity, 1970 .26
Table 13: Formal Sector Wages and Salaries, 1965, 1970 ........... . 27
Table 14: National Labor Productivity, 1965, 1970 .27
Table 15: Alternative Values for the Marginal Productivity of
Capital .28
Table 16: Lending Interest Rates in the Ivory Coast, 1973 29
Table 17: Alternative Values for the Marginal Productivity of
Capital, q .30
Table 18: Growth in Real Per Capita Consumption, 1967-1972 ........ 33
Table 19: Marginal Consumption Distribution Weights, d, in
Ivory Coast .35
Table 20: Distribution of Household Expenditure in the South-
East Region, 1963-1964 .36
LIST OF FIGURES (Continued)Page No.
Table 21: Value of Public Income: Simple Case ..................... 38
Table 22: Private Sector Marginal Propensity to Consume 1965-1972 .... 38
Table 23: Value of Public Income: With Reinvestment .... ............ 39
Table 24: Value of Public Income (v) Given Different CriticalConsumption Levels ..................................... 42
Table 25: Critical Consumption Levels Implied by AlternativeValues of v ............................................ 42
Table 26: Value of Pubi-ic Income: Consolidated Results .... ........ 44
Table 27: The Accounting Rate of Interest for Different Valuesof v, s, and q ........................................ 44
Table 28: SWR for Urban Unskilled Labor ........................... 48
Table 29: Data for the Computation of SWR ......................... 49
Table 30: Data for the Computation of the SWR for Non-IvorianRural African Labor ................................... 53
Table 31: SWR of Rural Non-Ivorian African Labor,as Proport:Lon of Wage ................................. 53
Table 32: Conversion Factors for Capital and Operating Costs ...... 58
Table 33: Grand Bereby Rubber Estate Project: Internal Rates ofReturn ................................................ 59
Table 34: Third Highway Project: Rates of Return .... ............. 62
Table 35: Cocoa Project: Rates of Return ......................... 67
LIST OF FIGURES
Figure 1: The Consumpt:Lon Distribution Function in the Ivory Coast. 41
I. INTRODUCTION
1. The purpose of this paper is to demonstrate in a specific countrycontext the application of the methodology for economic and social analysisof projects developed for the World Bank by Lyn Squire and Herman G. van derTak. 1/ Their approach consists of two separable steps in project evaluation:efficiency analysis, which is designed to select projects compatible withan optimal allocation of economic resources, but not allowing for distributiveor growth implications of project selection; and social analysis, whichexplicitly allows for the distributive and growth objectives of policymakers, by attaching appropriate weights to private costs and benefits.
2. This study presents estimates of the parameters required forefficiency and social arLalysis and applies the results to three projectswhich were approved by the World Bank for Ivory Coast,in recent years. Asa desk study, this exercise suffers from shortcomings which might have beencorrected with more direct exposure to country information and access tooriginal material in the country itself. Nevertheless, the present studyshows that the proposed shadow pricing methodology can be implemented ina specific country context, and that it offers a reasonable and systematicframework for assessing the trade-off between growth and income distributionin project selection. The information need is increased by the proposedmethod, but it does not appear to go far beyond what a rigorous applicationof conventional cost-berLefit analysis would require of the country andproject economists.
3. After a brief introduction to the principles of efficiency andsocial analysis of projects in Part II, Parts III and IV derive all the mainshadow prices (except the shadow wage rate) required for analysis, Part IIIpresenting those estimat:es which are necessary for efficiency pricing (i.e.the conversion factors and the marginal productivity of capital) and Part IVthe additional estimates required for social pricing (i.e., the consumptionrate of interest, the distribution weights, the value of public income,and the accounting rate of interest). In Part V we present estimates of fourshadow wage rates for Ivory Coast. In view of the overlap in analyticalarguments needed to derive the efficiency and social pricing components ofthe shadow wage rates, it proved convenient to discuss them jointly. Part VIconcludes the paper by applying the national parameters and the shadow wagerates to three projects in order to illustrate the implications of theproposed methodology for project selection.
1/ Lyn Squire and Herman G. van der Tak, Economic Analysis of Projects(Baltimore: Johns Ilopkins University Press, 1975). Extensive referenceto this publication will be made throughout the present study, whichshould help those readers not thoroughly familiar with the methodologyhere applied.
- 2 -
II. ECONOMIC AND SOCIAL ANALYSIS OF PROJECTS 1/
4. Economic analysis of projects has long been an integra:L part
of project appraisal procedures in the World Bank, although the techniquesactually employed have gradually become more refined. This section briefly
spells out the past and current piractices employed in this area by Bankeconomists, and gives an introduction to the principles of social project
analysis proposed by Squire and van der Tak.
Adjustment for Transfer Payments
5. At a minimum level of sophistication the project economist nets
out all tariffs, duties, and taxes from the cost and benefit streams of
his project, thus eliminating all "transfer payments" which do not reflect
economic opportunity costs or benefits. He then applies the official ex-
change rate to the c.i.f. price of imported inputs and to the f.o.b. priceof exported outputs to convert them into domestic currency terms, Net
benefits in year t are then derived according to the following equation:
NBt OE t t tNBt = (OER) z Ei - (OER) E M. - Dk
i j kk 1
where NBt net benefits in year t
OER official exchange rate
Et incremental export of commodity i produced by thei project in year t at constant f.o.b. prices (in dollars)
Mt incremental import of commodity j resulting from the
J project in year t at constant c.i.f. prices (in dollars)
Dk costs of locally produced inputs at local market prices. 2/
From the net benefit stream NBt for all years t the internal rate of return(R/R) is then calculated. If the R/R is greater than the cut-off rate,
1/ This section draws on a paper presented by the author at the International
Conference on Financing and Appraisal of Investment Projects, organized
by the Middle East Technical University in Istanbul, Turkey, June 1-4, 1976.
2/ For simplicity it has been assumed that all project outputs are exported.
A further term could be added to the equation showing domesticaJly consumed
product output, which would be treated in terms identical to Dt.k'
-3-
1/generally equated with the opportunity cost of capital in domestic prices,then the project is deemed acceptable; if it is below the cut-off rate,the project is rejected. 2/
Application of the Shadow Exchange Rate
6. Generally Bank economists have gone beyond netting out transferpayments by applying a shadow exchange rate (SER) to the foreign exchangecomponents of the project cost and benefit streams: 3/
NB = (SER) E - (SER) M - D (2)
In principle the purpose of this adjiustment has been to allow for tariff-induced distortions between domestic and border prices. But in practicethe SER has frequently been meant by project analysts to allow for the"over-valuation" of the domestic currency, for a "foreign exchange premium",etc.
Application of the Shadow Wage Rate
7. In some projects, a further, and until recently final adjustmentin the cost streams was made: Domestic costs were broken down into laborinputs (L), and non-labor domestic inputs (NL). Where judged appropriatelabor costs were then inultiplied by the shadow wage rate (SWR), which hadthe purpose of allowing for the fact that the project draws on unemployedor underemployed labor resources, and that therefore the opportunity costof employment in terms of output foregone elsewhere is below the marketwage paid to project labor:
NB = (SER) E - (SER) M - (SWR) L - NL (3)
8. Until recently then, the Bank's practice was to translate allproject costs and bene-fits into domestic prices and to reflect the economic(opportunity) cost and benefit of project outputs and inputs by applicationof the SER and the SWR where judged appropriate.
1/ The problems of e,timating the oppottunity cost of capital are wellknown and are not specific to the Bank approach. They are thereforenot further discussed in this paper. But see J. Schmedtje, "OnEstimating the Economic Cost of Capital", IBRD Report No. EC-138,October 21, 1965, and Shu-Chin Yang, "Social Rate of Return forProject Evaluation; An Estimate for Yugoslavia", World Bank StaffWorking Paper No. 205, June 20, 1975, for some work carried out byBank Staff in thi, area.
2/ Throughout this paper the R/R will be used as the salient measure ofproject acceptabi:Lity, which indeed has generally been the practicein the Bank. Of course, it is well-known that for a ranking ofmutually exclusive projects the net present value approach is thecorrect procedure for project selection.
3/ For simplicity, sunmmation signs, subscripts and superscripts havegenerally been om:itted in equation (2) and further below.
- 4-
Efficiency Pricing: The Squire/van der Tak Approach
9. After extended study of past Bank practices and alternative approachesthe Bank has now adopted a "new" methodology of economic evaluation ofprojects based on the work of Squire and van der Tak. 1/ In the firstinstance, this approach differs from the previous Bank practice by convertingall costs and benefits into border prices, rather than domestic prices.Returning to equation (2) above, this may be achieved by multiplying bothsides by the term OER
SERP'
NB' = (OER) E - (OER) M - cD (4)
where NB' denotes net benefits in terms of border, rather than domesticprices.
10. So far, of course, nothing has changed in substance betweenequations (2) and (4). However, the advantage of equation (4) is that itsystematically allows for a further step of refinement in project: analysis.The factor c may be viewed as a summary conversion factor, which translatesthe entire basket of domestically produced items from domestic to borderprlces, adjusting for distortions in each individual input price. Such
distortions may arise for a variety of reasons and may affect dif'ferentcomponents of the domestic items differently. To allow for this, oneshould rewrite equation (4) as follows:
NB' = (OER) E - (OER) M -EkCkDk (5)
where ck is the conversion factor converting each particular input (orhomogeneous group of inputs) from domestic prices to border prices.
11. Some of the Dk may consist of inputs which, although domesticallyproduced, are tradable commodities, i.e. close substitutes to comimoditiesactually imported or exported. Therefore their border price equivalentis obtained by allowing for tariff distortions and other trade restrictions.Other components of Dk will be non-tradables, such as the primary factorinputs labor and land. Their specific conversion factor ck may be derivedby considering the value of the output (in border prices again) f'oregoneby employing their inputs in the particular project. 2/ Finally, some of
1/ Squire and van der Tak, op. cit.; their approach in turn drawsheavily on the Little-Mirrlees/OECD Manual methodology.
2/ For instance, if in a cocoa-project labor is withdrawn from other cocoaestates and cocoa is an export commodity subject to a 100% export tax,the ck for labor is 0.5. If, however, under similar circumstancesotherwise, only 50% of the labor force is drawn from other cocoaestates, while the rest comes from an entirely unemployed group(a zero weight is given to their leisure), then the convers:ion factorck for labor is 0.5x0.5=0.25.
- 5 -
the Dk may consist of other non-tradables, such as construction, transport,etc. In this case conversion factors may be derived iteratively bybreaking down these inputs into their tradable and non-tradable componentsuntil only tradables and primary factors remain.
12. From this discussion it should now be apparent that the sumnmaryconversion factor c is niothing but the weighted average of the conversionfactors ck for individual inputs, weighted by their share in total domesticinputs as they occur in the context of the particular project being analyzed:
, =k cDk - OER (6)Ek Dk SER
13. The fact that c is in principle project specific is important,since it serves to clar[fy the limitations of the common use of the SERconcept. The SER has generally been calculated not on the basis of projectspecific weights, but on the basis of nationwide weights, i.e. using theweights of tradables and non-tradables 1/ in GNP. Therefore, only if itmay be assumed that the share of each domestically procured input in totaldomestic inputs for the project closely approximates the inputs' share intotal GNP, it is appropriate to use a nationwide SER, or its equivalent inthe Squire/van der Tak inethodology, the Summary Conversion Factor (SCF), whichwas labeled c in equation (4).
14. Indeed, the fact that traditionally labor was given specialconsideration by computing a SWR even under traditional Bank practices,indicates that the application of a nationwide SER was not alwaysconsidered sufficient in particular project contexts. It is now anempirical question to determine to what degree the use of project specificconversion factors rathier than summary, nationwide measures, such as theSER or SCF, actually affect project selection, or rather in what countriesand which types of projects the use of project specific conversion factorsis desirable. The SBank.'s current implementation exercise of the Squire-van der Tak methodology is designed to shed some light on this question.
15. One further advantage of the more explicit approach now adoptedby the Bank is that it will hopefully do away with some common misconceptionsregarding the purpose and use of the SER. As the discussion in the previousparagraphs made clear, the purpose of the factor c, or its counterpert, theSER, is to allow for the distortions in prices between domestic market andborder prices. It is not the purpose of c, or SER, to allow for a "foreignexchange premium" in developing countries, where foreign exchange is ascarce commodity, in that its availability permits more rapid growth thanwould otherwise be the case. This particular consideration goes beyond therealm of "efficiency prLcing", i.e. the adjustments aimed at correcting fordistortions in the relative prices, but falls into the realm of "socialpricing", where it is riecognized that, for instance, higher than actual
1/ The latter were in -iractice often neglected.
- 6 -
growth of GNP is regarded as desirable by a government and where thereforein project appraisal foreign exchange generation is accorded a premiumover the generation of consumption benefits. In order to keep the analysisclear, the SER should therefore not be taken to allow for such "social"considerations, but rather be restricted to correct for price distortionsintroduced by market imperfections, such as tariffs, quotas, etc. Instead,a separate analytical framework must be utilized to incorporate such socialobjectives as a higher than actual rate of growth, or a more equal incomedistribution. The next part of this paper deals with this type ofapproach, which is presently used on an experimental basis in soame Bankprojects.
Social Pricing: An Experimental Approach
16. The contrast between "efficiency" and "social" methodologies ofproject evaluation is not that the former is value-free while the latter isvalue-laden. Rather it is a constrast between a particular set of valuejudgments embodied in the efficiency approach on the one side; and themore general framework of the social analysis on the other side, whichpermits systematic consideration of alternative sets of value jucigments,one of which is in fact the particular set embodied in the efficiencyapproach.
17. There are four value judgments implicit in the efficiency approach:
First, the country's actual growth rate is optimal, i.e. desired savingsequal actual savings (investment); second, the distribution between publicsector and private sector expenditure is optimal; third, the economy isnot faced with a foreign exchange constraint; and finally, the existingdistribution of income is optimal, such that an additional unit ofconsumption for the rich is valued equal to an additional unit for the poor.Rather than presuming these rather extreme value judgments implicitly,the new methodology proposed by Squire and van der Tak allows for alternative,and less extreme sets of values as generally held by policy makers. 1/
18. To demonstrate the essence of the Squire/van der Tak methodologyassume that at the margin all public sector income is invested and allprivate sector income is consumed. 2/ Then take a project whose netefficiency benefit is NB',as derived from the analysis in the previoussection. The project induces a net increase in total consumption in theprivate sector equivalent to
C = i ci (7)
where'Ci is the consumption increment of the ith income group. 3/ Atborder prices, the consumption increment is equal to SC or eiCi, where
1/ A similar approach is adopted in I.M.D. Little and J.A. Mirrlees,Project Appraisal and Planning for the Developing Countries (London:Heinemann Educational Books, 1974).
2/ Obviously this represents an over-simplification; for a ful] demonstrationof the new methodology, where this assumption is not made the readeris referred to Squire and van der Tak, op. cit. especially pp. 49-62.
/ rItis of course&possible-that a project induces'consumption lossesamong some income groups for some of the years of the project life.In this case Ci is negative.
- 7 -
6 is the summary conversion factor which translates the national consumptionbasket from domestic into border prices. 1/ Therefore, out of total netbenefit, only NB' - BC) accrues to the public sector, and is thus availablefor investment. Assuming then that the government,wants to increase therate of growth and therefore places a premium on investment, net socialbenefits (NSB) may be written as follows:
NSB = (NB' - $C) v + C (8)
where v is the shadow price of public income (investment) at border pricesin terms of private consumption at domestic prices. Squire and van der Takgo one step further, in translating the analysis from a unit of account(numeraire) of private consumption as in equation (8), to a numeraire ofgovernment income in border prices (foreign exchange). This is achievedby dividing equation (8) on both sides by v:
NSB _ NSB' = NB' - aC (1 - v) (9)
Equation (9) shows that as long as public income (investment) is at a premiumrelative to private consumption, i.e. v$ is greater than unity, NSB' isless than NB'. Squire and van der Tak have suggested estimation proceduresfor the derivation of v which are based essentially on a comparison betweenthe marginal productivity of capital (investment) and the consumption rateof interest (CRI), i.e. the discount rate which translates futureconsumption into present value terms. 2/
19. So far the analysis has allowed for the suboptimality of GNPgrowth. 3/ A further step takes into consideration the suboptimality ofthe existing income distribution. Assume that the government values anadditional unit of consumption accruing to the poor more highly than anadditional unit accruing to the rich. This may be represented by attributinga weight di to the consumption of income group i, choosing the average income
1/ This parameter is generally calculated on a nationwide basis like theSCF. But note that under certain circumstances it may differ betweenincome classes, and even between, say, rural and urban projects.
