Publications  Income Distribution





In examining the impact of the specific variables on the mixed economies (excluding Eastern Europe) it is most useful to concentrate on the range of per capita income between $100 and $400. There are very few countries below $100 and the problem of absolute poverty is somewhat less serious once per capita income exceeds $400.


Display Table 7.

new window


Table 7 gives the range of predicted Gini coefficients and shares for the poorest 40 percent for the variables that are generally significant. The variable which could be altered most readily in the medium term is spread of education. Policy can also reduce reliance on primary exports. However, it is usually not desirable to go very far in this direction. A country with a very high primary-export ratio is likely to be natural-resource-rich. The low relative share of the poor may then be compensated in part by a relatively high absolute income. It would normally not make much sense to forgo the high absolute income which some primary exports can generate. Iraq, Venezuela, Iran and so on are in this group and while a more egalitarian strategy might consciously try to raise the importance of industry in the economy, it would generally not be very wise to bring the share of primary exports down from the 50—70 percent typical of these countries to the 1 15 percent typical of the resource-poor East Asian countries.

In a sense, primary exports and dualism are proxies reflecting the willingness and ability of the elite to appropriate a larger-than-average share of income. Another policy alternative, therefore, would be to deal with income distribution directly. The hypothesized reason for the impact of primary exports on inequality is the concen­tration of income from these exports. Fiscal policy or other steps could ensure the wider distribution of the resources from primary exports. Even a dualistic socio­political system is subject to change.

If one assumes for illustrative purposes that policy can change the education and primary-export variables by about one-third of the range, then the impact of the four variables significant in mixed economies is as shown in Table 8.


Display Table 8.

this window

new window


These magnitudes are not all that great. But at a per capita income of $100, a 2.5-percent increase in the share of the poorest 4 percent (say from 13.5% to 16%) would mean an increase of $34 to $40 in annual per capita income. This is a notice­able increase, an absolute amount equivalent to a 2-percent increase of per capita income over 8 years. But the main conclusion which seems to emerge is that none of the factors examined, by themselves, make a great deal of difference in the income distribution of non-dualistic, non-East European countries. But, in combination, the variables can make a difference. The share of the poorest 40 percent would be about 10 percent of national income in dualistic societies with reasonably high primary exports and low education and 60-75 percent higher in non-dualistic ones with reasonably low primary exports and high education.


Display Table 9.

this window

new window


One can similarly trace the effect of different significant variables on the absolute income (Table 9). The most important conclusion is that the absolute in­come of the poorest 40 percent rises quite dramatically, as per capita income in­creases from $80 to $400 despite the decline in their income share. Even in dualistic societies, their absolute income rises by 85 to 95 percent each time per capita income doubles from $100 to $200 and then to $400. There is no question that development on the whole has been highly favorable for the absolute income of the poor, even if their share declines slightly as per capita income rises initially.


Lessons from the Unusual Cases:

The Role of Influential outliers

Since the model explains only about half the variance, an examination of outliers, of countries whose income distribution is not well explained by the factors examined, may shed further light on what causes differences in income distribution.

The Appendix gives influence statistics for all influential observations. From these one can see which countries are outliers in the usual sense and what leverage they have on estimated parameters. These data also facilitate an analysis of the extent to which particular results stem from the unusual influence of outliers for which data may be incorrect. The statistics for DFBETAS are the most useful for that purpose. RSTUDENT indicates which observations have the largest residual, which are furthest from the fitted values; COVRATIO indicates influence; and DFBETAS in effect combines the two and indicates how the estimated parameter would change if the particular observations were removed.


