Economists have long been interested in the factors determining how a society divides its income proportionally between consumption and saving. In the past thirty years theoretical and empirical investigation of these factors has been focused by the concept of the consumption function—a list of the variables that influence consumption, together with the direction and magnitude of their effects. Income itself is, of course, high on any such list; and much of recent investigation has concerned the nature, reliability, and measurement of the dependence of consumption on income.
Since saving is by definition the difference, positive or negative, between income and consumption, whatever relationships are summarized in a consumption function can be summarized equally well in a saving function. It is a matter of indifference whether attention is directed to the determinants of consumption or to the determinants of saving; the results of one approach can always be translated into the other. For example, the proposition that U.S. households in the aggregate spend for consumption 95 per cent of their income after taxes can also be stated by saying that they save 5 per cent of the income at their disposition. In either form, incidentally, the statement is no better than a good first approximation.
Why have economists been interested in the division of social income between consumption and saving? There are two main reasons.
The first principal reason is the importance of saving for accumulation of the wealth of nations and for growth in their capacity to produce goods and services. Broadly speaking, consumption uses productive resources in the present, while saving enlarges the resources available for production and consumption in the future—by increasing stocks of finished goods or materials, productive plant and equipment, and net claims on foreign countries.
In practice it is difficult to measure these concepts of social accounting—net national product or income, consumption, saving, and wealth—in ways that correspond to the fundamental distinction between provision for the present and provision for the future. Very likely, saving, as measured in our national accounts, understates our society's provision for the future. One problem is to measure the value of the services rendered by existing stocks of producers' and consumers' durable goods, and to allow for their depreciation and obsolescence. This is difficult enough for business plant and equipment. Housing is the only consumers' durable asset for which such accounts are estimated; for other consumers' durable goods, consumption is simply equated to new purchases. And no attempt is made in many national accounts (including those of the United States) to estimate the accumulation of real wealth by governments, even before allowance for depreciation.
An even greater practical and conceptual problem, which has attracted considerable research interest recently, is to identify and to measure the saving embodied in human beings in the form of increased education, new skills, and greater capacities. The national accounts do not now count any educational outlays as saving or associate with them any increase in national wealth [seeCapital, human].
In spite of these and other difficulties, there can be no doubt of the importance for economic growth of the processes of individual and social decision that determine the share of saving, as conventionally measured, in net national product. This has been underscored in recent years by the shortages of capital confronting the less developed countries.
The second principal reason for concern with the consumption function is an outgrowth of the depression of the 1930s and of the revolutions in economic thought and policy to which it led. An economy will not produce at the rate that its manpower and capital resources permit unless total effective demand for goods and services suffices to purchase its capacity, or “full employment,” output. If private consumption demand falls short of capacity output, the difference must be made up by nonconsumption spending, that is, private investment at home or abroad and government expenditure. If these sources of demand do not absorb the saving that the economy would perform at full employment, then output, employment, and the use of industrial capacity will all fall short of their full employment levels. The national propensity to save will not be realized in additions to national wealth but wasted in unemployment and idle capacity.
The observation that saving is not always an unmixed blessing has a long history, associated with “underconsumption” explanations of the business cycle. But it received its most sophisticated, convincing, and influential expression in J. M. Keynes's General Theory of Employment, Interest and Money (1936). Keynes emphasized that consumption is on the whole a predictable and reliable component of aggregate demand. He saw investment spending, in contrast, as inherently volatile and unstable. And in the periodic failure to use the potential saving of the economy he found the main reason for the repeated lapses of capitalist economies into recession and depression. He feared, further, a chronic tendency in wealthy countries for full employment saving to outrun investment, leading in the absence of corrective policies to chronic unemployment.
The same theoretical apparatus can be applied to the opposite problem, excess demand and inflation, although this was not the economic disease originally at the center of Keynes's attention. Inflation occurs when nonconsumption spending, by governments or private enterprises, more than fills the gap between full employment output and the consumption expenditure that it normally induces.
In recent decades Keynes's theory has been the major impetus to research on the consumption function. This theoretical impetus was reinforced by the simultaneous development of national accounts providing the statistical raw material for empirical research on the subject. Note that the Keynesian motivation for understanding the social choice between consumption and saving does not require, to the same degree as the growth motivation first mentioned, close identification of consumption with the present and saving with the future. What is more important for Keynes's purpose is identification of consumption with the predictable, and investment with the volatile, elements of nation expenditure.