2/ One simple formula which has shown to yield useful estimates inpractice is v = q/l3CRI, where q is the marginal product of capital.This of course assumes no reinvestment, and may thus be taken as aminimum value. Nolte here that the CRI is defined as CRI=ng + Pwhere n is the elasticity of marginal utility with respect toconsumption, g the growth rate of per capita consumption, and p therate of pure time preference. See Squire and van der Tak, p. 140.
3! Although in a rather simplistic fashion, it also has allowed for sub-optimal a:Llocation of resources between the public and the privatesectors. Furthermore, to the degree that the foreign exchangescarcity is reflected in insufficient domestic savings, this has alsobeen allowed for. For a more complete discussion on this, see Squireand van der Tak, op. cit. pp. 50-51. Note finally that the term"suboptimality" is here used to indicate that the actual growth rate(and/or income distribution) differs from that desired by thegovernment.
- 8 -
in the country as a basis for comparison. 1/ Returning then to equation(8) this weight may be applied to the private consumption increment of
group i: 2/
NSB = (NB' - aCi) v + diCi (10)
After algebraic simplification and transformation to the numeraire ofpublic income in border prices, this becomes
NSB = NSB i NB' - d-
20. Equation (11) is essential to the understanding of how thesocial analysis of projects works. 3/ Note that there is a consumptionlevel at which di/vS is exactly equal to unity, and where as a result theentire second term on the right hand side of equation (11) vanishes. Atthat level of consumption, therefore, net social benefit equals netefficiency benefit. This consumption level is labeled the "CriticalConsumption Level " (CCL) by Squire and van der Tak, and it representsthat level at which the government judges private consumption just asvaluable as public income (investment). For project beneficiaries withincomes above (below) the CCL, d./vS is smaller (larger) than unity, andtherefore net social benefit is slmaller (larger) than net efficiencybenefit. The same of course holds for the rate of return calculated onthe basis of the net benefits.
1/ Squire and van der Tak derive these weights from a constant-elasticityutility function, for which marginal utility u'= c-n, where c is theconsumption level and n the elasticity. The weights di can then bederived as di = (E/ci)n, where f is the average consumption level,Note that the utility function is equated by Squire/van der Tak with thegovernment's objective function, and that therefore the distributionweights, derived from it, express the government's preferences forredistribution; ibid., pp. 63-66.
2/ Assuming for the sake of simplicity that all beneficiaries have
approximately equal consumption levels and may thus be lumped togetherin one group. See footnote 3 below for a relaxation of this assumption.
3/ The equation may be expanded to allow for beneficiaries at different
income levels. In that case one may write:
NSB' = NB'- $EC (1 - ECidi/EC.
This shows that the summary consumption weight in the bracket is theaverage of all individual weights, weighted in turn by the share ofthe consumption increment for group i relative to the total consumptionchange.
- 9 -
21. The CCL is "critical" in another respect. It has been shown in
practice to be an extremely useful device for cross-checking the estimatesof what are obviously rather elusive parameters, namely the consumptiondistribution weights (di) and the value of public income (v). Given theinitial estimates of these parameters one may compute the CCL which theyimply. The cross-check then consists in considering governmental policieswhich can provide a direct estimate of the consumption level at whichthe government judges its own income and private consumption as equallyvaluable. Cut-off levels of consumption subsidies, of the income tax,etc. may provide the informational input for this direct estimate of theCCL. 1/
22. In concluding this section, a few common criticisms of this typeof social pricing methadology may brLefly be anticipated: First, projectanalysis should be value free. This criticism was dealt with at theoutset of this section, where it was pointed out that efficiency'analysisincorporates indeed very drastic value judgments particularly regardingthe optimality of growth and income distribution. In terms of the Squire/van der Tak methodology we can now see that these judgments may be inter-preted as setting all di and Sv equal to unity, in which case net socialbenefit equals net efficiency benefit. Of course, it should be clearthat only few governments judge their countries' growth performance andincome distribution as optimal, and thus the assumptions of the efficiencyanalysis are overly restrictive.
23. A second criticism stipulates that alternative policy tools,particularly fiscal instruments such as progressive income or wealth taxes,rather than social project analysis should be applied to achieve optimalgrowth and redistribution. In the abstract this is true, in the sensethat costless, non-distorting lumpsum transfers would be the best way ofachieving redistribution. But in practice the limitations of fiscalinstruments in LDCs are well known, in that they are both costly anddistorting. This is best reflected in the fact that as far asgqvernments are concerned growth and income distribution remainsuboptimal in many countries.
24. The last set of criticisms relates to the practical aspects ofimplementing the social pricing methodology. The estimation of the requirednational parameters and their application is difficult at best, and more-over it is costly in terms of additional staff time required. Furthermore,so the argument goes, social analysis will not affect actual project selectionsignificantly, since governments and aid organizations take their growthand income distribution objectives a:Lready into consideration, thoughpossibly in a less explicit and systematic fashion. Criticisms such asthese are in fact the ones which are most cogent to the practitioners ofproject analysis, and it is their concerns which the experimental applicationof the new methodology in the World Bank is designed to meet. The presentstudy is one component of this exercise and aims at testing the usefulnessof social analysis of projects in the context of a case study of Ivory Coast.
i/ For a good practical example see Shu-Chin Yang, op. cit., pp. 7-12.See also pp. 42-44 below.
- 10 -
Summary of Estimation Results for Ivory Coast -
25. Table 1 illustrates the estimates of the main country parametersrequired for social cost-benefit analysis. For all conversion factors itwas found that reliance on simple estimation formulae may lead to misleading
results, unless the appropriate formula is chosen carefully in line with
the conditions prevailing in a specific country. Where possible, oneshould avoid having to rely on the simplified formula, but should insteadestimate the necessary spending propensities for individual commodity
groups entering the conversion factor. The standard conversion factor(SCF) may be regarded as the inverse of a shadow exchange rate (SER) sothat the estimate of 0.83 for the SCF implies a SER of 1.2 times the
official exchange rate.
26. Two alternative estimation approaches were used for the marginalproductivity of capital (q): The macro-economic approach utilizing
aggregative output, investment and employment data, and the micro-economicapproach, based on interest information and industry profit data. Thereis good reason to believe that the former method substantially overestimates
the true value of q, while the latter probably somewhat underestimates it;on balance the likely range for q appears to be between 8% and 12% with a
central value of 10%.
27. Turning then to the social pricing Farameters, the objectivefunction parameters, n (the elasticity of marginal utility with respect
to consumption), and p (the rate of pure time preference) were derivedtogether with the consumption rate of interest (CRI) by reviewing the policydirections of the Ivorian government. Two cases of possible parameter
combinations were accepted for further work:
Case 1: n = 0.5, and CRI = 5.0%
Case 2: n = 1.0, and CRI = 7.5%
Given a growth rate of 3.3% for per capita consumption, the implied values of
p are 3.4% for Case 1 and 4.2% for Case 2. Case 1 represents governmental
value judgments which attribute a high priority to fast growth, and a low
priority to income distribution. In Case 2 moderate emphasis is placed on
both objectives. The former case was judged to represent more closely the
growth-oriented attitudes of the Ivorian government particularly during the1960s, while the latter appears to reflect more accurately the recent shiftsto a more balanced policy mix between growth and equity objectives.
28. Given these parameters, one can estimate the marginal distributionweights, d, for different consumption groups in the Ivory Coast, and the
summary distribution measure (D). The latter was estimated to be:
Case 1: D = 0.91
Case 2: D = 1.00
1/ All estimates are derived in detail in the main text below. There the
reader who is unfamiliar with the Squire/van der Tak method will also findbrief explanations of those concepts which have not been fully treated
in the preceding methodological introduction.
- il -
Table 1: SUMMARY OF COUNTRY PARAMETERS
Paraineter Case 1 Case 2
Standard Conversion Faci:or (SCF) 0.83 0.83
Conversion Factor for Consumption Goods 0.84 0.84
Conversion Factor for Capital Goods 0.90 0.90
Conversion Factor for Construction 1/ 0.77 0.77
Marginal Product of Cap:ital (%) 10 10
n (Elasticity of Marginal Utility) 0.5 1.0
CRI (Consumption Rate oi- Interest in %) 5.0 7.5
Growth Rate of Per Capil.a Consumption (%) 3.3 3.3
p (Rate of Pure Time Preference in %) 3.35 4.20
Summary Distribution Measure (D) 0.91 1.00
s (Marginal Propensity of Private and Public Savings) 0.4-0.5 0.4-0.5
Value of Public Income 2.5 1.7
ARI (Accounting Rate of Interest in %) 7.4 8.5
1/ Assuming ]abor is shadow priced at the value of the SCF. It is 0.73if labor is shaddw priced at 0.70.
29. The value of public income (v) was deduced from the value of publicinvestment on the assumption that public income is allocated optimally betweeninvestment and other uses. The resulting estimates of v, together with theestimated values of n, were cross-checkea against an independent estimate ofthe critical consumptiorn level (CCL). On the basis of this cross-check, thefollowing values for v were accepted for further work:
Case 1: v = 2.5 (implying a CCL of CFAF 15,414 atapproximately the 3rd percentile ofthe income distribution)
Case 2: v = 1.7 (implying a CCL of CFAF 47,987 atapproximately the 45th percentile of theincoine distribution)
- 12 -
30. Finally, the accounting rate of interest (ARI) was estimated asthe internal social rate of return on the marginally-acceptable publicproject:
Case 1: ARI = 7.4%
Case 2: ARI = 8.5%
The estimates of the ARI are based on a value of s = 0.5, where (1-s) isthe private marginal propensity to consume out of total returns on themarginally-acceptable project.
31. Table 2 présents estimates of shadow wage rates (SWR) for differenttypes of labor. The shadow wage for urban unskilled labor is based on theassumption that the urban rate of unemployment remains constant becauseof its role as an equilibrating mechanism. It follows that the SWR mustallow for the additional migrants who move to the urban sector in responseto the creation of one job. However, even allowing for migration, the SWRis still below the market wage, w.
Table 2: SHADOW WAGE RATES
SWR/wEfficiency
SWR/w Case 1 Case 2
Urban Unskilled Labor 0.31 0.60 0.59
Rural non-Ivorian Labor
i) Regional Strategy 0.33 0.50 0.02
ii) National Strategy 0.40 0.90 0.90
32. The SWR for rural non-Ivorian African labor depends on whether ornot benefits accruing to non-Ivorian labor are valued in the same way asbenefits accruing to Ivorian labor. If they are (i.e., the governmentadopts a "regional" strategy) then the SWR is consistently lower than if theyare not assigned any value (i.e., the government adopts a "national" policy).For other types of labor (e.g. skilled and rural Ivorian unskilled) the ratioSWR/w is set equal to the SCF on the grounds that the relevant :Labor marketsare operating reasonably efficiently so that the only adjustmeat involvestransforming wages from domestic terms into their equivalent in terms ofthe value of foreign exchange. Finally, for non-African expatriate laborthe ratio of the shadow wage to the actual wage falls between SCF (=0.83) andunity, depending on the proportion of income that is repatriated.
- 13 -
33. Three differernt project types were selected to illustrate the impactof the new methodology under different circumstances. Given the ex-postnature of the desk study and its purpose of illustrating the methodologyrather than reappraising the particular projects, a number of simplificationswere made, relating first, to the conversion procedure used to convert costand benefit streams froni domestic into foreign terms, and second, to thedistribution among the beneficiaries. However, the results of the analysisreflect the directions and the approximate degree of change introduced bythe application of the new methodology. Table 3 presents the main resultswhich are then reviewed briefly in the following three paragraphs.
Table 3: RESULTS OF PROJECT APPRAISALS
Internal Rest of Return (%)Grand Bereby Third Highway ProjectRubber Estate Routine Betterment Cocoa
Project 1/ Maintenance Works Project
Case 1n=0.5, CRI=5%, ARI=7.4% 12.8 37.1 -1.7 21.3
Case 2n=1.0, CRI=7.5%, ARI=8.5% 14.6 >100.0 26.2 25.1
Efficiency Casen=0, CRI=q = 10.0% 13.4 >100.0 >50.0 23.8
Bank Appraisaln=0, CRI = q = 10-12% 13.2 > 50.0 50.0 19.9
1/ In line with the Bank appraisal, the results for this project assume thatthe government adopts the regiona:Lemployment strategy.
34. The Grand Bereby Rubber Estate Project which was marginal undertraditional Bank economic justification methods easily passes the rate ofreturn critetion under the proposed methodology. The main reason for thischange is that the project involves a:Lmost exclusively public sector costsand benefits. Thus, a larger proportion of project returns goes to theGovernment than in the hypothetical marginal project, which determines theARI, i.e., the cut-off rate of return. The study also shows the impact onthe rate of return of using alternative SWR assumptions for the non-Africanrural labor component.
35. Rates of return for the Third Highway Project were substantiallyreduced by applying the proposed methodology. For one of the components(road betterment) the rate of return actually falls below the ARI, givenCase 1 assumptions. The attractiveness of the project is reduced becausenearly all benefits of the project are assumed to accrue to the privatesector in proportion to the existing income distribution. If one allows for
- 14 -
savings, and/or a better income distribution impact (since the projecttakes place mainly in the rural, poorer areas of the country), the projectremains acceptable (although at reduced rates of return) under all alter-native cases here considered. Thus the review of this project shows upthe potential importance of allowing for the savings component in privatebenefits, and the need to specify carefully the income levels of expectedbeneficiaries.
36. The application of the proposed methodology to the Cozoa Projectchanged only slightly the absolute level of the rate of return to thisproject, although it improved somewhat relative to the ARI. Th,s explanationis that the beneficiaries of the project happen to be close to the criticalconsumption level, so that their net benefit from the project is assignedthe same value as public income. Comparing social and efficiency pricingprocedures in their impact on the rate of return, one finds that they areroughly equivalent for the Rubber Project and the Cocoa Project, but fordifferent reasons. In the Rubber Project, the similarity of resultsis due to the fact that nearly all benefits as well as costs are public;in the Cocoa Project, however, it is due to the fact that the beneficiariesare close to the critical consumption level at which point the foreignexchange cost of private benefits (i.e., consumption) is exactly offsetby the social benefit of the same, so that one may proceed as if allbenefit accrue to the public sector.
- 15 -
III. EFFICIENICY PRICING PARAMETERS IN IVORY COAST
Conversion Factors
37. In this section we estimate the conversion factor for consumption,the Standard Conversion Factor, SCF; the conversion factor for capital goods;and the conversion factcor for one non-tradable, construction {S & T, pp. 88-97;i22-1j2} .1/
As a starting point, the general formula for conversion factorsis specified as
a = z a.aI/p {S & T, p. 128} (12)
where ai is the proportion of marginal expenditure devoted to the jth commodity,
a. 1, xA is the shadow price and pj the market price of the jth commodity.
If one may assume that export demand and import supply are infinitely elastic; 2/that marginal changes in expenditure on non-tradables can be neglected; 3/and that all income elasticities of spending are unity and (or) the relativesize of the average (marginal) propensities to spend on importables and onexportables are approxirmately reflected by the relative size of imports andexports, equation (12), reduces to the simple formula
M + X (13)M(l+t ) + X(l-t )
m x
where M(X) is the c.i.f. value of imports (f.o.b. value of exports) and tm(t x)is the average tax on iimports (exports). Alternatively, if one may assumethat all exportables are exported; and that income elasticities for importcommodities are all unii:y the formula for the conversion factor becomes
1 (14)
m
since the exportable terms drop out of equation (13) and the importable termscancel.
38. When these special assumptions are abandoned, 4/ the marginalpropensities to consume exportables and importables have to be estimated.
l/ For clarity and easy reference, the appropriate page numbers in Squire andvan der Tak (henceforth abbreviated as S & T)are provided in bracketé.
2/ This assumption wiLl be further discussed and relaxed below.3/ This assumption is justified if either aj for non-tradables is small, or
it À /p. for non-tradaDies is approximacely equal to the conversior factor.4/ Marginal changes in expetiditure on non-tradables continue to be neglected,
except in the case of the conversion factor for capital goods, whereconstruction is taken into consideration.
- 16 -
Where data for direct estimates are not available, income elasticitiesand average propensities to spend can be used instead to estimate marginalpropensities. In any case, since marginal and average propensities canbe measured directly only in domestic price terms, adjustments must be madeto transform them into foreign currency terms. Noting that
c*pAPC.a. = APC (15)
J = z jAPC.j J
where Ei is the income elasticity, and APC. is the average propen:sity tospend on the jth commodity, equation (12) iay be written as
E APC E APC
1 + tm 1 - tx(16)
EAPC + - APCm m x x
where subscripts m and x refer to importables and exportables respectively.
