The effect on income distribution of development, the growth rate and economic strategy

The effect on income distribution of development, the growth rate and economic strategy

Journal of Development Economics 23 (1986) 55--65. North-Holland THE EFFECT ON INCOME DISTRIBUTION OF DEVELOPMENT, THE GROWTH RATE AND ECONOMIC STRAT...

648KB Sizes 0 Downloads 23 Views

Journal of Development Economics 23 (1986) 55--65. North-Holland

THE EFFECT ON INCOME DISTRIBUTION OF DEVELOPMENT, THE GROWTH RATE AND ECONOMIC STRATEGY* Gustav F. PAPANEK and Oldrich KYN Boston University, Boston, MA 02215, USA Received March 1984, final version received August 1985 An analysis of cross-section and time series data for 83 countries confirms some, and contradicts other work on income distribution. New findings include: a dualistic socio-political system is highly unfavorable for equality. Neither the extent of government intervention in the economy nor the rate of manufactured exports are systematically related to income distribution. The analysis confirms that there is no systematic relationship between equality and the rate of economic growth. Educational participation and a reduction in the share of primary exports in G D P are both favorable for equality. There is some support for the Kuznets hypothesis that inequality increases as per capita income rises to about $400 and then declines, with further income increase, but the empirical support is not strong and may be weakening over time. These findings lead to more optimistic conclusions then other work: that rapid growth in a mixed economy is quite consistent with unchanged, or even improved, income distribution, even at early stages of development.

An analysis of income distribution data for 83 countries I contradicts some widely held views and comfirms others, about the relationship between income distribution and other variables. Our findings are based on a sequence of regressions nested into the following general model:

Y=a + ~fl~Di+ 7T+(6+dpT)LIN +(d/ + 2T)LIN 2 + ~ i X , + u , where Y Di T

LIN Xi

is a measure of income distribution, i.e., Gini coefficient (GINI) or the share of the poorest 40Y/o of population (SHARE), are corrective dummy variables for the differences in definitions and coverage of the left-hand variables, is time, is the logarithm of per capita GNP in 1964 U.S. dollars, are additional explanatory variables.

Table 1 reports results on the assumption that the Kuznets curve is stable over time or that cross-country and inter-temporal Kuznets curves are *This is a summary version. More complete information on methodology, data used, sources, definitions and further analysis can be found in Papanek and Kyn (n.d.). 1For 44 countries observations were for one year only, 39 had observations for two or more years. Pooling of cross-section and time series data resulted in 145 observations over 1952 to 1978. 0304-3878/86/$3.50 © 1986, Elsevier Science Publishers B.V. (North-Holland)

East Europe Dual society Government intervention Education

Social factors

Log (INC) 2

Kuznets curve Log (INC)

Constant

Left-hand variable

-0.124 (-2.95) 0.049 (2.73) -0.0005 (-0.96) -0.001 (-2.29)

0.191 (2.46) -0.016 (-2.55)

- 0.026 (-0.11)

-0.044 (-0.19)

0.189 (2.33) -0.018 (-2.63)

(2)

(1)

Gini coefficient

-0.122 (-2.87) 0.040 (2.17) - 0.0006 ( - 1.21) -0.001 (-2.18)

0.162 (2.00) -0.014 (-2.08)

(-0.18)

-0.041

(3)

-0.109 (-2.51) 0.033 (1.76) -0.0006 ( - 1.09) -0.0004 (-0.723)

0.183 (2.11) -0.016 (-2.24)

--0.065 (-0.25)

(4)

- 6.93 (-1.74) 0.647 (1.95)

32.79 (2.84)

(5)

1.62

( - 1.91) -0.0006 (-0.03) 0.047 (2.02)

-

6.94 (3.51)

-8.57 (-2.20) 0.703 (2.20)

37.33 (3.28)

(6)

Share of the pooresi 40%

Factors in income distribution (stable Kuznets curve)?

