The consequences of macroeconomic policies for income distribution and their impact on the poor is a topic that is attracting an increased amount of attention. Although the analysis of poverty and inequality has often been centered around microeconomic considerations, there has been, recently, a growing awareness of the macroeconomic dimension of distributional issues. This recognition has been strengthened by the need to implement severe stabilization programs in many developing countries, programs which have had an effect on economic activity, particularly in the short run, and which may have had a distributional spill-over. 1/
The channels through which macroeconomic policies affect income distribution are indeed intricate, with complexities arising not only from the difficulties in isolating, at the analytical level, the expected impact of individual variables, but also from the fact that these effects will differ depending on the nature and the composition of aggregate policy packages. Policies which could have, by themselves, clear-cut distributional consequences would, when combined with others, interact in a manner which blurs their ultimate impact. The effects of particular policies, therefore, cannot be firmly evaluated except in the specific context within which they are implemented, and such an evaluation could require, at the theoretical level, the design of quite complicated models.
A different but rather complementary alternative is to approach the distributional aspects of macroeconomic policies at the empirical level. This could facilitate the understanding of the policy tradeoffs arising in practice between macro-adjustments and income distribution and shed light on some of the dominant channels through which macro-policies and macro-variables affect income shares. Such an endeavor is complicated by statistical deficiencies and by the lack of available short-term data which would allow the tracing of the immediate as well as the lagged effects of macroeconomic policy. For this reason, only few empirical studies on this subject—which, indeed, has large political implications—have been carried out. Among these, Blinder and Esaki (1978), and Blank and Blinder (1985) have analyzed the effects of inflation and unemployment on income distribution in the United States, and the staff of the International Monetary Fund have undertaken two broad studies of the consequences of a number of Fund-supported adjustment programs on income distribution and on the poor (see IMF (1986); and Heller, et al. (1988)). 2/
This paper attempts to contribute to this rather under-researched area by considering an important case study—that of the Philippines in the period 1980-86. Relevant information is currently available from the integrated surveys of households, which were carried out beginning in 1978 by the National Census and Statistical Office (NCSO) of the Philippines. These surveys provide data on income, employment, and the earnings of individuals in the urban and rural sectors on a quarterly basis. Since quarterly national income accounts are available from 1980 onward, a quantitative approach could be adopted. The empirical study carried out here is undertaken within an analytical framework that intends to capture some of the specificities of the country studied and the policies adopted during the period, and attempts to trace the mechanisms through which macroeconomic policies and variables interact in affecting different income groups.
In the next section a simple theoretical model is presented from which an empirical specification is derived. In Section III the characteristics of the Philippine macroeconomic adjustment are described in some detail. Section IV discusses the empirical results and places them in the context of the Philippine developments. In Section V conclusions and some generalizations are presented.
II. Analytical Framework
1. The model
In order to provide an analytical framework for the quantitative assessment of the covariance between income shares and a number of indicators reflecting macroeconomic policies and to trace the possible channels through which these policies affect income distribution, a simple three-equation model is postulated here. The system describes the factors affecting relative real earnings and real wages for the higher and lower ends of the income distribution spectrum as well as the short-run determination of real income. The three basic equations are the following:
|Si/P||real earnings pertaining to income class i (in particular, L is the lowest and H is the highest income groups) 3/|
|wi||=||real wages of income class i|
|E||=||real exchange rate|
|g||=||real government spending|
|r||=||real interest rate|
|n||=||rate of inflation|
Equation (1) is an earning function which measures, for each income group, the importance of a number of variables which are assumed here to be the main determinants of the overall level of personal income. Earnings are determined, in proportions that differ among income classes and which are measured by the coefficients βi, by real wages, by changes in the real exchange rate, and by the level of overall economic activity. Specifically, increasing real wages and higher levels of economic activity are expected to have a positive effect on earnings (β1i, β3i > 0), while the influence of the real exchange rate could affect earnings of different income groups in different, even opposite, manners according to the composition of their activities in terms of traded and nontraded goods (β2i ≷ 0). Moreover, within-equation constraints could also be derived since it would be expected that earnings of low-income groups will be affected more by wages than by activity levels when compared with high-income groups, since self-employed income and capital income are probably a larger share of total earnings at high-income levels (β1L/β3L > β1H/β3H).
