Any successful consumption theory must accommodate three basic patterns: the rich save at higher rates than the poor; national savings rates remain roughly constant as income grows; and national consumption is more stable than national income over short periods.
The first two patterns appear contradictory: If the rich save at higher rates, savings rates should rise over time as everyone becomes richer. Yet this does not happen. [James] Duesenberry [author of Income, Saving, and the Theory of Consumer Behavior] suggested that the explanation of the discrepancy is that poverty is relative. The poor save at lower rates, he argued, because the higher spending of others kindles aspirations they find difficult to meet. This difficulty persists no matter how much national income grows; hence the failure of national savings rates to rise over time.
To explain the short-run rigidity of consumption, Duesenberry argued that families look not only to the living standards of others, but also to their own past experience. The high standard enjoyed by a formerly prosperous family thus constitutes a from of reference that makes cutbacks difficult, which helps explain why consumption levels change little during recessions.
Despite Duesenberry's apparent success, many economists felt uncomfortable with his relative-income hypothesis, which to them seemed more like sociology or psychology than economics. The profession was therefore immediately receptive to alternative theories that sidestepped those disciplines. Foremost among them was Milton Friedman's permanent-income hypothesis, which still dominates research on spending.
Friedman argued that a family's current spending depends not on its current income, but rather on its long-run average, or permanent, income. Because economic theory predicts that people prefer steady consumption paths to highly variable ones, Friedman argued that people would smooth their spending--saving windfall income gains and drawing down savings to cover windfall losses. Consumption should thus be more stable than income over short periods. Friedman also argued that a family's savings rate should be independent of its income, leading him to predict the long-run stability of national savings rates.
Friedman dismissed the high savings rates of the rich as a statistical artifact. Because many of those with high measured incomes in any given year will have enjoyed positive windfalls, their permanent incomes will be lower, on average, than their measured incomes for that year. So if they save windfall gains, they will save a higher portion of their measured incomes than of their permanent incomes. The converse holds for those with low measured incomes in any given year, who will have experienced a preponderance of windfall losses that year.
Although this is a tidy story, its fundamental premises are contradicted by the data. As numerous careful studies have shown, for example, savings rates rise sharply with permanent income. Friedman's defenders responded by arguing that rich consumers want to bequeath money to their children. But why should the poor lack this motive? Another problem is that people consume windfall income at almost the same rate as permanent income. To this, Friedman responded that consumers appear to have unexpectedly short planning horizons. But if so, then consumption does not really depend primarily on permanent income.
Strangest of all, Friedman's theory assumes that context has absolutely no effect on judgments about living standards. It predicts, for example, that an investment banker will remain equally satisfied with his twin-engine Cessna even after discovering that his new summer neighbor commutes to Nantucket in an intercontinental Gulfstream jet.
In light of the abundant evidence that context matters, it seems fair to say that Duesenberry's theory rests on a more realistic model of human nature than Friedman's. It has also been more successful in tracking actual spending. Under the relative-income hypothesis, for example, it is easy to understand why a majority of families experience significant retrenchments in living standards when they retire. Under the permanent income hypothesis, this observation is a jarring anomaly. And yet Duesenberry's relative-income hypothesis is no longer even mentioned in leading textbooks.