Keynes's consumption function is a simple relationship between national consumption—and accordingly national saving—on the one hand, and national income on the other. He called this relationship “the propensity to consume” and derived certain conclusions as to its form from what he asserted to be a “psychological law”—that the community will divide an increase in income in some regular proportion between an increase in consumption and an increase in saving. That is, both the “marginal propensity to consume” (mpc) and the “marginal propensity to save” (mps) are between zero and one. (By definition the two marginal propensities sum to one.) This is all his theory required, but Keynes went further and speculated that the “average propensity to consume” (apc), the share of national income consumed, would be found to decline with increases in total income. This decline could reflect either or both of the following: (a) the mpc falls with income; (b) a certain component of consumption expenditure is independent of income. This means that the apc will be lower for higher incomes even if the mpc is constant, as is illustrated in Figure 1. Evidently Keynes believed in both.
Figure 1 displays a Keynesian consumption function and in the lower panel the corresponding saving function. Net national income, Y, is measured along the horizontal axes. The vertical axis represents consumption, C, in the upper panel and saving, S, in the lower. The 45° line from the conventional origin in the upper panel is a reference line, showing how much consumption would be if it were always exactly equal to income. The other line,
more gently sloped, is the consumption function. It shows that below a break-even level of national income, Y∘, the nation will wish to consume more than its income, drawing on past savings. Above Y∘, however, consumption falls short of income, by an increasing margin. The mpc is the slope of this line, m, positive but less than one. Given a one-dollar increase in income, the community will increase its consumption by $m and increase its saving, or diminish its dissaving, by $(1 – m). In the diagram the mpc is constant; a diminishing rape would be pictured by a consumption function curving away from the 45° reference line, and by an upward-curving saving function. But even along the linear consumption function in Figure 1, the apc declines with income. Below Y∘ the ratio of consumption to income, C/Y, exceeds one; at Y∘ it is exactly one; above Y∘ it is below one.
The use that Keynes made of the consumption function in explaining unemployment can also be indicated in Figure 1. Net national product or income, Y, is three things at once: (a) The sum of the incomes earned for productive services (wages and salaries, interest, rents, profits), plus taxes paid by businesses prior to the payment or calculation of these factor incomes. (National accounting practice distinguishes net national product and national income by deducting these “indirect business taxes” from net national product in order to arrive at national income. This means that net national product measures the market value of net output, at prices that include these taxes, whereas national income values production at prices corresponding to the income payments to productive factors. This distinction is not made in this article; the two terms are used interchangeably to refer to the market value of net output.) (b) The total market value of goods and services produced, (c) Total purchases of goods and services, the sum of consumption expenditure, private investment, and government purchases. To sustain any given level of production and income, say Yf, total spending must add up to Yf. Since consumption spending will amount only to Cf, nonconsumption spending must make up the difference. If Yf represents the full employment potential of the economy, failure of nonconsumption spending to reach If would mean failure of production and income to attain Yf. The result would be unemployment and excess capacity. (Likewise, should nonconsumption spending exceed If, total demand would tend to exceed Yf, causing inflation.)
Further, the consumption function enabled Keynes to say by how much national income would fall short for a given shortfall in nonconsumption expenditure. A difference of a dollar in nonconsumption expenditure means a difference of more than a dollar in national income. The multiplier, which expresses this relationship, depends on the mpc. In Figure 1, for example, if nonconsumption spending is equal to zero instead of to If, income will be at the break-even level, Y∘, instead of at Yf. The difference in income, Yf – Yf, is 1/(1 – m) times the difference in nonconsumption spending If. The multiplier is l/(l – m) . Thus, if the national mpc is ¾, the multiplier is 4; if the mpc is f, the multiplier is 3; and so on.
The multiplier is a measure of the response of total production, income, and employment to changes in nonconsumption spending, whether these changes reflect natural volatility of private investment or conscious policy in public expenditure. It was in the latter context that the concept of the multiplier was first advanced, by Kahn (1931), as a means of estimating the response of employment to public works expenditure during the depression.
The dynamics of the multiplier can be exhibited by assuming, along with Kahn and many other writers, a lag in the adjustment of consumption to income. Suppose nonconsumption spending was initially zero and income and consumption were in equilibrium at C∘ = Y∘. Now nonconsumption spending rises to If and stays there. Initially, income and production rise by a like amount, that is, to Y1 = C∘ + If . But then consumption adjusts to Yt along the consumption function and rises to C1 . This raises income to Y2 = C1 + If. The process converges to income Yf, where the consumption function indicates consumption of Cf, leaving just enough room for the new level of nonconsumption spending If.
