Chapter

3 The Implications of Adjustment Programs for Poverty: Conceptual issues and Analytical Framework

Editor(s):
Ke-young Chu, and Sanjeev Gupta
Published Date:
April 1998
Share
  • ShareShare
Show Summary Details
Author(s)
Ravi Kanbur

During the past decade, many developing countries have faced severe macroeconomic imbalances that they have tried to correct through a series of programs. Many of the programs have been supported by the IMF. Over the years, concern has been expressed by international agencies and by the countries themselves that the adjustment programs have had an adverse effect on some poverty groups, and the issue of finding ways to minimize this effect, while still achieving the goal of macroeconomic balance, is now on the agenda. A prerequisite for such an exercise is a conceptual and analytical framework for discussion of poverty and the transmission mechanisms between measures in programs and their impact on the poor. The object of this chapter is to review the literature and to evaluate the frameworks currently available, the focus being on identifying methodologies that are or can be made operationally feasible.

Conceptual and Analytical Issues in Measuring Poverty

Standard of Living

Our concern in this study is with the impact of IMF-supported programs on the poor. This requires us to specify what we mean by the poor and by poverty. To do this, we must answer two questions: (1) What is the “standard of living” concept on which our discussions are to be based? and (2) Given this concept, how are we to delineate the poor from the nonpoor?

The definition of what constitutes the standard of living is not an easy task and the academic literature has often taken a philosophical turn on this issue (Sen and others, 1987). Even eschewing abstract discussions of what it means to have a certain standard of living, making the concept operational is difficult because of its multidimensional nature. A straightforward economic approach would focus on the consumption of goods and services. For market items, the many dimensions of the standard of living can be reduced to a single numeraire by using prices to convert quantities into expenditure. For items such as home-produced consumption, prices can be imputed and used in the same way. The same can be done for education and health, but the use of the market framework becomes more tenuous for these publicly supplied services and others, such as access to clean water and public transportation. In principle, of course, all these could be reduced to the same unit of account if we could find appropriate shadow prices. However, this is bound to be a controversial exercise. For this reason, from the operational point of view, it seems appropriate to separate private consumption from “basic needs” indicators such as health or education.

Monetary Measures

For operational purposes, then, it is useful to have a unidimensional monetary measure of the standard of living. But what should this be? Should it be income, expenditure, or something else? An important issue here is the time horizon. Conceptually, the ideal would be permanent income or expenditure. But what we have in most cases is a snapshot household survey with measured income and expenditure for a year (or an even shorter period). If one is interested in current consumption then expenditure is a better indicator to use. In fact, given the difficulties of measuring income in rural areas of developing countries, it can be argued that total expenditure is probably a better track of “permanent income” than is measured income. Some have argued that a narrower category—food expenditure—would be even better (Anand and Harris, 1985). If we focus on expenditure, then we need to allow, of course, for price differences within a country and over time. The extent to which this can be done depends entirely on data availability.

Given the focus on real expenditure, there is still the question of whose real expenditure we are interested in—the individual, the family, the household, or the extended household, and so forth. From the normative point of view, it can be argued that we should ultimately be interested in the welfare of individuals and that larger groupings are relevant only insofar as there is income sharing. In fact, data availability largely forces upon us the use of household expenditure—the only question being how this is to be corrected for household size and composition. While there is a large literature on the use of adult equivalent scales to adjust for household composition, this literature is controversial (Deaton and Muellbauer, 1986). In any case, equivalent scales will have to be country specific and their calculation is a major research effort.

An additional issue, which has generated much literature in recent years, is that of intrahousehold inequality. The use of adult equivalent scales corrects for different needs of adults and children (or men and women), but their use in a normative context implies the assumption that consumption is distributed according to these needs. There is some evidence that there may be discrimination against female children in some parts of the world (Kynch and Sen, 1983), while other attempts to identify such discrimination have concluded that the evidence does not support this (Deaton, 1987). At the theoretical level, much work is now under way in modeling intrahousehold allocation decisions in the framework of noncooperative game theory. However, we must conclude that the literature as it currently stands does not provide clear guidelines on how intrahousehold inequality is to be incorporated into poverty measurement, particularly if, as is almost always the case, the only data available are income or expenditure at the household level.

Thus, while there are clearly a number of important issues unresolved in the literature, and the analyst should be aware of these, our recommendation is that for operational purposes real household expenditure (or income) per capita be used as the measure of individual welfare.

Inequality Versus Poverty, and the Poverty Line

Given that what we are interested in is the distribution of individuals by household expenditure per capita, the next question is, What features of this distribution are we most interested in? In the literature on income distribution, a distinction is made between a concern with inequality, which has to do with the distribution as a whole, and a concern with poverty, where the focus is on the lower end of the distribution. Here is not the place to get embroiled in the debate on which of these is the more important, or the conditions under which these two concerns are consistent with each other (in general they are not, but see Atkinson, 1987). Rather, we base our argument on the current policy concerns with poverty groups. In addition, it may be easier to arrive at a consensus on poverty alleviation as an objective as opposed to inequality reduction—the latter might involve, for example, weighing the relative incomes of the rich and the super rich.

Accepting a poverty focus, this still leaves open the question of where a poverty line should be drawn. This is related to the question of whether a poverty line (along the real expenditure per capita dimension) should differ from country to country, and whether it should change over time. This is in turn related to the absolute versus relative poverty debate.

Clearly, the nature and meaning of poverty are country and culture specific, so that there is an undeniable relativistic element to it. It is neither inconsistent nor incoherent to say of two individuals, one in the United States and one in Côte d’Ivoire, that both are in poverty even though the real income of the former (after making all relevant price corrections) is far greater than the real income of the latter. Moreover, as the income and structure of a country change, what constitutes poverty also changes, although this change may occur over a long period—too long to be of relevance to an adjustment program of, say, three years.

Ideally, what one would like is a specification of a basket of goods and services that an individual should be able to purchase to be considered not to be in poverty. This would include not only basic food and nutrition but also clothing and housing. It is in the specification of these and other items that differences may arise between countries, However, one would like an operational shortcut to arrive at the poverty line that would be reasonably applicable in a range of situations.