Equation (2), the real wage function, is a version of the Phillips curve which postulates a negative relationship between labor market conditions as measured by underemployment and the level of real wages. It also includes a measure of overall labor productivity, which is expected to have a positive effect on real wages. The magnitude of the effects of ρ and U on w at the different income levels are measured by the λi coefficients.
Equation (3) is an extended version of an IS curve, which relates aggregated real economic activity to fiscal and monetary variables as represented by the level of government spending and by the real rate of interest. In addition, it incorporates a time trend and the rate of inflation, which provide an additional variant of a short-term, augmented Phillips curve. 4/
Given the lack of infomation about real wages by income class, the structural equations cannot be directly estimated. To obtain a reduced-form version for the model—which in its present version is a recursive model—we substitute the log of equations (3) and (2) into equation (1). Then the ratio between the low-income group earnings and the high-income group earnings is calculated. The resulting equation is the logarithm of the ratio of the earnings of the low-income to the high-income groups as a function of productivity, underemployment, government spending, the real exchange rate, the real rate of interest, and the rate of inflation, with a constant and a time trend. 5/ The coefficients are the following:
For the most part, the β coefficients in the reduced form represent the net difference between the effect of a variable on the earnings, or on the income shares, of the low-income group and its effect on the high-income group. A positive coefficient is “progressive” in the sense that it indicates a relative improvement in the share of the low-income group with respect to the high-income group, while a negative coefficient is “regressive” in the same sense. Take, for example, the coefficient of the rate of underemployment, β1λ2 = (β1Lλ2L - β1Hλ2H) ≷ 0. It reflects the relationship between the underemployment semi-elasticities of real wages for each income group (the slope of their Phillips curves) weighted by the importance of wages in the total earnings of each group. If β1λ2 < 0, an increase in underemployment tends to worsen the relative share of the low-income group, reflecting a higher sensitivity of low-income wages to the employment conditions and/or a higher share of wages in the total earnings of the lower-income group. 6/ Similarly, a positive sign for the coefficient of the real interest rate variable, β3λ3 > 0, implies a negative impact of interest rates on economic activity combined with a stronger effect of economic activity (when employment effects are accounted for independently) in determining the level of real earnings for the high-income groups compared with the low-income groups (β3H > β3L).
2. Graphical presentation
The functional relationships described by equations (1) - (3), as well as the interactions among them, are illustrated in Figure 1. The upper quadrants I and II represent the relationship between real wages and earnings for the low-income and the high-income groups, respectively. The slope of the curves is given by the βi coefficients, while the parameters determining the location of each curve are real income (y) and real exchange rate (E). If changes in real wages account for a larger portion of earnings fluctuations in low-income groups than in high-income groups, β1L > β1H, the earnings curve in quadrant II will be steeper than in quadrant I. An increase in real income will shift both curves upward, the extent of the shift being determined by the coefficients β3i, while changes in the real exchange rate may shift the curves in either direction, depending on the type of effects that earnings arising from tradable and nontradable goods may have on each income class. Quadrants III and IV represent the Phillips curves of the L and H groups, respectively, with the productivity level as a parameter. If unemployment (or underemployment) has a stronger impact on the level of real wages in the low-income than in the high-income bracket (that is, |λ2L| > |λ2H|), the Phillips curve in quadrant IV will be steeper than in quadrant III.
Given an initial position represented by a general level of unemployment equal to U0, real wages will be WL0 and WH0 for the low-income and the high-income groups, respectively. At these real wage levels, total earnings are determined along the respective earning curves with the low-income group at point AL and the high-income group at point AH. The difference between the two levels of earnings is represented along the common vertical axis by the shaded area to the right, with the length of the segment measuring the extent of distributional inequality. An increase in the rate of unemployment to U1 induces a movement along the Phillips curves (from TL to VL in panel III and from TH to VH in panel IV), reducing real wages to wL1 and wH1. At the lower wage levels, earnings will fall for both income groups (to BL and BH). Given the set of parameters assumed in Figure 1, however, the higher level of unemployment causes a larger fall in the earnings of the low-income group, and this leads to a widening in inequality as measured by the larger difference in earnings shares depicted by the shaded area at the left of the vertical line.