For an example with numerical concreteness, take m = 3/5 and suppose the sustained increase ∆I in nonconsumption spending to be 10. Then the process is as presented in Table 1, where ∆Y, ∆I, and ∆C are the increases in these variables over their initial levels.
|”Round”||∆1||∆Y = ∆1+∆C|
The dynamic formulation serves to make two points. First, stability requires that the national mpc be smaller than one. Otherwise the multiplier process would never converge. An initial stimulus would keep spending and income rising indefinitely. (It is a mistake, therefore, to conclude from the formula, multiplier = 1/(1 – m), that the multiplier is negative if m exceeds one. Rather the whole multiplier concept is inapplicable in that case.) Second, assuming the mpc to lie in the normal stable range, as Keynes assumed, a one-shot injection of investment or government expenditure cannot lift income more than temporarily. A sustained higher level of income requires a sustained higher level of nonconsumption spending. Multiplier theory offers no comfort for “pump primers.” A temporary stimulus can work only if there are some unstable elements in national spending that can be set into sustained motion by an initial surge in economic activity. According to Keynes, consumption is not an unstable element of this kind.
Price and population adjustments
The consumption-income relation is intended to link real per capita magnitudes, measured in constant prices. An increase in national income in Figure 1, for example, represents an increase in production and purchasing power rather than an increase in the prices at which production is valued. The community cannot be expected to react to a doubling of money income in the same way when it reflects a doubling of prices and wages (with real output constant) as when it reflects a doubling of output (with prices constant). Creating a new franc by deleting two zeros from the old one should not change anyone's basic behavior. Of course, actual price movements are not so uniform, universal, and neutral as de Gaulle's decimal reform. Consequently both the level of prices and their rate of change may affect the relation between real consumption and real income; some of these possible effects will be mentioned below.
For somewhat similar reasons, the variables in the consumption function must also be corrected for changes in population. The logic of the argument that the average propensity to consume is lower when society is better off implies that an increase in real national income accompanied by an equal proportionate rise in population would leave the average propensity to consume unchanged. Consequently the variables in Figure 1 should be regarded as per capita measures. (The theoretically appropriate correction for population change is somewhat more complicated. Presumably persons of different age, sex, and family status have different consumption requirements and should receive different weights, as in the “equivalent adult” procedures used in studies of food consumption.)
However, this gross correction for population change is not the end of the story, any more than simple “deflation” for price changes disposes entirely of price effects. As noted below, both price and population changes play important roles in recent theories of the consumption function.
The reliability of the multiplier
Although the propensity to consume and the multiplier have become standard items in the tool kits of economists, these concepts have not gone unchallenged. When quarterly income and consumption data became available in the United States after World War II, they showed erratic fluctuations in the response of consumption to quarter-to-quarter changes in income. And at times, notably during the Korean War, there have been wide swings in the propensity to consume. Some critics have questioned the view that consumption is a particularly stable or predictable element in national expenditure, that consumers respond passively to shocks originating in business investment and government expenditure. In their view the multiplier is not a reliable tool of analysis and prediction, certainly not one that can be applied mechanically. It should be noted that the multiplier multiplies error as well as truth in translating fluctuations of nonconsumption expenditure into predicted changes in national product.
In a series of statistical studies, Friedman and Meiselman (1963) have argued that multiplier analysis performs less well in explaining fluctuations of national product than an alternative model of equal simplicity relating national product to the quantity of money. In a different vein, Katona (1960) has contended that consumers in a modern rich society possess sufficient discretion and autonomy to originate, not just respond to, fluctuations in over-all economic activity. Consequently he has pioneered in conducting surveys of consumer intentions and attitudes, which can be used to improve the short-run forecasts of consumer behavior made from conventional economic variables (see Katona & Mueller 1953; 1956). The record suggests that these surveys contain useful forecasting information, at least for the short-run timing of the durable goods component of consumer expenditures. [SeeSurvey analysis, article Onapplications in economics.]