If an established poverty line already exists in a country (as it does, for example, in India and Sri Lanka, where it is based on nutritional standards), the analyst would do best to adopt this line but do a sensitivity analysis for variations in this line. The development of such lines, however, is not an easy task and controversies may take a long time to resolve (in the case of India, a high-level Committee of the Planning Commission arrived at the line). In the absence of any other widely accepted line, we would recommend the following operational procedure: Given a distribution of individuals by real household expenditure per capita, choose a poverty line that cuts off the bottom 10 percent of individuals in some base period. These become the poverty lines with which to evaluate changes in poverty over time and differences across regions at a point in time. It should be emphasized that the procedure is meant as an operational device, and one which may address some problems regarding differences across cultures in what is meant by-poverty. However, once chosen, these lines are to remain fixed in real terms, so that the poverty figures capture changes in absolute poverty over time. This device was used, for example, in Kanbur (1990).

Poverty Index and Poverty Profile

Given the distribution of real per capita expenditure and the poverty line, we still have the problem of how to represent the information on poor incomes in an operationally convenient and normatively significant way. There is now a large literature on axiomatic approaches to poverty measurement (Sen, 1976; Donaldson and Weymark, 1986). However, for operational purposes the chosen measure must be able to capture a range of value judgments on the significance of the extent and depth of poverty but be easy to handle and interpret. One measure that has been found to be useful in this context is that put forward by Foster, Greer, and Thorbecke (1984). If real expenditures or incomes are ranked as follows:

Y1Y2 ≤ … ≤ Yq < z < Yq+1 ≤ – – ≤ Yn

where z is the poverty line, n is the total population, and q is the number of poor, then the Foster, Greer, and Thorbecke measure is

What the measure does is take the proportional shortfall of income for each poor person, (zyi/z), raise it to a power α (≥0) to reflect concern about the depth of poverty, take the sum of these over all poor units, and normalize by the population size.

As α varies, Pα takes on a number of interesting features. When α = 0, so that there is no concern for depth of poverty,

This is simply the “head-count ratio”—the fraction of poor units in the population. This has been criticized (Sen, 1976) for focusing on the number of the poor and ignoring how poor they are. However, with α = 1,

where ȳp is the mean income of the poor and I (= z – ȳp/z is known as the “income gap ratio”—the average shortfall of income from the poverty line. As can be seen, P0 = H focuses on the number of the poor but not on the extent of their poverty, while I focuses on the average extent of poverty but not on the number of the poor. P1 combines these two and is perhaps a benchmark poverty index that should be the focus of interest in policy-oriented analyses. It should also be noted that nzP1 is, quite simply, the total amount of resources required to eliminate poverty (1) if there were no incentive efforts in transferring money, and (2) if targeting was preferred. As such, it gives a lower bound on the financial commitment required to eliminate poverty.

Although the P1 measure should become, in our view, a standard measure of poverty, it is insensitive to redistribution among the poor (since it depends only upon their mean income). However, with α = 2, this sensitivity can be ensured. Whereas with α = 1, a dollar gained by the very poor would have the same effect on poverty as a dollar gained by the moderately poor; with α = 2, those two would be differentiated in the increases, and more weight is given to the poorest of the poor.

Another important feature of the Pα measure is that it is subgroup decomposable. By this is meant that if we divide the population into mutually exclusive and exhaustive groups indexed by j, and if Pα, j is the poverty in the j th group, then

where xj is the proportion of total population in group j. The above expression is useful, since we can assign the “contribution” of poverty in group j to national poverty as

Such a decomposition of national poverty into regional, occupational, crop production, or other sectoral groups can help in developing a poverty profile for the country in question. An example of a regional profile for Côte d’Ivoire is given in Table 3.1. The poverty line is defined as that point on the distribution of individuals by real household expenditure per capita that cuts off the bottom 30 percent of the population. The data are for 1985 and described in Kanbur (1990). The regional division adopted is a fairly standard one in Côte d’Ivoire and points out the extreme poverty in the northern Savannah region. The incidence of poverty is 61.3 percent in that region, more than double the 30.0 percent incidence for all of Côte d’Ivoire. Moreover, it is not only the incidence but the depth of poverty in the Savannah region that is a problem. This is shown by the fact that as α increases from 0 through 1 to 2, the contribution of the Savannah region to national poverty increases significantly.

Table 3.1.Côte d’Ivoire: Decomposition of the Pα Class of Poverty Measures by Region
P0 ValueContribution to National PovertyP1 ValueContribution to National PovertyP2 ValueContribution to National Poverty
(In percent)(In percent)In percent)
Abidjan0.0523.30.0101.90.0031.2
Other urban0.1299.70.0296.40.0115.1
West Forest0.21110.60.0598.70.0247.3
East forest0.45637.50.15136.60.07035.3
Savannah0.61338.90.25146.40.13151.1
All0.300100.00.102100.00.049100.0
Source: Kanbur (1990).
Source: Kanbur (1990).

Nonmonetary Measures

As argued earlier, although in principle all aspects of the standard of living might be reducible to a single monetary measure, there are some dimensions in which it is difficult to so reduce. Access to education, literacy, quality of education, health care and its quality, drinking water, and basic housing amenities, and so forth, are all dimensions that seem to fall into the category. The term “basic needs” is sometimes used to capture the distinction between them and more conventional income-expenditure measures. However, it is to be expected that income-expenditure may well be correlated with achievements along the other dimensions. If the achievement, or lack of it, compounds a low income-expenditure measure, then this is significant from the policy point of view. What is also significant for policy is that the bulk of these basic needs is usually supplied by the government. Cuts in government expenditure in these areas—unless managed and targeted in particular ways—may well end up worsening basic needs achievements of the poor.

Table 3.2 summarizes some information on nonmonetary indicators for the poor and nonpoor in Côte d’Ivoire. Perhaps not surprisingly, the poor have very much lower literacy and school attendance rates than the nonpoor. The primary school age attendance rates, for example, are 36.5 percent for the hard core poor and 63.6 percent for all Ivoiriens. Similarly, the poor consult health personnel less frequently than the nonpoor, and they have lower rates of preventive consultation.