The effects of other macro-variables can also be analyzed using this framework. For example, if b3L < b3H, indicating a higher positive effect of economic activity on the earnings of high-income compared with low-income groups, an increase in government spending would tend to increase inequality by shifting upward the earnings curve more in panel II than it would in panel I, increasing, for a given level of real wages, the vertical difference between SH and SL. Similarly, if real devaluations are progressive, that is, they improve the earnings of the L group relatively more than those of the H group (if, for example, low-income farmers producing exportable crops are benefited by an improvement in their competitiveness), the earnings schedule in panel I will shift upward more than in panel II. Moreover, it is conceivable, as mentioned above, that both curves could move in opposite directions.
III. The Philippine Experience with Macroeconomic Adjustment in the 1980s
Following a decade of substantial growth, the Philippine economy reached a crisis situation in 1983. Although GNP and exports grew rapidly during the 1970s (see Table 1), manufacturing growth largely occurred in highly protected sectors and the incentive system was not conducive to a broad-based export expansion. Investment rates were about 30 percent of GNP, but did not result in an efficient allocation of productive capacity. At the same time, government intervention in productive activities was on the rise.
|Real GNP growth rate (%)||6.4||5.0||3.4||1.9||1.1||-6.8||-4.2||2.0|
|Current account deficit/|
|GNP (%) (- = deficit)||-4.6||-5.4||-5.4||-8.1||-8.1||-4.0||0.1||3.3|
|Real exchange rate 1/|
|(Index 1980 = 100)||93.2||100.0||84.8||87.9||73.8||72.9||79.2||61.2|
|Consolidated public sector|
|Real interest rates|
Period average; increase indicates real effective appreciation.
Period average; increase indicates real effective appreciation.
The deterioration in the terms of trade in the late 1970s, followed by rising interest rates in the early 1980s, brought out into the open these underlying structural weaknesses in 1983. Growing import requirements were not met with higher exports receipts and the current account deficit increased from an average deficit of 4.6 percent of GNP in the 1975-80 period to nearly 8 percent in 1982. As most of these deficits were financed by foreign borrowing, total external debt increased from US$5 billion at the end of 1975 to US$24 billion in 1982. The growing public sector gap resulted in the doubling of the consolidated public sector deficit from 3 percent of GDP in 1980 to about 6 percent in 1982. The adjustment continued to be postponed by the increasing short-term debt but the external debt problem became insurmountable as commercial lenders refused to roll over short-term credits or extend new loans. Capital flight followed the political crisis, and this further exacerbated the balance of payments problems. The option of continuing to postpone the adjustment was no longer available.
The Philippines started its stabilization process in late 1983 and, by 1985, the short-run external imbalances were brought under control. The current account reversed from a deficit equivalent to 8 percent of GNP in 1983 to a small surplus in 1985, with both imports and exports falling at the same time. This was achieved through a large drop in real absorption and a consequent sharp output contraction. Drastic cuts in public expenditures and restrictive monetary policies resulted in a drop in GNP of 6.8 percent in 1984 and 3.8 percent in 1985. The consolidated public sector deficit was reduced to 2.6 percent of GDP in 1985. Total investment fell from 27 percent of GNP in 1983 to 16 percent in 1985.
In contrast, the real exchange rate, which was initially depreciated, was back to its 1981 level by 1985. Furthermore, quantitative import restrictions, increases in import duties, and taxes on exports were also introduced as part of the Government’s stabilization program. The intensification of import controls at the time of the initial 1983 nominal devaluations, complemented with accommodating monetary policies at the time, resulted in an acceleration of inflation from 10 percent in 1983 to 50 percent in 1984, which contributed to the erosion of the effects of the nominal devaluations on the real exchange rate. Another 28 percent devaluation took place in June 1984 and a managed float was implemented shortly thereafter.
Monetary policy was highly restrictive in 1985, contributing to a nominal currency appreciation and to a sharp increase in real interest rates. Real lending rates went up from 7 percent in 1982 to 21 percent by the last quarter of 1985, while the stock of credit to the private sector decreased by 54 percent during the adjustment period.
The stabilization period, during which output and government expenditures contracted, affected all sectors of society but, clearly, its negative impact was not evenly distributed. In the next section, the reduced form derived from the analytical framework is estimated in order to evaluate the quantitative impact of the stabilization package on income distribution and the specific contribution of each of its components to the distributional outcome of the adjustment effort.