A doubt of somewhat different import concerns a possible ambiguity in the direction of causal influence between income and consumption. Keynes and virtually all subsequent writers on the con- sumption function take income to be the determining variable and consumption to be the determined variable. The income of an individual or household or corporation is assumed to be outside its own control, at least in the short run. It is a datum, determined by market forces, constraining consumption and saving decisions. But there are many opportunities in an advanced economy for households to adjust their incomes, by working more or less, or by movement of wives and other secondary earners in or out of the labor force. A study by Rosett (1958), for example, shows the influence of the financial position of the household on the labor force participation of wives. It is possible that many households adjust their incomes to their consumption standards, rather than vice versa. An analogous opportunity is presumably available to some corporations; that is, if they are in need of profits to pay dividends or to make investments, they can take steps to increase them.
National versus household propensities
Keynes was concerned with the propensity to consume of a whole society, the relation of its consumption to its net national product. The difference between these magnitudes takes various forms: household saving, the retained earnings of enterprises, the receipts of governments. Accordingly, the national consumption function reflects a variety of social institutions and patterns of behavior—a mixture of family, business, and political habits and decisions. If it is based on any psychological law, the law must be founded in social psychology rather than in the psychology of individual consumers and savers.
Personal saving proper is defined as the difference between private consumption and personal disposable income—the income that individuals and households have wholly free disposition of. This differs from net national product by the sum of (a) taxes and other government receipts, net of government “transfer payments,” which are in the nature of negative taxes (pensions, benefits, subsidies, etc., for which no current productive services are rendered), and (b) the net earnings, after taxes, retained by corporations. The national mpc depends not only on the response of consumers to an increase in disposable income, but also on the response of personal disposable income to an increase in net national product. The latter response, in turn, reflects (a) the sensitivity of government tax receipts (net of transfer payments)—state and local as well as federal—to increases in national product and (b) the share of corporate profits after taxes in such increases in national product, together with the “propensity” of corporate directors to retain these earnings rather than to disburse them as dividends. In the analytical framework of Keynes's General Theory, a change in any of this complex of relationships would shift the consumption function, changing its level and perhaps its slope and shape. For example, tax rate reductions, increases in public welfare payments, increases in corporate dividends, reductions in the share of profits in national product would all raise the national propensity to consume. This could happen even though the thriftiness of households, as measured by their propensity to consume from disposable income, remained unchanged.
The practical quantitative importance of this observation can be illustrated from recent experience in the United States. In the short run at least, both the apc and mpc from disposable income appear to be between .92 and .96. But the apc with respect to net national product is about .70 and the mpc about .55. Thus the gap between disposable income and net national product accounts for the major part of the gap between consumption and net national product.
Of course, not all the difference between net national product and private consumption is saving in the sense of augmenting national wealth and future consumption possibilities. On the expenditure side, the counterpart of this gap includes government purchases of goods and services as well as net private investment. Some government expenditure is investment in the future, and some is collective consumption. Unfortunately, our present government accounting techniques do not permit a useful quantitative classification. But, as observed above, this failure is not crucial for the purposes for which Keynes designed the consumption function.
Subsequent theoretical and empirical investigation has not followed Keynes in postulating a single stable relationship linking consumption to national product. Instead “the consumption function” has come to have a narrower meaning: the relationship of consumption to disposable income and other variables. The linkages between disposable income and national product have been set to one side for independent investigation. Despite the fact that they are of greater quantitative importance than household saving proper, less theoretical and statistical effort has been expended on them (see, however, Ruggles & Ruggles 1949).
Although this division of labor is probably an advance over Keynes's global approach, the propensity to save from disposable income is surely not entirely independent of corporate saving or of certain tax payments. Retained earnings do not enter disposable income, but they are reflected in the value of corporate stocks and thus in the wealth of households and individuals. Accumulation of wealth by this route may substitute for personal saving; evidence that wealth affects the household propensity to save is noted below. Tax payments connected with social security programs are excluded from both disposable income and personal saving, even though private withholdings from wages and salaries for similar purposes are counted as both income and saving. One may suspect that socialization of saving for retirement, medical emergencies, and unemployment would lower the apparent propensity to save from personal disposable income. But it is hard to support this suspicion empirically.
What follows will concern the consumption function in the narrower sense, the division of disposable income between consumption and personal saving.
The Keynesian consumption function, typified in Figure 1, fits admirably two kinds of empirical data. First, economy-wide time series of consumption and income for the period between the two world wars lie along a function of the Keynesian type; there are even observations of negative private saving in the depths of the great depression. Second, any cross-section survey of household budgets appears to confirm the “psychological law” at a microeconomic level. When households are classified into income brackets, and the average consumption for a bracket is plotted against income, the scatter of points traces out a path (Engel curve) like the consumption function of Figure 1. Dissaving in low brackets gives way to positive saving at higher levels. Indeed some survey evidence suggests that the mpc falls in high brackets.