Table 3.2.Côte d’Ivoire: Some Nonmonetary Indicators
Hard-Core PoorPoorAll
Literacy (reading) rate (in percent)16.220.736.0
Literacy (writing) rate (in percent)12.616.231.3
School attendance rate, age 7–12 (in percent)36.546.863.6
Percentage of ill individuals who consulted health personnel39.342.551.5
Percentage of ill individuals who consulted health personnel in a hospital19.223.634.6
Percentage having preventive consultation14.916.222.7
Source: Kanbur (1990).
Source: Kanbur (1990).

Of course, much more detailed analysis is required before the reasons for these achievements in health and education can be precisely identified. However, as an adjunct to the income-expenditure measures of poverty, figures such as those in Table 3.2 are useful in serving to remind us of other dimensions of poverty.

The Nature of IMF-Supported Programs, Their Macroeconomic Rationale, and Criteria for Judging Their Success

IMF-supported adjustment programs have many dimensions. IMF, Fiscal Affairs Department (1986) classifies the policy instruments in 117 Stand-By Arrangements and 23 Extended Fund Facilities, from January 1980 to December 1984, into 7 major categories (monetary and financial policies, public sector policies, external debt policies, exchange and trade policies, wages and prices, other structural adjustment measures, and others). Each in turn has several subcategories. For example, in public sector policies various measures to restrain central government current expenditures are included, such as freezing or reducing numbers of government employees and their wage increases, capping or reducing food and fuel subsidies, or improving expenditure control mechanisms. With more recent developments of the Structural Adjustment Facility and the Enhanced Structural Adjustment Facility, greater emphasis on structural adjustment measures, such as a shift of activities from the public sector to the private sector, or removal of relative price distortions, is to be expected.

However, despite the plethora of measures often listed in IMF-supported programs, it should be clear that some measures are more important than others, and that the measures have a rationale in terms of a particular view of the nature of macroeconomic imbalances and their cure. To cut through the complications that would inevitably arise in assessing the poverty impact of each of a detailed set of measures specified or to be specified in a program, the approach adopted here is twofold—first, to analyze the impact of what are held to be the most important measures, and second, to do so in light of how they are judged to improve macroeconomic imbalances.

The central macroeconomic imbalance that attracts attention in IMF-supported programs is, of course, the balance of payments. Typically, it is an unsustainable deficit in the external balance that prompts a country’s approaching the IMF. The macroeconomic framework that is used to analyze the problem is focused on the balance of payments (IMF, 1987), and the success of a program is judged by the restoration of external balance within a limited period of time while not sacrificing other macroeconomic objectives.

The macroeconomic framework underlying IMF-supported adjustment programs is too well known to require detailed rehearsal. It is nevertheless useful to have a brief statement of the overall framework as a backdrop to our subsequent discussion on poverty and adjustment. The nature of the macroeconomic problem that initiates a country’s approaching the IMF might be summarized as an excess of aggregate demand over aggregate supply. The manifestations of this are inflationary pressure in the economy and a balance of payments deficit. The balance of payments deficit is only problematic insofar as it is unsustainable. Many countries have historically run large deficits as they imported capital to aid their development—in other words, the deficits were part of a rational intertemporal allocation plan, which was sustainable because of the perceived productivity of the ongoing investment. However, it is believed that in many or most cases the deficits being run by developing countries are not at their sustainable levels. They therefore have to be cut—if they are not cut in a planned way, then a “hard landing” is likely, with unpredictable and possibly hazardous consequences.

The deficit can clearly be cut in one of two ways—reducing aggregate demand and increasing aggregate supply. The latter is clearly preferable of the two, but it is also the slower of the two routes. If the nature of the macroeconomic problem is to reduce the deficit within a short horizon, greater reliance will necessarily have to be placed on quickacting measures—in practice, this means measures that curtail aggregate demand. However, a given change in aggregate demand is consistent with many different compositions of the change, and the same is true for aggregate supply. Thus, a reduction in demand can come about through a reduction in the absorption of the public sector or the private sector, of consumption or of investment. A major focus in IMF programs is the reduction of public sector absorption through increasing revenues and reducing expenditure. Here again, any given total reduction in expenditure is consistent with a number of different patterns of reduction. Sometimes, for example, IMF-supported programs seek to cap expenditure on food and fuel subsidies.

On aggregate supply, there are similarly many ways of achieving a given increase in supply. One might attempt to eliminate inefficiencies in production directly, for example, by reforming the parastatal sector. Thus, the focus on structural reforms helps a country improve the efficiency with which it uses its scarce resources. Or one might attempt to improve allocative efficiency through relative price measures, for example, by reducing the wedge between consumer and producer prices for particular commodities (such as rice in West Africa). One important relative price is that between tradable and nontradable goods. As IMF (1987, p. 6) notes: “The dichotomy between “nontradable” and “tradable” goods has become a principal analytical tool for analyzing devaluation and other expenditure-switching policies.” Expenditure switching policies become important if there is a rigidity in the relative price between tradable and nontradable goods—perhaps because of a rigidity in wages. However, Like all relative price measures their success depends on elasticities of supply response, the short-run values of which may be low.

To summarize, the following seem to be the most important aspects of IMF-supported programs:

  • (1) Reducing absorption through monetary restraint;
  • (2) Reducing absorption through government expenditure restraint;
  • (3) Increasing the production of tradables by increasing their price relative to nontradabies;
  • (4) Increasing the production of tradables and nontradabies by eliminating price distortions and production inefficiencies within these sectors; and
  • (5) Mobilizing domestic and foreign savings to promote growth.

Analytical Framework for Assessing Poverty and Adjustment

What Are the Questions?