39. Equation 16 is the basic equation used below except thal it isfurther modified to account for less than infinite demand elastic:ities ofthose commodities in which Ivory Coast contributes a major share of totalworld trade, i.e., coffee, cocoa, and wood. For these exportables theborder prices in equation (16) have to be replaced by marginal revenueproducts. 1/ Short of estimating marginal revenue products separatelyone can assume that the export duties levied by Ivory Coast on thesecommodities are optimal and that, therefore,the duties equal the inverseof the export demand elasticities. As a result, domestic prices may beused instead of border prices in equation (16) for these exportables. 2/
40. Conversion factor for consumption (e). From import and exportdata averaged over the years 1968 and 1972 (Table 4) one can estimate thesimple conversion factor for consumption goods from equation 13, assumingthat the resulting conversion factor is 1.03. Alternatively, if one assumesthat all exportables are exported and all importabies imported (ecluation 14),the conversion factor becomes 0.80.
1/ Distribution effects of relative price changes are here neglected.2/ A further possible modification is the disaggregation of import flows by
source, considering separately imDorts from European countries, andthose from non-European countries. This allows explicitly for thepreferential trading practices between the Ivory Coast and the Europeancountries (in particular the EEC). It was found, however, that the effecton the conversion factors is negligible. Note also that equation (16)assumes that the Ivorian government will continue its existing patternof mutual trade preferences with the EEC; and that the IvoriaLn balanceof payments is in long-term equilibrium and that, therefore, no devaluationwill take place in the foreseeable future.
3/ These average tariff rates were derived from a rough inspection of thetariff structure and a judgment of the country economist; they are notdirectly related to the more detailed tariff estimates presented below.
- 17 -
Table 4: IM?ORTS AND EXPORTS OF CONSUMPTION GOODS(CFAF Billions)
5-Year19658 1969 1970 1971 1972 Average
Importa (M) 28.8 29.3 36.6 38.2 40.5 34.7
Exports (X) 70.1 73.7 90.2 85.9 85.1 81.0
Sources: World Banik staff estimates
41. For the more complex estimation procedure (equation 16), considerfirst the import side. Using 1970 data from the input-output tablepresented in the Ivorian Five-Year Plan 1971-1975, tariff rates wereestimated for individual commodity groups by dividing tariff revenueby import value for each group. Average expenditure propensities forimportables were also derived from the Ivorian input-output table as theshare of expenditure on a particular commodity group in total expenditure.The elasticities were derived from a household expenditure survey of ruralfamilies in the South--East Region of Ivory Coast, performed in 1964/65(Table 5 summarizes the computation on the importables).
42. On the export side, export duty rates were determined by majorexport commodities from inspection of the Tariff Code, while the averagepropensity to consume and the income elasticities were derived from theabove mentioned expendLiture survey. Input-output data could not be usedfor this purpose, since they are not detailed enough, and since they do notspecify revenues from export duties by commodity group. Table 6 summarizesthe export data. Note that it is assumed that coffee, instant coffee,cocoa, cocoa products, and wood face an inelastic foreign demand; and thatfor most of the export. products the average propensity to consume is zeroor negligible. The main exceptions are wood, bananas, and cotton, with thelast one having the largeàt overall weight.
Table 5: COMPUTATION OF MARGINA1 PROPENSITIES TO SPEND ON IMPORTABLES
Final Consumption Total Domestic Uses-Tariff rate.(tm) PC b! e ec emA PCm emAPCm ePCm/ ed/ emPPCM eAPCm
(%) _/ (%) 71 tmtl ('°) 1 + tm
Animals (and Products) 2.3 23.6 0.8 18.88 18.46 16.6 o.8 13.28 12.98Industrial Agriculture 44.5 1.2 0.9 1.08 0.75 3.0 0.9 2.70 1.87Fishing 3.1 3.3 0.5 1.65 1.60 2.1 0.5 1.05 1.02Grains and Flower 6.4 5.7 0.5 2.85 2.68 4.1 0.5 2.05 1.93Cans, Tea, Coffee, Cocoa 42.8 1.2 1.0 1.20 0.84 0.7 1.0 0.70 o.49Drinks, etc. 127.2 5.3 1.0 5.30 2.33 3.3 1.0 3.30 1.45Other Food 44.9 4.9 0.5 2.45 1.69 3.3 0.5 1.65 1.14Energy, Water 0.5 4.8 1.5 7.20 7.16 7.8 1.1 8.85 8.81Extractive Activities 8.3 0.2 1.5 0.30 0.28 0.1 1.0 0.10 0.09Metals 25.6 0.0 1.9 0.0 0.00 1.8 1.2 2.16 1.72Construction materials 15.8 0.0 1.9 0.0 0.00 1.8 1.2 2.16 1.87Fertilizer 0.0 0.0 0.0 0.0 0.00 0.3 1.2 o.36 0.00 HChemicals and Rubber 32.8 3.8 1.5 5.70 4.29 3.6 1.2 4.32 2.25Wood Products 42.8 1.2 1.5 1.80 1.26 2.1 1.2 2.52 1.76Vehicle Assembly and Repair 25.6 4.1 2.0 8.20 6.53 5.5 1.1 6.05 4.82Mechanical and Electr.Prod. 24.7 4.6 1.5 6.90 5.53 5.6 1.7 9.52 7.63Textiles 35.6 2/ 13.5 1.3 17.55 12.94 10.0 1.3 13.00 9.59Leather and Shoes 38.9 2/ 2.2 1.3 2.86 2.06 1.4 1.3 1.82 1.31Fats 17.7 2.3 1.0 2.30 1.95 2.1 1.0 2.10 1.78Rubber and Plastic Prod. 47.2 1.4 1.5 2.10 1.43 1.6 1.2 1.92 1.30Various Ind. Products 30.9 2.5 1.5 3.75 2.86 3.0 1.2 3.60 2.75Totals 92.07 74.82 83.21 66.56
Notes: l/ Final consumption plus intermediate uses.1 If allowance is made for the fact that a substantial proportion of imports of textiles and leather consists
of intermediate products which are further processed in Ivory Coast and are then subject to the valueadded tax (TVA), the actual leVel of protection on these commodities is, according to Gary Pursell (verbalcommunication) in the neighborhood of 53 percent and 85 nercent resnectively The conversion factor woll-dthen be 0.81.
Sources: a/ Input-Output Table for 1970 from Cote d'Ivoire, Plan Quinquennal, 1971-1975 pp. 88-89; obtained as aratio of tariff revenue over imports for each commodity group.
b/ ibid.; obtained by taking ratio of final (or total domestic) uses per commodity group over total final(domestic) uses.
c/ SEDES, Region du Sud-Est, Les Budgets Familiaux, Paris 1967; these are elasticities estirnated fromhousehold expenditure data for different income groups.
d/ for items 1-7, 17-19, where final use is predominant in total domestic use, these are as for finalconsumption; for item 8-16,20,21, where intermediate uses are predominant, elasticities were computed fromimport and GNP data on 5 yr average basis.
Table 6: COMPUTATION OF MARGINAL PROPENSITIES TO CONSUME EXPORTABLES
a b/ ~~~~~~~APC eDuty Rate, tx a/ AP / c/ APC e Implied(%) (W x - x 1 -tx Export Demand (Price)
Elasticity 3/
Coffee 23 0. / .5 0.00 2/
Cocoa 23 û.i1/ û. û0.00 - 4.3
Cocoa Products 16 O. i/ 0.5 0.00 - 2/ 6.3
Wood 27 1.2 0.5 1.80 - 2/ 3.7
Instant Coffee O o.01/ 0.5 0.00 -2/ œ
Bananas 12 >-9 0.5 2.45 2.78 o
Pineapple and Products 10 0.0-/ 0.5 0.00 0.001/ 0)5 00000Cola Nuts 14 0.0-/ o.5 o.oo o.o
Palm Oil and Kernels 7 0.7 0.5 0.35 0.38 C
Cotton (and Products) 5 4.3 1.3 5.59 5.ô8 8
TOTAL 10.19 9.C4
Sources: a/ Cote d'Ivoire, Code des Tarifs, 1973b/ SEDES, Region du Sud-Est, Les Budgets Familiaux, Paris 1967.c/ Assumed elasticity, with the exception of cotton for which the elasticity of textiles
from Table 5 is taken.
Notes: 1/ Zero or negligible.2/ Export commodity with finite foreign demand elasticity.3/ Computed as the ratio 1/tx on the assumption that tx, the export duty is levied at an
optimal rate for those export commodities in which the Ivory Coast contributes asignificant amount to world supply.
- 20 -
43. Using these import and export data, the conversion factor fromequation (16) was computed to be 0.838, which we may round off to 0.84. 1/Comparing these results with the conversion factors estimated from the
simple formulae it appears that in the case of the Ivory Coast the formulawhich assumes that all exportables are actually exported (equation 14)
provides an estimate much closer to the result obtained by the detailedanalysis, than the simple formula based on the proportionality oftradàbles and traded goods (equation 13). The reason is that with theexception of a few commodities (most importantly textiles) Ivory Coastis only a marginal consumer of its own export commodities. This situationwill be true for most less developed countries which have a low diversification
in their export structure, and which apecialize in primary products thatneither are part of the domestic consumption basket, nor enter as inter-mediate goods into the domestic production process. Ivory Coast is, of
course, a good example for this type of economy. In countries, where theimport substitution process has gone further and the export structure ismore diversified, containing a significant proportion of items which are also
domestically consumed, the simple formula based on the proportionalityassumption (equation 13) may lead to better results.
44. Standard Conversion Factor (SCF). The data for the two simple
estimation equations are shown in Table 7, where import and export datawere used together with total import and export duty revenues to c.alculatethe tariff rates. For the case where the proportionality assumption between
traded and tradable goods are made (equation 14), the SCF is 1.00. Forthe alternative assumption that all tradables are traded (equation 14),the SCF is 0.86.
Table 7: DATA FOR COMPUTATION OF STANDARD CONVERSION FACTOR(Simple Case)
5-Year1968 1969 1970 1971 1972 Average
Total Imports 77.9 86.3 107.7 110.8 114.3 99.4
Total Exports 104.9 118.2 130.2 126.6 139.5 123.9
Import Tariff Revenues 12.9 12.5 15.4 18.7 21.3
Export Duty Revenues 10.8 12.4 15.5 17.5 19.0
Average Import Tariff (%) 16.6 14.5 14.3 16.9 1.8.6 16.2
Average Export Duty (%) 10.3 12.3 11.0 13.4 1:3.3 12.1
Source: Bank estimates.- -
1/ Using an expenditure survey of rural households in the South--East Region of
Ivory Coast as an alternative data source for average propens:ities and
income elasticities, a conversion factor of 0.836 was derived.. Consideringt-hat both data sources lead to the same (rounded off) result, one can infer
that rural and national conversion factors for consumption are approximatelyequal.
- 21 -
45. For the less limited approach of equation 16,the Ivorian input-output data are again used for the import side, but this time, intermediateuses are considered in addition to final uses when computing the averagepropensities (cf. Table 5). On the export side the same data are used asbefore except that the intermdiate uses of wood are also considered, whichincreases the average propensity to spend from 1.2% to 2.1%. As regards theincome elasticity data, for those items where final uses constitute a majorpart of total domestic uses, the previously applied elasticities were employed.However, where intermedi.ate uses constitute an important share of totaluses, income elasticities were estimated from actual trade data on theassumption that the income elasticity of imports equals the overall incomeelasticity of importables. 1/ The resulting SCF is 0.830.
46. This is again more closely approximated by the simple formulabased on the assumption that all exportables are exported (equation l4).than by the simplification assuming proportionality (equation 13). Forthe present study, an SCF of 0.83 will be used in further estimationexercises. This is equivalent to a foreign exchange premium of 20% or aSER of 1.20. Considering that'a number of non-tariff protective measureshave not been incorporated into this analysis, an SCF of 0.83 may somewhatoverestimate the true SCF. But there seems little reason to believe thatit would be below 0.80. For the purpose of sensitivity analysis therange of 0.85 to 0.80 might usefully be explored. With respect to futuredevelopments in the SCF,, the Ivorian 5-Year Plan data from the Plan projectionsuggest that ,gith proportionality assumptions (equation 13), the simpleSCF for 1975 and 1980 l:ies at 0.95; if exports are given zero weight(equation 14), the SCF -Ln both periods is 0.80. The slightly lower valuesprojected for SCE in fut:ure years reflect somewhat higher levels of averageimport tariff rates. For practical purposes, it seems reasonable to assumethat the SCF remains approximately constant over the foreseaable future.
47. Conversion factor for capital goods. If non-tradables are neglectedthis factor may be approximated by unity, since Ivory Coast does not exportcapital goods, and in e-fect does not levy tariffs of any significance onthe imports of capital goods, due to pervasive exemptions from tariffs onimported machinery, etc., which are granted to most enterprises. However,for present purposes, non-tradables should not be neglected since constructionforms a substantial part: of capital formation, the marginal propensity tospend on construction iS not negligible, and since the ratioof accounting tonominal prices of construction may be different from the conversion factorfor capital goods in general.
48. This may be allowed for by assuming first that the income elasticityof all capital goods is unity, and then using the proportion of constructionin total capital goods to approximate the marginal propensity to spend onconstruction and other capital goods respectively. From the input-output
1/ -A...urther imior.tant asstimtion in this estimation is that there were norelative price changes over the years.
- 22 -
table for 1970 in the Ivorian 5-year Plan, it is found that the proportionof capital expenditure going to construction is 37.8%. Furthermore, theconversion factor for construction is estimated below to be 0.77 or 0.73,depending on the shadow price of labor. These various elements may then becombined to complete the capital goods conversion factor as follows:
1 (1.0 x 62.2 + 0.77 x 37.8) = 0.91 for SWR = SCF = 0.83;100 w
1 (1.0 x 62.2 + 0.73 x 37.8) = 0.90 for SWR = 0.70.100 w
Based on this rather crude estimate, it will be assumed for the ensuingestimation procedures that the capital goods conversion factor equals 0.90.
49. Conversion factor for construction. This factor for a non-tradablecan be estimated on the basis of the input-output data for Ivory Coast,including the import tariff information previously used for the computationof the conversion factors for tradable goods (cf. Table 5 above). Table 8shows the domestic values of the various inputs to the construction sector,including salaries and "other value added" items. For the tradable goods thetariff data may be used to compute individual conversion factors by applyingthe formula {l/(1 + tm)}. These conversion factors in turn are mnultipliedwith the domestic input values of tradables to obtain the border priceequivalent. For non-tradables and "other value added" the SCF was used.In the case of salaries and wages two alternatives were tried: first, usingthe SCF, i.e., assuming that the output foregone by the labor force employedin construction is valued at SCF to obtain the border price equivalent; andsecond, shadow pricing labor at 70% of the domestic value. This value waschosen between 60% (SWR for urban unskilled labor) and 83% (SCF), since onlypart of labor employed in construction is unskilled and drawn from the urbansector. Thus, the value of labor employed must lie between the value of urbanunskilled labor and the full value of labor in foreign terms. The resultingconversion factors for construction are 0.769 for the case where labor isshadow priced at 0.83, and 0.734, when labor is shadow priced at 0.70. 1/
The Marginal Productivity of Capital, g.
50. The estimation of the marginal productivity of capital may bebased on macro and micro data {S & T, pp. 110-112}. For Ivory Coast bothapproaches are explored. Taking the macro-economic estimation procedurefirst, one multiplies the incremental employment/capital ratio with themarginal product of labor and subtracts the resulting term from the incre-mental output-capital ratio. One thus obtains an estimate of the marginalproductivity of capital in domestic prices. In order to express it in foreigncurrency terms it must be multiplied by the ratio of the Standard ConversionFactor over the conversion factor for capital goods.
1/ Note that in a particular project context it may be necessary to estimatea project-specific conversion factor for construction, since it may not beappropriate to assume the factor input proportions applicable to the country'sentire construction sector to reflect the factor input proportions in theparticular project. Thisapplies in particular to low-cost housing, whichmay employ a relatively high proportion of unskilled labor and of domesticallyproduced inputs.
- 23 -
Table 8: VALUfATION OF INPUTS FOR CONSTRUCTION SECTOR
Domestic Value a/b/ Border Value-l
Importe (CFAF millioxli) Conversion Factor- - (CFAF million)
Energy 2,308 0.995 2,297
Matals 3,643 0.796 2,900
Construction Mat. 5,591 0.864 4,828
Chemicals, etc. 749 0.753 3/ 564
Wood Prod. 2,734 0.700 1,915
Vehicles 866 0.796 689
Mechaniical & Electr. 2,521 0.802 2,022
Textiles 2 0.737 1
Rubber 328 0.679 223
Various Indust. 640 0.764 489
Construction 530 0.830 440
Transport 1,965 0.830 1,631
Rant 495 0.830 381
Other Services 1,386 0.830 1,150
Salaries 13,463 0.830 (0.700)-1 11,174 (9,424)
Other Value Added 10,605 0.830 8,840
Taxes 3,610 0 0
Total 51,436 0.769 (0.734) 39,544 (37,749)
Notes: 1/ Computed as /'(l+t), or SCF for non-tradables.2/ Computed by multiplying domestic value with conversion factor.3/ An alternative, and probably more appropriate approach would have
used the export tax on Ivorian wood products of 27X (see Table 6),rather than the import tax on the relatively small imports of woodfurniture and related products. In that case the conversion factorfor wood prodiucts, l/(l-tx), would have amounted to 1.370, and as aresult the conversion factor for construction to O.8O4. I am indebt,edto Gary Purse]l for this observation.
