Analysing the Distributional Impacts of Stablisation Policy with a CGE Model: Illustrations and Critique for Zimbabwe

Transcription

1 Analysing the Distributional Impats of Stablisation Poliy with a CGE Model: Illustrations and Critique for Zimbabwe Sonja Fagernäs Eonomi and Statistis Analysis Unit April 2004 ESAU Working Paper 4 Overseas Development Institute London

2 The Eonomis and Statistis Analysis Unit has been established by DFID to undertake researh, analysis and synthesis, mainly by seonded DFID eonomists, statistiians and other professionals, whih advanes understanding of the proesses of poverty redution and pro-poor growth in the ontemporary global ontext, and of the design and implementation of poliies that promote these objetives. ESAU's mission is to make researh onlusions available to DFID, and to diffuse them in the wider development ommunity ISBN X Eonomis and Statistis Analysis Unit Overseas Development Institute 111 Westminster Bridge Road London SE1 7JD Overseas Development Institute 2004 All rights reserved. Readers may quote from or reprodue this paper, but as opyright holder, ODI requests due aknowledgement. ii

3 Contents Aknowledgements Aronyms iv iv Exeutive Summary Purpose Model and data desription Poliies and results Conluding remarks Chapter 1: Introdution 1 Chapter 2: The Model What is a CGE model? The IFPRI standard model 4 Chapter 3: Applying the Model to Zimbabwe: Bakground Zimbabwe s eonomi situation Data 10 Chapter 4: The Impats of Stabilisation Poliies: Theory 13 Chapter 5: Previous Researh with IFPRI Model 15 Chapter 6: Analysis of Simulation Results Devaluation Fisal poliy Combinations Summary 29 Chapter 7: Another Model for Zimbabwe 30 Chapter 8: Conlusions 32 Referenes 34 Appendix 1: Some key model equations 36 Appendix 2: Zimbabwe Soial Aounting Matrix (Zim$ billion) 38 Appendix 3: Additional Poliy Experiments 41 Tables Table 1: Eonomi struture of Zimbabwe, Table 2: Key elastiities 12 Table 3: Base ase (Zim$ billion) 17 Table 4: Devaluation 18 Table 5: Redution in government onsumption (by 9%) 21 Table 6: Inrease in diret tax rates (13%) 23 Table 7: Redution in government onsumption ombined with devaluation 26 Table 8: Inrease in diret tax rates ombined with a devaluation 27 Table 9: Summary 29 iii v v v vi vii

4 Aknowledgements This work was arried out as part of an ESAU projet on exploring the pathways to eonomi reovery in Zimbabwe. I would like to thank John Roberts, head of ESAU, for invaluable ideas and omments as well as enouraging me to finish this work. Speial thanks also go to Dr Jeffery Round for onstrutive peer review omments and Jane Northey for administrative support. Aronyms AfDB Afrian Development Bank CES Constant Elastiity of Substitution CGE Computable General Equilibrium CET Constant Elastiity of Transformation CPI Consumer Prie Index EIU Eonomist Intelligene Unit GDF Global Development Finane (World Bank) IDRC International Development Researh Centre IFPRI International Food Poliy Researh Institute IMF International Monetary Fund LES Linear Expenditure System MPS Marginal Propensity to Save OECD Organisation for Eonomi Cooperation and Development PSIA Poverty and Soial Impat Analysis SAM Soial Aounting Matrix WDI World Development Indiators (World Bank) iv

5 Exeutive Summary Purpose This paper looks at the use of a standard, stati omputable general equilibrium (CGE) model (IFPRI model) for analysing and omparing the distributional impats of different maroeonomi stabilisation poliies. The primary objetive is to improve the external balane of an eonomy in risis. The paper disusses how this an be ahieved in the ontext of the partiular model and identifies the limitations of the model. The ountry analysed is Zimbabwe, whih is faing deep eonomi and politial risis and is in aute need of maroeonomi stabilisation. Real GDP is in a ontinuous deline and inflation in 2003 hit over 600%. Nearly all setors of the eonomy have experiened a sharp downturn, agriulture in partiular beause of the land reform. The Zimbabwe dollar has been grossly overvalued, whih has led to the emergene of a large parallel exhange market. Donor support and apital inflows have ome to a halt and there is an aute shortage of foreign exhange. CGE models have been used for several deades to analyse the effets of poliy hanges and eonomi shoks. They are equilibrium models that over the hanges that arise from a move from one market equilibrium to another as a result of a poliy shok. CGE analysis and eonomi modelling have reently ome to the fore with the inreased advoay of Poverty and Soial Impat Analysis (PSIA) of eonomi poliies and reform in developing ountries. As studies using this methodology may inreasingly influene poliy, there should be general awareness among poliy-makers of the advantages and disadvantages of this tool and the situations for whih it is best suited. The paper onentrates on the following poliies: exhange-rate devaluation (expenditure swithing) and redution in government onsumption or an inrease in diret tax rates (expenditure redution). Usually both poliies an be used to alleviate an external imbalane. The paper formulates poliy reommendations, where possible by omparing alternative poliies. However, it also funtions as a ritique of features as the model arising from restritive assumptions. Expenditure redution alone does not improve the urrent aount balane in the model framework; this poliy does not therefore onstituted a omparable alternative to devaluation, and its distributional effets an be misleading. The paper therefore also looks at the impliations of ombining a devaluation with different expenditure redution poliies to arrive at more balaned outomes. By examining the results of speifi poliy hanges, it identifies the weaknesses and strengths of the model and assesses the reliability of the distributional impats. It also briefly disusses possible alternative assumptions. Model and data desription The model used in this study is the IFPRI standard CGE model. This is a stati, real eonomy CGE model that is mainly neolassial, but an inorporate a ertain degree of market rigidity. In the model, poliy shoks ause hanges in relative pries. Prodution is arried out by ativities that maximise profits subjet to a prodution funtion and given pries of inputs, outputs and fators. Total output is divided between domesti sales and exports, based on revenue maximisation. There is zero substitutability between fators of prodution (value-added) and intermediate inputs in prodution. Consumers maximise their utility (funtion of onsumption), subjet to given inomes and pries. Domesti goods and imports are imperfet substitutes. In CGE models maro losure rules govern the mehanism by whih the maro eonomi balanes the fisal balane, the urrent aount balane and the savings and investment balane are ahieved. In this study, government savings adjust to balane the government aount, and v

