The roles of Time and Cohort Effects in Rising Wealth Inequality
This paper gauges how inequality evolves over the life cycle when controlling for time and cohort effects. To differentiate between time and cohort effects we construct two alternative regression models. The first model is constructed with a full set of age and year dummies, which controls for time effects. The second model is constructed with a full set of age and cohort dummies, which controls for cohort effects. We construct age profiles and plot the corresponding coefficients for the age dummies from both regressions. We find significant differences in the trends of inequality when controlling for cohort effects and time effects. If we adopt the cohort view, trends in cross-sectional inequality are attributed to successive cohorts starting out more unequally. If we adopt the time view, trends in cross-sectional inequality are attributed to all cohorts experiencing an increase in within-cohort inequality. We follow Heathcote, Storesletten and Violante (2005) (henceforth HSV) framework for analyzing whether rising inequality in the United States is best explained by time or cohort effects.
Obtaining the relative importance of time effects and cohort effects is difficult because age, period, and cohort (APC) are linearly dependent.2 A solution to the latter is to impose structural assumptions to normalize and overcome the identification issue. Deaton and Paxson (1994) maintain that the rise in consumption and earnings inequality can be attributed to age and cohort effects. They normalize year effects to be orthogonal to a time trend to overcome the identification problem.3 On the other hand, HSV (2005) follow the method of (Juhn, Murphy Pierce 1993; Ameriks and Zeldes 2001) and assume that APC effects are additively separable and only two of the three are operative. Different treatments for overcoming the identification problem produces statistically different results when decomposing the APC effects for the rise in wage, earnings, hours, and consumption inequality (Slesnick et al 2004, HSV 2005).4 The rise of life-cycle inequality in wages, earnings, hours, and consumption provides information about households’ ability to insure labour market risks. The age profiles of inequality improve the tractability of life cycle models consisting of heterogenous agents and incomplete markets (Heathcote, Perri, Violante 2009; Heathcote et al, 2004a). Age profiles of inequality for wealth have been neglected in the literature, due to difficulty measuring wealth and the changing composition of wealth across age groups.5 The effects of APC decomposition are used to measure and compare the sources driving the rising inequality in wages, earnings, and consumption over the life cycle in the United States. To our knowledge, this paper is the first to present a measurement of observed increases in wealth inequality that are due to cohort effects and time effects. Specifically, how much of observed increases in wealth inequality are due to younger cohorts starting out more unequal, as opposed to inequality within all cohorts increasing over time. We apply the decomposition method in HSV (2005) using the Survey of Consumer Finances (SCF) from 1989-2016.
This paper provides a measurement of observed increases in wealth inequality and presents inequality trends while controlling for cohort and time effects. Though the existing literature has applied similar methods to decompose income and consumption inequality over the life cycle, to our knowledge this paper is the first to present such measurements for wealth inequality. We differentiate cohort and time effect contributions to inequality trends and capture their relative importance for explaining the dynamics of cross-sectional inequality for United States using the model proposed by HSV (2005). We use a time series of cross-sectional data to follow cohorts of individuals from 1989-2016.
In this paper, we investigate trends in life cycle inequality of wage income and wealth to determine whether our results are invariant to assumptions about time and cohort effects. We find that when we control for cohort effects, the rise over the life cycle wage income and wealth inequality are significantly larger by a factor of 1.4 and 3.16, respectively. These results further motivate us to examine the relative significance of time effects and cohort effects for understanding the dynamics of wealth inequality in the United States from 1989-2016. The method used to study the importance of time and cohort effects follows that presented by Heathcote, Storesletten, Violante (2005). The correlation coefficient between the average across all age groups of the respective within-age changes, and the average across all cohorts of the respective within-cohort changes is 0.655, compared to HSV’s 0.647. We find that our results are consistent with HSV (2005), such that the dynamics of cross-sectional wage income inequality across time and age groups is best explained with the presence of time effects and absence of cohort effects. We extend HSV (2005) model to investigate the age profiles of wealth inequality with time and cohort effects. We find that within-cohort wealth does not grow over time, thus supporting the cohort view model. We show that the dynamics of cross-sectional wealth inequality across time and age groups are consistent with the presence of cohort effects and absence of time effects. We conclude that the rise in wealth inequality is better explained through successive cohorts starting out more unequally as opposed to all cohorts experiencing an increase in within-cohort inequality. Mechanisms that generate our empirical results are consistent with inheritance of inequality through intergenerational transmissions of wealth. Intergenerational transfers such as bequest, gifts, and human capital from old to young cohorts are plausible mechanisms driving successive cohorts to start out more unequally. The view that rising income inequality solely drives wealth inequality through household savings behavior cannot merely explain our empirical findings. Our results point towards intergenerational wealth as the source driving inequality. Wealth endowments aimed at younger cohorts are plentiful for the rich and non-existent for the poor, thus widening wealth disparities between successive generations. Our empirical findings have important policy implication for alleviating intergenerational wealth disparities through bequest, gift, and estate taxation. We find that mechanisms generating wealth inequality are linked with intergenerational economic transfers rather than significant differences between each successive generation.