After World War II, however, several pieces of evidence combined to cast doubt on the Keynesian consumption function. The reappraisal that followed led to new theories and deeper empirical investigation of the determinants of household consumption.
First, extrapolations of statistical consumption functions based on prewar United States data to potential postwar income levels greatly underestimated the postwar propensity to consume. These extrapolations led some analysts to pessimistic views of postwar economic prospects in the United States, which were in the event quite unjustified. These extrapolations were based either on interwar time series or on Engel curves relating household consumption and income in prewar budget studies of 1935–1936.
The Keynesian propensity to consume is primarily a tool for the short run, for analysis of the determination of income and employment during a business cycle or a period of underemployment. The failure of the postwar forecasts did not impair its usefulness for this primary purpose. Moreover, Keynes certainly did not exclude shifts in the consumption function, and no doubt he himself would have understood the significance of the artificial shortages of consumers' goods and the abnormal accumulations of liquid savings that a great war leaves in its aftermath. At the same time, many of his obiter dicta suggest that he expected rich capitalist societies to face in the long run a chronic and increasing excess of potential saving at full employment. Experience since the war, persisting beyond its immediate legacy of backlogs and liquidity, compels a considerable modification of this expectation.
Second, just as the interwar Keynesian consumption function forecasts too much saving after World War II, it “backcasts” too little saving before World War I. Indeed it would not indicate any positive saving until about 1908, when per capita income reached the break-even point observed during the depression. One does not require data to know that U.S. economic growth in the nineteenth century and the early twentieth century was not generated from dissaving. But a statistical study by Kuznets (1946) provided the data, showing that the share of capital formation in U.S. output, averaged for overlapping decades, has been roughly constant since the Civil War. Moreover, Brady and Friedman (1947) took the trouble to look at earlier household budget studies, going as far back as 1901. They found that while any one survey indicates the same kind of Keynesian consumption-income relationship observed in the 1935–1936 study, the relationship shifts upward in successive surveys. A family with, say, an income of $3,000, corrected for price changes, will in general be observed to save much less in 1918 than in 1901, still less in 1935, and, as it has later turned out, less in 1950 than in 1935 or 1941. This finding, of course, undermines the simple use of the Engel curve of a cross-section survey for prediction or aggregation. The hypothesis that Brady and Friedman offered to explain the finding anticipates that of Duesenberry, which will be discussed below.
Recognition of these facts led a number of investigators to formulate and test hypotheses that would explain them—broadly speaking, hypotheses that would reconcile the short-run, or cyclical, success of the Keynesian consumption function with its long-run, or secular, failure. Indeed in the early 1940s Samuelson (1943) proposed a “ratchet” model: Consumption grows in the long run roughly in proportion to income; but during cyclical interruptions of long-run growth, consumers defend living standards already attained, and consequently consumption follows a flatter (lower rape) Keynesian path. In independent but similar contributions, Duesenberry (1948) and Modigliani (1949) formalized the ratchet idea and tested it statistically, making the ape depend inversely on the ratio between current income and previous peak income.
These formulations do the trick in a statistical sense. They do it by eliminating absolute real income from the determination of the long-run average propensity to consume, which becomes a constant. Why should this be so?
Duesenberry (1949) offered one explanation, his “relative income” hypothesis. Consumer utility depends, he reasoned, not on absolute amounts of consumption but on the relation of these amounts to the consumption of others with whom the consumer feels in social competition or under pressure to conform. This hypothesis has obvious support in many findings of modern sociology and psychology, and in some older ideas of Veblen. Duesenberry's hypothesis rationalizes the Brady-Friedman findings, in much the same way they themselves suggest. The pattern of any one cross section reflects the fact that the consumption leaders, in the upper ranks of the income distribution, can “afford” high propensities to save, while the followers in the lower ranks respond to social pressures by high propensities to consume. A general increase in absolute incomes, leaving the relative income distribution unchanged, will leave unchanged both these social pressures and the responses to them in terms of shares of income consumed. Most of the differences between successive budget surveys disappear when consumption ratios are plotted not against absolute incomes but against relative income positions. The broad conclusion is that while substantial changes over time in the inequality of incomes might alter the aggregate consumption ratio—contrary to the usual view, equalization would tend to increase the saving ratio—sheer growth in per capita income will not.