6 diret tax rates and government onsumption are fixed. We apply two different savings-investment losures. In the first ase savings are investment-driven (I) and marginal propensity to save (for private savings) adjusts to maintain the balane between savings and investment. In the seond ase investment is savings-driven (II) and marginal propensity to save is fixed. There are three savings aounts: private, government and foreign. Savings are a funtion of a number of fators, but investment in this model depends simply on total savings. For the external balane we use the losure, in whih the real exhange rate is fixed and foreign savings adjust to maintain the balane. We use the IFPRI 1991 soial aounting matrix (SAM) for Zimbabwe as the base, but with a few alterations. The number of eonomi ativities has been aggregated to 7 and the value of prodution for eah ativity and thus their value-added shares of total output have been adjusted in the light of events sine There are 5 household groups: large-sale farm owner and manager households, large-sale farm worker households, smallholder farm households, urban high-inome households and urban low-inome households. Other institutions are enterprises (non-agriultural), government and the rest of the world. Poliies and results Standard eonomi theory suggests that there are at least two ways to onfront a urrent aount defiit: expenditure swithing and expenditure redution poliies. The latter will also diretly target the fisal balane. The aim of expenditure swithing is to produe a diret impat on exports, for instane via hanges in the exhange rate or tariffs. Devaluation funtions as a tool to boost exports and alleviate the hroni shortage of foreign exhange in the eonomy. The impat of expenditure redution on the urrent aount an be more indiret, but by means of a fall in the prie of domestially marketed goods it an improve export inentives and thus the urrent aount balane. Expenditure redution an also enhane the hanes of ahieving other poliy objetives by improving the budget defiit and restoring redibility. Often the two types of poliies need to be implemented simultaneously to ensure suessful stabilisation. The first experiment onduted is a real exhange-rate devaluation (inrease in domesti pries of tradable goods and foreign transfers). The results suggest that devaluation is likely to be ontrationary. Depending on our assumptions about investment, either private onsumption or investment will fall substantially. GDP ontrats, agriulture and export setors benefit, but others lose out. The same holds for households: real inomes rise for agriulture-related households and fall for urban households, espeially when we assume more market flexibility. The effets of a devaluation are somewhat extreme sine, beause of the nature of the maroeonomi balane rules, the adjustment to a hange in the urrent aount balane mainly takes plae via a rise in private savings or a fall in investment. However, they are still quite plausible. The results suggest that, in Zimbabwe, the devaluation should be omplemented with a poliy that both improves the fisal balane, whih otherwise worsens slightly, and redues some of the negative distributional impats on the non-export setors. However, the devaluation an lead to a higher rise in the prie of neessities (e.g. fuel) than an be demonstrated by this model. Seondly, the apaity of the export setors to respond to the devaluation may, at least in the short run, be limited. We then experiment with a few expenditure redution poliies. Expenditure redution may not have as large an impat as devaluation on the urrent aount balane, but we ould expet a small improvement. However, this does not our in the model. The ritique in the paper has foussed on the fat that standard expenditure redution poliies may not alleviate a urrent aount defiit in the IFPRI model. The resulting maroeonomi responses are unrealisti and ast doubt on the model s distributional preditions. The results suggest that the model does not allow higher taxes or lower government expenditure to exert a neutral or negative effet on private setor absorption. With investment-driven savings, an improvement in the government balane (ahieved by reduing government onsumption or inreasing taxes) is translated into a lower marginal propensity to save, lower private savings and a vi