Tablel

-

1.38

( - 1.56) 0.004 (-0.15) 0.041 (1.74)

-

6.72 (3.33)

-6.98 ( - 1.68) 0.569 (1.67)

33.73 (2.85)

(7)

6.78 (3.14) - 1.63 ( - 1.68) -0.008 (0.32) 0.014 (0.53)

- 8.08 ( - 1.73) 0.697 (1.83)

36.62 (2.73)

(8)

g,

gr

i

t%

t~

0.348 0.089 9.09

0.541 0.076 12.99

0.552 0.076 10.59

0.0006 (0.25) 0.0001 (1.67) 0.0002 (0.35)

0.623 0.071 10.89

0.036 (1.39) 0.129 (3.61) 0.003 (0.12) 0.069 (2.04)

-0.00001 (-0.04) 0.0003 (0.47) 0.0002 (0.29)

0.359 4.05 8.66

at-statistics in parentheses. Not shown are the six corrective dummy variables.

SER F

R2

Statistics

West Asia & North Africa

Asia

Regional factors South & Central America Africa sub-Sahara

Economic factors Growth rate Primary exports Manufactured exports

0.537 3.50 11.58

0.545 3.51 9.35

-0.045 (-0.39) -0.039 ( - 1.34) 0.004 (0.15)

0.578 3.44 8.16

0.32 (0.23) -2.69 ( - 1.43) 1.21 (0.78) - 1.21 (-0.69)

-O.37 (-0.31) -0.069 (-0.64) 0.007 (0.22)

L~

58

G.F. Papanek and O. Kyn, Effects on income distribution

identical, i.e., y = ~b=2=0. As groups of variables for social, economic and regional factors were sequentially added to corrective dummies and Kuznets curve variables, L I N and L I N 2, the following findings resulted: 1. The rate of growth has no systematic effect on income distribution. The argument has been made [Sheehan (1980), Griffin and Khan (1972)] that a high rate of growth increases inequality because it requires great rewards for such well-to-do groups as investors, managers and land owners. That conclusion is not borne out by our results, nor by the similarly inconclusive results of studies based on a smaller sample [e.g., Ahluwalia (1976a, b), Chenery et al. (1974), Cline (1975), Fields (1981), Papanek (1975)]. The reason may be the increase in labor income with rapid growth. Where rapid growth is not due to income from primary exports, a separate variable in our analysis, it is usually caused by growth in labor intensive activities. That labor intensive development is favorable for income distribution remains a hypothesis, with preliminary supporting evidence l-see Papanek (1980, 1979a, b)]. 2. Primary e x p o r t s 2 a r e assumed to generate a less equal income distribution because they produce concentrated rents. Regressions in table 1 support this hypothesis only mildly. The effect has the right sign but is slight and with low statistical significance. When regional dummies are introduced statistical significance disappears, probably because of high correlation between primary export dependence and regional location. 3. Exporting manufactured goods a at a high rate, in contrast, is assumed to be favorable to an egalitarian distribution for two reasons. First, such exports usually are competitive in the world market only if produced by labor intensive industries [Chenery and Syrquin (1975)]. The consequent increase in demand for labor will raise wages. Second, economies competing in world markets may have fewer price distortions than those emphasizing import substitution and therefore generate fewer windfall gains for the rich IPapanek (1978)]. Contrary to these arguments, generally based on the experience of East Asia, in our regressions a high rate of manufactured exports has no significant effect on equality. A plausible explanation is that some countries have achieved exports of manufactured goods by indirect subsidies quite as large as those provided to import substituting industries, with comparable distortions in factor and product markets. 4. Government intervention in the economy, often justified as designed primarily to increase equality, is generally assumed to achieve that purpose 2The variable used is the percent of primary exports

to GDP. 3The variable is the percent of manufactured exports to GDP.

G.F. Papanek and O. K yn, Effects on income distribution

59

[Adelman and Taft Morris (1973), Papanek (1975)]. But our results do not support the contention that intervention increases equality in mixed economies. Our proxy for the extent of government intervention was the share of public investment in total investment. Its use resulted in coefficients that are small and not significant. A plausible explanation [see Papanek (1979a)] is that intervention often benefits not the poor, but another part of the elite: the political, bureaucratic and military leadership, worlers in public enterprises and favorite businessmen.

