The paper reconsiders the motivation for the PIH and shows that alleged inconsistencies between household and aggregate spending estimates are derived from non-comparable data and econometric misperceptions. The actual cause of the conflict is that annual shifts in the means of the household data are incorporated in the aggregate data which in turn causes aggregate marginal propensities to reflect not only income effects but also shifting effects. This latter effect can be eliminated by recalculating the household data in deviation form.
Another old idea is the suggestion that household data can be grouped by criteria other than income to eliminate the problem of transitory income influences on spending behavior. Unfortunately, appropriate data was not available when this regrouping was proposed in the mid1940s. However, Consumer Expenditure Survey data compiled since 1984 by the U.S. Bureau of Labor Statistics allows household data to be grouped not only by income level and income quintile but also by occupation, education and age.
Results from estimating simple Keynesian type consumption functions indicate that transitory influences are significant because mpcs found with income level data are 70 to 80 percent lower than those found when data is grouped by some non-income criteria. Cross sectional and time series differences disappear when the household mpcs are estimated with data in deviation form. Household data aggregated over 27 years yield mpcs approximately equal to the average of those estimated 27 times in cross section.
Overall, differences between household and aggregate estimates of household spending behavior are explained by differences in the data used to determine the estimates rather than by differences in behavior.