7 rise in private onsumption. This is not entirely plausible. The outome would appear more realisti, if the urrent aount were to absorb some of the adjustment required to keep investment fixed. When investment is savings-driven, an improvement in the government balane is translated into higher (publi and private) investment. This is also slightly ounterintuitive. It indues a boost to the domesti setors that produe investment goods. Foreign savings adjust minimally and often in an unexpeted manner. An improvement in the external balane should however be possible, however, in the urrent IFPRI framework. The fall in demand for domesti goods should then redue their domesti market pries. This would lead to an inrease in the relative prie of export goods as ompared with domestially marketed goods and ould thereby even inrease the value of exports. This in turn, would improve the urrent aount balane. However, in the model there is no suh impat, or it is overshadowed. In aggregate, the prie for domestially marketed goods atually rises, as inreases in onsumption or investment inrease demand for these goods and imports. Exports may even fall. The results suggest that expenditure redution as a tool to improve the external balane often annot be analysed with the IFPRI model. This illustrates the limitations in using this type of model for omparing alternative stabilisation poliies. We therefore annot ompare a devaluation and expenditure redution on their own as tools to alleviate the external balane for Zimbabwe. In the model, expenditure redution is likely to improve the external balane only if the prie of exports is able to hange, for instane as a result of a real exhange-rate depreiation. A hange in the exhange rate will alter the prie of exports relative to domestially marketed goods. The results therefore indiate that, in the IFPRI model, for expenditure redution to generate a reasonable outome, it needs to be ombined with a devaluation. This is arried out for a few different poliy versions to arrive at more balaned outomes and to ompare a few poliy alternatives. A ommon, agriulture-friendly presription appears to be the ombination of a devaluation with a ut in both government onsumption and in tax rates. However, the results of a ombined poliy still rely predominantly on the outomes of single poliies, and it is not reasonable to interpret these results without understanding the impats of individual poliies. Conluding remarks This analysis reminds us that assumptions play a vital role in CGE models and need to be taken into aount arefully before drawing onlusions about the usefulness of a poliy. These are important issues also in the PSIA ontext, espeially if the results of CGE models are being used for poliy purposes. This paper has highlighted that restritive assumptions about maroeonomi balaning mehanisms an limit the possibility to omparing alternative poliies. However, some poliy reommendations an still be derived from the IFPRI model; for instane, the results of a devaluation are fairly plausible. The paper also briefly disusses alternative losure rules and another more strutural CGE model for Zimbabwe that might produe somewhat different results. A losure option, in whih both the real exhange rate and the urrent aount balane an adjust ould be an improvement for our purpose of omparing the outomes of poliies aimed at improving the external balane. Another useful improvement would be more disaggregation on the publi setor side, suh as a government apital aount. vii

8

9 Chapter 1: Introdution 1 This paper looks at the appliation of a standard, stati omputable general equilibrium (CGE) model for analysing and omparing the impat of different maroeonomi stabilisation poliies for a ountry in eonomi risis. The ountry hosen for this purpose is Zimbabwe and the main poliy target is the external balane. The aim is to assess the suitability of a speifi model, the standard CGE model (the International Food Poliy Researh Institute), for the purpose of omparing the distributional outomes of two types of poliies; expenditure redution and expenditure swithing. The paper has both an illustrative and poliy presription purpose. It illustrates how a widely used model behaves in different instanes and assesses whih poliy presriptions have a realisti effet in this framework. The main argument is that, in the model, expenditure swithing leads to reasonable outomes on inome distribution and diretly alleviates the external defiit, whereas expenditure redution does not. This limits the range of poliy options for improving the external balane that an be ompared within the model framework. The ritique fouses on one type of model and therefore annot be generalised to hold for all types of CGEs. This model, however, derives its basis from the familiar neolassial-struturalist framework of Dervis et al. (1982). CGE models have been used for several deades to analyse the effets of a wide range of poliy hanges and shoks for both developing and developed ountries. A few examples are adverse terms-of-trade shoks and hanges in taxes, subsidies, trade poliy or in domesti eonomi and soial struture (tehnologial hange et.). CGE models are equilibrium models that over the hanges required to move from one equilibrium to another as a result of an eonomi or poliy shok. One advantage of these models is that they enable ounterfatual analysis. In the real world a variety of poliies and shoks influene eonomi outomes and it is diffiult to isolate the impat of a single poliy. A CGE model enables us to analyse the effets of a speifi poliy hange, isolate them from other hanges and, importantly, ompare the impat to the situation without the partiular hange. CGE models were first used to analyse less developed ountries by Adelman and Robinson (1978) in Korea (Sadoulet and de Janvry, 1995). Despite having been used for deades to analyse the distributional impats of poliies, CGE models have reently re-emerged under the spotlight in poliy irles, due to the inreased advoay for Poverty and Soial Impat Analysis (PSIA) of poliy reforms in developing ountries. The World Bank PSIA website 1 desribes PSIA as an analytial approah that aims to offer guidane in poliy design by arrying out an ex ante analysis of expeted poverty and the soial impat of a poliy reform. It also enourages monitoring of the impats and arrying out an ex ante impat assessment. Several studies undertaken in the PSIA ontext, in addition to the speifi World Bank 123-PRSP model designed for simple PSIA analysis, rely on CGE models. 2 CGEmodelling is enouraged as a PSIA tehnique, as it enables ex ante analysis and unlike, partial equilibrium or several other less sophistiated models, takes into aount the indiret effets of poliies. Importantly, it also inorporates prie effets. As studies using CGE methodology may inreasingly influene poliy spheres, there should be general awareness among poliy-makers of the advantages and disadvantages of this tool and the situations to whih it is best suited. Various reviews and ritiques have been written throughout the years about CGE models (for a few see de Maio et. al (1999): Robinson (1989; 2003). The fat that results based on the CGE model are sensitive to assumptions, and that assumptions an be restritive is well reognised. It is ommon for CGE-based work to ompare the outome of different senarios aimed at the same goal. The poliies are usually either quite similar (for example, the effet of redistribution measures to target a ertain group or fisal measures to redue the budget defiit) or are arried out with a wider aim in the form of a poliy pakage (for example, general maroeonomi stabilisation). What this paper emphasises is that the ability of a For PSIA toolkit, see