Display Table A1.

this window

new window

Display Table A2.

this window

new window


For Table 1, the intertemporal Kuznets Curve, influence statistics are given in Appendix Tables Al and A2. For a particular observation to have a major in­fluence on the results, the DFBETAS should be about 2.0 or above. It can be seen immediately that there are none, and that there are only a handful DFBETAS that are even above 0.4. The most interesting result in these tables is that the exclusion of data for Pakistan from the analyses with the share of the poorest 40 percent as the dependent variables makes very little difference to the results. Pakistan is an outlier, but not an influential one. Two of the annual observations for Taiwan represent influential outliers and the decision to include that country in the analysis does have a small effect on the results. It weakens the Kuznets Curve because income distri­bution became more equal as per capita income rose. But, as already noted, there is ample evidence that Taiwanese data reflect reality and that it represents a “good” outlier. On the other hand, the two outliers that strengthen the Kuznets Curve, Mali in 1958 and India in 1955, may be of more dubious reliability. Both represent early efforts, when the data collection machinery may not have been well developed. But they were retained in the analysis to reduce any possibility of our inadvertently bias­ing the results otherwise.


Display Table A3.

this window

new window

Display Table A4.

this window

new window


Tables A3 and A4 report the influence statistics for the most com­plete regressions for Tables 2 and 3, which are based on the combined cross-country/intertemporal Kuznets Curve. Again, DFBETAS are not large, with only a few in the 0.5 0.6 range. A brief discussion of the most influential outliers may shed some light on the reliability of our results and on the possibility of excluded variables.

Our conclusion on the weakness of the Kuznets Curve is, if anything, strengthened by this examination. There are three observations from Pakistan that would be influential against the Kuznets Curve, with share as the dependent variable, if they had been included in the analysis. So, by excluding Pakistan, we weakened our hypothesis that the Curve does not exist. Most of the other outliers strengthen the Curve and are all from Africa (Chad, Gabon, Sierra Leone are outliers for both dependent variables). We have no basis for judging how good these observations are.


Our conclusion that rapid growth can occur without any deterioration in equality may also be strengthened by examining influential outliers. Rapid growth was followed by less equality, mostly in countries where inequality was less due to rapid growth than to a very unequal historical distribution of assets, especially land (Iran), or extreme dualism (Zambia in 1959), or other excluded variables (Taiwan in 1953, when substantial inequality may have reflected the difference between wealthy newcomers and poorer indigenous residents).

The most important outliers contradicting our conclusion that a high rate of primary exports and inequality go together are Mali and Zambia in the 1950s. This may be related to their classification as dualistic and to our inability to measure the degree of dualism, but this is speculation.


Singapore is influential (for the Gini in 1975) against our hypothesis that a high rate of manufactured exports and equality go together. Actually, equality increased steadily over time as manufactured exports increased, but the country started out with substantial inequality. That may have been due to socio-political dualism before independence and its after-effects. In intertemporal analysis Singapore would have supported the hypothesis but with combined cross-country/intertemporal data it does not. On the whole, though the influential outlier data do not really add much to our attempt to explain why manufactured exports and equality are not related.

The influential observations weakening the conclusion that education and equality are related can largely be explained by excluded variables, such as the low participation rates of girls in some Islamic countries (Libya in 1962, Pakistan) or our inability to measure the strength of dualism (Ecuador, for Gini in 1970).


That inability may also explain some of the observations pulling against our conclusion that dualism explains inequality to a considerable degree. Sierra Leone probably suffered from some measure of dualism in 1969, but we did not categorize it as dualistic, perhaps out of ignorance of the true situation. Similarly, neither Gabon (in 1968) nor Venezuela (in 1962 and 1971) was classified as dualistic. The former may be an error, if the economy continued to be French-dominated even if the country was independent, while in the latter case the dualism may be subtler or non-existent (as defined). On the other hand, Surinam, classed as dualistic, may have been very weakly so. If the strength of dualism could be measured, our con­clusion on its importance might well have been further strengthened.

Finally, there are three influential observations acting against our conclusion that greater government intervention (more public investment) does not affect equality: Chad, Iraq and Surinam. The latter two lack actual observations for public investment, and so had to be estimated. Their influence on the results might, there­fore, well be disregarded. Chad appears as an influential outlier for many variables. Together with the inherent difficulty of surveying, income distribution in that country may justifiably lead one to doubt the reliability of its income-distribution data. With possible doubts on the reliability of data for all three outliers pulling against our conclusion, that conclusion can be seen as further strengthened.