4/ Alternative SITR
Sources: a/ Cote d'Ivo:ire, Plan Quinquennal 1971-75, pp. 88-89.b!/ Table 5
- 24 -
Table 9: OUTPUT/CAPITAL RATIO,. 1965-1972(CFAF billions)
a/ a/ Real a/ / Real AOutputYear Output Deflator Output AOutput ACapital Deflator ACapital ACapital
1965 236.8 78.6 301.3 43.6 82.7 52.7
1966 257.3 81.9 314.2 12.9 44.6 85.0 52.5 0.24
1967 274.4 83.7 327.8 13.6 45.9 84.8 54.1 0.26
1968 325.1 88.3 368.2 40.4 54.0 85.3 63.3 0.75
1969 364.0 92.1 395.2 27.0 61.8 90.3 68.4 0.43
1970 414.0 100.0 414.0 18.8 83.8 100.0 83,.8 0.27
1971 445.1 99.2 448.7 34.7 92.4 105.2 87.,8 0.41
1972 480.0 99.5 482.4 33.7 93.4 104.7 89.2 0.38
Average: 0.39
Sources: a/ Bank estimatesb/ Bulletin Mensuel de Statistique; price index for exports of
machinery from France, based on Franc currency value; 1970=100.
Table 10: ALTERNATIVE ESTIMATES OF OUTPUT/CAPITAL RATIO
A O>utputA Capital ICOR
1965-1972 0.39 2.56
1970-1972 0.35 2.86
1970-1979 1/ 0.31 3.25
1/ Bank projections
- 25 -
51. The incremental output/capital ratio: This is the inverse of themore conventional ICO)R and may be estimated from output and investment datafor Ivory Coast. The cost of living index for Africans was used to deflatethe output figure`s,lJ' while a price index of exports of machinery from Francewas used for capital goods, considering that most capital goods are importedfrom France. Table 9 presents the necessary data and transformations andindicates that the avrerage value of the incremental output/capital ratio forthe period 1965 to 1972 was 0.39 (equivalent to an ICOR of 2.56). If onlythe last three years are taken, a lower value (0.35) may be observed; Bankprojections for the period 1970-79 are even lower (0.31). These values aresummarized with their corresponding ICOR values in Table 10. For our presentanalysis, it has been assumed that the central value of the ratio is 0.35,while sensitivity analysis will consider values of 0.4 and 0.3.
52. The incremental employment/capital ratio: Interpolating from thegrowth of Ivorian employment from 1965 to 1970 data, one may estimate this
Table 11: EMPLOYMENT/CAPITAL RATIO, 1965-1972
a/ b/Employment A Employment Real A Capital A Employmentt x 10 6
Year (in thousand) (in thousand) (CFAF bill) à Capitalt-i
1965 1,880 52.71966 1,935 55 52.5 1.041967 1,991 56 54.1 1.071968 2,048 57 63.3 1.051969 2,108 60 68.4 0.951970 2,165 57 83.8 0.831971 2,228 63 87.8 0.751972 2,292 64 89.2 0.73
Sources: g SETEF, Lt3-mage Base 1970 Emploi: Vol. I and Annex (Paris,1973), us-ng the l965 ànd 1970 employment figures andextrapolating the other years by applying the annualaverage compound rate of growth of 2.9%.
b/ T'able 9.
1/ Cost-of-living indices tend to be unreliable and are not necessarilyappropriate GNP def3ators; in the present case, hawever, it is theonly available price deflator.
- 26 -
ratio for the period 1965 to 1972. Table 11 indicates that the incrementallabor/capital ratio has continuously dropped over the recent years. This isfurther considered in conjunction with the values for the marginal productivityof labor estimated in the next paragraphs.
53. The marginal productivity of labor: Aggregate labor productivityestimation is difficult in the case of Ivory Coast, since a large proportionof employment is in the informal, i.e., non-wage sector of the economy. Incombining informal sector income estimates with modern sector wage informationonly a rough approximation of the average national marginal productivity oflabor can be hoped for. Starting with informal sector incomes, availabledata show that in 1970 the annual average income per worker was CFAF 73,000(derived by dividing informal sector earnings by informal sector employment;see Table 12). This calculation is based on the assumption that informalsector incomes correspond quite closely to national minimum wages. Afurther assumption is made that real labor productivity did not changebetween 1965 and 1970, and that as a result the CFAF 73,000 figure appliesfor both years.
Table 12: INFORMAL SECTOR PRODUCTIVITY, 1970
AverageAnnual Earnings Employment Total Wage Bill(in CFAF thous.) (in thousand) (in CFAF mill.)
Rural (Primary) Sector 67 1,452 97,284
SMIG (Construction andServices) 121 79 9,559
White Collar (Industrial) 156 60 9,360
All Informal Sector 73 1,591 116,203
Source: SETEF, op. cit., assuming that SMIG wages approximate informalsector incomes.
54. In the formal sector, wage data, although imperfect, may berelied upon to provide proxies for the marginal productivity of labor.The wage data are presented in Tàble 13, deflated to reflect real wages.
- 27 -
Table 13: FORMAL SECTOR WAGES AND SALARIES, 1965, 1970
Annual RealAverage Wag3_ b/ Average Wage
Year (CFAF thousands) Price Deflator (CFAF thousands)
1965 326 78.6 415
1970 465 100.0 465
Sources: a/ SETEF, op. cit.b! Table 9.
55. Combining the informal and formal sector productivity data intoa weighted national average, it is found that the national marginalproductivity of labor increased between 1965 and 1970 from CFAF 150,000p.a. to CFAF 179,000 p.a. (cf. Table 14).
a/
Table 14: NATIONAL LABOR PRODUCTIVITY, 1965, 1970
(Earnings in CFAF thousand p.a., Employment in thousands)
Informa] Sector Formal Sector TotalYear Earnings Employment Earnings Employm. Earnings Employm.
b/ c/1965 73 1,468 415 425 150 1,893
1970 73 1,591 465 590 179 2,181
Sources: a/ Tables 12 and 13.
b/ Assumes that informal sector productivity remained unchangedbetween 1965 and 1970.
c/ SETEF, op. cit.
56. Multiplying i:he 1965 and 1970 marginal labor productivity datawith the corresponding values of the employment/capital ratio one findsthat labor's share in marginal output remained virtually constant over these
- 28 -
five years at 0.15.1/ Subtracting then the share of labor from theincremental capital/output ratio one obtains the marginal productivity ofcapital in domestic terms. Multiplication with the ratio of the SCF overthe conversion factor for capital (0.83/0.90 = 0.92) yields the marginalproductivity of capital in foreign currency terms. Table 15 summarizesthe possible values, which fall into the range of 13.8% (for a low output/capital ratio) and 23.0% (for a high output/capital ratio).
Table 15: ALTERNATIVE VALUES FOR THE MARGINALPRODUCTIVITY OF CAPITAL
A OutputA Capital 0.30 0.35 0.40
q 0.138 0.184 0.230
57. For the micro-economic estimation procedure two sources ofinformation on the marginal productivity of-capital are available: theinterest rate structure in the country, and evidence on profits in Ivorianindustry. Looking first at the interest rate structure, Table 16 presentsvarious rates in force during 1973. This includes an estimate of the realinterest rate, obtained by deducting the average rate of increase in thecost of living index over the years 1965 to 1972, which amounted to 3.5%p.a. (cf. Table 9). Since these rates are largely managed by central bankoperations, they cannot be taken to reflect accurately the marginal returnsto capital in the country. However, they may be used in connection withthe data on Ivorian industry profits to obtain a more accurate picture ofthe value of the opportunity cost of capital.
58. Actual pre-tax profits on equity in 1971 ranged from -5.1% to39.8% between 14 industrial sectors covering 114 firms, with an averageprofit rate of 15.3%. 2/ Adding to this an average return on equity fromroyalties of 2.0%, a total average return on equity of 17.3% is obtained.But considering that Ivorian industry was financed to 48% by equity, 18%by long-term borrowing, and 34% by medium-term borrowing, one can compute aweighted rate of return on all invested capital of 9.0%: 0.48 x (17.3 - 3.5) +0.18 x 5.0 + 0.34 x 4.5 = 9.0. The value of 9% for the marginal productivity
1/ This leads to the interesting conclusion that the share of labor in totalincome remained virtually unchanged over the years, with increases inproductivity being offset by a declining labor-capital ratio. One implica-tion of this is that if one projects a constant share of labor, and afalling incremental output/capital ratio, one also has to project a fallingq. Here it is assumed that all three are constant at the estimated values.
2/ Information provided by Gary Pursell.
- 29 -
Table 16: LEàDING INTEREST RATES IN THE IVORYCOAST, 1973 (IN %)
Short-Term Medium-Term Long-Term
Discountable at BCEAO 7.50 - 9.50 7.25 - 8.25 7.50- 9.25
Non-discountable at BCEAO 8.00 -11.00 11.00 8.75-10.25
Real Rate (for 3.5% inflation) 5.00 4.50 5.00
Note: Although the 3.5 percent inflation rate is taken from a seven-year
average of inflation rates for 1965-1972, the real interest rate
approximately reflects the 1971 rate of increase in prices, whichis to be used below in conjunction with 1971 profit date.
-Source: Bank estimates for nominal rates;. the real rates are computed
for discountabiLe (risk free) loans
of capital (in domestic terms) is likely to represent an underestimateof the true value of the opportunity cost of capital in all sectors of
the Ivorian economy, sinice it may be argued that the average rate of returnon equity in industry is not fully representative of the opportunity cost
of capital in the Ivory Coast. World Bank estimates show that the average
pre-tax returns on equity are depressed by the poor returns for some ventures,
reflecting avoidable failures due to lack of adequate project preparation,
insufficient knowledge about the market and deficient management; and that
industry may be one of the less profitable sectors in the Ivory Coast. 1/
Finally, the Ivorian Government has regarded gross profits on equity between
15 and 30% as normal. This suggests that if one were to consider as the
opportunity cost of equity finance the return on equity funds, a (nominal)rate of 25Z may be regarded as quite possible. Using this rate together
with the interest rate figures, inflation rates, and financing ratios above
specified, one obtains a rate of return to capital of 12.7% (for 17% and 20%
return on equity, the opportunity cost of capital is respectively 8.9% and
10.3%). 2/
59. Further information may be obtained from Bank financed projects and
from past Bank estimates of the opportunity cost of capital. The following
list summarizes the estLmated internal economic rate of return for some recent
Bank projects:Cocoa I: 20 - 35%Oil Palm Il: 16 - 18%Highway III: 1.5 - 30%
Highway IV: 16 - over 50%
Rubber I: 13.2%
i/ The only factor in the opposite direction,i.e., tending to lower the
value of the marginal productivity of capital, is the fact that before
1973 the interest rate structure was somewhat lower than shown in Table
16, due to lower discount rates of the regional central,bank (BCEAO).
2/ See also Pursell, op. cit., p. 51.
- 30 -
Moreover, the World Bank has in the past used the range of 10 - 12% asan estimate for the opportunity cost of capital. Thus it does not seemunreasonable to conclude from the micro-data presented that the marginalproductivity of capital falls within the range of 9-13%. For the purposesof our present analysis, this estimate, however, still has to be translatedinto border price terms by applying to it the ratio of the SCF over theconversion factor for capital.l/ Table 17 shows the resulting values forq, given alternative values for the marginal productivity of capital indomestic terms:
Table 17: ALTERNATIVE VALUES FOR THE MARGINALrKODUCTIVITY OF CAPITAL, q
Marginal Productivityof capital, %(in domestic terms) 9 il 13
q (%) 8.3 10.1 12.CI
60. From the micro-data, the range of reasonable values for the marginalproductivity thus seems to be between 8.3% and 12%, which differs substantiallyfrom the range obtained from the macro-economic estimation procedure. Thereasons for this discrepancy lie probably mostly in the limitations of thesimple macro-formula used, which in particular omitted the returns to otherfactors of production and the increased productivity resulting from technolo-gical progress.2/ Moreover, the difficulties in estimating reliable valùesfor the macro-economic parameters are such that not much confidence can beplaced in the resulting values of q. It is, therefore, suggested, thatprimary reliance be placed on the estimates from micro-economic data and10% be taken as the most probable central value.
1/ Note that the marginal productivity estimate applies not to industry,but to the Ivorian economy as a whole, and it is, therefore, appropriateto apply the ratio of conversion factors as specified in this paragraph.
2/ An alternative macro-economic approach (not pursued here due to datalimitations) would have estimated a two factor production function,allowing for technological progress. In other country studies performedfor the IBRD, this has been found useful particularly where preexistingproduction function estimates could be relied upon.
- 31 -
IV. SOCIAL ERICING PARAMETERS IN IVORY COAST
Value Judgements: The CRI, n, and p
61. The consumption, rate of interest is defined as:
CRI = i = ng + p {S & T, p. 140} (17)
where n: elasticity of marginal utility with respect toconsumption {S & T p.63 }
g: growth rate of per capita consumptionP: rate of pure time preference.
This section will attempt to estimate the CRI and n, while p is determinedas a residual (within reasonable limits). The estimation of n is based onthe perceived value judgements of the Ivorian Government regarding the intra-temporal and inter-temporal tradeoff between consumption groups, while thesize of the CRI will be judged mainly on the basis of whether or not the govern-ment is oriented towards high economic growth. It is of course notoriouslydifficult to determine unequivocally a government's set of value judgements,partly because one has ta distinguish between policy priorities and actualpolicies as determined by numerous constraints under which the governmentoperates in the application of its various economic policy instruments.
62. Beginning with objectively observable policy instruments, one mayfirst consider the income tax in Ivory Coast. The rate structure of anyincome tax reflects not necessarily current policy objectives, but insteadthe cumulative outcome of a whole host of past policy decisions. Furthermore,not only the equity objective determines the rate structure, but also otherobjectives, such as the fear of disincentive effects of progressive rates.However, despite these limitations, which are offsetting each other to someextent, one may infer with some confidence that n is greater than zero fromthe fact that Ivory Coast employs a progressive income tax structure, andthere is some reason to believe that it falls in the neighborhood of unity.l/
63. Second, industrial minimum wages were increased at higher ratesfor lower as compared with higher wage categories in Ivory Coast in theyears 1970, 1973 and 1974. For the lowest category of industrial workers,for instance, minimum wages were raised by 25 percent in 1970 and 1973, andby 20 percent in 1974; while for the highest categories, the incrementsamounted to only 8 percent and 6 percent respectively. From this evidenceone may again infer that n is greater than zero, and falls probably in theneighborhood of unity.2/
1/ For the US, an estimate of n equal to 1.5 from the income tax structure wasmade by K. Mera, "Experimental Determination of Relative Marginal Utilities,"Quarterly Journal of Economics, 83 (1969), 214-252. Fo*r the Ivory Coasta value of n equal to 1.0 was inferred tentatively from very scantyinformation on effective income tax rates.
2/ A graphical presentation of the inverse of the rates of increases atvarious levels of minimum wage rates for 1974 shows a scatter diagramapproximating closely an only slightly concave hyperbola through the origin.This supports the conclusion that n is in the neighborhood of unity.
- 32 -
64. Third, during the early 1970s agricultural pricing policy,particularly for cocoa, coffee, and cotton has stressed the improvement ofagricultural incomes, aiming at preventing a continued increase in therural-urban income gap. Finally, the regional distribution of plannedpublic investment programs favors particularly the South West and theNorth of Ivory Coast, of which the latter is the poorest region in thecountry, and the former attracts migrants from the North.
65. Qualitatively, the Ivorian Development Plan for 1971-1975 andthe Investment Program (Loi-Programme) for 1975-1977 explicitly state thatimproved social conditions and a better income distribution are among thebasic objectives of national development policy. This is reflected in theprogram aimed at improving the regional balance, infrastructure development,housing, health, and education. At the same time there can be no doubt thatthe Ivorian Government continues to pursue a policy oriented towards vigorousgrowth, the objective stated first and foremost in the National Plans for thecurrent decade.