10 2 partiular type of CGE model to ompare the effets of alternative poliy outomes an be limited when the poliies implemented are relatively different in nature, but in pratie should lead to similar outomes. The IFPRI standard CGE model was hosen beause it is publily available in GAMS-programming ode with detailed desriptions on the IFPRI web-site. 3 This model forms the basis of muh of the CGE-based poliy analysis arried out at IFPRI and some others in reent years. As already mentioned, the model derives from a ertain tradition, whih means that a similar framework (details and number of losure options may vary) has also been used by others. 4 Zimbabwe is a ountry in need of signifiant maroeonomi adjustment and stabilisation. A ruial problem is the lak of foreign exhange and thus the external defiit. Inflation has peaked to three digit levels, the offiial exhange rate is highly overvalued, a parallel market is thriving and the eonomy has been badly hit by the eonomi onditions, agriulture in partiular, by the land reform. Eonomi reasoning suggests that there are at least two ways of onfronting a urrent aount defiit: expenditure swithing and expenditure redution poliies. This assumption is entailed in basi Keynesian adjustment models, whih assume that if net national saving (savingsinvestment) is not suffiient to finane a budget defiit, it will be finaned by foreigners and vie versa for an improvement in the budget balane. The model is a real eonomy model, and therefore only ertain types of stabilisation poliies an be examined. The starting point of our analysis is a situation where politial stability has been restored, sine without it stabilisation efforts are unlikely to take plae or to sueed. The paper onentrates on analysing and omparing the impats of a real exhange-rate devaluation (expenditure swithing) and expenditure redution in the form of uts in tax rates or a redution in government onsumption. The latter also diretly aims to improve the budget balane, whih has deteriorated severely in Zimbabwe. Expenditure swithing should produe a diret impat on exports, for instane via hanges in the exhange rate or tariffs. The impat of expenditure redution on the urrent aount is more indiret. Expenditure redution should lead to a fall in the demand for imports and domestially produed goods, whih would improve the urrent aount balane. A redution in domesti expenditure might also indue a swith in prodution from domestially marketed goods towards exports, assuming the demand for exports is elasti. Expenditure swithing and expenditure redution often need to be implemented together to ensure suessful stabilisation. The results suggest that a devaluation that removes the urrent aount balane an be ontrationary. The agriultural and export setors benefit, but others lose out and in aggregate GDP falls. Real inomes rise for agriulture-related households and fall for urban households. The result entails the assumption that the agriultural setor has the apaity to respond to the devaluation stimuli, whih in reality may be limited. In pratie, this will first require a politial hange and investment in the setor. As already mentioned, it turns out that, beause of model features, we are unable to ompare expenditure redution and swithing poliies as tools to improve the external balane in the model framework, sine the former is unlikely to affet the external balane as expeted. The result arises due to assumptions about the maroeonomi balane, espeially a strong reliane on the assumption that government savings rowd out/in private savings. The model seems to ignore some eonomi effets, whih results in quite extreme outomes, and asts doubt on the distributional effets of expenditure redution. This part of the analysis therefore serves as a ritique and not for poliy reommendation purpose Large number of individual papers, but for another eonomy-wide modelling forum, see the MIMAP projet, Siddiqui and Kemal (2002) and Pradhan and Sahoo (1998). MIMAP stands for Miro Impat of Maro Adjustment Poliies, a projet funded by IDRC, Canada,

11 3 For the model to generate more balaned outomes and be used for poliy presription, expenditure redution should be ombined with expenditure swithing that diretly hanges the pries of tradables, for instane a devaluation. The paper moves on to ompare briefly the impats of various ombinations, suh as a tax ut ombined with devaluation or redution in government onsumption and devaluation. The outomes of a ombined poliy still rely predominantly, however, on the outomes of single poliies. Alternatively, the losure rules should be hanged to ahieve more balaned outomes. The paper disusses a few options and briefly desribes the results of a struturalist (in ontrast to the more neolassial assumptions of the IFPRI model) model for Zimbabwe that behaves somewhat differently. Some aution is needed about the magnitudes of hange when interpreting the results. The original data set (see note 3 for data soure) was based on the eonomi situation in As there have been hanges in the eonomi struture, this data set was altered slightly to take this into aount. However, as the data only provide an approximation of the situation, the results need to be interpreted approximately as well. The other onern is that the model does not aurately take into aount the rationing, restritions and shortages that haraterise the eonomy. The paper starts out by explaining the foundations of CGE models and desribes the IFPRI model in some detail (Chapter 2). It ontinues with a brief aount of the eonomi situation in Zimbabwe and desribes the data set to be used (Chapter 3). Chapter 4 briefly disusses the eonomi theory behind the poliies to be examined and Chapter 5 gives some of the poliy results obtained by IFPRI for Zimbabwe. Chapter 6 desribes the simulation results in detail. Chapter 7 briefly desribes an alternative model and Chapter 8 onludes.