IV. Empirical Results
The empirical implementation of the model centers around the estimation of equation (4) for the period 1980-86. The data used are discussed in the Appendix. The dependent variable, log (SL/SH), is calculated as the ratio of the share of income of the lowest three deciles of income earners to the share of income of the highest decile. The reasons for this split are that a very high proportion of income is received by the top 10 percent and that the incidence of poverty can be better captured by focusing on a wider proportion of earners such as the one selected here. 7/8/
The results of estimating equation (4) are reported in Table 2. 9/ The fit of the regression is very good, with all the variables yielding statistically significant coefficients. The coefficients of labor productivity, the real exchange rate, the real interest rate, and the time trend are positive, while those of underemployment, the inflation rate, and government spending are negative.
Dependent Variable: log (SL/SH)
|ρ - Productivity (log)||2.713||3.89||1.003|
|U - Underemployment||-0.202||6.25||-0.915|
|E - Real exchange rate (log)||1.500||2.19||0.362|
|g - Government spending (log)||-3.639||4.46||-0.613|
|r - Real interest rate||0.021||2.94||0.548|
|π - Inflation rate||-0.041||2.41||-0.532|
|t - time trend||0.962||7.56||1.731|
Thus, the estimated coefficients indicate that, over the sample period, gains in productivity, probably by inducing real wage improvements, tend to reduce inequality. A rise in the real exchange rate also tends to reduce dispersion, supporting the view that low—income groups tend to benefit relatively more from gains in external competitiveness than the high-income groups. The positive sign of the real rate of interest may arise from its negative correlation with economic activity and probably stems from the fact that β3H > β3L, or that short-run improvements in economic activity benefit high-income groups more than low-income groups. 10/
The negative signs for underemployment is consistent with the findings of Blank and Blinder (1985) for the United States. In the case of the Philippines, the size and significance of the coefficients strongly suggest that underemployment, probably through its depressing effects on wages, affects low-income groups more adversely than high-income groups. Regarding inflation, and unlike in the United States, it is found that it is indeed a regressive tax which probably arises, in addition to its effects through economic activity, from the better ability of the upper-income groups to protect their real earnings from being eroded by price increases.
A somehow striking result arises from the sign of the government expenditure variable. 11/ It is apparent from the results that government spending has had regressive redistributive effects. This could be caused by the positive relationship between y and g (if β3H - β3L > 0), and also from a government expenditure composition which has a Low (or falling) component of social expenditures. This type of result, although not clearly expected, is not necessarily surprising. Tanzi (1973) reports a number of similar results arising from studies of government expenditure incidence in Latin America, and discusses at length the reasons for, and the consequences of, nonprogressive patterns of public spending in developing countries.
Because the units of measurement for the dependent variables are different, one way of determining the relative influence of the explanatory variables on income shares is to calculate the Beta coefficients which are also reported in Table 1. 12/ We can observe from those coefficients that labor market conditions are the most important variables in explaining inequality because the higher Beta values correspond to productivity and underemployment (abstracting from the time trend). The negative effect of government spending seems to follow in importance. Inflation and the real interest rate have a similar (although opposite) impact while the real exchange rate, although significant, is substantially less quantitatively important in explaining the changes in income inequality.
How could these results be interpreted within the specific Philippine context? Clearly, the recent macroeconomic adjustment negatively affected the poorer sectors mainly through the increase in underemployment that followed the stabilization program. Increases in underemployment have a high social cost because they have a larger effect on those workers who have the least specific skills and, therefore, are the first ones to become underemployed when aggregate employment falls. The low brackets were also hit by the increase in inflation and by the overvaluation of the exchange rate. The poorest segments seem to bear most of the burden of the inflation tax since both the urban and rural poor cannot protect their real incomes because they do not have indexed wages, and they seldom have any other assets that keep their real value in times of inflation. Inflation was indeed found in this study to be a regressive tax.
The lack of a significant real devaluation during the stabilization program had a negative effect on the poor because they are mostly involved in the production of tradables, and export production is intensive in labor, their main resource. A devaluation, on the other hand, does not have serious adverse effects on the purchasing power of the poor since their consumption basket contains a small share of imported goods. 13/ In the medium term, the landless workers will also benefit from a real devaluation, to the extent that this expands employment opportunities in rural areas.