Before specifying a framework for linking macro adjustment to micro poverty, we must specify the questions to which the framework is meant to provide an answer.1 Here are some questions which might be of interest to the IMF:

(1) What has happened to the extent and nature of poverty over the life of an IMF-supported program? This purely factual question could be answered directly if, for example, there was an annual household income and expenditure survey in the country. Such data simply do not exist in many developing countries, although a major effort is now under way in the World Bank to create the institutional capacity for collecting such data in sub-Saharan Africa. In the absence of household-level data, we may use more aggregated data on sectoral incomes to get a feel for what has been happening. It should be noted, of course, that what happens to the extent and nature of poverty over the life of a program is a product not only of program policies, but also of exogenous events (e.g., external terms of trade developments, natural phenomena) and preprogram policies.

(2) What would have happened to the extent and nature of poverty in the absence of an IMF-supported program? This question immediately raises further questions, since an answer to it requires not only a model of the economy but also a counterfactual on what would have happened to policy instruments without an IMF-supported program. Notice, however, that these issues arise not only for poverty but for any economic variable. Even for a basic macroeconomic variable such as the balance of payments, a counterfactual assessment of what would have happened without an IMF-supported program requires a model and a view on what would have happened to key policy instruments. The problem of the counterfactual is not peculiar to the poverty question.

(3) What would have happened to the extent and nature of poverty if the IMF-supported program had been different? It is important to specify the sense in which the IMF-supported program could have been different. Clearly, if more resources had been available, then the expenditure reduction would have needed to be less drastic, and poverty would have been less adversely affected. The appropriate comparison is clearly with an alternative program that achieves the same adjustment with the same resources over the same period. If this is accepted then the focus is on alternative compositions of expenditure reduction or supply enhancement that lead to the same deficit reduction over the same period. Could some alternatives be better than others from the poverty point of view?

(4) What might happen in the future, in the absence of an IMF-supported program? There are several routes that a country might take. One of them is to, in fact, enact the measures that would have been in an IMF-supported program anyway. But without the injection of resources that an arrangement with the IMF makes possible, such a “go it alone” strategy may cause more hardship. Or the country might try a different combination of instruments, accepting that it would not have access to IMF resources. It might, for example, impose or intensify quantitative controls to deal with the external deficit. The consequences of this for poverty should also be worked out, since it is an alternative to the IMF package.

(5) What should be the nature of a future IMF-supported program that has the least adverse effects on poverty? This question is closely related to the last one, and in many ways is the central question for policy. It requires a stipulation of the adjustment strategy, a macro model of the economy that links instruments to outcomes, and a menu of instrument combinations, all of which give rise to the same adjustment in the balance of payments deficit but (it is to be hoped) different poverty consequences. It should be clear, therefore, that if one is analyzing the impact of an instrument one should focus not only on its impact on poverty but also on its impact on the deficit—it is the poverty impact “per unit of deficit reduction” that is important. Throughout our discussion, this aspect will be stressed.

The Basic Approach

The basic approach adopted here is that of viewing the macroeconomic instruments in IMF-supported programs as affecting household real incomes either directly or indirectly through their effect on the production structure of the economy. We view households as drawing their incomes from different production activities, supplying factors of production to different sectors. As the composition and level of national output and income change, so will the distribution of household income. It is important, therefore, to have a characterization of how households in general, and poor households in particular, tie in with the production structure of the economy. This is where a poverty profile comes in, providing a policy-relevant guide to how sectoral changes in income might feed through to household incomes.

It should be emphasized that the level of production disaggregation required in an operational setting will depend upon the instrument in question and the policy being considered. Thus, if the instrument is the nominal exchange rate and the policy is devaluation so as to increase the output and income of the tradable goods sector as a whole, it would seem appropriate to use the tradable/nontradable disaggregation. However, if the policy instrument is tariff reform, then the disaggregation should be importables, exportables, and nontradables. Since the macroeconomic analysis must have made (implicitly or explicitly) the relevant disaggregation in order to arrive at forecasts of the impact on the balance of payments, it must also give some indication of the outputs and incomes of the different sectors after the reform. It is these outputs that can be used to generate implications for the household distribution of incomes and for poverty, provided certain assumptions are made.

Another basic question that arises in analyzing the micro impact of macro changes is the extent to which feedbacks should be taken into account. Our approach is driven by the need to translate IMF-supported measures, which are primarily directed toward improving the balance of payments, into impacts on poverty. Our approach is thus to take feedbacks into account only insofar as the macroeconomic analysis does so. Presumably, the forecasts of balance of payments based on the implementation of an IMF-supported program take feedbacks into account to the extent they are considered relevant, and to the extent it is feasible to do so. We take the output of the macroeconomic analysis as given—our task is to convert it into implications for poverty.

Poverty Decomposition and Some Impacts on income Poverty

Aggregative Impact of Absorption Reduction

Absorption reduction measures reduce total domestic expenditure. They act quickly and for this reason are important in adjustment programs, A very simple and highly aggregative approach to assessing the impact of this on poverty is to assume that all household incomes are reduced in the same proportion as the absorption reduction. If we have the current distribution of individual real income per capita, we can shift this accordingly, and for a fixed poverty line, we can calculate the new numbers in poverty and the extent of poverty. This is bound to be higher relative to the current position—somebody has to lose during an austerity program, and all we are doing is developing a benchmark by assuming that everybody loses proportionately to their income. Of course, it is the absorption reduction that gets you the balance of payments deficit reduction—it is the trade-off that is important.

One way of quantifying this trade-off is to use the Pα measure and to see how that changes with absorption reduction under the assumptions made. If all incomes fall by a factor θ, then for small θ it can be shown that

Notice that this result does not depend on the specific functional form of the income distribution, only on the form of the poverty index and on the assumption that the reduction in total income does not change the relative distribution of income among income groups. For α = 1, we get an approximation

where I is the income gap ratio defined previously. Thus if I = ⅓, which is close to the figure for Côte d’Ivoire from Kanbur (1990), it follows that a 100 θ percent fall in absorption will lead to a 200 θ percent increase in the P1 measure (as a first order of approximation). Notice that the impact is inversely proportional to I, but upon reflection, this is not at all paradoxical—if the extent of poverty is already great, then a given decrease in incomes will have a proportionately smaller impact.