12 4 Chapter 2: The Model 2.1 What is a CGE model? As already mentioned, CGE models are used to analyse the effets of poliy hanges and different shoks to the eonomy. Basially an extension of multi-setor input-output and fixed prie models CGE models inorporate the indiret effets and prie effets of poliies. They over the time period it takes for an eonomy to arrive at one equilibrium from another after a shok, so that all markets reah an equilibrium. A stati model then generates a medium-term solution - a situation after initial disequilibrium, but before dynami effets set in (Sadoulet and Janvry, 1995). The data requirement is a soial aounting matrix (SAM), whih is a square matrix that groups together both input-output aounts national inome and produt aounts. The olumns of a SAM traditionally refer to the expenditure and the rows to the inome of eah aount. An important fat to aknowledge about these models is that the results rely quite heavily on assumptions; this will be refleted in our study. There are different shools of modelling. Traditional neolassial CGE models are based on Walrasian general equilibrium theory. Firms maximise profits, and wages and pries adjust to equate supply and demand in fator and produt markets. Fators of prodution are therefore fully employed. Struturalist models differ from neolassial ones in assumptions about maroeonomi balaning mehanisms and in the way markets are leared. They inorporate features of short-run maromodels, where there an be wage or prie rigidities and an equilibrium without market learane. In these models an equilibrium an therefore be demand-driven. In real side models there are no asset markets, money is neutral and pries are relative. In neolassial CGE models only relative pries are determined and some prie or prie index is hosen as a numéraire against whih all other pries (wages, rental rates et.) are valued. Often this is the aggregate domesti prie. Struturalist models often hoose the nominal wages or the exhange rate as a numéraire. (Robinson, 2003). 5 The IFPRI model desribed below inorporates features from both struturalist and neolassial frameworks. Some CGE models also inorporate asset markets and finanial effets. One example is the OECD maquette model (see Bourguignon et al., 1991 and Robinson, 1991); the more reent World Bank IMMPA model (Agenor et al., 2003) also relies on this type of bakground. Thissen (1999) surveys this literature. Some dynamism and expetations an also be introdued into CGE models, whih equip them to deal better with issues suh as eonomi growth or investment. 2.2 The IFPRI standard model A detailed desription of the IFPRI model an be found in Löfgren et al. (2002). The model is presented only briefly here. Some of the key model equations related to prodution and pries an be found in Appendix 1. In the model, poliy shoks ause hanges in relative pries. An adjustment proess leads to a new equilibrium, where demand equals supply in eah market. There is no inflation as the onsumer prie index (CPI) is the numéraire, but relative pries an hange. The CPI is based on omposite ommodity pries (a omposite of imports and domesti goods). Prodution is arried out by ativities that maximise their profits, subjet to a prodution funtion and given pries of their inputs, outputs and fators. There is perfet ompetition. Within eah 5 See this referene also for a onise aount of a few standard CGE modelling approahes with referenes to appropriate literature and a brief disussion on the benefits and shortomings.

13 5 ativity, top-level prodution is determined aording to a Leontief funtion. This implies zero substitutability between fators of prodution (value-added) and intermediate inputs in prodution, whih introdues some rigidity to the model. Substitutability between intermediate inputs is also haraterised by a Leontief funtion, but between fators by a onstant elastiity of substitution (CES) funtion (see Appendix 1, equations 3 and 4). The first step in the prodution proess is to generate aggregate marketed domesti output from the output of different ativities. This aggregation is based on a CES prodution funtion, and here the demand for output of eah ativity is based on minimising the ost of supplying a given quantity of aggregated output. Ativity-speifi ommodity pries will lear the market for eah disaggregated ommodity. At the next stage aggregate domesti output is alloated between domesti sales and exports based on suppliers revenue maximisation. Profits are maximised subjet to given pries, fixed quantity of total output and imperfet transformability between domesti sales and exports expressed by a onstant elastiity of transformation (CET) funtion (see Appendix 1, equations 2, 5, 6). Export supply is therefore determined by the domesti prie of exports relative to the supply prie of domesti goods. The domesti prie of exports is simply the world prie of exports adjusted by export taxes and the exhange rate. The world prie is given. As there are no export taxes and the real exhange-rate is exogenous in our set-up, the domesti prie of exports stays fixed unless the real exhange rate hanges. The ountry is thereby haraterised as a smali, open prie-taking eonomy. In the model an inrease in the export-domesti prie ratio generates an inrease in the export-domesti demand ratio. Export demand is infinitely elasti at the given world prie. Domesti demand onsists of demand by households, government onsumption, investment and intermediate inputs. If a ommodity is imported, domesti demand is for a omposite ommodity made up of imports and domesti goods. The optimal mix between the demand for imports and domesti output is based on ost minimisation (see Appendix 1, equations 1, 8, 9 for first order onditions for minimisation). Domesti output and imports are imperfet substitutes again expressed in the model via a CES aggregation funtion. This so-alled Armington assumption generally allows a ertain amount of independene of the domesti prie system from the international and hinders unrealisti export and import responses to poliy hanges. The supply of imports is infinitely elasti at given world pries. The domesti prie of imports is the given world prie adjusted by the exhange rate and tariffs, both of whih are exogenous in our set-up. In the domesti markets, flexible pries equate demand and supply. Households reeive inome from fators of prodution and transfers from institutions namely the government and other households. The government reeives inome from taxes and transfers from the rest of the world, whih it spends on onsumption and transfers. The household saves, onsumes, and pays diret taxes and transfers. The behaviour of the diret tax and savings shares is linked to hosen losure rules for government and savings-investment balane (see below). Household onsumption is alloated aross different produts aording to a linear expenditure system (LES), 6 whih in this ase simply means that onsumption of individual ommodities by eah household is a linear funtion of total onsumption expenditure. Different households onsume different shares of eah ommodity and have different elastiities of market demand for eah ommodity. More details on the institutions will be provided in the setion desribing the data (Chapter 3). The key elastiities are those for fator and intermediate input substitution in prodution, the elastiity of transformation (CET) for alloation of output between domesti sales and exports and the elastiity of substitution (CES) between demand for domesti and imported goods. The elastiity values used in this analysis are shown in Chapter 3. The maro losures in CGE models govern the mehanism by whih the three maro balanes the urrent government balane, the urrent aount balane and the savings and investment balane are determined. The hoie of model losure rules, in other words beliefs about how an 6 This refers to the first order onditions that result when households maximise a Stone-Geary utility funtion subjet to a onsumption expenditure onstraint.