Permanent and lifetime income
In independently developed formulations, Friedman (1957) and Modigliani and Brumberg (1954) offered alternative explanations of the same phenomena. Students of household consumption behavior have long observed that incomes other than current incomes affect current consumption. Consumers suffering income declines resist departures from their previous consumption standards; and those enjoying income gains generally consume less than other households who long ago achieved the same income level. Similarly, the investigations of Katona and his colleagues (Katona & Mueller 1953; 1956) at the Survey Research Center at the University of Michigan indicate that optimistic expectations of future incomes encourage current consumption, while pessimistic expectations have the reverse effect.
According to Friedman's “permanent income hypothesis,” the consumption of a household is proportional to its permanent income, that is, the average income it expects to earn over its planning horizon. Friedman is not definite about either the factor of proportionality—which might vary with the household's stage in the life cycle, its wealth, the interest rate, and other variables—or about the length of the planning horizon. On these matters the lifetime income hypothesis of Modigliani is much more explicit, as will be seen below. In any case, Friedman employs his hypothesis to explain both the evidence of cross-section budget surveys and the ratchet effect observed in aggregate time series.
In budget surveys the low-income brackets are bound to be abnormally loaded with families temporarily below their permanent incomes, while the high brackets naturally have more than their share of families temporarily above their permanent status. This is the reason that the low brackets in any cross section exhibit a higher propensity to consume, relative to measured current incomes, than the high brackets—even though the propensity to consume from permanent incomes, unobserved, may be the same for all groups. When the households of a cross section are classified, not by current incomes but by other observed attributes— occupation, educational attainment, residence, age —better correlated with permanent income status, the scatter of points relating average consumption to average income gives a better approximation of the long-run consumption function. It starts from the origin—after all, groups with low permanent income status cannot do much dissaving—and shows a higher marginal propensity to consume than the usual Engel curve (see Watts 1958). Whether this mpc is a constant or diminishing function of income is, however, an open empirical question.
As for the ratchet effect, many households are below their permanent incomes in recessions and depressions; therefore their consumption is high relative to their current incomes. Like the “previous peak” hypotheses discussed above, this explanation is more satisfactory for recessions clearly believed to be temporary than for prolonged depression periods like the 1930s when the memory of the preceding peak grows faint.
The Modigliani-Brumberg model is in the same spirit, but by being bolder in its assumptions it is more specific in its conclusions. In its most stark formulation, the planning horizon of the individual consumer is his whole lifetime. And the factor of proportionality between consumption and permanent income is simply one. Individuals are assumed to plan no net lifetime saving; they transfer to their heirs no more and no less than they inherited. Subject to this constraint, they try to spread their lifetime consumable resources evenly over their lives. In particular, they seek to accumulate enough savings during their earning years to maintain the same consumption standard during their years of retirement. The division of life between work and retirement is taken, somewhat implausibly, as an institutional fact, a constant independent of national and individual incomes. The model has several interesting implications:
(a) In a society with stationary population and income, aggregate net personal saving would be zero. The dissaving of the retired would exactly offset the saving of the workers, whose only purpose in saving is to provide for their own future retirement.
(b) In a society with growing population or growing per capita income or both, aggregate net personal saving will be positive. Indeed, the higher these rates of growth, the higher will be the ratio of saving to aggregate income. For in a growing economy, the retired of the future always exceed— in number or in lifetime income, or in both—the retired of the present. Consequently the saving necessary to provide for future retirements always exceeds the dissaving of the currently retired. (See Modigliani & Ando 1963 for precise calculations.)
(c) Changes in this year's income affect this year's consumption only to the extent that current consumption benefits, equally with consumption in all future years, from a general recalculation of lifetime consumable resources. This effect will not be large unless current income changes are regarded as permanent. This is the counterpart of Friedman's more extreme contention that all temporary income gains will be saved and none consumed, and symmetrically that temporary income losses will be wholly offset by dissaving.
