Publications  Income Distribution

 

 

THE RESULTS

 

After reporting (in Table 1) the results for the intertemporal Kuznets Curve, the bulk of the paper analyses a combination of the effects of the inter-temporal and cross-country Kuznets Curves. These results, with the Gini coefficient as the dependent variable, are. given in Tables 2 and 5; for the share of the poorest 40 percent they are given in Tables 3 and 6. Tables 2 and 3 show a stable Kuznets Curve while Tables 5 and 6 introduce the time shift in that curve. The results overall are quite good, although the statistical significance of some estimated parameters is low. Regressions with all right-hand variables explain more than 60 percent of the variation in the Gini coefficient and nearly 60 percent of the variation in the share of the poorest 40 percent. However, some of the hypotheses are not supported.

 

 

Hypothesis 1:

Per Capita Income -The Kuznets Curve

 

One of us had argued earlier that the widespread statistical support for the Kuznets hypothesis was due to the exclusion from models of those variables which represent the true causes of the variation in income distribution, most notably the export of primary products and a dualistic socio-political structure (Papanek 19]). Because these variables are most likely to be correlated with the level of development, per capita income served as a proxy for the omitted true explanatory variables, thus producing an artefact which appeared to confirm the Kuznets hypothesis.

The structural changes resulting in the Kuznets Curve were set out originally by Kuznets  and elaborated by others e.g. Cline, Ahluwalia, Bacha. A clear theoretical basis could have been provided by the Lewis and Fei-Ranis  model: the poorest countries would include all labor surplus eco­nomies, where the real wage, or real labor income, remains unchanged in early stages of development.8 Therefore, as development proceeds, initially all the addi­tional income would accrue to owners of physical or human capital. Real labor income would rise only when enough labor has been transferred from agriculture to industry to raise the marginal product of labor in agriculture to the level of the agricultural wage, which has been kept above the marginal product by institutional factors. In this model, the faster the rate of growth, the more rapid the decline in the relative share of the poor, since their income is derived from labor at an unchanged wage.

 An alternative model has been suggested (Papanek; Manove and Papanek) in which labor income is related to the average product in work-and-income-sharing activities. It can therefore increase even if the marginal product of labor remains below the wage (and may be zero). In that model, the change in income dis­tribution depends on the relative rate of change in income from capital (physical and human) and from labor. Some preliminary evidence has been advanced that the real wage changes with the average product in agriculture. This model, therefore, provides plausible theoretical and empirical reasons for hypothesizing that the Kuznets Curve does not exist at an early stage of development.

 

 

Since there are plausible reasons for both sides of the argument, we turn next to empirical tests. The evidence is quite mixed, but tends to suggest that our hypothesis should be rejected and that the Kuznets Curve may exist.

Most empirical studies of the Kuznets Curve have relied on cross-country data, for lack of adequate time-series. Since the Kuznets Curve is supposed to describe a temporal relationship, this is less than satisfactory. We do not have enough observa­tions for any country to use a pure time-series regression, but we have at least two observations at different times for 39 countries (34 in the case of shares), although per capita income and time interval between observations differ for different coun­tries. These data were used to investigate the existence of the inter-temporal Kuznets Curve. The estimates of model (2), described in the Methodology section, are presented in Table 1. 

Display Table 1.

this window

new window

 

 

 

But instead of using the country dummy variables, we run the regression with both the left-hand and right-hand variables expressed as deviations from country means. The resulting estimated parameters for the two Kuznets Curve variables and also their standard errors and t-statistics are exactly the same as if the regressions were run on the original variables but with country dummies. How­ever, R-squares are different. The F-statistic tests the joint hypothesis that the coefficients for both Kuznets variables are simultaneously equal to zero. The results in Table 1 do not support the Kuznets hypothesis.

For the regressions for all countries (regressions 1 and 3) the Kuznets Curve coefficients have the wrong signs for both the Gini coefficient and the share of the poorest 40 percent. For the Gini their significance is weak (jointly 8.9%), but for the Share it is considerable (jointly 5.4%, individually below 2%). Only if Taiwan is excluded, which is not really justified (see above), do the Kuznets Curve variables have the right sign. Even then the Curve is flat and either completely insignificant (Gini) or only somewhat significant (for Share the joint significance is 6.9 percent, although separately the coefficients are insignificant). Excluding other outliers does not really change the results. The net effect of these regressions is to reject the existence of the inter-temporal Kuznets Curve. However, the data base is poor, because we have very few observations for each country.9 Inter-temporal analysis is potentially more significant than the pooled cross-country time-series data, since the Kuznets hypothesis is important largely because it is assumed to indicate what is likely to happen when per capita income rises with development. At the very least, therefore, these findings suggest that it is highly desirable to apply further tests for the Kuznets hypothesis as additional time-series data on income distribution accumu­late. If taken at face value, they imply the absence of the inter-temporal Kuznets effect.