14 6 eonomy funtions, has an important effet on the model results. Below is a brief desription of these rules. i) Government balane The following equations relate to the government balane: (1) R G = E G + S G (2) E G = C G + transfers where RG = government revenue, EG = government urrent spending, CG = government onsumption and SG = government savings There are basially two losure rule options for the government balane. One is a rule where diret tax rates are fixed, but the real fisal balane adjusts if government revenue reeipts hange. The other option uses flexible diret tax rates and fixed government savings. For a developing ountry the latter may be unrealisti (at least in the short run). This study uses the first option, as it is more realisti and beause the assumption of flexible tax rates does not allow for simulations, where tax rates would be hanged. The results are not fundamentally affeted by this hoie of losure. Government onsumption is exogenous, but does not inlude government investment. ii) Savings-Investment balane The savings-investment balane is (3) SNG + SG + SF = I where S = savings, NG =non-governmental, G = governmental, F = foreign, I = investment There are various options for this losure in the IFPRI model, but investment will always depend solely on savings. This is harateristi of stati neolassial CGE models. Savings, on the other hand, are determined by various fators. Beause of the sensitivity of the results to losures, we apply two different investment losures. In the first ase (assumption set I) we assume investmentdriven savings and that marginal propensity to save (private) adjusts to maintain the equality between savings and investment. In the seond ase (assumption set II) we assume savings-driven investment and fixed marginal propensity to save. In other words, real investment expenditure adjusts to equal the volume of savings available to finane it. An additional savings-investment losure option in the model is to assume fixed apital formation as well as fixed government onsumption absorption shares, but flexible marginal propensity to save (not used in this study). Work arried out by IFPRI on Zimbabwe mainly uses the savings-driven investment losure (see, for example, Bautista et al., 2002). Depending on the poliy, the results an differ onsiderably under the two losures. As de Maio et al. (1999) laim in a ritique of a CGE model of Afrian eonomies, different losures lead to quite different welfare effets. These authors strongly ritiise the use of flexible investment and mobile, fully employed fators, espeially in less developed Afrian ountries. Although these assumptions may apply in the long run, in the short run they are muh less likely to be valid. The welfare impats from adopting more rigid losures and assumptions are more adverse. It is often beause of short-run rigidities that muh of the harm of eonomi reform is done to the poor. Thus, in addition to the medium- and long-term effets, it is also important to look at the short-run impats, if possible. iii) External balane Foreign savings are the differene between reeipts and expenditures of foreign exhange. Unless the exhange rate hanges, the only variables that affet the balane are exports and imports.

15 7 Transfers between the rest of the world and domesti institutions and fators (net transfers) are fixed in foreign urreny. In the model either the real exhange rate or the urrent aount defiit adjusts to maintain the external balane. If we wish to examine the impat of a devaluation, the real exhange rate needs to be set exogenously and the urrent aount will move to balane the external aount. The external balane is determined in relation to the rest of the world as (4) M + net transfers - XI = SF where M = imports of goods and servies, X = exports of goods and servies, and SF = foreign savings (trade defiit) iv) Fator markets Fator market losures determine the mehanisms that equilibrate the supply and demand of fators. In line with general equilibrium theory, eah ativity uses a set of fators up to the point where the marginal revenue produt of eah fator equals its wage. There are two wage variables: the eonomy-wide wage, and an ativity-speifi wage that is the produt of the eonomy-wide wage and an ativity speifi wage or distortion term. The wages are expressed subjet to the numéraire and are thereby real. This means that whenever ertain wages are fixed in our analysis later on, these refer to the real wage. There are basially three losure variations: a fator is fully employed and mobile, is fully employed and immobile, or is mobile but an be unemployed. In the first ase the quantity of eah fator supplied is fixed. The eonomy-wide wage rate varies to equate the supply and demand for fators. In the seond ase, eah ativity is fored to hire the observed quantity of a fator, whih means that the demand for the fator is fixed. The eonomywide wage is fixed, but the fator supply and the ativity-speifi wages are flexible. In the third ase, supply and demand are flexible, whereas both the eonomy-wide and ativity-speifi wages are fixed. The fator market assumptions used in this study are desribed in Chapter 6. The model is a single-period omparative stati model that ompares one equilibrium with another. It is not possible to separate short- from long-run effets, but by applying more rigid onstraints we an simulate a shorter-run impat. Some ombinations of the losure options may not however, produe realisti results in a stati model. As an example, let us onsider an external losure of flexible foreign savings and a fixed real exhange rate. With a single-period model, we annot aount for negative welfare effets that may be aused in the future by rising foreign debt (if foreign savings adjust by an inrease).