The above assumes that the reduction is proportionate to current household income. To get an idea of how a more unequal sharing of the burden could affect matters, consider the case where the θ percent reduction in total consumption is divided equally between households—so that the poor bear a disproportionate share of the burden. It can be shown that, for small θ,

where z is the poverty line and M is mean income. For α = 1, we get as an approximation

Thus, if z = (½) M and I = ⅓ (approximately the Côte d’Ivoire’s figures again),

so that the proportionate increase in poverty is six times the proportionate decrease in income.

Unemployment Effects of Absorption Reduction at the National Level

The foregoing analysis assumed that expenditure reduction affected all incomes uniformly, whether additive or multiplicative. However, if there is wage rigidity then a contraction in aggregate demand will be accompanied by an increase in unemployment. The extent of this increase is, of course, a matter for country-specific analysis, and one of the outputs of the modeling underlying the macroeconomic analysis of an IMF-supported program should be the impact on unemployment.

Let Pα,E be poverty among the employed and Pα,U the poverty among the unemployed. Assuming these to be mutually exclusive and exhaustive categories, if XU is the unemployment rate and XE is the employment rate, (XU + XE = 1), then national poverty is

The impact on poverty of an increase in the unemployment rate following from a change in policy instruments is given by

where π is the policy instrument, and we assume that the newly unemployed take on the poverty characteristics of the existing unemployed.

The differential in poverty between the unemployed and employed is thus a key parameter that should be borne in mind in assessing IMF-supported programs. If we take P0 as the measure, then, on the basis of 1981–82 data for Sri Lanka, the differential incidence of poverty among individuals living in households where the head was employed or unemployed was 11.2 percentage points. Thus, under the assumptions made, every percentage point increase in the unemployment rate would lead to an increase in the national incidence of poverty of 0.112 percentage points.

Income Effects of Absorption Reduction at the Sectoral Level

The analysis has so far assumed a certain uniformity in the pattern of absorption reduction across the different sectors of the economy. In fact, in most IMF-supported programs there is an attempt to control the government’s absorption, and this has certain specific effects. First, there is an attempt to reduce the number of government workers and to restrain their wages. To discover the possible impact of this on poverty, we need to identify the extent of poverty in this sector. Table 3.3 presents poverty figures for individuals living in households headed by government sector employees in Côte d’Ivoire in 1985. The incidence of poverty among these individuals is 3.1 percent, compared with 30.1 percent for all Ivoiriens. Now, a 1 percent cut in the incomes of these individuals would, using the formula in equation (7), lead to a 5.6 percent increase in the P1 measure in this sector, but since this sector only contributes 0.8 percent to national poverty, the impact on the national P1 would be only 0.04 percent. Of course, these figures are only approximately true, but they should not be surprising given the pattern in Table 3.3. While there may be some poor government workers badly affected for whom special compensatory measures might be instituted, the overall effect on poverty at the national level is likely to be small.

Table 3.3.Côte d’Ivoire: Decomposition of the Pα Class of Poverty Measures by Socioeconomic Group
P0 ValueContributionP1 ValueContributionP2 ValueContribution
(In percent)(In percent)(In percent)
Export crop farmers0.36522.30.11420.40.05018.8
Food crop farmers0.49559.00.18464.10.09065.9
Government workers0.0311.30.0020.20.0020.1
Formal private sector0.0611.90.0090.80.0030.6
Informal sector0.19315.50.06214.50.03014.6
All0.301100.00.103100.00.049100.0
Source: Kanbur (1990).
Source: Kanbur (1990).

Absorption Reduction: Food and Fuel Subsidies

Capping or reducing food and fuel subsidies is another major component of some IMF-supported programs. This method of absorption reduction has very particular effects depending on the commodity involved and the nature of the subsidy. However, the question still remains—what is the poverty impact of different schemes per unit of expenditure saved? Besley and Kanbur (1988a) provide a general theoretical analysis of this question. Here, we provide an operationally oriented application of those results.

If the subsidy is given in the form of a general ration, which is subsidized below market prices, and if there are no obstacles to the resale of the ration, then the scheme is equivalent to an income transfer equal to the ration multiplied by the subsidy relative to the free market price. This closely approximates the system in Sri Lanka before 1977. Since the ration and the subsidy would be the same for all in a universal scheme, the government would effectively be engaging in an equal income transfer to everyone with ration cards. A cutback in either the quantity of the ration or the subsidy would be equivalent to a cutback in the income transfer. Thus, the formula in equations (8) and (9) would be relevant in gauging the percentage poverty impact per 1 percent cutback in expenditure.

On the other hand, if the subsidy was given in the form of an import subsidy, then upon its removal consumers lose in proportion to their current consumption, while producers gain in proportion to their current production. For small changes, the overall cut in subsidy to consumers is the cut in subsidy per unit multiplied by total consumption, and the overall improvement for producers is the increase in price multiplied by their net marketed surplus. In evaluating the poverty impact, therefore, we need to know the poverty configuration among producers and consumers of this commodity, as shown in Table 3.4, which gives the net consumption of a subsidized commodity by various groups. It is assumed that poor net producers supply an amount Q11 of the commodity, while nonpoor net producers supply Q12. Poor net consumers consume Q21, while nonpoor net consumers consume Q22. To a first order of approximation, a one-unit reduction in the price subsidy saves the government (Q21 + Q22) − (Q11 + Q12), but it costs the poor Q21Q11. If consumption by the poor dominates production by the poor, then removing the subsidy has net poverty costs that have to be set against the deficit reduction.

Table 3.4.Net Consumption of a Subsidized Commodity, by Net Producer/Net Consumer and Poor/Nonpoor
PoorNonpoorAll
Net producerQ11Q12−(Q11 + Q12)
Net consumer+Q21+Q22+(Q21 + Q22)
AllQ21Q11Q22Q12(Q21 + Q22) − (Q11 + Q12)
Source: Kanbur (1990).
Source: Kanbur (1990).

