Introduction

Understanding the impacts of the macroeconomic structural adjustment programs (SAP) on income (and wealth) distribution and poverty is important because of the vulnerability of the poor as a group in developing economies. There is much evidence that economic and financial crises often hurt the poor who have few cushions to protect themselves when a downturn occurs. There is also growing evidence that particularly for poverty reduction objectives, there are dynamic trade-offs in the implementation of SAPs (Agenor 2002; Khan 1997, 1996). For instance, it is well known now that budgetary retrenchments associated in many cases with the SAPs have fallen largely on various types of social expenditures leading to a short-run worsening of the poverty situation in the absence of countervailing programs. In the medium to long-run, however, the SAPs are expected to bring down inflation, lessen credit rationing through lower borrowing rates for all by ending financial repression, and increase economic activities leading to sustained growth. To the extent that the poor are also beneficiaries of these outcomes, poverty is expected to decline.

This paper has two related goals. The first and the main aim is to survey selectively and analytically the implications of the various (macroeconomic) computable general equilibrium (CGE) models constructed for the purpose of integrating poverty analysis with the usual macroeconomic variables and relationships. Taking stock of our existing knowledge in this area will help clarify the relationships between macroeconomic policies and poverty reduction objectives, if and when such relationships are postulated to exist. Such a survey will also lead to a second, operationally relevant research question. Are there intermediate models---generic models, so to speak--- that can be used or constructed for the purpose of identifying the poverty impacts of policies both qualitatively and quantitatively? The second goal of the paper is to explore this question. It should be said at the outset that the answer to this question does not appear to be either an obvious "yes" or an obvious "no". If such models can be identified or constructed their operational relevance for Asian Development Bank lending operations can be significant.

The emphasis on poverty reduction at the national and international levels as embodied for example, in the Millennium Development Goals, calls for a careful methodological approach to the estimation of the poverty reduction impacts of macroeconomic and other policy variables. A recent report produced at the Asian Development Bank sorts out many of the complex issues involved at the macro-, meso- and micro-economic levels and pinpoints the need for further conceptual and modelling work at the appropriate levels of (dis)aggregation (Bolt et.al.2003). The identification of the three different levels and treating the meso-economic level as the (institutional) link between the other two levels are encouraging in terms of understanding the complex causal relations that are involved in understanding and reducing poverty.

There are both econometric studies at the aggregate level and some economy-wide Social Accounting Matrix (SAM)-based CGE models that have attempted to depict the impact of policy on poverty. However, the former are usually "kitchen sink" variety regressions without clear theoretical elaboration. The SAM-based CGE models are detailed but are usually without clear expositions of the causal connections between policies and poverty reduction (Azis, 2002). This paper will explore the strengths and weaknesses of the various models on offer and try to identify the uses to which they can be put for understanding the poverty reduction implications of macroeconomic policies.

At this stage, we need to know at what level of aggregation we can pose meaningful questions regarding the impact of policy on poverty reduction. In particular the impact of policies on the poor households as well as the near-poor through both direct and indirect causal channels will be examined within the context of the various macro-models that have tried to include poverty analysis.

There are at least two aspects of any poverty impact analysis for a particular policy. These are: i) the impact on economic growth; ii) the impact on income and asset distribution. The growth effect on poverty reduction is then given by some estimated growth-poverty elasticity. In the second case, a more (less) favorable income/asset distribution for the poor may reduce (increase) poverty. A distributional neutrality assumption in a model simply allows one to look at the growth aspect by itself. Here, too, different sectoral growth rates and different sectors themselves may affect poverty differently. (Quibria 2002; Khan 1999; Thorbecke and Jung 1996).

A related issue is the heterogeneity of the poor households/individuals. The distinction between chronic and transient poverty (Jalan and Ravallion 1998 a and b; Hulme and Shepherd 2003) is important here for assessing the poverty impact of policies. Poverty severity differences among households (Thorbecke and Jung 1996; Khan1999, 1997) are also important to keep in mind in assessing the impact of policies on different types of poor households.

Partly following from the above considerations, the selection of a particular poverty index or poverty line can bias policy analysis as well. Some analytical effort needs to be devoted towards the clarification of these and related issues in the context of a particular class of macromodels. For example the headcount ratio, the Sen index and the FGT (Foster-Greer-Thorbecke) measures may lead in different directions as to which are the most appropriate groups/ geographical regions etc. for policy interventions. The across the board growth-poverty elasticity approaches via the headcount ratios given by one dollar/ two dollars a day poverty lines may be too crude for meaningful impact analysis. Clearly, these can be good starting points in the absence of further information, but good policy impact analysis needs to go much further.