16 8 Chapter 3: Applying the Model to Zimbabwe: Bakground 3.1 Zimbabwe s eonomi situation Zimbabwe ahieved independene in 1980 and set out with the objetives of national unity and eonomi growth with greater equity. However, sine 1997 the ountry has been in steady and aggravating eonomi deline and today real inome per apita is lower than before Independene. Unemployment, aording to some estimates, is as high as 70%. A reent Eonomist Intelligene Unit (EIU) ountry report (2003) shows that, in terms of GDP per apita (in 2001 around US$480) Zimbabwe has beome one of the poorest ountries in the Southern Afrian Development Community. An eonomi outlook (ADB and OECD, 2003) indiates that GDP per apita has fallen to about half the Afrian average, ompared with an equal level in Real GDP per apita growth was -5% in 2000, -8% in 2001 (WDI) and is predited to be -13% in 2003 (EIU). There is hyperinflation. Towards the end of 2003, inflation was over 600% and is expeted to rise in 2004 (EIU). Real interest rates are negative and prie ontrols are in plae on a variety of produts. The Zimbabwe dollar has been grossly overvalued for a long time, whih has led to the emergene of a large parallel exhange market. As the foreign-exhange shortage persists, the parallel rate flourishes at around Z$6500:US$1 (EIU, 2004). Foreign exhange ontrols whih were abolished by the strutural adjustment programmes ( and ) have been reinstated to regulate imports. (Afdb - OECD, 2003.) The offiial exhange rate of Z$55:U$1 was still in plae at the end of 2003, but a separate rate at Z$824:US$1 was set for seleted exporters. All other export inentives (tobao subsidies, floor prie of gold) have been suspended. A new system of foreign-urreny exhange has been set up reently, by whih 25% of foreign-exhange earnings (previously 50%) is to be surrendered to the entral bank with the offiial exporters' rate, and a further 25% is to be sold at an aution organised by the entral bank (EIU, 2004). An important reason for linging on to an overvalued exhange-rate has been that it gives the government the opportunity to import neessities suh as food and fuel at a heaper rate. Land reform and unfavourable onditions have driven an agriultural ountry to the brink of having to rely on imported food. Other neessities are also in short supply due to lak of foreign exhange. There are also fears that a devaluation might lead to even further hyperinflation. In terms of eonomi ativity, almost every setor but that of real estate has experiened a sharp downturn. Real estate is blooming beause money that might normally have gone into savings is being hannelled into property, due to low rates of return. Table 1 shows the aggregate eonomi struture in 2001 (WDI, 2003). Agriultural prodution is estimated to have delined by 13% in 2001 and by another 22% in 2002 (partly due to a drought). Aution sales for the major export rop tobao were as low as 80 million kg in the 2003 season, ompared with 167 million kg in 2002, and thus the lowest level of prodution for 20 years (EIU). The land reform has brought onsiderable hanges and has aused the deline of export-oriented agriulture. Aording to Games (2002), by the end of Otober 2002 around 97% of the formerly white-owned ommerial land had been seized by the government. Only about 600 of the 4,500 ommerial farmers had been left on their land. The volume of industrial prodution is also falling and the manufaturing setor s share of GDP has dereased from 22% in 1995 to 14% in 2001 (see Table 1). Currently revival is diffiult, one reason being that pries often annot be inreased by firms due to the ontrols on a variety of industrial produts. In addition, beause of the slowdown in tourism, the servie setor may have ontrated signifiantly. The mining setor has already been in risis for some years, and gold prodution is falling. Due to a fairly reasonable tobao rop and revenue from gold exports, foreign-exhange earnings were still holding up in 2002, but sine then they have fallen beause of the dramati redution in

