Consumption Expenditure Inequality and its Relation to the Factors that Drive the Macroeconomy (1990 - 2019)

Abstract

This paper offers two sets of empirical facts regarding (I) the evolution of various dimensions of consumption expenditure inequality and (II) the manner in which these inequality variables are affected by key macroeconomic fluctuations, from 1990 to 2019 in the United States. First, using the Bureau of Labor Statistics’ Consumer expenditure Surveys, we construct 24 time series of different expenditure-based, crosssectional, and nationally representative inequality measures. Namely, we compute inequality in household expenditures on durable goods, nondurable goods, and services, plus a heretofore unstudied measure of welfare inequality that relates a household’s well-being to how it allocates purchases across luxuries and necessities. For each of these four, we compute percentile-ratio-based measures of top-half, bottom-half, and overall inequality at both the quarterly and monthly frequencies, yielding the 24 series. We then document the series’ long-run trends, seasonal properties, and unconditional ‘cyclical’ correlations with primary macroeconomic variables. This first set of exercises produces a wealth of interesting facts, among the most notable of which are the findings that (i) top-half inequality in durables expenditure substantially exceeds all other variables in both levels and growth during our sample period, (ii) across categories, consumption expenditure inequality exhibits marked seasonality that differs from seasonality in the corresponding aggregates, and (iii) recessions tend to reduce most inequality measures on impact. Second, using 114 monthly-frequency and 205 quarterly-frequency time series of U.S. macroeconomic indicators that we obtain from the FRED-MD and FRED-QD databases, we estimate via principal components analysis an approximate static factor model of the macroeconomy for each of the two sets of series. We then broadly characterize each factor on the basis of the kinds of macroeconomic time series on which it primarily loads (e.g., the ‘interest rate’ factor). Lastly, we estimate a battery of distributed lag models in which each inequality measure is regressed on each named macro-factor and its lags. This set of exercises also produces an abundance of noteworthy facts, such as evidence that (i) durables, non-durables, and welfare inequality are pro-cyclical to varying degrees while services inequality is counter-cyclical, (ii) greater exposure to asset price fluctuations at the top does not account for a substantial fraction of the decreases in certain types of inequality during downturns, but that this phenomenon may be driven in part by disproportionate cutbacks in luxury purchases at the top, (iii) interest rate movements affect durables inequality more than other inequality measures, and (iv) high-consumption households may be more forward looking regarding expected future economic conditions than low-consumption households, especially in durables expenditure.