5.1.3. The Relation Low Income - Fast Growth Hold in the Long-Run

In his study of industrial distribution of income by states, Kuznets (1958, p. 42) writes that “the states with the highest rates of growth in per capita income had low per capita incomes in the initial year; and those with lowest rates of growth in per capita income had initially high per capita incomes (…). In other words, in the country’s growth during recent decades interstate inequalities in per capita income were reduced.”

However, Kuznets analyzed only five years (1919-21, 1929, 1940, 1950, and 1955) to draw his conclusion on. Using the series of the state average of per household income, we oppose here the number of low-income states with fast growth rates to the number of low-income states with slow rates of growth. ²⁶ Again, the data shed light on this point. The two panels of the figure below provide a visualization of the results. The first one utilizes the full size of the sample (1913-2003), and the second duplicates the former for selected years only (1919-21, 1929, 1940, 1950, and 1955 from Kuznets’ perspective).

Figure 5.2 suggests convergence because the low-income states recorded fast growth rates 52 times out of the 90 years considered. This number outbids the number of years when low-income states recorded slow growth rates (32 years out of 90). The remaining 6 years were featured by the neutrality of the relation between income levels and growth rates. It is not surprising that both Kuznets and Barro concluded to convergence because the years they considered displayed more states where initial incomes were negatively correlated to growth rates.

Figure 5.2 reveals that the negative correlation switched to the positive correlation in recessionary years: the aftermath of 1929’s black Friday, the early 1950s, early 1960s, almost the full decade of the 1980s, and the second half of the 1990s. At these points in time, there were more rich states growing fast than rich states growing slow. ²⁷ This remark suggests that the β convergence does not hold anymore in the downward phase of the Juglar’s cycle, whose complete occurrence usually takes eight to ten years.

Note the difference in income measurement, though. Kuznets used per capita income (and not per household income), thereby making comparisons less unequivocal. Note also that Kuznets’ main argument was not to demonstrate the links between growth rates and income differentials, but to inform on the industrial distribution of income among the sectors of agriculture (A), manufacturing (M) and services (S), at a very refined degree of disaggregation. For 1919-1921, 1929, 1940, 1950, and 1955, he arrayed the percentage increase in states per capita income in descending order, and partitioned them into six groups of 8 states each. He thoroughly performed both cross-section and trend over time analyses, and concluded: ²⁸

‘The share of the A sector declined in all six groups of states; but the decline was greater in the states in which growth in per capita income was greatest. (…) Thus the results for the share of the A sector conform to the expectations derived from the cross-section analysis.’ ‘The changes in the share of the M sector also agree with our expectations. The share rose in all six groups, but it rose more in the groups in which per capita income grew at a higher rate.’ ‘The share of the S sector also rose fairly substantially in all six groups; but the association between the rate of growth of per capita income and the increase in the share is not as clear-cut as for the A and M sectors.’

The industrial distribution of income across states is a topic where boundaries lie beyond the scope of the present work. The next step is to address the same issue of inequality and growth, through the traditional concept of the β convergence.