Regarding the high-income group, its share was affected negatively primarily by increasing real interest rates and cuts in government spending, both leading to the sharp drop in economic activity. There are two components to this result. One, which has been observed in other countries, is that periods of macroeconomic difficulties are usually accompanied by a reduced concentration in the distribution of income (although certainly not by a reduction in the incidence of poverty). This seems to be confirmed here by the significance of the time trend. A channel for this effect to take place is the buoyancy of nonmarket and informal market activities in the face of a deteriorating formal economy. A second reason for the results observed in the Philippines is the type of government spending that took place in the past. It benefited mainly the high-income group through large capital-intensive projects, and through a social service system that did not reach the poor.
At the normative level, the results indicate that, although a drop in absorption was indeed necessary in order to correct the large external and internal imbalances, the impact of such an adjustment on income distribution could have been perhaps softened, if there had been a larger change in the real exchange rate and if the drop in government spending had been accompanied by a drastic change in the composition of public expenditures.
The results reported in Table 2 present evidence on the effects of macroeconomic activity on income inequalities rather than on poverty levels. Obviously, however, in a country like the Philippines, the impact of macroeconomic policies on the absolute incidence of poverty could be of equal or more importance for policy makers. But even if one is only interested in poverty, a case has been made 14/ that the share of income received by families below some particularly defined poverty line (or some other measure for the bottom part of the distribution) is at least as good an indicator of advances made against poverty as is the absolute count below an official poverty line. Still, the measure used in the estimations above, being an indication of the relative performance of high-income and low-income groups, may not reveal the full extent of the impact of macro-variables on the poorest groups of the population, or the results may be misleading in their respect, if they mainly reflect redistributions between middle-income and high-income classes. In order to shed additional light on the distributional incidence of macro-policies, a specification similar to equation (4) was estimated, with the share of income received by the lowest quintile of households as a dependent variable (S20). 15/ The results of the estimation are reported in Table 3.
Dependent Variable: log (S20)
|ρ - Productivity (log)||3.921||4.93||2.544|
|U - Underemployment||-0.175||4.77||1.396|
|E - Real exchange rate (log)||2.165||2.82||0.919|
|g - Government spending (log)||-5.128||5.54||-1.515|
|r - Real interest rate||0.027||3.31||1.221|
|π - Inflation rate||-0.075||3.89||-1.700|
|t - time trend||0.061||4.20||1.941|
The results in Table 3 are very consistent with those obtained in Table 2. Clearly, underemployment, the rate of inflation, and government spending are highly regressive. However, contrary to equation (4), and as indicated by the Beta coefficients, the importance of inflation seems to exceed that of underemployment in affecting negatively the income share of the poor, which strengthens the observed regressive aspect of the inflation tax. As in equation (4), productivity gains and real exchange rate depreciations improve the relative lot of the poor. As a whole, the results in Table 3 strongly confirm the inferences made on the base of the ratio of income shares regarding the significance, direction, and size of the effects of all the variables considered.
V. Concluding Remarks
This paper studies the experience of the Philippines in the 1980s, focusing on the effects of the stabilization program initiated in 1983. A simple analytical framework is developed to examine the evidence on the distributional impact of macroeconomic variables. The results show that underemployment and inflation are strongly regressive in the Philippines. The same is true for the level of government expenditure, reflecting the specific composition of public spending during the period. Productivity levels, the real interest rate, and the real exchange rate were found to be progressive instruments, since gains in these variables improved the relative income shares of the poor.
While the evidence also shows that, when accounting for employment effects, the income distribution narrows when the economy shrinks and widens when it expands, such distributional “improvements” certainly could not be construed to be welfare-increasing. The fact that a fall in real activity may have hurt higher-income groups relatively more should not distract from the need to devise policy packages that soften the brunt of the adjustment on those groups that are largely deprived in absolute terms and therefore lack the ability to protect themselves during periods of austerity.
On the sobering side, the Philippine experience seems to indicate that “good” policies also have a distributional payoff. Reducing inflationary pressures, avoiding real exchange rate overvaluation, and attaining positive real interest rates all have a desirable incidence, while undiscriminating expansionary fiscal policies, with no attention to public expenditure composition, will probably result in a higher skewness of the distributional curve.