 

 

All the following tests of the Kuznets hypothesis will assume the combined inter-temporal and cross-country Kuznets Curve. The results of this regression analysis, shown in Tables 2 and 3, tend to lead to a rejection of our hypothesis and to the conclusion that the Kuznets Curve exists, especially for the Gini coefficient as a dependent variable. Considering the lack of evidence for the inter-temporal Kuznets effect, this result seems to be due to the cross-country Kuznets effect. Although most of the additional variables are significant, the explanatory power of the Kuznets Curve did not disappear when they were added.

Display Table 2.

this window

new window

Display Table 3.

this window

new window

For the Gini (Table 2), the Kuznets Curve parameters have the right sign and are significant, both individually (t-statistics over 2) and jointly (probability values ranging from 1% to 5%). For the share of the poorest 40 percent (Table 3), the co­efficients have the right sign and are reasonably stable, but the t-statistics are only between 1 .0 and 1 .8. The joint significance of the Kuznets Curve variables has a probability value of only 16 percent to 26 percent, that is barely significant to insignificant.

 

 

 

The pooled cross-country/time-series results for the Gini are consistent with other studies, all but one of which (Papanek [18,19] is the exception) support Kuznets hypothesis. But the weakness of the Curve should also be noted. When the definitional variables are added to the Kuznets variables, as they should be, the R2 is still only 0.35. Other significant variables raised that figure to about 0.6. As can be seen from Table 4,  higher per capita income has relatively little effect in worsening income distribution once it exceeds $100 (in 1964 prices). While inequality conti­nues to increase until it reaches about $300 (for the Gini) or $400 (for the Share both in the more complete regressions), the estimated deterioration is small. With a complete regression, including all variables tested (except regions), the drop in the estimated share of the poorest 40 percent between $100 and $400 is 1.2 percent; the rise in the Gini is 0.02. The estimated deterioration in the share of the poorest also exists as countries move from $70 or $80 per capita to $100, but there were only 10 countries in our group of 85 which had such low incomes in the past and several of these have now moved beyond this category (even in the 1964 dollars), e.g. Indonesia, India, Pakistan, and Malaysia, so theirs is not a very widespread problem.10 The initial improvement in income distribution beyond $400 to $1 ,000 per capita is also not large, reaching only 0.02 of the Gini and 0.4 percent in the share of the poorest 40 percent. However, as per capita income continues to rise, the estimated improve­ment becomes quite large, with the share of the poorest at $5,000 at 3.1 percent above $1,000. However, only 5 countries in our sample had per capita incomes above $2,000 during the period under review.

Display Table 4.

this window

new window

The real concern has been with the presumed deterioration in distribution as per capita income first rises with development. But except for the handful of countries still below $100 (in the 1964 prices), the evidence is that the deterioration is rather small. These results provide little comfort to governments or societies that claim that an unequal income distribution is not their responsibility, but is due to the inevitable increase in inequality which accompanies development. Nor do they provide support to those who argue that massive government intervention is neces­sary to prevent a severe deterioration that would otherwise inevitably take place. In other words, the Kuznets Curve exists but appears to be quite flat in the relevant range. And once a country has passed through the plateau between $200 and $400, the Kuznets Curve works in its favor according to this analysis: other things being equal, income distribution will tend to become more equal. Altogether these are more optimistic conclusions than are generally drawn on the basis of the original Kuznets hypothesis.11

When the results of both tests are taken into account, one can reasonably con­clude that the cross-country Kuznets Curve may well exist, but the limited evidence that we have for the inter-temporal Kuznets effect is negative. In any case, the effect of the Kuznets Curve on income distribution seems to be quite weak. 

 

 

 

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

okyn@bu.edu --- okyn@verizon.net

This website contains the following sections:

General  Economics:

http://econc10.bu.edu/GENEC/Default.htm

Economic Systems:  

http://econc10.bu.edu/economic_systems/economics_system_frame.htm

Money and Banking:

http://econc10.bu.edu/Ec341_money/ec341_frame.htm

Past students:

http://econc10.bu.edu/okyn/OKpers/okyn_pub_frame.htm

Czech Republic

http://econc10.bu.edu/Czech_rep/czech_rep.htm

Kyn’s Publications

http://econc10.bu.edu/okyn/OKpers/okyn_pub_frame.htm

 American education

http://econc10.bu.edu/DECAMEDU/Decline/decline.htm

free hit counters
Nutrisystem Diet Coupons