66. As a result of these observations it appears reasonable to placethe value of n between 0.5 and 1.0. The latter value is judged to reflectmore appropriately current government objectives, represented by a balancedpolicy as between the growth and equity objectives. The lower value for nis taken mainly for sensitivity analysis, and allows for the possibilitythat, contrary to its presently stated intentions, the government may infact continue to apply the basic policy mix of the sixties which was gearedpredominantly to achieve a rapid growth rate without paying much heed to theincome distribution objective. Similar considerations suggest that the valueof the CRI falls between 5.0 percent and 7.5 percent. The former value reflectsa continuation of the past emphasis on growth, while the latter is based onthe present policy pronouncements which also appear to give some weight tothe benefits of current consumption. Further, given an estimate of thegrowth rate of per capita consumption kg), the value for the rate ofpure time preference (p) is then implicit (in equation 6). One can estimatethe per capita consumption growth rate for the Ivory Coast from three sources:First, averaging past growth experience, it is found that g was :3.9% p.a.between 1967 and 1972 (cf. Table 18). Second, Bank projections for IvoryCoast GNP and domestic savings growth indicate a projected growth in consumptionof 3.3% p.a.; finally, from the 1971-75 Plan projections, an average annualgrowth in per capita consumption of 4.2 percent p.a. may be derived. Wehave adopted here the low value of 3.3%, especially considering that theimpact of the oil crisis will probably further dampen the growth performanceof the Ivory Coast.
67. This leads us to retain the following combinations of valuejudgements for further consideration:
Table 18: GROWTH IN REAL PER CAPITA CONSUMPTION, 1967-1972
5-Year1967 1968 1969 1970 1971 1972 Average
Private Consumption (CFAF bill) 178.7 200.9 210.4 236.2 261.5 283.0
Public Cons. (CFAF bill) 38.8 42.8 53.4 63.5 74.6 85.3
Total Cons. (CFAF bill) 217.5 243.7 263.8 299.7 336.1 368.3
Population (thousand) 4,586.0 4,738.0 4,890.0 5,065.0 5,232.0 5,405.0
Per Capita Cons. (CFAF) 47,427.0 51,435.0 53,947.0 59,171.0 64,239.0 68,141.0.
Price Index (1960=100) 124.6 131.4 137.1 148.9 147.7 148.2
Real Per Capita Cons.(1960 CFAF) 38-,063 39,144.0 39,349.0 39,739.0 43,493.0 45,979.0
Change in Real P.C. Cons.(1960 CFAF) 1,081.0 205.0 390.0 3,754.0 z,486.0
Growth in Real P.C. Cons. (%) 2.8 0.5 1.0 9.5 5.7 3.9
Source: Bank estimates
- 34 -
Case 1: CRI = 5.0% and n = 0.5, and p = 3.4%Case 2: CRI = 7.5%, and n = 1.0, and p = 4.2%i/
From the evidence reviewed above it appears that the current valuejudgements of the Ivorian Government are more closely representedl byCase 2. However, both cases will be carried through the remaindear ofthis study in order to test the sensitivity of further results tovariations in the value judgments.
The Consumption Distribution Weight, d {S & T,pp.63-65; 102-104; 136-137}
68. The consumption distribution weight, d, for a particularbeneficiary group will usually have to be estimated in the specificproject context where non-marginal changes in consumption are the rule.Only in those cases where it may be assumed that the project results inmarginal consumption changes for people with comparable levels of percapita consumption, can the marginal consumption distribution weight beapplied. Table 19 shows the average per capita consumption levels forfive percentile groups of the population in Ivory Coast and the associatedvalues of the marginal consumption weight.
The Summary Distribution Measure {S & T, p.66-67;LO4; 137-139}
69. The summary distribution measure which is to be applied to minoror non-attributable benefits is defined as:
D = a ( - 1) { S & T,p. 138} (18)(n + a- 1)
where Gini coefficient = 1/(2a- 1). A value of 0.43 for the Gini coefficientwas computed for Ivorian income distribution (1959 data; national coverage)2/.For the South-East Region household expenditure survey data were used here tocompute the Gini coefficient according to the formula:
m-1Gini coefficient = 1- 1 (f i+l- fi) (Yi i+l)
1/ Somewhat higher estimates of p have been derived from private sector con-sumption and savings choices. But it is not clear whether a government orthe Bank would be well advised to follow individual time preferences as aguideline. Presumably, private pure time preference derives from generalizedrisks facing the individual, e.g. death, political instability etc. Forsociety as a whole some of these risks do not apply, in the sense thatsociety as such does not cease existence except under very extreme andunlikely circumstances. On the other hand, different political systemsmay result in different rates of time preference for the policy maker inline with his dependence on popular support.
2/ See Jain, cited in lable 19.
- 35 -
Table 19: MARGINAL CONSUMPTION DISTRIBUTIONWEIGHTS, d, IN IVORY COAST
Percentage Average Per CapitaPopulation of Total Consumption in Marginal 3/Percentile Consumption- Percentile Group Distribution Wleights
(%) (%) (CFAF) 2/ n = 0.5 n = 1.0
0 - 20 7.5 25,553 1.63 2.67
21 - 40 12.0 40,884 1.29 1.67
41 - 60 15.0 51,106 1.15 1.33
61 - 80 25.0 85,179 0.89 0.80
81 - 100 40.5 137,985 0.70 0.49
Total Population 100.0 68,141
Note: 1/ This column is derived from S. Jain, Size Distribution of Incomet; ACompilation cf Data, (Baltimore: Johns Hopkins Press, 1975), byslightly increasing the percentages for the lower income groupsand reducing those for higher income groups. This was necessarysince we assume here that the Gini coefficient of consumptiondistribution is 0.40, while Jain has estimated a Gini coefficientof income distribution equal tô 0,43,
2/ This column is derived by dividing the proportion in totalconsumption (CFAF 368.3 billion) of a particular incomegroup by the proportion in total population (5,405 thousand)of the same group.
3/ The distribution weights are derived by applying the formulad = (B/c)n, where c is the national average per capita incomelevel (CFAF 158,141) and c is the average per capita incomelevel in the particular percentile group.
- 36 -
Table 20: DISTRIBUTIUN UF-HUUSLIduLV EXPEi4bITU-REIN THE SOUTH-EAST REGION, 1963-1964
Cumulative CumulativeTotal Annual Expenditure Population Shares (f.) Expenditure Shares (yi)
per Person (CFAF) (%) (%)
0 - 14,999 11.1 3.6
15,000 - 19,999 30.4 13.7
20,000 - 24,999 44.4 22.4
25,000 - 29,999 57.4 32.6
30,000 - 34,999 71.9 46.4
35,000 - 39,999 77.9 53.1
40,000 - 49,999 87.3 65.5
50,000 - 69,999 94.7 79.0
70,000 and over 100.0 100.0
Source: SEDES, op.cit., p. 33
- 37 -
where f is the cumulatLve population share of the ith observation;
Yi is the cumulat!Lve expenditure share of the ithobservation; and
m is the number oif observations.l/
70. The value of 0.40 for the Gini coefficient is used in thederivation of D, on the grounds that on the one hand national consumptionexpenditures are likely to be less equally distributed than the expenditureof the rural population of the South-East Region, while on the other hand,the national consumption distribution is likely to be more equal than theincome distribution, and thus the value of 0.43 is too high. Using thevalue of 0.40 for the Gini coefficient one can compute D from the formulain paragraph 49: for n 0.5, D = 0.91;and for n = 1.0, D = 1.0.
The Value of Public Income, v {S & T,pp.67-68; 104-106; 140-142}
71. The value of public income is estimated in terms of the con-sumption stream generated by a'unit of public investment, assumingoptimal allocation of public funds. The plausibility of the resultingestimates are then checlced by relating them to estimates of the criticalconsumption level (CCL)., i.e., cha't`level where the government judgesprivate consumption as valuable as public income.
72. Consider firsi: the estimation of the value of public investment.When neglecting reinvescment of the returns to the public sector, the valueof public income and investment may be determined by use of the simple formula
v = q/Si. {S & T,p.106} (19)
Using equation (19) for different combinations of values for q and i, withe = 0.84 throughout, corresponding values can be estimated for v (Table 21).If one lets the range oi- CRI = i be 5% to 7.5%, and lets the range of q be8% to 12%, then the possible values of v fall between 1.27 and 2.86, implyinga premium on public income over average private consumption ranging from27% to 186%. For the central value of q = 10%, v = 2.38 with i = 5%, andv = 1.59 with i = 7.5%.
73. In the more complex case, where it is assumed that the govern-ment and the private sector save and reinvest some of the returns frompublic investment,y the value of public income (investment) may beestimated by using the following equation
V=q - sq T, p.105 } (20)i - sq 6 2/
wnere (1-s) is defined as the portion of q diverted to private consumption.
1/ Taken from Jain, op. cit.2/ Note that equations (19) and (20) are based on the following assumptions:
Public income, public investment, and private savings are all equallyvaluable; all consumxption benefits accrue to the man at the averageconsumption level; and all parameters remain constant over tiine.
- 38 -
Table 21: VALUE OF PU1tLIC INUUME: SlMPLE CAbS
v = q/ei; -0.84
5.0 7.5
8 1.90 1.27
10 2.38 1.59
12 2.86 1.90
The value of (1-s) may be approximated by estimating the private sectormarginal propensity to consume out of total GDP, using historical data,Bank projections, and Five-Year plan estimates. Table 22 presents thehistorical data, which show an average value for (1-s) of 0.57 for the
last seven years; for the last three years it lay at a higher averagevalue of 0.65. The Bank projections in turn provide a lower estimate of
Table 22:.PRIVATE SECTOR MARGINAL PROPENSITY TOCONSUME 1965-1972 (CFAF BILL)
Private .Cpns.Year GDP GNP Consumption Cons. AGDP
1965 236.8 153.71966 257.3 20.5 163.0 9.3 0.451967 274.4 17.1 178.7 15.7 0.921968 325.1 50.7 200.9 22.2 0.441969 364.0 38.9 210.4 9.5 0.241970 414.0 50.0 236.2 26.0 0.52
1971 445.1 31.1 261.5 .25.3 0.811972 480.0 34.9 283.0 21.5 0.62
7 year average 0.57last 3 year average 0.65
Source: Bank estimates
- 39 -
0.55; while Plan data suggest that (1-s) = 0.63. An intermediate valueof 0.60 is here chosen for further estimation uses, which implies thats = 0.40. This value probably represents a low estimate of the truevalue of s, since public investment is likely to generate on averagehigher public income t,hàn other investment activity, given that thefinancial profits from public investment accrue directly ta the publicsector. As a result, the following estimates of v may be on the low sideas well, since q > i. 1/
74. The value of public income may then be estimated using thevalues of s = 0.40, and e = 0.84. Table 23 summarizes the values forv given alternative assumptions concerning q and i. For low values ofi and high values of q, v becomes unrealistically large, and even forthe central value of q = 10% and i .= 5%, the premium on public incomeamounts to 610%. For lower values of q, and/or higher values of i moreacceptable results are obtained, e.g., a premium of 209% for i = 5% andq = 8%, and a premium of 31% for i - 7.5% and q = 8%.
Table 23: VALUE OF PUBLIC INCOME:WITH REINVESTMENT
v = q e l = 0.84i-sq s = 0.40
-ZO \ 5.0 75
8 3.09 1.31
10 7.10 2.03
12 35.50 3.09
75. The low estimate of s notwithstanding, equation (20) probablyoverestimates the true value of v for two main reasons. First, the formulaassumes that all the parameters remain unchanged over time. But second,and possibly more important, contrary to what has been assumed here, publicinvestment and consumption may not be equally valuable. In Ivory Coast itappears that the governnent-is constrained in its use of public funds, andfinds itself forced ta divert more of its funds to current expenditures (inparticular civil servant salaries, etc.) than is thought optimal. This is
1/ Ideally one should cross check the macro-economic estimate with itsmicro-economic counterpart. This was not possible given the availabletime and resources.
- 40 -
witnessed by its stated desire to reduce the rate of growth in civilservant employment and salaries, and to increase the savings component ofthe (semi-) autonomous public enterprises. Moreover, the Ivorian taxlaws specify certain business profit tax remissions, if profits are re-invested in Ivory Coast. This and the explicit policy that public invest-ments are to be increased -/ indicates that the government cannot at presentobtain all the public investment it desires. Thus the value of public income(which must be viewed as being the weighted average of the values of publiccapital and current expenditures) must be below the value of public invest-ment estimated here. If these considerations are taken into account, itappears reasonable to expect that the value of public income actually comescloser to the value estimated by the simple .method first explored. However,before accepting this conclusion, given the uncertainty about the value ofpublic income as derived from macro-economie data, one must first cross-checkthe plausibility of these results by relating them to estimates of the criticalconsumption level. {S & T, pp. 106-1081
76. The critical consumption level (CCL) is that level of per capita con-sumption at which the government regards private consumption just as valuableas public income {S & T, pp. 70; 106-1081. As a first step in the deter-mination of the critical consumption level it will be useful to review theavailable information on the relationship between per capita consumption andpopulation percentiles. It is possible to derive point estimates of populationpercentiles and their associated per capita consumption level by using theaverage consumption value for percentile population ranges found in Table 19to approximate graphically a smooth function relating population percentilesto per capita consumption (see Figure 1). For example, 40% of the populationhave a consumption of less than CFAF 46,000 per head.
77. As a second step one can derive alternative values of public income(v), associated with assumptions concerning the critical consumption level.Table 24 shows different population percentiles and their associated percapita consumption level, as read off from Figure 1. Assuming that the CCLis set alternatively at each of these consumption levels one can compute theimplied value of public income (v) from the condition that at the criticalconsumption level d = v = (C/C*)n, where C* denotes the CCL at the ithpercentile {S & T, p. 107, fn. ±2 }.
1/ Ivory Coast, Five Year Plan 1971-75, p. 10.
- 41 -
Figure 1
The Consumption Distribution Function in the Ivory Coast
Per Capita Consumption (CFAF Thousand)
85
80
70
60
50 4
40
30 2~~~~~~~~~~~
I Average_ ! ~ ~ _ ____ L _ _ .. _j -;
10 20 40 50 60 80 Population Percentile
(%)
Note: The height of the bars indicates the average per capita consumptionlevel for the population percentile range, as read off from Table 19.The smooth function was then derived by hand, by approximatedly makingthe shaded area above the curve equal to the shaded area below thecurve for each percentile range. This derivation procedure is veryrough, and should only be used where detailed information on consump-tion (income) distribution by small percentile ranges is not available.It was moreover assumed that the subsistence consumption level lies atCFAF 10,000 per annum.
- 42 -
Table 24: VALUE OF PUBLIC INCOME (v) GIVEN DIFFERENTCRITICAL CONSUMPTION LEVELS
C * Value of Public Income (v)Population Percentile (CFAF) C/C (n = 0.5) (n = 1.0)
At Subsistence(approx. 0-lst. percentile) 10,000 6.81 2.85 8.11
At Average of0-20th percentile 25,553 2.67 1.78 3.15
At 10th percentile 27,000 2.52 1.73 3.00
At 20th percentile 36,000 1.89 1.50 2.25
At 40th percentile 46,000 1.48 1.33 1.76
At 50th percentile 50,000 1.36 1.27 1.62
78. Next, one may derive the CCLs implied by the values of v
previously computed from the public investment approach. Table 25 showsthe various central values of v, given different assumptions concerningvalue judgements and degree of reinvestment of returns from public projects.The corresponding CCLs are computed from the equation
C = (ve) 1/nC {S & T, p. 107, fn. 12}
and the population percentile levels can be read off from Figure 1, giventhese CCLs.
Table 25: CRITICAL CONSUMPTION LEVELS IMPLIED BYALTERNATIVE VALUES OF v
1/(as estimated from the value of Public Investment)
v C* PercentiLeCase 1
No Reinvestment (n = 0.5, i = 5.0%) 2.38 17,207 4
Case 2(n = 1.0, i = 7.5%) 1.59 50,851 52
Reinvestment Case 1 2/(n = 0.5, i = 5.0%) 7.10 1,918- 0
Case 2(n = 1.0, i = 7.5%) 2.03 39,849 25
Note: 1/ Assuming q = 10%.2/ This consumption level is far below the subsistence level of
CFAF 10,000.
- 43 -
79. Finally, one has to assess whether any evidence exists whichwould permit an independent estimate of the critical consumption level.Consumption subsidies do not appear to be used in Ivory Coast, althoughin the national government budget some minor transfers are recordedunder the heading of "Assistance and Subventions," comprising such itemsas "Help for the Needy," payments to orphanages, etc.l/ These payments maybe assumed to approximate consumption subsidies, and they appear to bedirected mainly at beneficiaries at or near the subsistence level.Considering this information one can return to Tables 24 and 25, whereone finds that for Case 1 and a CCL near subsistence (CFAF 10,nO0a value of v of 2.85 is obtained, while the value of v of 2.38 as deriveafrom the public investment formula without reinvestment is associatedwith a CCL of CFAF 17,207 falling on the 4th percentile level. Thesetwo sets of values are close enough to allow a reasonable consolidationto an intermediate value of v = 2.5 implying a CCL of CFAF of 15,414falling at about the 3rd percentile level.