The NCSO integrated survey of households contains quarterly data on the earnings of individual earners for 19 quarters from 1980/III to 1986/IV. The dependent variables, log (SL/SH) and log (S20), were calculated as follows: data on the earnings distribution of individual workers were derived from Table 6 of the Integrated Survey of Households Bulletin. This includes earnings from all sources, including those derived from capital holdings but excludes income from financial assets (and probably from all assets held abroad). This table reports the number of days worked against the earnings from the primary job of those in the sample. 16/ In order to derive decile earnings distribution, the income ranges cum number of workers data from Table 6 were converted to actual income earned for each income class. The earnings classes in Table 6 range from no earnings, to less than P 500 per quarter, to the highest and open-ended class of P 50,000 and over. The NCSO did not report the actual averages within each earnings class as a basis for calculating incomes earned in each class, but the overall average income for all workers is reported. Thus, information for the overall income was derived by multiplying the total number of workers by the average income.
As a first approximation, the midpoint of each earnings class range was used as the average for all classes except for the open-ended highest one. The overall income less the sum of all the other classes provided an estimate for income earnings in the highest class. Adjustments were then made on the income averages for each class, deducting some from the lower income classes until reasonable values for the average income in the highest class could be derived. “
The resulting shares of income and their corresponding fraction of total income earners for each income class were then converted to a decile basis using a linear interpolation of a Lorenz curve. The shares of income by decile formed the basis for the dependent variables.
a. Labor productivity (ρ) was derived by estimating the total number of actual days worked by earners in the sample, by multiplying the midpoint of the number of days worked by the number of earners working within the corresponding bracket of number of days worked. The actual productivity variable was the ratio of the quarterly GDP to the estimated number of hours worked as derived from Table 6, NCSO.
b. Underemployment (U) was measured as one minus the ratio of the actual number of days worked (estimated as in the previous paragraph) to the potential number of days worked (estimated by assuming that all workers worked 65 days in the corresponding quarter).
c. The inflation rate (π) was represented by the quarterly rate of change of the consumer price index as reported in the IMF, International Financial Statistics. In estimating equation (4), to allow for lagged effects, πt-1 was used.
d. The real rate of interest (r) was estimated as the difference between the nominal rate (i) at the beginning of the period and the inflation rate over the period. The nominal rate (i) was represented by the domestic 91-day treasury bill rate.
e. Government spending (g) was represented by real government consumption expenditures, as reported in the quarterly national income accounts. Using government construction expenditures, or the sum of government consumption and government construction, does not materially change the results.
f. The real exchange rate (E) was computed by deflating the nominal peso to the dollar exchange rate by the corresponding consumer price index. In the estimations, the one-period lagged rate (Et-1) was used.
AhluwaliaM.S. and F.J.Lysy “Employment, Income Distribution and Programs to Remedy Balance of Payments Difficulties” in W.R.Cline and S.Weintraub (eds.) Economic Stabilization in Developing Countries (Washington: Brookings Institution1981) pp. 149–90.
BlankRebecca M. and Alan S.Blinder “Macroeconomics, Income Distribution, and Poverty” NBER Working Paper No. 1567 (Cambridge, Massachusetts: National Bureau of Economic Research1985).
BlinderAlan S. “The Level and Distribution of Economic Well-Being” in M.S.Feldstein (ed.) The American Economy in Transition (Chicago: University of Chicago Press1980) pp. 415–79.
BlinderAlan S. and Howard Y.Esaki “Macroeconomic Activity and Income Distribution in the Postwar United States” Review of Economics and Statistics (November1978) pp. 604–09.
BrunoMichael “Income Distribution and the Neoclassical Paradigm” Journal of Development Economics Vol. 6 No. 1 (March1979).
DemeryLionel and TonyAddison “Stabilization Policy and Income Distribution in Developing Countries” World Development Vol. 15 No. 12 (Oxford: Pergamon PressDecember1987) pp. 1483–98.
HelleinerGerald K. “Stabilization, Adjustment, and the Poor,” World Development Vol. 15 No. 12 (Oxford: Pergamon PressDecember1987) pp. 1499–1513.