5. 7he Eastern European countries, with the most extensive government intervention in the sample record an unusually egalitarian income distribution (using a dummy variable). This is consistent with other results [e.g., Kyn (1978)]. It is difficult to determine to what extent recorded equality is due to a different structure of asset ownership and to what extent to different statistical conventions and the failure to take full account of income in kind, various fringe benefits and unofficial forms of income. 6. A dualistic socio-political structure, with an elite drawn from a different ethnic or racial group than the poor majority, makes for an unequal income distribution. One would expect a 'foreign' elite to be forceful in using its political power to assure itself a disproportionate share of income. Sociopolitical dualism differs analytically from economic dualism, usually defined in terms of a great gap between a high income modern sector and a low income traditional sector. Given this definition one would expect a near perfect correlation between economic dualism and inequality, but this conclusion approaches a tautology [see Adelman and Taft Morris (1973)]. Countries with socio-political dualism include: South Africa, Rhodesia, and several in Latin America where the elite is of European origin, the majority Indian, Mestizo and Black. The effects of socio-political dualism has been examined [Papanek (1975, 1978) and Bacha (1977)] but the variable has not been clearly defined or rigorously tested. In our regressions it appears to be a very significant (dummy) variable, but loses some of its importance and significance when economic and regional factors are added. A plausible explanation is that dualistic societies are concentrated in Africa and Latin America and affect income distribution in part through such indirect means as education and the allocation of the concentrated resources generated by primary exports. Therefore dualism may lose statistical, but not necessarily causal, significance when other variables are added. 7. Education, measured as a percent of school enrollment, is significantly related to equality in this as in other studies [Adelman and Taft Morris (1974), Slama (1978), Ahluwalia (1976a, b)]. The regions differ in educational level. So when regional dummies are introduced education declines in

60

G.F. Papanek and O. Kyrg Effects on income distribution

statistical significance. The coefficients for education are low. Comparing countries with 109/o and 909/0 enrollment ratios - few countries exceed those limits, since secondary schools are included - the estimated GINI differs by only 0.03, but the SHARE of the poorest 40~ differs by an estimated 3.39/0 to 3.69/o. This suggests that the spread of education benefits particularly the poorest, because in most countries much of the middle class is already educated.

8. Regions share a variety of attributes. If all relevant variables were included in the regressions, pure geographic location should be irrelevant. But two of the five regions differ significantly from the other three in income distribution for GINI, after taking account of a number of region-related variables - education, socio-political dualism, communist governments and structure of exports. This may be partly due to inadequate measurement. We use a dummy variable for dualism, and therefore fail to capture differences in degree of dualism. There may be similar qualitative differences in education and other variables across the regions. In addition there appear to be excluded variables. The distribution of wealth, particularly of land, undoubtedly affects income distribution, but is not included in our analysis, for lack of a good measure.

9.

7he Kuznets hypothesis (1955) that as income per capita rises over several

decades income distribution would first become less equal and then more equal has been supported by a large array of empirical studies [e.g., Bacha (1977), Ahluwalia (1976a, b), Chenery et al. (1974), Adelman and Taft Morris (1973), Cline (1975), Paukert (1973)]. Testing for the Kuznets curve in standard fashion by using cross-country data, on the assumption that crosscountry and inter-temporal Kuznets curves are identical, the Kuznets hypothesis is confirmed, the coefficients remain relatively stable and do not lose much of their significance even after social, economic and regional factors are introduced in regressions. However, the Kuznets curve itself explains only a relatively small part of the total variation in income inequality. The sequence of nested regressions allowed us to use F tests for the joint explanatory power of the Kuznets curve and of the groups of variables added to the equations, as is common in the analysis of covariance. The hierarchical approach favors the factors which enter early into analysis, unless the factors are completely orthogonal. The 'regression' approach may underestimate the significance of all factors. In the hierarchical test implied by the sequence of regressions in table 1 the joint explanatory power of the two Kuznets curve variables was quite significant (computed F-statistics were 13.9 for GINI and 5.8 for SHARE). The joint significance of the social factors was even higher, but economic factors were insignificant. Regions were significant for GINI (F of 5.95) even when entered last but not for SHARE. When the