80. However, there is other evidence which indicates that the govern-ment puts the critical consumption level above mere subsistence, althoughthis may not be reflected in consumption subsidies. For instance, exemptionsfrom income taxes are granted for per capita incomes below CFAF 150,000.Assuming the marginal prcpensity to save equals 25 percent at that incomelevel, the corresponding per capita consumption level, at which the govern-ment is willing to foregco public revenues is CFAF-112,500. If one interpretsthis as the critical consumption level, it would fall at about the 80thpercentile, and one would derive a value for public income below unity.This extreme result may be explained by the fact that tax exemptions forlow incomes reflect not cnly the equity objective, but also such consider-ations as collection costs, disincentive effects, etc. The actual criticalconsumption level is therefore'likely to fall somewhere between the subsistencelevel and the consumptior, level at which income tax exemptions begin to apply.This is confirmed by the fact that public investment programs in Ivory Coastare at least in part aimed at providing consumption benefits to the poorerregions which have, however, an average per capita consumption level abovesubsistence. If one therefore assumes that the critical consumption levelfalls at about the 40th percentile, one obtains by implication a value forpublic income (1.76, Table 24) which falls approximately between the valuesfor v computed from the two alternative investment formulae in paragraph 58(1.59 and 2.03). For purposes of further work v is taken to equal 1.7 whichimplies a critical consumption level of CFAF 47,987 falling at about the 45thpopulation percentile.
81. In summary, one finds that although in the case of the Ivory Coastno reliable precise estimate of the critical consumption level is possible,the evidence supports the previous conclusion that the parameter values ofCase 2 (n = 1.0, CRI = 7.5%) reflect approximately current government value
1/ Direction des Budgets et Comptes, Budget General de Fonctionnement,Gestion 1973 p. V-3, !'Depenses d'assistance et subventions diverses."
- 44 -
judgements. Case 1 (n = 0.5, CRI = 5.0%) clearly represents a lower bound,implying as it does a near subsistence critical consumption level. Byimplication, the value of v = 2.5 associated with Case 1 can safely beassumed to represent an upper bound on the value of public income. Table26 summarizes the parameter values for Case 1 and 2 as well as used in theremainder of this paper.l/
Table 26: VALUE OF PUBLIC INCOME: CONSOLIDATEDRESULTS
i7 n v
Case 1 5.0 0.5 2.5
Case 2 7.5 1.0 1.7
The Accounting Rate of Interest, ARI {S & T, pp. 75-76; 113-115; 142}.
82. The accounting rate of interest is estimated from the following
equation:
ARI = sq + (1-s) q/vs {S & T, p. 114} (21)
Table 27 shows the alternative values for ARI computed from equation (21)for the two cases of v derived in the previous section. Different valueswere also tested for s and q.
Table 27: THE ACCOUNTING RATE OF INTEREST FORDIFFERENT VALUES OF v, s, AND q
ARI(%) for q-l0% ARI(%) for s=0.5v s=0.4 s=0.5 s=0.6 q=8% q=10% !i=12%
Case 1 2.5 6.9 7.4 7.9 5.9 7.4 8.9
Case 2 1.7 8.2 8.5 8.8 6.8 8.5 :L0.2
1/ For the DRC Ivory Coast model, L. Goreux reports shadow prices onpublic funds and additional investments of 20 to 70%. Although some-what lower than the here derived results, they are of the saiae generalmagnitude and also tend to confirm the appropriateness of Case 2. Notethat the DRC model concentrates on evaluating the effects of alternativesavings constraints and considers income distribution only in its inter-regional aspects.
- 45 -
Taking the central value of q = 10%, the minimum likely value for ARI is6.9% in Case 1 and 8.2% in Case 2; the upper bound on ARI is q itself.Considering that s=0.4 is a low estimate, s=0.5 was taken to approximatethe value of s for the marginally-acceptable project, whose internal socialrate of return equals thE ARI. Testing the effect on the ARI of switchingto the higher and lower bound of q, 1/ given s=0.5, one finds that with q=8%,ARI = 5.9% for Case 1 anc. ARI = 6.8% for Case 2. When q=12%, the correspondingvalues for ARI are higher, viz. 8.9% and 10.2% respectively. The intermediatevalues for the ARI for q=10% will be taken as the baseline of comparisonduring the remainder of this study; i.e., ARI=7.4% for Case 1, and ARI=8.5%for Case 2.
1/ Changing q also impl:Les changing v, if the public investment approachis chosen. However, for the present purposes it is assumed that thecritical consumption level estimates of v continue to apply, even asq changes.
- 46 -
V. THE SHADOW'WAGE RATE, SWR
83. In most general terms the SWR may be defined as
So-cial Price = Efficiency Price + C(B-w)> f & T, p. 8:3} (22)
where 6 indicates foreign exchange cost and w indicates the socialbenefit ot the increase in consumption, C. In.this part we will derivethe SWR for four types of labor in the Ivory Coast: Ivorian labor,excluding the urban unskilled, but including all rural Ivorian labor;urban unskilled labor; non-Ivorian rural African labor; and non-African expatriate labor. The Ivorian economy is sufficiently laLrge,and the labor market sufficiently heterogeneous, to warrant thisdisaggregation for the purpose of estimating the SWR.
Ivorian Labor (Excluding Urban Uriskilled)
84. It is here assumed that all Ivorian labor is generally
fully employed year-round._/ The only exception are the urban unskilled,who are discussed separately. Given this full employment assumpt:ion, theoutput foregone elsewhere may be assumed to equal the wage paid 1to theworker in his new position, i.e., he is drawn from a comparable employ-ment sector. The SWR then equals the efficiency price since there is noincrease in consumption resulting from the new job:
SWR = Efficiency Price = nw (23)
where a is the accounting ratio used to obtain the border value ofthe foregone output. For the purpose of the project analysis in PartVI below we assume that the accounting ratio a equals the SCF, andtherefore
SWR/w = SCF - 0.83. (24)
This result is subject to a number of limitations: First, althoughthe condition of full employment may be justified in general, rurallabor requirements in certain areas of the country are subject toseasonal fluctuations; where this is the case, shadow pricing of labormay have to allow for this. Second, there is evidence that wages inthe modern private sector and in the public sector are above marginalproduct, which would imply that the use of w as a proxy for output fore-gone in the employment of labor in these sectors leads to an upwardbias on the efficiency price component. Finally, it is possible thatlabor employed in the industrial sector is drawn from lower prodtuctivityrural employment, in which case again the use of w in the context ofindustrial projects would not be justified. If any one of these ractorsapplies, it is likely that not only the efficiency price componentchanges, but that there is also a consumption change associated with theadded employment, which must be taken into account in the computationof the social price of labor.
1/ This assumption is probably not entirely appropriate for the ruralIvorian labor force in the North. A special shadow wage estimatewould be required for projects using Ivorian rural labor there.
- 47 -
Urban Unskilled Labor
85. The urban unskiLled labor market in LDCs typically involves highunemployment and high rural-urban migration rates, which are generally takento be caused bv a persistent gap in the real income levels between urban andrural inhabitants. However, other factors also play a role, such as theurban labor market structure, the financing of the urban unemployed, andfinally the absorptive capacity of the city population in the case wherethe unemployed are financed by their urban receptors. Numerous alter-native migration models may be formulated and applied to the determinationof the SWR for urban unskilled labor;, 1/ their degree of acceptabilitywill vary depending on the particular socio-economic conditions. Onecharacteristic common ;to many of these models is that they imply a constantrate of urban unemployment, which in turn implies that the migration response,M, i.e., the nimber of migrants who will come to the city as one additionaljob is created, can be expressed as the ratio of total urban labor force, L,to total urban employmient, N:
2/m = L/ (25)
86. Assuming a migration effect of an added urban job as specifiedin equation (15), consider SWR for urban unskilled labor. If wageearners do not save and their disutility of effort is valued at zero,the SWR is
SWR = am + (w'm)(e - d/v), [S & T, p. 84] (26)
where m is the output foregone and a is the conversion factor for theforegone output; in the present context the efficiency price componentin equation (26) has to be modified to allow for the fact that with anadded job in the city M migrants are attracted, reducing agriculturalproduct by amM; and tha consumption component must be adjusted to allowfor the changes in coflsumption indue(ad by the job creation and resultingmigration for each incone group affected. Assuming that the consumptionconversion factors for all consumption groups are identical, one can thenwrite the entire SWR formula as follows:
SWR - amM + d.C (6-d /v) (27)
where ACi is the change in consumDtion of the it consumer group, and
(e- di) the correspondLng weight.v
87. The remaining step is to specify carefully the changes inconsumption resulting from employment and migration. For this purpose itis necessary to use a particular model context, since the consumptioneffects will depend on who finances the urban unemployed and to what
For a discussion of a number of alternative models cf.D.Mazumdar,"The Rural-Urban Wage Gap, Migration, and the Shadow", IBRD 13ank StaffWorking Paper, No. 197.
2/ Ibid.
- 48 -
extent, In Ivory Coast, available evidence on migration motivation,migration flows, and reception practices support. the use of a moclalwhich assumes that the unemployed rural-urban migrants are financedby their urban receptors who share their incomes, y., equally withthe new arrivals; however, they will accept immigrants only as long asthe expected net change in urban incomes resulting from migration isnon-negative. Three consumer groups may then be distinguished: First,the existing urban population experiences a change in consumption(income), to be weighted by (e - d1/v), amounting to (w - Myc), i.e.,the wages earned in the new job, w, minus the consumption of theimmigrants, Myc. Second, each of the M migrants experiences a discreteincrease in consumption from ya,the average agricultural consumption,to yc, weighted by (3 - d2/v). Finally, the remaining rural populationexperiences a change in per capita consumption, weighted by (3 - d3/v),i.e., the consumption of the migrant minus his marginal product lost,multiplied by the number of migrants, M. The three consumption effectscan then be added up with their appropriate weights, resulting in thefollowing equation for the SWR;l/
SWR = an + (w - My)(-l) _ M(y )y)( --2) + M(ya - m)(i3 - d3) (28)c v a - C v v
88. In order to estimate the urban SWR, as specified in equation 28,the parameters were derived from the limited amount of data available in adesk study. Table 29 presents the data base, and the resulting estimatesof the SWR, summarized in Table 28, should be considered as indicative only.
Table 28: SWR FOR URRAN UNSKILLED LABOR (CRAP THOUSAND)
Efficiency SocialPrice Price M
Case 1 SWR 49.4 96.4 )(SWR/w) (0.31) (0.60) )
) 119Case 2 SWR 49.4 95.0 )
(SWR/w) (0.31) (0.59) )
_ It is useful to cross-check the correct specification of the consumptionchange by ensuring that the unweighted elements of the change in con-sumption collapse into the aggregate consumption change (w-mM). Notealso that when using the marginal distribution weights it is assumedthat the changes in consumption are spread equally over the rural andurban populations respectively. An alternative approach would assumethat the consumption gain or loss is restricted to a particular familyunit, in which case one would have to estimate the average number ofmembers per urban and rural family, and then distribute the consumptionchanges on a per capita basis per family. In that case discrete distri-bution weights have to be used.
- 49 -
Table 29: DATA FOR THE COMPUTATION OF SWR
m = 50,000 CFAF p.a.; derived as 250 x 200 CFAF, the dailyaverage wage in agriculture for fieldwork; average product in agriculture isabout 67,000 CFAF, i.e., larger than, andtherefore consistent with the value of m(Bank estimates).
w = 160,000 CFAF p.a.;this is the average wage for non-IvorianAfrican labor in construction, i.e., reflectsthe wage of unskilled and semi-skilled labor(1970 data; from SETEF, op. cit.).
Yc = 60,00E0 CFAF p.a.; interpolation of urban per capita incomesin Abidjan and the Center, 1965 (Bankestimate); this may underestimate the percapita income in the urban sector in 1970,due to inflation and real income growth.
y = 20,000 CFAF p.a.; interpolation of rural per capita GDP inCenter and North (Bank estimate); this alsomay underestimate 1970 incomes for the samiereasoris cited for Yc
U = 80,0C0 (1965); = 140,000 (1970); urban unemployment; SETEF,op. cit.
L = 363,640 (1965);= 519,520 (1970); urban labor force; ibid.
N = 282,640 (1965);= 379,520 (1970); urban employment; ibid.
e = 0.84
a = SCF - 0.83.
Distribution Weights:n = 0.5 n = 1.0
dl: Marginally evaluated at yc 1.07 1.14
d2: For discrete change from Ya to Yc 1.37 1.87
d3: Marginally evaluated at Ya 1.85 3.41
v : 2.5 1.7
Note: No allowance has been made for general price level differencesbetween rural and urban sectors.due to the absence of data.
- 50 -
Two comments are in order concerning these results. First, in thismigration model the overall consumption effect raises the social costof additional urban employment above its efficiency cost. Second, itis accidental that the two cases with different distribution parametersshould have roughly equal social SWRs; offsetting factors do notnecessarily always cancel out, as in this case.
89. Some limitations of this analysis of the SWR of urban unskilledlabor should be noted. On the empirical level, differentials in the cost-of-living between urban and rural sectors and migration costs have beenneglected; the assumption of equal sharing of income among rural andurban families is probably extreme, although it approximates the experiencecommonly observed in Ivory Coast. Finally, the data probably have to beadjusted to be appropriate for a particular project appraisal. On themethodological side, the limitations of the narrow income maximizationapproach to migration are well established; and furthermore, in the caseof Ivory Coast, the fact that urban migration involves both Ivorians andnon-Ivorians must not be neglected. Strictly speaking, the above analysisonly applies to Ivorian rural-urban migrants, but could be extended toincorporate foreign migration.
Non-Ivorian Rural African Labor
90. The SWR of non-Ivorian African Labor drawn into rural employmentin Ivory Coast depends on which of two possible policy alternatives isadopted, viz.a "regional" policy or a "national" policy. The former assumesthat the decision maker tries to maximize welfare in the region as a whole(i.e., including Mali and Upper Volta) through his employment policy; whilethe latter assumes that the decision maker is primarily, if not exclusively,concerned with the welfare of Ivorian nationals.
91. If the regional policy is chosen, the consumption costs (andbenefits) to the whole region resulting from additional employment in IvoryCoast must be allowed for in the SWR. For this purpose two groups shouldbe considered: First, the migrants drawn from the neighboring countries(where the marginal product of labor amounts to only about 40% of the wagepaid in rural Ivory Coast) 1/ increase their earnings, but remit on average40% of their income to their families abroad. The migrant thereforeincreases his consumption by the difference between his retained wage (60%of the wage received) and his consumption before he migrated, which is hereassumed to equal per capita income in a rural family in Mali or Upper Volta.Second, the migrants' families remaining behind face a change in consump-tion, consisting of the increase due to the remittances from the migrants,the decrease due to lost output of the migrants, and the increase due to thedeparture of the migrants leaving fewer mouths to feed. In formal termsthis may be expressed as follows:
Migrant's change in consumption: w - t - yN
Family's change in consumption: t - m + yN
1/ The specification of the analysis draws on the data collected for appraisalof the World Bank Rubber Estate Project in Ivory Coast, and thus applies tothe South-West Region.
- 51 -
where w is the project wage
t is the transfer abroad
YN is the per capita consumption of the non-Ivorian rural family
m is the marginal product of the migrant in his previousoccupation abroad.
The shadow wage for immigrant labor under such a regional policyis then: Il
SWR acm + (w - t - YN)(s -:1) + (t - m + YN) (6 v (29)v
This indicates that the SWR for foreign rural labor consists of the outputforegone abroad, multiplied by the conversion factor which transforms itinto border prices; plus the net consumption cost which results from thechanges in consumption of the migrant and of his family abroad, appropriatelyweighted to express it iin terms of government income. di is the distribu-tion weight given to the consumption change of the migrants -- a discretejump from yN to (w - t) -- while d2 is the distribution weight given to thechange in consumption by the families abroad evaluated marginally at thelevel of consumption YN. This SWR is appropriate if the decision makertreats all residents as if they were Ivorians and is concerned also withthé impact abroad from additional employment in the project. Note that inthis case there is no "excess" migration induced by the project, in contrastwith the assumptions made earlier with respect to urban migration. Notealso that the parameters and conversion factors which are used to weight theconsumption changes were estimated on the basis of Ivorian data only and noton the basis of regional conditions. 2/
92. The alternative, a national employment policy, is concernedexclusively with Ivorian benefits from development. In that case the costof employing an additional foreign worker consists of the amount of foreign
1/ Assuming there is no saving and tliat all the changes in effort arevalued at zero.
2/ If a regional analysis is generally applied, a consistent applicationof the methodology would require that all parameters be derived fromregional, not Ivorian data.