Clearly, the answers to the question of the poverty cost per unit of reduction in expenditure depends very much on the commodity in question. The above schema can also be applied if producer price and consumer price are different. The poverty cost per unit of expenditure reduction on producers is then given approximately by Q11/(Q11 + Q12), and the corresponding expression for consumers is Q21/(Q21 + Q22). An illustration of the use of these expressions using household-level data on rice production and consumption in Côte d’Ivoire is to be found in Kanbur (1990).

Of course, the analysis up to now assumes no behavioral responses. This may be valid in the short run, but not so in the long run, where net producers may become net consumers, and quantities produced and consumed may change. The degree of complication introduced depends on how widely we wish to model the complete effects. If we focus purely on own-price effects, then the entries in the matrix in Table 3.4 can be simply adjusted for own-price elasticities of supply and demand. If we wish to take into account the full demand system, then the techniques developed in Ravallion and van de Walle (1988), which rely on an estimation of the equivalent income function, may be used. Substitution in production can also be introduced, using the “multimarket” methodology of Braverman and Hammer (1988). We would suggest, however, that as a first cut it would be useful to produce Table 3.4 for all commodities being considered for policy reform.

In fact, the ratios calculated from Table 3.4 can be helpful in comparing the poverty impact of price changes of alternative commodities. For simplicity, let production be zero. Then, Table 3.4 suggests that an approximate indicator of poverty impact per unit of deficit reduction is simply Q11/(Q11 + Q12), This can be calculated from household income expenditure surveys. Notice, however, that the critical ratio is not the fraction of their total expenditure that poor households spend on this commodity, but the fraction of the total expenditure in the economy on this commodity that is accounted for by poor households. The latter is what is needed. Yet, one often finds the former purporting to play a similar role in many poverty profiles that have been produced.

Expenditure Switching: Effect of a Change in Relative Price of Tradables and Nontradables

The central reallocation attempted in IMF-supported programs is that between tradables and nontradables through a change in their relative price, and this in turn is attempted through a nominal devaluation. Much of the literature discusses various nominal feedbacks through wages and other variables. Whether or not a nominal devaluation will actually succeed in raising the price of tradables relative to nontradables is an empirical issue, and is as much a matter of political economy as macromodeling. However, for our purposes, it will suffice to assume that a given change in the relative price has indeed taken place, and then to inquire as to its effects on poverty.

Since we are interested in expenditure switching, let us assume that total expenditure is kept constant. Then, a shift in the relative price, if successful, will alter the composition of national output in favor of the tradable sector. The impact on poverty depends on how exactly this alteration affects factor incomes and therefore household incomes. Over the very long term, we know that under assumptions of competition in product and factor markets, those factors used intensively in the tradable sector will benefit in terms of an increase in their incomes. Thus, if the tradable sector is relatively labor intensive, then the returns to labor will rise. To some extent, the computable general equilibrium models are capable of quantifying this result, since they in effect solve for the new price and quantity equilibrium following a policy change (Bourguignon, Branson, and de Melo, 1989). This change in wages can then be applied to an assumed distribution of the total wage bill to derive poverty impacts.

At the other extreme, in the very short run, the change in relative price will simply affect entrepreneurs’ incomes in the two sectors. Tradable sector entrepreneurs’ incomes will rise, so that if a large number of these are poor, there will be a positive impact on poverty. But nontradable sector entrepreneurs’ returns will fall. As they adjust by lowering output, there will initially be an increase in unemployment if wages are rigid downward. This will cause hardship temporarily, for which compensatory measures may be required. In any case, those who receive incomes from nontradable sector activities will be made worse off in the short to medium term. To get a quantitative expression for the poverty impact, first we need to get a forecast for the value of each sector’s output in the short to medium term following the policy shift. This output can then be distributed across household income groups according to patterns revealed in a base-year household income and expenditure survey.

For simplicity, suppose that household participation in the two sectors is mutually exclusive and exhaustive. Then, if output changes translate into income changes in fixed proportion, it can be shown that pure expenditure switching that increases the share of the tradable sector, πT, leads to a poverty impact of approximately

where M is mean income, and subscripts N and T refer to sectors. This expression, which was introduced in Kanbur (1987) and can be derived, can be recognized to be the development of equation (6) in the two-sector case. If output changes in a sector translate into income changes in additive fashion, that is, all incomes in that sector gain or lose equally, then

These expressions are, in principle, calculable from household income and expenditure surveys. What they highlight is the need to be able to develop poverty profiles along the tradable-nontradable dimension. This could be attempted using information on occupation of the head of household, for example, or information on the main crop grown. An attempt to develop a production-sector-oriented poverty profile is illustrated in Table 3.3. Export crop farmers, food crop farmers, and formal private sector workers might be grouped together into the tradable sector, while government workers and the informal sector (largely urban) could be classified as nontradable. Using the information in Table 3.3, the expressions in equations (13) and (14) can be calculated.

From equation (14), it can be seen that if α = 1, then expenditure switching reduces poverty if the incidence of poverty in the tradable sector exceeds that in the nontradable sector. It seems to us that attempts to establish whether this is actually so in specific cases will have a high payoff. However, a number of objections can be raised against the foregoing analysis. First, it takes an overly simple view of how greater or lesser profitability in a sector translates into factor incomes. If the tradable sector is dominated by smallholders, then an increase in the price of tradables translates directly into their incomes. However, if there is a large landless class, then the extent to which they benefit may be circumscribed by the workings of the local labor market. Thus, if there is essentially surplus labor in this sector, then all the gain will be appropriated by landlords, and the impact on poverty will be minimal. Similarly, if there are wage rigidities in the labor market serving the nontradable sector, then unemployment may result, so that there will be an unequal distribution of the income loss. We might attempt to model these directly, but a shortcut might be to assume that income loss is additive and income gain is multiplicative, that is, the distribution of gains and of losses is not equalizing. Then we get a hybrid expression

which could, in principle, be estimated from data.

Finally, we have focused on sources of income and not on uses of income. A change in the relative price of tradables and nontradables will also affect households as consumers. But the effects of these on poverty can be analyzed in similar fashion to the price changes accompanying food subsidy removals, provided of course that a link can be made between available data on commodity demands on the tradable-nontradable divide.