G.F. Papanek and O. K yn, Effects on income distribution

61

Kuznets curve entered after the other factors, equivalent to a 'regression' approach, it remained significant for G I N I when entered after the social factors but lost its significance when entered after the economic and regional factors. For S H A R E the Kuznets curve lost its signficance as soon as education entered the regression. 4 10. The temporal K u z n e t s curve. Tables 2 and 3 report regressions designed to test whether the temporal and cross section Kuznets curve are identical, as is usually assumed. They would not be identical if the relationship of income and income distribution has changed over time. Bacha (1977) summarizes the arguments of the (Latin American) structuralist school that benefits of growth in the recent past accrued primarily to the developed countries and the elite in the less developed countries. As a result, the trend toward equality in Western Europe found by Kuznets could now be postponed for an indeterminate period. Conversely one could argue that some of the same socio-political factors which made for greater equality in Europe are felt in the less developed countries at lower per capita incomes. Whether the Kuznets curve has changed slope over time is tested in table 2 which adds time shift variables (i.e., time and interactions of the Kuznets curve with time) to the variables of table 1. Any time shift would imply that the results from a cross section of countries could not be used to predict the secular development of a country. The results indicate a relatively fast flattening of the curve. The joint significance of the three time shift variables for the Kuznets curve (based on an F-test) appears to be significant for G I N I but not for the S H A R E when entered early. Its marginal significance when entered after other factors is low. The evidence on the stability of the Kuznets curve is therefore ambiguous. However, it seemed worth testing since the possible weakening of the Kuznets curve has never been raised, finds some support, could be important and could readily be tested in a few years. If the estimated time shift continues, the Kuznets curve would have all but disappeared by the early 1980s. Table 3 reports a direct test of the intertemporal Kuznets curve. On a subsample of countries with observations at two or more points in time, it is possible to estimate the inter-temporal Kuznets curve assuming that all countries have parallel Kuznets curves with identical slopes but individual, country specific, levels of income inequality. 5 Again the model was estimated with corrective dummies, two Kuznets curve variables L I N and L I N 2 but no time or time interactions (i.e., restrictions 7 = ~b= 2 = 0 were imposed) and all the Xi variables were replaced by a set of country specific dummy variables. This implies that although the regression uses pooled cross-section and time 4For details of these F-tests, see the basic paper. SParallel Kuznets curves is, of course, also the standard implicit assumption of all crosssection tests, the great bulk of empirical work.

J.I).E.

C

62

G.F. Papanck and O. Kyn, Effects on income distribution

Table 2 Factors in income distribution (shifting Kuznets curve)? Gini coefficient Left-hand variable

Share of the poorest 40%

(9)

(10)

(11)

1.468 ( - 2.13)

-0.0891 ( - 1.46)

-0.826 ( - 1.47)

Shifting Kuznets curve Time 0.086 (2.143)

0.052 (1.495)

0.695 (2.916) -0.030 ( - 2.22) - 0.061 (-3.01) 0.0026 (2.260)

Constant

Log (INC) T* log (INC) Log

(INC) 2

T ' l o g (INC) 2

-

Social factors East Europe Dual society Government intervention Education

(15)

67.93 (2.22)

60.81 ( 1.91)

0.053 (1.470)

-3.42 (-- 1.64)

-2.12 (-- 1.15)

-1.92 ( - 1.01)

0.488 (2.317)

0.460 (2.096)

-24.54 (--2.06)

- 18.20 (-1.70)

- 15.27 (-1.36)

-0.018 ( - 1.49) - 0.041 (-2.30) 0.0015 (1.498)

-0.018 ( - 1.45) - 0.39 (-2.07) 0.0015 (1.450)

1.16 (1.50)

0.67 (1.06)