- 52 -
exchange which is lost to the country due to transfers and the cost of hisincreased consumption. If the immigrant's consumption is considered entirelya cost, with a zero weight given to his improved living standard, 1J theSWR of foreign labor under a national policy becomes
SW = t + (w - t) 6 . (30)
93. The SWR as formulated for these two alternative policies may then beestimated, using the parameters presented in Table 30. Table 31 summarizesthe results, which depend of course on the values of v and n that arechosen. Table 31 also shows the efficiency price of labor, i.e., the SWRexcluding consumption costs.
94. The highest SWR is obtained for the national policy alternative,when consumption costs are considered. In fact, the SWR in domestic terms,when divided by the SCF, is above the actual wage rate. Substantially lowerSWRs are obtained for the regional policy alternative. Thus, the regionalapproach encourages use of foreign labor while the national approach treatsforeign labor as more costly than domestic labor and so discourages theemployment of foreigners. Changing value judgements sharply affect the out-come of the regional policy by lowering the SWR with the increase in nand the decline in v (i.e., when proceeding from Case 1 to Case 2). Con-sequently, for Case 1 the regional social SWR lies above the efficiencySWR, while for Case 2 it is lower. This shows that the consumption weightsin Case 1 result in a positive consumption cost, while in Case 2 the consump-tion weights bring about a negative consumption cost (i.e., a benefit) whichpartially offsets the loss in output under the regional policy option, andthe loss in foreign exchange under the national policy option.
1/ This may be viewed as an extreme form of national employment strategy.A less extreme national policy might still attach distribution weightsto the consumption changes experienced by the migrant; i.e., once themigrant is in the country he is treated like a national, while the lossof output and change in consumption abroad are not taken into account.The SWR for foreign labor would then be
SWRN = t + (w-t)o - (w-t-yN) (dl/v). (31)
This policy alternative was also tested and found to yield SWRs betweenthe regional and the "extreme" national cases. Since the range is thusdefined by these two alternatives, the intermediate case of the moremoderate national policy was not further considered in detail.
- 53 -
Table 30; DATA FOR THE COMPUTATION OF THE SWRFOR NON-IVORIAN RURAL AFRICAN LABOR
w = 50,000 CFAF p.a.; derived as 250 x 200 CFAF, the daily wage for projectworkers; Bank estimate;
m - 20,000 CFAF p.a.; derived as 250 x 200 CFAF, the estimated income(monetary and non-monetary) from production peremployee in Upper Volta; Bank estimate;
t/w = 40%; Bank estimate;
YN= 8,333 CFAF p.a.; derived from m, assuming that the average family unitin rural Upper Volta and Mali (as in the Ivory Coast)consists of six members, 2-1/2 of whom are employed;
SCF = 0.83
= 0.84
Distribution Weights: n = 0.5 n = 1.0
d1 : for a discrete change from yN to (w-t): 1.34 1.88
d2 : marginally evaluated at yN: 1.95 3.81
v : 2.5 1.7
Note: These distribution weights are computed for an average regional con-sumption level of CFAF 31,786 (derived from aggregated per capitaconsumption data found in the most recent economic reports on theIvory Coast, Mali, and Upper Volta).
Table 31: Smi OF RURAL NON-IVORIAN AFRICAN LABOR,AS PROPORTION OF WAGE
Regional NationalPolicy Policy
Case 1 0500 0.904
Case 2 0.018 0.904
Efficiency 0.333 0.404
- 54 -
Non-African Expatriate Labor
95. For expatriate project staff the wage paid in domestic: terms maybe taken as the appropriate SWR, if it is assumed that all incomne of expatriatestaff is either remitted abroad or spent on tariff-exempt import:ables.Alternatively, the component of expatriate income which is spent domesticallycould be weighted by an appropriate conversion factor, e.g., the conversionfactor for general consumption, and would then be added to the tnweightedtransfer component. In any case it is probably reasonable to assume thatthe government follows a national policy towards expatriate consumption gains,thus not weighting them as a benefit. Note that the ratio of the SWR to theexpatriate wage falls between the value of e and unity, depending on theproportion of remittances in total income. In any case, it lieEs above theSCF which was used to convert Ivorian labor into foreign terms, and thusfavors substitution of Ivorian for expatriate labor.
- 55 -
VI. IMPLICATIONS FOR PROJECT SELECTION
96. In this part, three World Bank projects which were recently approvedfor Ivory Coast are reviewed by applying to them the efficiency and socialprices which were derived above. The exercise concentrates entirely on re-formulating the economic justification of the projects without attempting toreview the financial conditions or the basic cost and benefit assumptions,except to the extent required by the social pricing analysis. Also, due todata limitations inhereint in a desk study, shortcuts are frequently employedwhere a genuine project appraisal would go into greater detail. It is there-fore not a full reappraLsal of the particular projects to show in retrospectthat they were (or were not) socially justified, but rather an illustrationof the application of proposed methodology and the qualitative results thatcan be expected.
The Grand Bereby Rubber Estate Project, 1973
97. The appraisal report summarizes the Grand Bereby Rubber Estate Projectas followst.
The principal objectives of the project are to increaseand further diversify agricultural production and exports, and toestablish a focus of development in the hinterland of the new portof San Pedro.
The project comprises the first phase of development,1972 through 1979, of a 13,500 ha. rubber estate. In this firstphase all 13,500 ha. would be planted but another 5 years wouldbe required until all the rubber came into production. Fulldevelopment would be completed in 1988 with the final expansionof processing facilities to meet project production needs.
The Grand Bereby estate would be owned by SOCATCI(Societe des Caoutchoucs de Cote d'Ivoire). SOCATCI would bea state-owned company charged with industrial rubber develop-ment in the Ivory Coast. SOCATCI would sign a technical assist-ance and management contract with Michelin, the French rubbermanufacturer, to manage the project which would be SOCATCI'ssole operation at present.
The economic rate of return is estimated at 13.2%. Ina sense this is a "regional" rate of return since about 80% ofproject employees are expected to come from Mali and Upper Volta,countries with poor development and employment prospects, and ashadow rate has been used in assessing the opportunity cost ofthis labor. CostLng labor fully, the rate of return is estimatedat 11.5%.
- 56 -
98. Besides these assumptions concerning the costing of labor, theproject's economic analysis is based on the following considerations:First, the life of the project extends over 41 years, with the first rubbersales occuring in the eighth year. Second, all foreign costs and benefitsare shadow priced at a rate of 1.25 above the official exchange rate. Inthe following analysis we shall use instead the conversion factors appliedto derive border prices for domestic costs. All other costs are chargedfully to the project, except for hospital and school construction, and forhousing costs. The original appraisal omits the former on the assumptionthat the benefits are at least equal to the cost of these social infra-structure works. In our analysis, strictly speaking, one should considersuch costs and benefits explicitly, since the costs are borne by the govern-ment, while the benefits accrue to private consumers, and thus have to beweighted differentially. However, the appraisal assumption was carried overinto the present study, by assuming that the benefits of these works (weightedto reflect the value of public income and the distribution weights) are atleast equal to the costs to the government. 1/ Labor housing costs areshadow priced at 50% in the appraisal, on the grounds that the opportunitycost of the present housing of the labor force is only 50% of that whichwould be constructed under the project. Although this rationale does notseem entirely clear on strict economic grounds, it was also carried overinto the present analysis, mainly in order to preserve the basic coststructure to be used and thus to show how the new methoao±ogy <instead ofchanged costing assumptions) leads to different results from the traditionalBank approach. 2/ Third, the benefits are defined by the appraisal toconsist only of the f.o.b. sales value of rubber produced on the estate,including the export tax levied on it by the government. The same practiceis followed here. Fourth, all taxes (with the exception of the export taxon rubber) are netted out in the computation of the economic rate of return.This,too, is done in the present analysis.
Il Note that the simple approach now often chosen for omitting certainnon-quantifiable benefits (and costs) in Bank projects, which assumesthat benefits at least equal costs, needs to be modified under the newmethodology, since the weight attached to costs and benefits may con-siderably alter their relationship with each other.
21 As with hospital and school construction, housing benefits the privatesector while the costs are borne by the government; thus again, theyought to be weighted differentially. The approach here chosen (follow-ing the original appraisal) would perhaps best be justified by, claimingthat the consumption benefits to labor from its project housing amountto 50% of the cost to the government if properly weighted by the valueof government income and distribution weights.
-- 57 -
99. Since the project agency is a public, government owned organization,which receives its capital from the government and contributes all surplusesto the governnient sector, one can assume that, except for the labor component,all costs and benefits remain in the public sector.I/ The cost and benefitstreams therefore do not have to be weighted, except that it is necessary toexpress them in foreign instead of domestic terms, as discussed in the nexttwo paragraphs. With respect to labor costs, private sector benefits occur inaddition to public sector costs when consumption changes take place as a resultof the employment on the project. This should be reflected in the SWR usedfor evaluating labor co3ts.
100. In the cost and benefit streams of the rubber project as originallyappraised under traditional Bank practices, capital and operating costs arepresented as a mixture of domestic and border prices. Imported inputs areexpressed in border prices, since import tariffs are deducted from the domesticvalue of the commodities; together with all other taxes. But all other tradableinputs (i.e., domesticaLly produced importables and all exportables) and allnon-tradable inputs are expressed in domestic prices. The new methodologywould want to express aLl inputs in foreign terms right from the beginning bytaking border prices fo'r all tradables and applying conversion factors fornon-tradables. In the present exercise, however, it was necessary to rely ona rough approximation which consisted in separating out the foreign exchangecomponent of all inputs according to summary proportionality factors providedin the original appraisal report (on p. 9). The foreign exchange componentwas then taken at full cost while the domestic component was expressed inborder prices by the application of an appropriate conversion factor. Table32 summarizes the necesr,ary data and shows the total project conversion factorwhich was derived for each cost component. These conversion factors are thenapplied to the cost streams to obtain the border price equivalents. For thecosts of management and extension services, the following conversions weremade: For expatriate st:aff full domestic cost was taken. Although this isprobably based on somewliat extreme assumptions it represents a reasonablefirst approximation. For local staff the SCF is applied. Management operat-ing costs are also valued at SCF, while the Michelin contract fee, which is atransfer of foreign exchange to a company abroad, is valued at full domestic(and foreign exchange) cost.
101. This leaves the labor component on the cost side. Twenty percent of thelabor force is Ivorian, and its wage cost is converted into foreign terms by apply-ing the SCF. The remainiing 80% of the labor force are drawn from neighboringcountries, especially U'pper Volta and Mali, and the SWR applied to them dependson whether the employment strategy followed by the government is considered to
1/ This is also based on the assumption that private benefits derived fromhospital and school construction and from labor housing have implicitlyalready been account:ed for according to the conditions specified in theprevious paragraph.
- 58 -
Table 32: CONVEPSION FACTORS FOR CAPITAL AND OPERATING COSTS
ConversionFactor for Total Project
Foreign Exchange Domestic ConversionInputs Component (%).a/ Component Factor Il
Equipment and Materials(incl. Maintenance) 35 SCF 0.83 0.89
Vehicle Capital 85 0.796_/ 0.83
Vehicle Operation O SCF n 0.83 0.83
Buildings, Construction, etc. 45 Const. Conv.Factor=0.772/ 0.87
Factory Capital incl. Spares 100 1.00
Seeds, Fertilizers,Insecticides 35 SCF = 0.83 0.89
Electricity, Water O SCF = 0.83 0.83
1/ Computed by suimming the foreign exchange component with the product ofthe domestic component and the conversion factor for the respectivedomestic component; applicable only to this particular project's coststreams.
2/ The use of the general construction conversion factor for the domesticcomponent of construction inputs can be taken only as an approximation.Two offsetting factors are at work: First, if labor is priced at fullcost in domestic terms (equivalent to an SWR of 0.83), the conversionfactor for the domestic component lies above the conversion factor forthe entire industry, since labor constitutes a larger proport:ion indomestic than in total construction costs. To take an extreme case,assume that labor makes up the entire domestic cost component (in factit only accounts for about 26% according to I-O information), the project'sconstruction cost stream would have to be adjusted by a factor of 0.91instead of 0.87. But second, if labor is shadow priced below full costin domestic terms (because of urban unemployment or immigration fromabroad), then the project's construction conversion factor would belower than 0.91, even if labor made up the entire domestic cost component.For instance, with labor shadow priced at 0,70, the total conversionfactor for the project's construction cost stream would be 0.84. Thus0.87 may be taken as a reasonable intermediate value.
Sources: a/ Bank estimatesb/ Table 8.
-. 59 -
be regional or national. The total wage bill may then be shadow priced byadding the foreign labor component to the domestic share of labor:
SWR - w x SCF(I/].) + SWRN(N/L) (31)
where I and N are t'he number of Ivorians and non-Ivorians in the totalproject labor force L, and SWR is the shadow wage applied to the non-Ivorians (see Table 31). The comptutation of the labor cost stream assumedthat the conversion factors and SWRs remain unchanged for all years. Forthe conversion factors this implies constancy of the consumption patternsand tariff barriers; while for the SWRs it is based on the assumption thatdecreases in value of public income are offset by increases in the wage rate,the opportunity cost of labor or in the proportion of wages which migrantssend abroad.L/
102. All converted costs streams can be added together and compared with
the stream of benefits, which is derived simply by taking the sales of rubber,including the export tax when levied; conversion of the original projectappraisal figures is not necessary. From these cost and benefit streamsthe internal rates of r-eturn (R/R) can be computed, for different policyassumptions, and compared with the corresponding ARIs (see Table 33).
Table 33: GMAND BÉREBY RUBBER ESTATE PROJECT:INTERNAL RATES OF RETURN (IN %)
Regional NationalPolicy Policy ARI
Case 1 12.8 11.4 7.4
Case 2 14.6 11.4 8.5
Efficiency 13.4 13.2 îo.0oi
Bank Apprai:sal 13.2 - 10-12
Note: 1/ This is the value of q.
1/ Such assumptions are recommended by Little and Mirrless, op. cit., p. 283.
- 60 -
103. Looking at Table 33 the most important conclusion is that the pro-ject passes the R/R test quite easily under any of the policy combinationshere considered. In the original appraisal the project was marginal in thesense that its R/R (13.2% with shadow pricing) lay only little above theopportunity cost of capital suggested by the region for use as a discountrate in the Ivory Coast (10-12%). In other words, a project which undertraditional Bank appraisal practices was marginal turned into a non-marginalproject, especially if the regional policy option is considered appropriate.The explanation for this result is that the project produces almost exclus-ively public sector income which has a relatively high social value in IvoryCoast (v = 2.5 in Case 1 and v = 1.7 in Case 2). This factor was not con-sidered in the original. project evaluation. Note also that although theabsolute levels of the R/Rs of the project are quite close to the originalR/R in this particular case, the greater acceptability of the project ismainly due to the lower value of the ARI, i.e., the cut-off rate.
104. Changes in value judgements cause little change in absolute levelsof R/Rs in this case, since they affect only the SWR, and since labor costconstitutes only a small part of the total cost stream. Only if a significantproportion of the net benefits goes to the private sector will the change invalue judgements affect the project outcome appreciably. The actual differ-ences in R/R between the various cases and policy options can be explainedentirely in this case by the differences in the SWRs applicable to each case(see Table 31).I/ The higher the SWR the lower is the R/R. This means thatthe regional policy makes the project appear more attractive than the nationalpolicy; the net benefits of the migrants and their families are valued in theformer case, while only the cost of increased migrant consumption is accountedfor in the latter case (in addition to the foreign exchange cost of remittances).Since the consumption cost is not accounted for with efficiency pricing, theR/R with social pricing is below the R/R for efficiency pricing in the case ofthe national policy. For the regional policy option the efficiency SWR liesbetween those of Case 1 and Case 2, and accordingly the efficiency R/R alsohas an intermediate value.
The Third Highway Project, 1972
105. The Third Highway Project in the Ivory Coast consists of (a) theimplementation of a four-year highway maintenance and betterment program(1973-76); (b) the strengthening of about 110 km. of paved roads, includingpartial realignment and widening; (c) the construction of the bridge on theBoubo river; and (d) the study of urban infrastructure development in
1/ The slight difference between the R/R for efficiency pricing and the R/Rof the original project appraisal is caused by two offsetting factors;the lower SWR in the former case tends to raise the R./R, while the lowerimplicit SER tends to lower it. On balance these two effects almostcancel each other.
- 61 -
Abidjan..!/ Only the first of these four components is reviewed here. It inturn consists of two separate parts, a maintenance program, and a bettermentprogram. As far as one can tell from the original appraisal report, noshadow pricing was used, neither with respect to foreign exchange nor withrespect to labor.
