Hypothesis 2:
The Time Shift in the Kuznets Curve 
There has been some discussion of a time shift in the Kuznets Curve. It has 

Conversely, one could argue that some of the same sociopolitical factors which made for greater equality in Europe are increasingly felt in the LDCs. After countries gained independence, some governments became more responsive to the demand of the poor majority, albeit with a lag (e.g. Tanzania, Sri Lanka). In other countries, land reform or the transfer of assets by populist or revolutionary regimes (e.g. Peru, Iraq, Libya) produced an irreversible change in income distribution. In still another group of countries, wages rose more rapidly than per capita income, as the result of increasing average product in the incomesharing sector. One can therefore hypothesize that there will be a lessening of any tendency towards greater inequality in the early stage of development. 



.
The tests of the alternative hypotheses are reported in Tables 5 and 6. The regressions on which they are based are the same as in Tables 2 and 3, but with the three timeshift variables added. However, the estimated parameters for variables other than those for the Kuznets hypothesis are not reported. The effect of the timeshift variables on the significance of these other parameters can be seen from the Fstatistics testing the joint significance of groups of variables. 

For the Gini coefficient (Table 5), all five variables of the shifting Kuznets Curve are individually significant and their coefficients decrease only slowly as economic factors and education are added in regressions 7 and 8. The values and significance of the three timeshift variables indicate relatively fast flattening (that is declining importance) of the Kuznets Curve over time. However, when the three sociopolitical variables are added (regression 9), the three timeshift variables lose significance. Since there has been no increase over time in the number of Communist countries in the data set, one of the sociopolitical variables, and the share of government investment (another variable) is not significant in any regression, it is most plausible that this loss of significance is related to the third variable, that for sociopolitical dualism. One can speculate that a decline in dualism, with the ending of colonialism in a few countries and of neocolonialism in a few others, is a factor in the flattening of the Kuznets Curve over time. We have so far not tested that speculation. 

The joint significance of the timeshift variables is pretty low, even in regression 6 where only the Kuznets variables are included. Significance is further reduced as other variables are added and disappears when the sociopolitical variables are introduced. This is one of the rare cases where a group of variables is jointly less significant than each of them separately. The low Fstatistic casts doubt on the significance of the time shift. The hypothesis of a stable Kuznets Curve can not be rejected. Moreover, when the share of the poorest is the dependent variable (Table 6), the time shift is only weakly significant in the regression with only the Kuznets variables (6) and becomes weaker as other variables are added. The timeshift variables are never jointly significant. 



There is, therefore, little statistical support for the hypothesis that the effect of per capita income on income distribution may be weakening over time. The facts that coefficients all have the right sign, are reasonably stable and are significant in some regressions suggest that the flattening of the Kuznets Curve during the 20 years observed, illustrated in Fig. 1 and Fig. 2 (where fitted values are plotted), warrants further exploration as data on income distribution become available for the mid and late 1970s for additional countries. That is especially desirable, because if one were to extrapolate the estimated linear trends, the Kuznets Curve would have all but disappeared, or even reversed itself, by the early l980s. Given the lack of statistical support, neither extrapolation nor firm conclusions are warranted on the gradual flattening of the Kuznets Curve over time on the basis of our data, but it is a possibility worth investigating further. 