0.60 (0.90)

2.08 (2.05) - 0.095 ( - 1.54)

1.45 (1.59) -0.052 (--0.97)

1.20 (1.25) -0.045 (-0.81)

-0.126 ( - 2.96) 0.48 (2.584) -0.0005 (-0.87) - 0.001 (-2.14)

-0.124 ( - 2.98) 0.039 (2.063) -0.0006 ( - 1.119) - 0.001 (-2.062)

7.01 (3.48)

6.81 (3.33) 1.53 (1.69)

Manufactured exports 0.374 0.089 7.22

(14)

84.45 (2.48)

Economic factors Growth rate Primary exports

Statistics R2 SER F

(13)

0.550 0.076 10.53

-

1 . 7 7

(-2.02) - 0.0042 (-0.170) 0.045 ( - 1,69)

-

- 0.0004 (-0.017) 0.040 ( - 1.64)

0.00002 (0.007)

-0.069 (-0.42)

0.00103 (1.682) 0.00024 (0.3777)

-0.040 ( - 1.36) 0.0016 (0.05)

0.561 0.076 8.927

0.376 4.04 6.63

0.546 3.51 9.37

0.554 3.52 7.88

aNot shown are the six corrective dummy variables or regressions (12) and (16) with regional variables.

G.F. Papanek and O. Kyn, Effects on income distribution

63

Table 3 The inter-temporal Kuznets curve."

GINI Constant t-statistics Log (income) t-statistics Log (income) 2 t-statistics R2 SER Degrees of freedom F for joint significance

SHARE 0.02 (0.06)

11.5 (0.6)

0.16 (1.5) -0.01 ( - 1.7) 0.83 0.05 54.00 2.38

0.47 (0.2) -0.11 ( - 0.2) 0.81 2.6 46 0.46

"Not shown are the 35 country dummy variables for the regression with GINI, 32 country dummies for the regression with SHARE, and six dummy variables to correct for differing definitions/coverage.

series data it does not assume that cross-sectionally the countries lie on the Kuznets curve. For GINI the Kuznets curve coefficients have the right signs and surprisingly similar values to those i n table 1. They are, however, significant only at 20% and 10% levels. For SHARE they have the wrong sign and are insignificant. These are not very strong tests, because we have few observations for each country. But the inter-temporal analysis provides only partial support for the Kuznets hypothesis. 6 11. Conclusions and implications. There is enormous variance in income distribution at all levels of income. Even if one leaves out extreme values, of doubtful reliability, the share of the poorest 40% in LDCs ranges between 6% and 23% and the GIN! between 0.2 and 0.55. Very little of this variance is explained by the Kuznets curve: as per capita-income rises from $100 to $400 the maximum deterioration in S H A R E is 1.3% as a result of the Kuznets effect. The change in GINI is even smaller in percentage terms. Table 3 brings some evidence that the Kuznets effect may not hold for SHARE and table 2 indicates that the intensity of the Kuznets effect, however measured, may be declining. Each of the two variables under government control has about the same effect on income distribution as the maximum predicted effect of the Kuznets curve. A 30% change in educational participation and in the share of primary exports in GDP, in combination 6Ahluwalia [in Chenery et al. (1974)] and Fields (1981) also conclude that time series show no deterioration in the share of the poor.

64

G.F. Papanek and O. K yn, Effects on income distribution

increase the predicted SHARE from 12~ to 169/o while the maximum change due to the Kuznets effect is 1.39/o. Contrary to expectations, the share of manufactured exports in GDP does not significantly affect income distribution. A dualistic socio-political structure is highly correlated with an unequal income distribution. Neither the rate of growth, nor government intervention are related to income distribution. Regardless of socio-political systems, policies or economic structure, the absolute income of the poor triples or quadruples as per capita income moves from $100 to $400, even on the most pessimistic assumptions about the importance of the Kuznets effect. On the whole development has been highly favorable for the absolute income of the poor. There is no clear trade-off between a higher rate of growth and greater government intervention on the one hand and a more equitable distribution of income on the other. More rapid growth therefore contributes to the more rapid alleviation of poverty. These are far more optimistic inferences about reconciling growth and equity than those of other analysts [e.g., Adelman (1975)].