However, the operational needs of the multilateral banks and data constraints on the ground may not leave much room or time for detailed classification of poverty, comparison of various indexes and further analysis of static vs. dynamic poverty and related issues. Nevertheless, it will be useful if our survey of models can lead towards the identification of simpler models or approaches that can stand at an intermediate level between large SAM-based economy-wide CGE models, for instance, and the existing ADB practice in many instances of fairly vague statements regarding poverty reduction impacts of policies (Bolt et. al. 2003).

The structure of the paper is as follows. In the following section I discuss the general macroeconomic policy issues arising out of program lending and their relevance to the poverty reduction strategy in the "post-Washington consensus" policy environment. This raises---among other things--- the questions regarding the measurement of poverty and the nature of macroeconomic environment in the developing economies. Consequently, in the two sections following immediately), I discuss these issues in the context of developing economies. In section 3, I deal with some fundamental issues for the measurement of poverty. This is followed up in section 4 by a discussion of some issues regarding the general structure of macro-models. In particular, the possible uses of SAM-based fixed price multiplier models are discussed. Section 5 then takes up the issue of CGE modelling for developing economies. Section 6 explores specifically the questions related to income distribution and poverty in CGE models for developing economies. In the penultimate section (section 7), I discuss the structure of what has been termed the "dual-dual" model In the concluding section I raise the question of how applicable these are for low and middle income Asian economies with large pockets of poverty. I end with some tentative suggestions regarding poverty analysis in an "extended dual-dual" framework for a low-income Asian economy such as Bangladesh as a first stage in building models that are applicable to Asian economies.

At the outset it is fair to mention that even the poverty ‘incidence analysis’ at the micro level is not as straightforward as it seems. For example, even cash transfers may modify behavior. Such modifications can lead to general equilibrium effects in an economy wide set of repercussions. Typically, of course, most transfers are made indirectly--- through public spending and indirect taxation. The allocation rules are not always transparent and implementation is incomplete or distorted (Bourguignon et. al. 2002). More relevant to our purpose here, often macroeconomic and structural adjustment instruments and outcomes are also involved. The declared purpose of such reforms is to enhance economic activity and long-term rate of growth. In the short-run, however, the effects may even run in the opposite direction. A careful specification of the macro-models and the macro-micro linkage is thus a prerequisite for any meaningful and policy-relevant economic analysis.

Essentially there are three levels that such a relatively complete analysis of poverty reduction impacts of macro-policy changes would involve. First level includes the macroeconomic tools and models that will allow us to estimate and evaluate the impact of various exogenous shocks and policies on macro or aggregate variables such as the GDP/capita and its macro-components, the rate of interest, inflation/deflation via changes in the aggregate price level, the exchange rate and so on. The time frame must also be made explicit. At the second level we need to have tools and procedures for disaggregating the values of the variables obtained through our modelling and estimation exercises at the first level. Thus, at the end of our procedures at this level we will have at our disposal a disaggregated picture of the effects of policies on sectoral activities, and returns to factors and households at the appropriate levels of disaggregation. The last, bottom layer usually consists of a micro-module where an ‘incidence analysis’ can be carried out through the manipulation of household micro data with the help of relevant theories of distribution, household income generation and consumption.

Anticipating the results of our review of the CGE models in particular, it will be seen that for developing economies these models can be conveniently categorized in three "generational" classes.¹ The first generation, starting with the pioneering works of Taylor and Lysy (1980) and Adelman and Robinson (1979) in the late 70s and the 80s focused increasingly on trade policy issues. The second generation in the late 80s and 90s made income distribution in the context of structural adjustment policies as the main focus, although it must be added that the pioneering works in both the Lysy and Taylor volume, and the Adelman-Robinson volume did not neglect distribution. The main difference is the explicit reckoning with Structural Adjustment Programs (SAPs). In the late 90s, explicit attention began to be paid to the poverty impact of SAPs within a CGE modelling context. In this context, with the Work of Decaluwe et. al. (1999), we seem to be in the third generation of CGE models where poverty impact has been modeled explicitly. At the end of the paper I will make some suggestions for perhaps a "fourth generation" of models for poverty in general equilibrium setting.

²

Download this Discussion Paper [ PDF 469.8KB| 73 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.