![]()
CHAPTER IV
Consistency of the Permanent Income Hypothesis with Existing Evidence on the Relation between Consumption and Income: Budget Studies
ONE of the aims of this and the next chapter is to document the consistency of our hypothesis with a number of broad empirical findings that strikingly contradict the impression that consumption is a stable function of absolute real income and that this function can be identified with the regression of consumption on income computed from either budget studies or time series. These findings, cited at the outset, are: (a) the rough constancy of the average propensity to consume in the United States over the past half-century, as measured by time series data, despite a substantial rise in real income; (b) the rough similarity of the average propensity to consume in budget studies for widely separated dates, despite substantial differences in average real income; (c) the sharply lower savings ratio in the United States in the period after World War II than would have been consistent with the relation between income and savings computed from data for the interwar period. Another finding of the same kind is (d) the apparent decline over time in the inequality of income despite the possibility of interpreting the consumption-income relation from time series or budget data as showing that the rich are getting richer and the poor, poorer.
But consistency with such broad findings would by itself not be much of a recommendation. Accordingly, these chapters examine the consistency of the permanent income hypothesis also with more detailed evidence on consumption behavior. This chapter compares it with evidence from budget studies; the next, with evidence from time series. Although the empirical evidence examined is by no means exhaustive, it covers a fairly wide span of time and a fairly broad range of phenomena. Its chief defect is that so much of the evidence is for the United States. This is partly because more empirical work has been done for the United States, particularly in recent years; partly because my knowledge of the work that has been done in other countries is more limited, and I have relied mainly on material that was fairly readily available.
As noted in the preceding chapter, the permanent income hypothesis accounts for the broader features common to observed regressions of measured consumption expenditures on measured income computed from budget data: the tendency for expenditures to exceed income at low incomes, and to fall short of income above some point; the uniform tendency for the ratio of consumption to income to be lower, the higher the income. We turn now to consider the consistency of the hypothesis with (1) temporal changes in the inequality of income; (2) differences among regressions for (a) widely spaced dates, (b) different countries, (c) farm and nonfarm families, (d) different occupational groups and (e) Negro and white families; (3) the relation between savings and age; and (4) the effect of changes in income on the relation of measured consumption to measured income.
1. Temporal Changes in Inequality of Income
A tendency for the inequality of income to increase over time has frequently been inferred from the regressions of consumption on income computed from budget data. These regressions show savings to be negative at low measured income levels, and to be a successively larger fraction of income, the higher the measured income. If low measured income is identified with “poor” and high measured income with “rich,” it follows that the “poor” are getting poorer and the “rich” are getting richer.
The identification of low measured income with “poor” and high measured income with “rich” is justified only if measured income can be regarded as an estimate of expected income over a lifetime or a large fraction thereof. One step in this direction is taken when the computed regressions are regarded as estimates of a stable function relating consumption to income, since, in our terminology, this is equivalent to regarding them as estimates of the relation between the permanent components of consumption and income. It requires only the additional step of giving “permanent” the special meaning of expected income over a large fraction of a lifetime to make increasing inequality of income a valid inference from the observed regressions.
On the permanent income hypothesis, the observed regressions give no evidence on the secular behavior of the inequality of income. Negative savings at low measured incomes reflect precisely the fact that measured income is not a valid index of wealth; that many people have low incomes in any one year because of transitory factors and can be expected to have higher incomes in other years. Their negative savings are financed by large positive savings in years when their incomes are abnormally large, and it is these that produce the high ratios of savings to measured income at the upper end of the measured income scale. The existence of large negative savings is a symptom that the observed inequality of measured income overstates substantially the inequality of permanent income. It is not a harbinger of a widening gap between rich and poor.
Empirical data show no tendency for inequality of income to increase. If anything, inequality seems to have been decreasing in recent decades.1 This fact has been widely recognized, and so has the possibility of reconciling it with the observed regressions of consumption on income by considerations very similar to those embodied in our hypothesis. But this has not prevented these regressions from being treated as describing a relation between permanent components in other contexts where the contradiction with available evidence is less obvious.
2. Consumption-Income Regressions for Different Dates and Groups
Table 1 summarizes some of the salient findings of budget studies for a wide range of dates and groups of consumer units. The studies summarized in this table are a major source of the evidence examined in this section.
For the United States, the average propensity to consume is remarkably similar for different studies covering either nonfarm groups alone or all consumers (Table 1). Eight of nine values are between .89 and .92, and these are for years as much as six decades apart; the one deviant is for 1944, and reflects the extraordinarily high level of savings during World War II. Time series data show both the same similarity of the average propensity for different periods of time and an average propensity very nearly the same in magnitude (see Table 12 in the next chapter). On a definition of consumption like that used in these budget studies—one which treats as consumption, expenditures on consumer durable goods other than housing—th...