References Adelman, Irma, 1975, Development economics - A reassessment of goals, American Economic Review 65, no. 2. Adelman, Irma and Cynthia Taft Morris, 1973, Economic growth and social equity in developing countries (Stanford University Press, Stanford, CA). Ahluwalia, M., 1976a, Inequality, poverty and development, Journal of Development Economics 3, no. 4. Ahluwalia, M., 1976b, Income distribution and development: Some stylized facts, American Economic Review 66, no. 2. Bacha, Edmar L., 1977, The Kuznets curve and beyond: Growth and changes in inequalities, Development Discussion Papers no. 29 (Harvard Institute for International Development, Harvard University, Cambridge, MA). Chenery, Hollis and M. Syrquin, 1975, Patterns of Development, 1950-1970 (Oxford University Press, New York). Chenery, Hollis, M.S. Ahluwalia, C.G. Bell et al., 1974, Redistribution with growth (Oxford University Press, New York). Cline, W., 1975, Distribution and development: A survey of the literature, Journal of Development Economics 1, no. 4. Fei, John C.H. and Gustav Ranis, 1964, Development of the labor surplus economy (Irwin, Homewood, IL). Fields, G.S., 1981, Poverty, inequality and development, Journal of Policy Modeling 3, no. 3. Griffin, K. and A.R. Khan, 1972, eds., Growth and inequality in Pakistan (Macmillan, London). Kuznets, S., 1955, Economic growth and income inequality, American Economic Review 45, no. 1. Kuznets, S., 1957, Quantitative aspects of the economic growth of nations: II, Industrial distribution of national product and labor force, Economic Development and Cultural Change 5, suppl., July. Kuznets, S., 1963, Quantitative aspects of the economic growth of nations: VIII, Distribution of income by size, Economic Development and Cultural Change 11, no. 2. Kuznets, S., 1966, Modern Economic Growth (Yale University, New Haven, CT). Kyn, Oldrich, 1978, Education, sex and income inequality in Soviet-type Socialism, in: Z. Griliches, W. Krelle, H.-J. Krupp and O. Kyn, Income distribution and economic inequality (Campus, Frankfurt and Wiley, New York).

G.F. Papanek and O. K yn, Effects on income distribution

65

Lewis, W. Arthur, 1954, Economic development with unlimited supplies of labor, The Manchester School, May. Papanek, Gustav F., 1975, Distribution of income, wealth and power, in: Y. Ramati, ed., Economic growth in developing countries (Praeger, New York). Papanek, Gustav F., 1978, Economic growth, income distribution and the political process in less developed countries, in: Z. Griliches, W. Krdle, H.-J. Krupp and O. Kyn, Income distribution and economic equality (Campus, Frankfurt and Wiley, New York). Papanek, Gustav F., 1979a, Real wages growth, inflation, income distribution and politics in Pakistan, India, Bangladesh, Indonesia, Discussion paper no. 29 (Department of Economics, Boston University, MA). Papanek, Gustav F., 1979b, Methodological and statistical appendix, Discussion paper no. 30 (Department of Economics, Boston University, MA). Papanek, Gustav F., 1980, The Indonesian economy (Praeger, New York). Papanek, Gustav F. and Oldrich Kyn, n.d., The effects on income distribution of development, the rate of growth and economic strategy: Flattening the Kuznets curve, Pakistan Development Review, forthcoming. Paukert, F., 1973, Income distribution at different levels of development: A survey of evidence, International Labor Review 108. Sheehan, John, 1980, Market-orientated economic policies and political repression in Latin America, Economic Development and Cultural Change 28, no. 2, 267-291. Slama, Jiri, 1978, A cross-country regression model of social inequality, in: Z. Griliches, W. Krelle, H.-J. Krupp and O. Kyn, Income distribution and economic equality (Campus, Frankfurt and Wiley, New York).