The response of household incomes to stock price and GDP growth by income quantile

Abstract

How household incomes respond to GDP and stock price growth is important for an understanding of the economic costs of business cycles and the driving forces of income inequality over time. This article examines to what extent household incomes react differently across income distribution quantiles and time. It employs U.S. Panel Study of Income Dynamics data for the period 1979–2000 and quantile regression techniques. Significant differences are found in how household incomes respond across income quantiles. For the same income quantiles, large differences are identified when the time period 1979–1987 is compared to 1988–2000.

†The content of the article reflects the personal opinion of the authors and should not be associated with FedEx Services or any of its affiliated companies.

Acknowledgements

The authors thank Stuart Fowler and Albert E. Deprince, Jr., for helpful comments. The comments of an anonymous reviewer are greatly appreciated.

Notes

†The content of the article reflects the personal opinion of the authors and should not be associated with FedEx Services or any of its affiliated companies.

¹ Business cycle cost is defined as utility loss from consumption volatility. Households are assumed to prefer a smooth consumption profile based on their intertemporal choices to maximize lifetime utility.

² For example, as children grow up and enter the labour force, household income may increase substantially; and when they set up their own family, household income may decrease substantially.

³ The PSID calls it the head and wife's taxable income; this is because the PSID always assigns the male as the household head.

⁵ Quantile regression automatically sorts the dependent variable (hwty) to determine the distribution quantile point. Without averaging the head and spouse's taxable income, households of singles would cluster at the low end of the income distribution.

⁴ The PSID definition of income is as follows. Labour income equals the sum of the components labour part of farm income, labour part of business income, wages, bonus, overtime, commission, professional practice, trade, labour part of roomers, etc. Capital income includes the components asset part of farm income, asset part of business income, asset part of roomers, etc., rent interests dividend, etc.

⁶ The growth rates of GDP and the S&P500 index, which also appear in Table 1, are not part of the PSID.

⁷ Each of the three education variables is constructed as the sum of the equivalent education indicator variables for head and spouse to avoid collinearity problems.

⁸ The quantile points are derived for the pooled PSID data, which means that each quantile contains observations from various years.

⁹ Time subscripts are left out for simplicity.

¹⁰ A notable exception is the coefficient of the variable hage.

¹¹ The opposite signs for the sub-samples explain why the coefficient of GDP growth is not statistically different from zero in the regression on the complete sample.

¹² For the OLS estimates of Table 4, hage is even positive for the period 1979–1987.

¹³ This is consistent with recent trends identified by the College Board (2005).

¹⁴ The quantile estimates of Tables 6 and 7 come from the qreg estimator of the statistical package STATA.

¹⁵ The coefficient values of Table 8 are the differences between the estimated coefficients of the high and the low-income quantiles as reported in Tables 6 and 7. The t-values are based on bootstrapped standard errors. The iqreg command in STATA is employed to derive the results presented in Table 8.

¹⁶ Similar conclusions are reached by Hartog et al. (2001), who apply quantile regression to investigate the returns to education relative to the wage distribution. Sakellariou (2004) is another interesting application of quantile regression to the returns to education.

¹⁷ This is reflected in Table 8 by a smaller negative coefficient for urban for the 1988–2000 period than the 1979–1987 period.

¹⁸ Table 8 reflects the fact that age raises the income disparity between high- and low-income quantiles: the coefficients are positive. The coefficients also rise from the 1980s to the 1990s.

¹⁹ We note that this result provides some justification for unemployment compensation and other government programs that assist people in regaining employment.