Another objection is that some households may have sources of income both in the T and in the N sectors. In classifying households as definitely belonging to one sector or another, one may be making a major miscalculation of the impact of expenditure switching, since poor households may well be “diversified” with respect to incomes from these two sectors. We do not have good information on this, and it is an important item on the research agenda.

Other Relative Price Reforms

Most other types of relative price reform in IMF-supported programs can be analyzed using one or the other of the techniques mentioned above. For example, a tariff reduction would alter the relative price between exportables and importables, as well as between nontradables and importables. Any forecast of the effects of these on the balance of payments must contain an account of how the output levels in these three sectors change. Once we have these, then, making assumptions on factor markets, the price changes can be translated into income distribution effects.

Imposition of Tariffs and Quantitative Controls

One route that is often taken by governments when faced with an unsustainable balance of payments deficit is to impose quantitative controls on foreign exchange and on imports. This is, in fact, an alternative to an IMF-supported program that some countries have followed. What is the poverty impact per unit of deficit reduction of such a strategy?

Let us, first of all, take the case where a tariff is imposed on an imported luxury item largely consumed by the rich. On the consumption side, this is a good way of reducing the deficit without a significant poverty impact. In addition to reducing the deficit, the government raises revenue. However, domestic producers of close substitutes for this product benefit as their prices increase. The empirical question in this case is the distribution of income from production of substitutes for luxury imports. Motor cars, for example, are capital intensive. However, for smaller personal items the substitutes could well be labor intensive.

The use of a quota rather than its tariff equivalent has a similar economic effect, except that the income that previously accrued to the government as tariff revenue accrues to those who have the quota as rent. There can be little doubt that these rents do not go to the poor (except perhaps through the demands of the recipients of these rents for services), although it would be difficult to establish this on the basis of household income-expenditure surveys. However, it might perhaps be stated as a reasonable generalization that the replacement of quotas by their tariff equivalents, which is the first stage in many liberalization sequences, is by and large not regressive in its distributional implications.

Once one moves away from luxury imports to commodities consumed further down the income distribution, then tariffs on these to protect the balance of payments begin to have greater impact on consumer poverty, an effect that can be quantified to a first order by the ratio of poor consumption to total consumption. However, one needs to know how the production of import substitutes, which is made more profitable relative to exports, is linked to poor incomes. But data on production structure are rarely available in such detail that we can separate importables from exportables—this is a problem for quantification.

Some Further Considerations

The foregoing analysis has focused on the parts of an IMF-supported program instrument by instrument, as it were. What of the total package? In principle, the different effects can be combined into a coherent whole. Thus, if there is expenditure reduction that leads to a reduction of all incomes and an increase in the share of the tradable sector by dπT, the net effect on poverty will, to a first order of approximation, be given by a combination of equations (6) and (13), say, if the basic model is that all income change is multiplicative in nature. Various other combinations are possible, and their usefulness can be judged on a case-by-case basis.

Throughout this chapter we have focused on poverty impact per unit of deficit reduction. To complete the picture, we would need the consequence for poverty if the deficit is not reduced. For example, among the most important considerations in the reduction of subsidies are the adverse implications of the government deficit, monetary expansion, and inflation. To complement the analysis, we would need to think through the poverty impact of inflation.

One issue that arises in combining the effects of different instruments is that some effects would take longer to materialize than others. In simply combining the different effects discussed in earlier subsections, we would have to make the assumption that we were looking at the situation after the full effect had worked through. This is not just a problem for poverty analysis, of course. It is as much a problem for forecasting macroeconomic effects. The strategy suggested here is that the poverty analyst rely on the macroeconomist for the magnitudes of the macro effects over a given time horizon (e.g., the extent of expenditure reduction and the extent of change in the composition of output as between tradables and nontradables), and then use this information to arrive at poverty impacts.

Compensatory Measures and Targeting

Up to this point we have attempted to derive, under various assumptions, the poverty impact of adjustment measures. As we have seen, there are bound to be gainers and losers from attempts to alter the structure of the economy and to curtail aggregate demand. Our focus has been on the impact on poverty at the national level, that is, the net effect of the increase in poverty among losers and the decrease among gainers. However, it has been argued that this masks the dire position of those whose income is drastically cut, or those who face dramatically higher prices or unemployment, as a result of the adjustment process. Can there not be short-term compensation in these cases? The question arises not only in terms of the very poor, but also the not so poor who are articulate enough to oppose the adjustment strategy and do so successfully. However, in this discussion, we focus on the poor.

The crucial question is whether the compensatory measures can be targeted to the poor at little cost, and whether they can be designed so that their operation does not in fact impede the adjustment process itself. Another important question is whether the resources for the compensatory measures are to be made available externally or are to come from within the country. Let us suppose that someone previously employed in a nontradable or import-competing activity loses his job as a result of a devaluation or liberalization of the trade regime. If he stays in the labor market associated with the activity, we may suppose that he will eventually get employment at a lower wage as the market adjusts. He has the option of moving into an exportable producing activity, but (1) he may lack the necessary resources for migration if this activity is in the rural sector, or (2) he may lack the necessary training even if no migration is involved. In any case, his short-term position is made difficult by the lack of employment. Compensatory schemes that are temporary and directed to such individuals would not necessarily impede adjustment while providing short-term respite. Indeed, by retraining, they may well aid adjustment.

Compensatory measures may be used when a generalized food subsidy is removed for fiscal or efficiency reasons. A certain sum could be made available to be used to grant food subsidies targeted to the poor. One of the well-known examples of this is the retargeting of the Sri Lanka rice ration program after 1977. In fact, this method could be used not only to protect the poor who are in danger of losing a specific commodity subsidy but also to create a new subsidy for those out of work because of adjustment.