HellerPeter S. and others “The Implications of Fund-Supported Adjustment Programs for Poverty: Experiences in Selected Countries,” IMF Occasional Paper No. 58 (Washington: International Monetary FundMay1988).
International Monetary FundFund-Supported Programs, Fiscal Policy and Income DistributionIMF Occasional Paper No. 26 (Washington: International Monetary FundSeptember1986).
KanburS.M.R. (1987a) “Measurement and Alleviation of Poverty: With an Application to the Effects of Macroeconomic Adjustment,” Staff PapersInternational Monetary Fund (Washington) Vol. 34 No. 1 (March1987) pp. 60–85.
KanburS.M.R. (1987b) “Structural Adjustment, Macroeconomic Adjustment and Poverty: A Methodology for Analysis,” World Development Vol. 15 No. 12 (Oxford: Pergamon PressDecember1987).
KnightJ.B. “Devaluation and Income Distriubtion in Less Developed Economies,” Oxford Economic Papers Vol. 28 No. 2 (July1976) pp. 200–27.
PapanekGustav F. and OldrichKyn “The Effect on Income Distribution of Development, the Growth Rate, and Economic Strategy,” Journal of Development Economics Vol. 23 (1986) pp. 55–65.
Ms. Guerrero is an Economist in the Country Operations Division of the Asia Regional Office of the World Bank. The authors benefited from a number of valuable comments from colleagues in the World Bank and in the Asian Department and Fiscal Affairs Department of the IMF.
For a review of the literature on the subject, see Demery and Addison (1987). See also Helleiner (1987). Kanbur (1987a, 1987b) discusses in detail the methodological issues regarding poverty measurement in the context of macroeconomic adjustment. Ahluwalia and Lysy (1981), Bruno (1979), and Knight (1976) concentrate on issues related to functional distribution.
See also Papanek and Kyn (1986) for an empirical analysis of the effects of growth on income distribution.
L and H do not necessarily exhaust the totality of income groups but, rather, stand for representative income groups at both ends of the distribution.
Obviously, variables such as underemployment and inflation are interrelated and they cannot be taken realistically as exogenous. However, given the focus of the study and the data limitations, they are assumed to be independently determined with respect to the dependent variable.
Notice that, given total income as a common denominator, the log (SL/SM) can also be interpreted, and measured, as the logarithm of the rate of the low-income shares to the high-income shares in total income. This interpretation is given in some of the following discussions.
Since β1i > 0 and λ2i < 0, β1λ2 < 0 when |β1Lλ2L| > |β1Hλ2H|.
Within the data set used, the average fraction of income received by families in the lowest 30 percent was 6.3 percent, while the highest 10 percent received 44.6 percent.
As explained in the Appendix, a limitation of the data is the lack of information on returns from financial assets. This probably affected SH much more than SL but could not be corrected on the basis of the available information.
The estimation is performed using OLS with a Cochrane-Orcutt Iterative adjustment to account for serial correlation. It was not felt that there was a need for additional complexity in the estimation techniques since no serious reverse causation from relative shares to independent variables could be expected, and heteroskedasticity is not likely to be present given that most of the variables (except, of course, t) show much of a time trend.
Clearly, the fact that earnings in the survey used (see Appendix) do not include returns to financial assets may affect this conclusion. Increases in real interest rates probably benefit more the higher decile. This particular conclusion could change depending on the magnitudes involved. However, the lack of information on returns to financial assets does not allow us to incorporate this effect in our calculations.
Although government consumption is used in the estimation as a proxy for government expenditures, adding construction expenditures did not alter the results.
Beta coefficients measure the change in income shares, other things being equal, for a unit change in each of the independent variables. Since all the variables are expressed in standard deviation, the Beta coefficients are independent of the unit of measurement, and can thus be compared directly within the equation.
In the Philippines, 72 percent of the expenditures of the poor are on locally produced cereal and cereal preparations, fish, and fruits and vegetables.
See Blinder (1980).
This group, the poorest of the poor, is considered to be below subsistence income levels.
Data on Table 6 after 1980 have not yet been published by the NCSO but were taken from special tabulations. Professor M. Montes, from the University of the Philippines, provided technical assistance on data analysis and on the empirical specification of the variables.