U.S. growth and income inequality: evidence of racial differences

Abstract

Numerous studies have assessed the relationship between income inequality and per capita income (i.e., the Kuznets curve) using various data and empirical techniques. In this study, we employ a gradual switching regression model to uncover structural change in the Kuznets curves for the U.S. aggregate, Black American, and White American segments of the population. Similar to prior studies, we find that the Kuznets curves have switched to a U-shape. Our results further show that structural change of the Kuznets curve has been gradual, generally beginning in the 1950s, and that the adjustment process has varied across race.

Notes

1 The opinions expressed are those of the authors and do not necessarily reflect views of California State University at Sacramento nor the California Department of Education.

1 Many other studies find empirical support for the current positive relationship between income and income inequality (Braun, 1988; Levernier, Partridge, & Rickman, 1998; Partridge et al., 1996; Ram, 1991, 1997).

2 Since several studies find income inequality to be sensitive to various factors, such as geography (Braun, 1991), marriage rates (Cancian & Reed, 1999;Levernier et al., 1998), education levels (Partridge et al., 1996), and the percentage of the population that is unionized (Levernier et al., 1998), provided the distribution of Black and White Americans differs across one or more of these factors, then income inequality may differ by race. For example, given that income inequality tends to be higher in regions where the percentage of female-headed families is higher (Levernier et al., 1998), coupled with the higher percentage of female-headed families among Black Americans (Darity, Myers, & Chung, 1998), this offers one explanation for the higher income inequality within the Black American population.

3 Our approach also differs from studies (e.g., Eusufzai, 1997) that assume the relationship between income inequality and income abruptly changes. Instead, by allowing for the possibility of a gradual adjustment in the curve, our model has nested within it various approaches. In particular, if it is found that t₁ coincides with the first period and t₂ coincides with the last period, then similar to some studies (e.g., Papanek & Kyn, 1986;Zang, 1998) the right-side variables simply interact with a linear function of time. Alternatively, regardless of t₁, if it is found that t₂=t₁+1, then the model is similar to others (e.g., Eusufzai, 1997) in that the structural change is immediate.

4 Although this approach does not pinpoint specific events that caused structural change, given that many U.S. policies have likely affected the distribution of income (such as policies addressing civil rights, poverty, and taxation), coupled with the re-distribution of income resulting from such policies likely taking some time, the use of a generalized procedure to allow for a gradual adjustment of the Kuznets curve is appealing.

5 In terms of the Gini coefficient, its value lies between zero and one, with values closer to zero (one) signifying a more equal (unequal) distribution of income. For example, if the Gini coefficient equals one, then this is tantamount to one person having all the income in the economy; whereas a value of zero for the Gini coefficient implies that everyone has an equal share of the income earned.

6 We also estimated Eq. (1), treating income inequality as a cubic function of income. However, for each inequality measure, the coefficients associated with the cubic term were insignificantly different from zero. Hence, we report the quadratic results in Table 3.

7 It is interesting to note that structural change began to occur in the period coinciding with initial efforts to address racial inequity (as evidenced by, among other events, the desegregation of the U.S. military in 1948 and the Brown v. Topeka Board of Education decision in 1954). Moreover, although many social policies were enacted in the 1960s, changes in income inequality began to occur prior to this period.

8 To check whether or not the results are sensitive to differences in the time span of the data across the three groups (i.e., 1947–1987 for Black Americans and 1947–1998 for White Americans and the aggregate), Eq. (1) was also estimated for White Americans and the aggregate with the data truncated to 1947–1987. The results were similar to those provided in Table 3 in that (1) the coefficients associated with the structural change remained significant, (2) t₁ remained the same for each regression, and (3) for Ratio t₂ remained the same. However, for the Gini coefficient, t₂ equals 1987 for both the White American and aggregate groups, suggesting that structural change continued to occur after 1987 (which is consistent with the results in Table 3). These results are available from the authors upon request.

9 Also, when evaluated at current per capita income levels, the slope of the Kuznets curve in the second regime, (α₁+γ₁)+2(α₂+γ₂)LY_t, is larger for Black Americans than for White Americans. Hence, not only is income inequality higher for Black Americans in recent years, but similar to findings of other studies (e.g., Tribble, 1996, 1999) the difference between Black American and White American income inequality is also increasing.