106. In the present exercise all domestic costs are converted intoforeign terms by applying the standard conversion factor. This is a simpli-fication; in a full appraisal it would be pLeferabie to use border pricesfor all tradabLes and to decompose all non-tradables. Here, however, decompo-sition stopped at the local and foreign cost components shown in the originalappraisal report. The latter are already expressed in border prices, sinceall tariffs were netted out from the imported inputs.
107. On the benefit side, the savings in vehicle operating costs, whichin the appraisal report are expressed as private benefits in domestic terms,have to be translated inito the numeraire, i.e., uncommitted foreign exchangein the hands of the government. The reduction in operating costs, consistingmainly of reduced fuel, oil, and tire use and in reduced automobile deprecia-tion, provides a saving to the government in terms of foreign exchange equi-valent to S(1-t), where S is the dcomestic (tax inclusive) value of the costof savings, and tS is the proportion of cost savings accounted for by thetax component in vehicle operating costs.A/ In addition, one has to accountfor the increase in consumption resulting from the reduction in road usercosts and the associated increase in private disposable income. The cost ofthis increase in consumption may be expressed as S(,3 - D/v), if we assumefirst, that all increases in disposable income are actually consumed, andsecond, that the cost savings are distributed according to the existing incomedistribution in the country. Whether these two assumptions are acceptablewill be reviewed in some more detail below. Combining these elements, i.e.,foreign exchange savings to the government and (net) consumption costs, thetotal benefit in any year may be formulated as follows:
B = S(l-t) - S(e - D/v) = S(1 - t + D/v) (32)
l/ The foreign exchange component of the two programs combined amounted to51.3% of total project cost.
2/ Using (1-t) to obtain S in foreign exchange terms, it is implicitlyassumed that S consists only of imported commodities. This shouldbe close to actual fact.
- 62 -
108. From pre-appraisal data it was established that the average valuefor t in the Ivory Coast is 0.229.1/ 8 was previously estimated as 0.84.This leaves the determination of D/v. D was estimated as 0.91 where n = 0.5(i.e., Case 1) and 1.0, where n - 1 (i.e., Case 2). The value of v is 2.5in Case 1 and 1.7 in Case 2. Multiplying the estimated value of savings inprivate vehicle operating costs (S) with (i-t-e + D/V) zives the bp.nefit streamsfor the two cases. The benefit stream on the basis of efficiency pricing isidentical to that presented in the original Bank annraisal.
109. The R/R derived from these various cost and benefit streams areshown in Table 34.
Table 34: THIRD HIGHWAY PROJECT: RATES OF RETURN(IN %)
Routine BettermentMaintenance Works ARI
Case 1 37.1 -1.7 7.4
Case 2 >100 26.2 8.5
Efficiency >100 61.2 10.0
Bank Appraisal > 50 50.0 12.0
It suggests that the maintenance component of the project remains clearlyacceptable under all value judgement combinations here considered. Thebetterment component, however, becomes non-acceptable for Case 1 assumptions(n = 0.5, CRI = 5%), although it remains acceptable under Case 2 assumptions(n = 1.0, CRI = 5%). Since we concluded earlier that Ivorian government valuejudgements are closer to Case 2, both components remain acceptable, althoughonly marginally in the case of the betterment works. However, before accept-ing this conclusion we should reconsider the two assumptions made aboveconcerning the income distribution impact and private sector savings be-havior.
1/ SETEC, Ivory Coast Transportation Survey, Vol. I, General Report, Phase I,July 1969.
- 63 -
110. First, the above estimation assumed that the effects of reducedoperating costs were spread so as to leave the existing income distributionunaffected. A full prcoject appraisal should try to determine project bene-ficiaries in greater dUtail than can be done here; but as a first step, itmight be argued that, since the road improvements under the Third HighwayProject take place pred'ominantly in the rural sector of the country, thecost savings are going to be distributed mainly to the poorer rural inhabit-ants. Assuming the population benefitting from the project has a per capitaconsumption of some 75% of the average national level, d can be estimatedto equal 1.15 for Case 1, where n = 0.5. Using this value the benefitstream can be recomputed. The assumed change in beneficiaries raises theinternal rate of return. from -1.7% to 10.7%, i.e., a project which origin-ally was not acceptable, given Case i value judgements becomes acceptable withthe revised assessment of who benefits from the project. The proper specifi-cation of the distributional impact of a project can therefore be vital toits acceptability, especially when it is on the borderline and when a largeproportion of the benefits goes to the private sector.l/
111. Second, the assumption was made above that no part of the increasein disposable income resulting from :Lower operation costs will be saved.Especially since these benefits were assumed to be distributed proportionallyit is reasonable to assume that the beneficiaries have a positive marginalpropensity to save, probably equal to the marginal propensity to save out ofdisposable income in the private sector. The data available in Bankeconomic reports do not permit an estimation of this parameter, but Bankprojections show an overall domestic marginal propensity to save out of GDPof 18% for 1970-75. If we assume that the private marginal propensity to con-sume is lower than that of the public sector, the former might then be esti-mated, say at 10%. Adjusting the benefit formula on that basis to includesavings by private beneficiaries,2/ one can recompute the benefit streams.For Case 1 of the betterment works the rate of return is increased from-1.7% to 4.6% and the project remains non-acceptable, since the ARI is 7.4%.However, the result shows that the correct specification of the privatesector savings propensity may have an important impact on the rate ofreturn to a project, particularly where a significant proportion of thebenefits is reaped by the private sector.
1/ This raises of course the larger question of how to assess expenditurebenefit distribution.
2/ The annual benefit may then be defined as: B = S (i-t) - S(l-s)(8 -D/v)where s is the private marginal propensity to save. Note that thisformulation assumes that private savings is as valuable as publicincome.
- 64 -
The Cocoa Project, 1970
112. The project consists of the new planting of about 19,000 ha, ofcocoa and the rehabilitation of about 38,000 ha, of cocoa not yet in fullproduction. Since the economic analysis of both components is essentiallyidentical, only the new planting is here reviewed. The project is carriedout on small family farms and includes provision of improved seeds; creditto famers for seasonal inputs and equipment and the cash required for hiredlabor; extension services; training facilities; a warehouse; vehicles; pro-specting and survey and a study for a second phase cocoa planting program.New planting (and resulting cocoa production) proceeds in four successivecycles, starting at yearly intervals during the first four years of theproject with 3,790 ha. in the first year, and additional 4,820 ha. in thesecond, 4,900 ha. in the third, and 5,320 ha. in the fourth. The originalproject appraisal justified the project on the grounds of substantial foreignexchange earnings and increases in incomes of participating farmers. Theeconomic rate of return, without shadow pricing of labor or foreign exchange,was estimated at 20% for the new plantation component.i/
113. The project thus involves private and public costs and benefits:The private farmer?s costs consist of on-farm cash costs including hiredlabor, plus the opportunity cost of foregone income through the employmentof family labor in the project. The farmer's benefits consist of the saleof his products at the government controlled price. Public costs consist ofthe government's initial project costs and its subsequent administrativecosts; public benefits result from the difference between the price whichthe government pays the farmers for their product and the price at which thegovernment can sell the product on the world market.!/ Further complicationsare that the government provides the farmers with subsidies during the yearof planting and extends credits to them to cover on-farm costs during theinitial years of the project, to be repaid later.
114. In order to evaluate the sectoral costs and benefits, one shouldassess the resource costs and benefits to the country resulting from theproject, as well as the costs and benefits arising from the induced consump-tion change. The resource cost consists of the sum of private and publicexpenditure on project inputs (including imputed family labor costs), modi-fied by the appropriate conversion factor to express them in terms of foreignexchange. The consumption cost (or benefit) consists of the change in con-sumption of the private sector (here set equal to the net project incomes to
1/ The project appraisal report is reviewed and partially reprinted inGittinger, Economic Analysis of Agricultural Projects (Johns HopkinsUniversity Press, 1972).
2/ The Government benefits may actually be further split into the receiptsfrom the export tax, which go to the central government directly, andthe receipts of the CSSPPA (a price stabilization fund), a public agency.
- 65 -
the farmers, assuming that marginal propensity to save to be zero), multi-plied by the appropriate weight, i.e., (e - d/v). The resource gains tothe country;consist of the increase in sales at border prices. Subsidiesand credit payments and repayments need not be further specified in thisparticular formulation or the problem: The resource cost computation takesthem implicitly into account by valuing the entire private on-farm costs asa cost (even though part:Lally paid for by a governnent transfer); creditpayments and repayments are accounted for in the effective change in incometo the private sector and hence-in the ensuing (net) consumption cost to thegovernment in terms of government income.i/
115. Following this approach one can convert the cost and benefitaccounts of the original appraisal into the terms required by the new method-ology. On the benefit s-Lde, the sale of cocoa produced by the project isalready evaluated at border prices and needs no adjustment. On the costside, however, the original input data have to be transformed into borderprices; as previously, the adjustment is made by taking the foreign exchangecomponent at full cost s-ince all import tariffs have been subtracted fromimported inputs.2/ Impul:ed family labor cost is converted into foreign termnsby simple application of the SCF. Total resource cost consists of the valueof all project inputs regardless of whether they are paid for by the publicor the private sector. The original project appraisal includes governmentcredits under private on--farm costs, and presents private on-farm costsinclusive of the credit element, but not of first year subsidies; these there-fore have to be added heretto private on-farm costs, in order to obtain totalresource cost.
116. Computation of the consumption cost (and benefit) associated withthe project is complicated by the fact that the original appraisal does notgpecify the expected number of participating families, i.e., the number ofbeneficiaries. Nor doees it set up an overall income account for the familyunit which may be used to compute the changes in income per head. Insteadall computations are set up on the basis of per hectare unit cost and bene-fits. Consequently, a number of intermediate steps must be taken to derivethe consumption cost of,t:he project. But first, to facilitate this derivation,assume that the government keeps the farm family's income (with the project)from falling below the incoôme which the family could have earned without theproject, by applying an appropriate-subsidy and credit policy during the
1/ Note that here as in the previous cases, strictly speaking only the tariffcomponent has been netted out, while the distortions of domestic fromborder prices due to quantitative restrictions have not been allowed for.The implicit: assumption is therefore that none of the inputs are subjectto quantitative restrictions.
2/ The proportionality factor of domestic cost in total cost used for thispurpose is 56% for public, and 44% for private costs; it is purelyaccidental that these two figures add up to 100%.
- 66 -
initial years of the projectJI/ In other words, the family has a constantincome up to the point where the cash returns from the project begin toexceed the sum of' on-farm costs and the opportunity cost of family laborworking on the project; after this point, there are no credits and subsidiesgiven and family income and consumption rise with the increase i.n net cashrevenue. This increase in family income can then be derived from the perhectare data provided in the appraisal report by going through the followingsteps.
117. Step 1 computes the number of family workers associated with ahectare of project !and. Assuming that the entire family labor force isemployed in the project in the peak labor demand year (the second year ofeach of the four production cycles), the total number of family workers perhectare is derived by dividing the per hectare family labor cost for thatyear by CFAF 62,500, i.e., the annual per worker labor cost.2/ Thus, onefinds that in year 2, 0.529 family workers are associated with the projectper hectare. Step 2 derives the number of family members per hectareinvolved with the project, assuming that the ratio of family workers tofamily members is 2.5/6 = 0.417.3/ Hence the number of family members perhectare of project land is 0.529/0.417 = 1.269. Step 3 computes the percapita value of net private profits for each year, by dividing the per hectarenet private profit figure by 1.269. All this assumes implicitly that the sizeof the family varies in proportion with the size of the farm, which appearsjustified in the case of the Ivory Coast, where "land within community bound-aries is allocated to farmers by the Chiefs according to each family's laborpotential." (IBRD Report No. Pa-41a). Step 4 computes the value of thedistribution parameter for the increases in income each year using thenational average consumption level (CFAF 68,141) and the without-project percapita income (CFAF 26,042) as benchmarks for comparison. Step 5 then computes( - d/v) for each of the four planting cycles, where e equals 0.84, and vis set equal to 2.5 in Case 1 and equal to.l.7 in Case 2. Step 6 derives theannual consumption cost for each production cycle by multiplying total annualnet private profits by the appropriate weight (3 - d/v). Step 7 obtainstotal consumption cost for each year by summing'across all four cycles.
1/ Note that in a full appraisal a careful specification of the farm familybudget in line with actual government subsidy and credit policy would benecessary; it could show that during the first years of the project thefarmers actually experience a drop in their total income with the project,as compared to a situation without the project.
2/ Derived as CFAF 250 per day for 250 days per year.
3/ This ratio is based on Bank estimates for the average rural family.
-- 67 -
118. The economic rate of return can then be computed in terms ofefficiency prices; and the social rate of return on the basis of socialprices, including conswunption costs. Two cases are here considered: Case1 (n = 0.5 and CRI = 5,0%)Iand Case 2 (n = 1.0 and CRI = 7.5%). Since forthese two cases the value of ( - d/v) differs, different consumption coststreams are obtained. Note that the consumption cost stream for Case 1 ispositive, i.e., the change in consumption results in a cost to societyexpressed in terms of uncommitted foreign exchange in the hands of thegovernment, while for Case 2, consumption costs are negative, i.e., theincrease in consumption is valued as a benefit to society. The reason forthis is that the value of (~- d/v) is positive in the former, but negativein the latter. In otheir words, for the low value of n and the high valueof v (Case 1) the beneficiaries of the project lie above the criticalconsumption level, while they lie below this level in Case 2, when n takeson the higher value and v the lower value, respectively.
119. Table 35 sumimarizes the rate of return results for the two casesof social pricing, for efficiency pricing, and for the original projectappraisal.
Table 35: COCOA PROJECT: RATES OF RETURN(%)
Internal Rate Discountof Return Rate
Case 1 21.3 7.4
Case 2 25.1 8.5
Efficiency 23.8 10.0
Bank Appraisal 19.9 10-12
The rate of return in the original appraisal is lower than the R/R withefficiency pricing since the original project appraisal did not shadow priceforeign exchange, as is done in the case of efficiency (and social) pricing.Comparing the social returns with the efficiency returns, one finds thatCase 1 yields a lower return, due to the positive consumption cost; whileCase 2 yields a higher return, due to the consumption benefit (i.e., negativeconsumption costs). Relative to the ARI, however, both social returns aremore favorable than the efficiency return, since the benefits which are notretained by the government accrue to a relatively poor segment of thepopulation.
- 68 -
120. The impact of social pricing on the project returns is not very
large, given the assumptions made here concerning the (net) consumptioncosts. Two factors in particular contribute to reduce the impact of theconsumption cost element: First, the assumption that consumption dloes notchange during the initial project years, but is effectively maintained by thegovernment at its pre-project level, reduces the influence of actual con-sumption changes during these years. More detailed knowledge of the actualtime path of gains and losses in consumption might give a different coutcome.Second, and more generally, the consumption cost is quite sensitive to changesin the value of d, which in turn depends on assumptions made with respect tothe consumption levels of the beneficiaries. Given our particular assumptions,the weighted consumption costs (benefits) are relatively small, because thevalue of d/v is fairly close to the value of S. In other words, thebeneficiaries happen to be close to the critical consumption level at whichgovernment values public and private income and consumption equally. In thatcase social pricing does not produce significant differences as compared withefficiency pricing, where by definition all benefits and costs are valued equally.However, a more detailed project appraisal might conclude that the beneficiariesare actually better (or worse) off than here assumed, in which case social pricingwould produce quite different results from those of efficiency pricing.
121. The calculations presented in this exercise can only be taken as anindication of the kind of information required by the project economist forhis social pricing computations. They should nàt be taken as a model withsufficient degree of accuracy for actual project appraisal, where it may notbe too difficult to determine the income and consumption levels and expectedchanges therein resulting from the project in greater detail than was herepossible.!/ More detailed work than presented here would be required of theproject economist also in the derivation of the border price equivalent ofproject costs. The method used here throughout is acceptable only as afirst approximation for a desk study. Since the project economist disposesof more detailed cost data and can ascertain border prices directly in thefield during the appraisal mission (or have them ascertained by thie consult-ants) this should not cause great problems.
1/ In fact it may be argued that it would be good practice to make such com-putations in any case. It may be important for the success or failure ofa project to what degree the private farmer incurs losses init:ially whichhe may find difficult to sustain; or to what degree he is faced with risks,due to initial losses, which he may not want to shoulder, thus resultingin unwillingness to participate in the project. In the absence of calcu-lations of private cost and benefit streams on a per capita basis andrelative to non-project related income, Bank project appraisals in thepast have had no factual basis for judgement in this respect.
PUB HG3881.5 .W57 W67 no.253Linn, Johannes F.Economie and social analysis
of projects : a case studyof Ivory Coast /
PUB H C38 8 1,5 . 5 6 1 . 5
Linn Joann W57 W67 no.253Linn, Johannes F.Economlc. and Social analysi,0f pr>jec : a case study
Of Ivory Coast /