In sum, the following seem plausible explanations for the divergences from expected values.

1. The distribution of assets, especially of land, reflects past history and current power-relationships. It would obviously affect income distribution, but was not included in the regression analysis because we could not find a good measure of asset distribution. This could affect the less-egalitarian-than-expected income distri­bution of such countries as Iraq (1956), Iran and the Dominican Republic, and the more egalitarian distribution in Taiwan (1972), Libya and Israel (1957).

2. The degree of dualism was not reflected in the regression since a dummy variable had to be used, which only indicates whether a country is dualistic, not by how much. As a result, the extent of inequality was not accurately predicted in countries with considerable dualism such as Gabon (1968) and Sierra Leone (1969).

3. Nor do the regressions take account of government policies limiting wages (or profits), or raising wages, or providing high windfall gains to particular segments of the elite. Only a great deal of knowledge about individual countries would enable one to take account of such factors in income distribution, which could have affected such countries as Brazil and Taiwan in 1953 (less egalitarian), and Israel (more egalitarian).

There may well be serious errors in some data, such as the astoundingly high Gini coefficient of 0.61 for Sierra Leone or the low 0.27 for Libya and 0.3 for Guatemala. Finally, in some countries more than one of the excluded variables probably come together: dualism, land tenure and policy in Rhodesia or land tenure and policy in Taiwan and Korea.

In any case, data on outliers suggest the direction of worthwhile further re­search. Since plausible explanation can be given for many of the outliers, their analy­sis may increase confidence in the basic results.






There is great variation in income distribution. Even if one leaves out extreme values because of doubt about their reliability, one finds that the share of the poorest 40 percent of the population in LDCs ranges between 6 percent and 7 percent and 22 or 23 percent of national income. Per capita income, or the Kuznets Curve, explains about 1.5 percent of the 16 percent of the range at low income levels, but its effect may be declining over time. Socio-political dualism an elite drawn from an ethnic minority explains about as much, but if it were possible to measure the degree of dualism, its explanatory power would most probably increase further.

Two variables which can be changed to some degree over several years also contribute significantly to variation in income distribution: the coverage of the edu­cational system and the degree of reliance on primary exports. Since massive primary exports can provide a powerful stimulus to growth, there may be a trade-off here between growth and equity. But for all the countries under examination there appears to be no conflict between the objectives of rapid growth and an egalitarian income distribution, a rather optimistic conclusion. Nor is there a clear trade-off between greater government intervention in mixed economies and greater inequality.

The large number of observations in our study to a substantial extent compensate for the unreliable nature of many income distribution data. As a result of that large number, even the most influential outliers do not appear to have a major impact on the results, which increases confidence in the validity of the conclusions. New findings of this analysis include:

·   some support for the argument that the inter-temporal Kuznets Curve does not exist;

·   good evidence that even in cross-country analysis its effect is rather weak: income distribution deteriorates only moderately as per capita income rises and the effect may be weakening over time;

·   the absence of any statistically significant relationship between the importance of the role of government in the economy and equality more interventionist governments do not achieve greater equality; and

·   the great importance of socio-political dualism in explaining inequality.

All these run counter to much of previous work. Consistent with previous findings are:


·   the relationship between the spread of education and equality, although the effect appears to be weak; and

·   the effect of a large role for primary exports on greater inequality.

A good deal of variation about half remains to be explained. Some of that is now picked up in regional variables. The distribution of land and government policies may be among the variables requiring further investigation. But it is encouraging that a significant deterioration in income distribution does not necessarily follow from rising incomes or more rapid growth. Also, contrary to some widely accepted beliefs, the absolute income of the poor rises with average per capita income (economic development), even if their share declines slightly.








OK Economics was designed and it is maintained by Oldrich Kyn.
To send me a message, please use one of the following addresses: ---

This website contains the following sections:

General  Economics:

Economic Systems:

Money and Banking:

Past students:

Czech Republic

Kyn’s Publications

 American education

free hit counters
Nutrisystem Diet Coupons