17 9 tobao and gold exports. As the eonomy ontrats, export earnings are expeted to fall further in The EIU s estimate for the urrent aount defiit in 2003 is 10.4% of GDP, and it is predited to be around the same level in This hroni shortage of foreign exhange suggests that it is ruial to take measures to improve the external balane. Devaluation has therefore been hosen as one of the poliy reforms to be looked at in this study. An EIU report of Deember 2003 desribes how the tobao farmers have warned the government that, without a devaluation, what is left of the industry will soon ollapse. Some laim, however, that a devaluation may not be suffiient to inrease export inentives and that there is a need to fous on reviving produtivity (Chitiga and Mabugu, 2001). Table 1: Eonomi struture of Zimbabwe, 2001 Value added as % of GDP Agriulture 18 Industry 24 of whih Manufaturing 14 Servies 58 Soure: World Development Indiators (WDI) 2003 Donor support annot be relied upon as a means of generating foreign exhange, until there is a hange in the politial sene. Currently soures of funding for Zimbabwe are very limited. Multilaterals suh as the World Bank and the International Monetary Fund have suspended projet finane and balane-of-payments support to Zimbabwe (Games, 2002). The IMF has already adopted a poliy of non-ooperation, beause of the ountry s overdue finanial obligations. The government has borrowed to finane the budget defiit, whih has only aelerated inflation. In addition to politial tensions and the land reform, many see fisal mismanagement as one of the key ontributors to the eonomi deline and the failure of the adjustment programmes to deliver growth and poverty redution. The fisal problems have a long history. Zimbabwe is highly indebted beause of an aumulation of debt during the losed eonomy period of the 1980s and post-independene restruturing. The 1990 figure for total debt as a perentage of exports was 160% that for % (GDF). Estimates of the budget defiit vary. After grants it reahed 21% of GDP in 2000; the figure for 2003 was lower, around 7.5% of GDP. However, offiial figures for defiits tend to be underestimated as they do not inlude off-balane-sheet finaning, suh as loans to parastatals or the urreny subsidies urrently offered to exporters (EIU, 2004). The fisal situation means that, in addition to improving the foreign balane and the availability of foreign exhange, efforts to improve the fisal balane are also required. However, the government has a number of fisal obligations. Importantly, it will need to devote funds to resettled farmers in order to revive agriultural prodution. The agriultural setor annot respond easily to export inentives without resoures, and the produtive apaity of the manufaturing setor has been in deline for years. There is therefore a need to seek new areas of ativity and speialisation as well as enouraging existing ones. Severe unemployment and dereasing real inomes also put a onstraint on domesti demand that an delay adjustment. Investment opportunities may be onstrained in the short run due to the foreign-exhange shortage. Before a number of more strutural issues are, however, onfronted, the immediate objetive should be maroeonomi stabilisation. It is widely agreed that a politial hange is required for a redible stabilisation programme to be implemented. This desription illustrates that there may be some diffiulties in using a CGE model to analyse poliy impats in Zimbabwe, as there are supply onstraints and parallel exhange markets. A market equilibrium approah may seem unrealisti. This is however, a normal problem with equilibrium models. We take these problems as more or less given, and will onentrate on some of the other harateristis of the model and the entral question of how well it is suited to

18 10 omparing expenditure swithing and redution polies for improving the external balane. This was taken as the main poliy objetive in this paper, although expenditure redution will also diretly redue the fisal defiit as required for stabilisation. 3.2 Data This hapter offers a brief desription of the Zimbabwe soial aounting matrix (SAM) to be used in this study. We use the IFPRI 1991 SAM, but with a few alterations. This is the most reent SAM produed by IFRPI for Zimbabwe. The year 1991 was hosen as the base year, sine the SAM was primarily to be used to analyse the impat of the strutural adjustment programmes that started in 1991, and beause the following two years were haraterised by a severe drought. The original IFPRI matrix is quite detailed, as it inludes a large number of setors (and thus goods). This holds espeially for agriulture. It also inorporates some informal ativity, and takes smallholder unmarketed prodution as well as marketing margins into aount. All these are important features to ater for when analysing developing ountries, but they are often exluded. In partiular, marketing margins are reasonably new additions to CGE models. Although this study uses the IFPRI 1991 SAM as a basis, the data have been readjusted to orrespond slightly better to the present situation. The original SAM was obtained from IFPRI, 7 but this paper uses a simplified version of this SAM as a basis, 8 in whih the 27 setors and produts have been aggregated to seven. The simplified SAM also does not inlude marketing margins or non-marketed prodution. The data were first adjusted manually, but the hanges made were not radial. They were then realibrated using the IFPRI SAM balaning programme to ahieve a balane in all aounts. 9 The values of prodution for eah ativity were adjusted to resemble the eonomi struture of 2001 (as in Table 1), so that the value-added share of a ertain ativity of total value-added (by all ativities) hanged. The shares of speifi types of labour or apital in the total prodution of an ativity remain losely alike, as do the onsumption patterns of households. However, payments to different fators as a share of total fator payments (of all ativities) hange as the value of prodution and thus the values of fator and intermediate inputs hange within eah ativity. The adjustment does not take into aount patterns of hange that may have ourred in the share of informal labour in eah ativity, whih may have risen. Thus the adjustment is only a partial attempt to apture hanges in the eonomy. However, this does not weaken the purpose of assessing the suitability of the model to ompare poliy options. The urrent situation in Zimbabwe does have some similarities to that in The eonomy had only just embarked on its adjustment programme. During the 1980s Zimbabwe had a large budget defiit (on average 8.4 % of GDP annually), heavy external borrowing, slow export growth, importprotetion poliies, and an administered and overvalued exhange rate. (Bautista et al., 2002). A detailed desription of the original Zimbabwe data set an be found in Bautista and Thomas (1999). The SAM we use overs the following aounts: Setors and produts: Agriultural exports (mostly large-sale prodution) All other agriultural ativity Mining Food proessing Manufaturing 7 From Marelle Thomas 8 On the IFPRI website (footnote 3) as a part of a set of model input files. 9 The objetive funtion of the SAM balaning programme is a ross-entropy distane from the initial SAM oeffiients for the entire matrix rather than for olumn sums. The method approximates to the restrited additive Shwarz (RAS) proedure exept that the only restritions are that olumn and row sums be equal. It is not neessary to know the olumn sums.