However, the problems with such compensatory schemes should not be overlooked. These are discussed in Besley and Kanbur (1988b) and are to be found in all attempts to “fine-tune” transfers to particular target groups. We may be able to identify the characteristics of losers on the basis of a priori reasoning supported by household surveys, but it is quite a different matter to identify those characteristics and monitor the scheme on the ground. The administrative costs of attempts at fine targeting can be large, as has been seen in developed countries (unfortunately, there is very little analysis of the costs of administering such schemes in developing countries). However, as discussed before, more generalized targeting, such as switching subsidies to foods consumed by the poor, may be possible. If external resources were available to achieve such retargeting it would certainly ease the pain of adjustment, but detailed country-specific analysis would be required to identify possibilities for such support.

Conclusions

The object of this chapter has been to sensitize the macroeconomist to the sort of issues that might arise in an analysis of the complications of adjustment programs for poverty. However, our focus throughout has been on identifying methodologies that have some hope of being made operationally feasible. The central methodology suggested here is one based on a policy-relevant poverty profile that can be used to translate the impact of policy instruments on the size and composition of national income into the consequence for poverty. In this context, the use of subgroup decomposable poverty indices is recommended, and we have given some examples of how they might be used. The extent to which feedback effects should be taken into account, that is, the extent to which the analysis should be general equilibrium in nature, is a question faced by the macroeconomist as much as the poverty analyst. The same is true of the extent of disaggregation entertained. However, our approach to both has been to accept the norm and findings of the macroeconomic analysis of IMF-supported programs, and then to use the output from this analysis in the expressions we have derived for poverty impacts. Of course, it would always be better, ceteris paribus, to have as complete a model as possible. But in the operational context, the shortcuts suggested here may prove useful.

Our approach does depend on the availability of data on the basis of which poverty profiles can be constructed. In the absence of such data, crude poverty rankings of sectors and socioeconomic groups based on secondary information might suffice. But it is a useful truism to note that without the data for poverty analysis, an analysis of poverty impacts cannot be done. However, it should be noted that the excuse of a lack of good-quality data to do poverty analysis is likely to become less and less acceptable over the next few years. The World Bank’ Africa Region has launched a major program of data collection on the basis of household surveys that will eventually cover about 25 African countries. Côte d’Ivoire already has such data; Ghana and Mauritania have completed one year’s survey activities; and Senegal is due to begin soon. Those involved in the design of IMF-supported programs will have considerable interest in this information as they begin to systematize and analyze.

References

    AnandSudhir and ChristopherHarris1985“Living Standards in Sri Lanka, 1973–1981/82; An Analysis of Consumer Finance Survey Data,”Applied Economics Discussion Paper No. 84 (Oxford: University of Oxford Institute of Economics and Statistics).

    • Search Google Scholar
    • Export Citation

    AtkinsonA.B.1987“On the Measurement of Poverty,”EconometricsVol. 55 (July) pp. 74964.

    BesleyTimothy J. and S.M.Ravi Kanbur1988a“Food Subsidies and Poverty Alleviation,”Economic JournalVol. 98 (September) pp. 70119.

    • Search Google Scholar
    • Export Citation

    BesleyTimothy J. and S.M.Ravi Kanbur1988b“The Principles of Targeting,”Policy Research and External Affairs Working Paper No. 385 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    BourguignonFrançoisWilliam H.Branson and Jaime DeMelo1989“Macroeconomic Adjustment and Income Distribution: A Macro-Micro Simulation Model,”Technical Papers (International) No. 1 (Paris; OECD Development Centre, March).

    • Search Google Scholar
    • Export Citation

    BravermanAvishar and JeffreyS. Hammer1988“Computer Models for Agricultural Policy Analysis,”Finance & DevelopmentVol. 25 (June) pp. 3437.

    • Search Google Scholar
    • Export Citation

    DeatonAngus1987“The Allocation of Goods Within the Household: Adults, Children, and Gender,”LSMS Working Paper No. 39 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    DeatonAngus and JohnMuellbauer1986Economics and Consumer Behavior (New York; Cambridge: Cambridge University Press).

    DonaldsonDavid and John A.Weymark1986“Properties of Fixed-Population Poverty Indices,”International Economic ReviewVol 27 (October) pp. 66788.

    • Search Google Scholar
    • Export Citation

    FosterJames and JoelGreer1984“A Class of Decomposable Poverty Measures,”EconometricaVol. 52 (May) pp. 76166.

    International Monetary Fund Fiscal Affairs Department1986Fund-Supported Programs Fiscal Policy and income Distribution IMF Occasional Paper No. 46 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    International Monetary Fund Fiscal Affairs Department Research Department1987Theoretical Aspects of the Design of Fund-Supported Adjustment Programs IMF Occasional Paper No. 55 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    KanburS.M. Rav61i1987“Measurement and Alleviation of Poverty: With an Application to the Effects of Macroeconomic Adjustment,”Staff Papers International Monetary FundVol. 34 (March) pp. 6085.

    • Search Google Scholar
    • Export Citation

    KanburS.M. Rav61i1990“Poverty and the Social Dimensions of Structural Adjustment in Cote d’Ivoire,”Social Dimensions of Adjustment in Sub-Saharan Africa Working Paper No. 2 (Washington; World Bank).

    • Search Google Scholar
    • Export Citation

    KynchJoceIyn and Amartya K.Sen1983“Indian Women: Well-Being and Survival,”Cambridge Journal of EconomicsVol. 7 (September/December) pp. 36380.

    • Search Google Scholar
    • Export Citation

    RavallionMartin and Dominique vande Walle1988“Poverty Orderings of Food Pricing Reforms,”DERC Discussion Paper No. 86 (Warwick: University of Warwick Development Economics Research Centre).

    • Search Google Scholar
    • Export Citation

    SenAmartya K.1976“Poverty: An Ordinal Approach to Measurement,”EconometricaVol. 44 (March) pp. 219232.

    SenAmartya K.1987The Standard of Livinged. byGeoffreyHawthorn (New York; Cambridge: Cambridge University Press).

Note: Originally prepared for presentation at an International Monetary Fund seminar on “The Implications of Adjustment Programs for Poverty,” November 1988.
1An annex describing the derivation of several of the key mathematical expressions presented in this section is available from the author on request.

    Other Resources Citing This Publication