This implication of the two models has attracted widespread attention and controversy. One reason is that it casts doubt on the efficacy of temporary changes in tax rates or income transfers as measures of economic stabilization. Some caution is required in drawing this inference from the permanent income or lifetime income hypotheses. Strictly speaking, these theories concern consumption rather than expenditures on consumers' goods, for example, the use of an automobile rather than its purchase. It may be that temporary windfalls are used to purchase durables but do not appreciably increase their use; the purchaser uses the newer product with the same intensity. But from the standpoint of economic stabilization, the stimulus of such expenditure is equally good whether it is properly classified as saving or as consumption. The pure theories assume, moreover, that there are no obstacles in financial markets to borrowing against expected future incomes. In actual fact many households are at the limits of their lines of credit and are therefore prevented from adjusting current consumption fully to their permanent incomes. Temporary windfalls may increase their consumption expenditures simply by adding to their liquid resources.
Empirical measurement of the mpc from temporary income has not been conclusive. But Bodkin (1959) observed that the households of veterans who received unexpected national service life insurance dividends during the 1950 budget survey spent significantly more than households similar to them in income and in other respects. Using different methods and data, Watts (1958) estimated the mpc from temporary income to be significantly positive, although perhaps only half of the mpc from permanent income. Watts's results also indicated that household reaction to temporary income changes may not be symmetrical for positive and negative changes.
(d) An individual whose life proceeds according to plan will gradually build up his wealth, so that at any age his wealth plus his remaining expected earnings during working years just suffice to maintain his consumption throughout his lifetime, including his years of retirement. To every age there corresponds a normal ratio of wealth to income. But unexpected income changes, consumption emergencies, or capital gains and losses on past savings may cause an individual's wealth to deviate from its normal relation to income. A high wealth-income ratio permits the individual to consume more, both now and in retirement, while a low wealth–income ratio requires him to consume a lower fraction of his current income in order to provide for undiminished consumption in retirement. Extended to the whole population, this argument suggests that the aggregate saving ratio depends inversely on the wealth-income ratio. Along with the effects of growth rates mentioned in b, this is the major aggregate implication of the lifetime income hypothesis.
Wealth and the propensity to consume
The importance of wealth in the consumption function had been urged by several earlier writers. They based their view on more general considerations than those involved in the Modigliani-Brumberg model. In particular, saving may be motivated by the desire to make bequests or other transfers of wealth to the next generation, as well as by the need to even out consumption over the savers' lifetimes. The adequacy of wealth to meet a bequest target—which may itself be a function of lifetime income, of course—will then be one of the determinants of current saving and consumption.
Pigou (1947) emphasized the wealth effect in the course of an abstract theoretical attack on the whole structure of Keynes's General Theory. Keynes had contended that in certain circumstances no amount of wage and price deflation could restore aggregate real demand to full employment levels. Pigou pointed out that the real value of private wealth can be indefinitely increased by price deflation. The reason is that deflation would increase the purchasing power of gold, government-issued currency, and interest-bearing government debt, all of which are fixed in money value. As the owners of these assets become saturated with real wealth, their propensity to consume is bound to increase. Via what has come to be known as the “Pigou effect,” the absolute price level, because of its effect on the real value of private wealth, has a bearing on the propensity to consume.
Quite apart from the theoretical context of Pigou's argument, a general wealth effect is an alternative explanation of some of the same phenomena explained by the other hypotheses under review. Over the long run disposable income and private wealth tend to grow in step, with wealth five to six times income. Recent empirical calculations (Modigliani & Ando 1963) suggest a marginal propensity to consume from income of .5 to .7 and from wealth of .05 to .07. When income and wealth grow in step, estimates in these ranges explain how consumption can consistently take about 90 per cent of increases in disposable income. But the apparent mpc from income can be quite different when cyclical fluctuations break the normal long-run linkage of wealth and income. Assuming that the real value of wealth fluctuates less than real income, consumption will appear less sensitive to income in the short run than in the long run.
At the level of individual households in budget surveys, the dissaving recorded for the lower brackets is of course supported by past savings. As pointed out by Tobin (1951), differences in the wealth available to households at a given real income can help to explain the differences observed in different surveys in their propensities to save or to dissave. This explanation is not inconsistent with those by which Duesenberry, Friedman, and Modigliani explain the same phenomenon.
Sometimes the wealth effect on saving and consumption has been attributed to a particular kind of wealth, liquid asset holdings, rather than to total net worth. In principle net worth is the relevant variable. This is the constraint that the past and the market impose on consumers, a constraint that they can change only slowly and to which they must meanwhile adjust. In contrast, many households can, within wide limits, decide for themselves at each moment of time how much of their wealth they wish to hold in liquid form; and they can implement such decisions very quickly by purchases or sales of other assets and debts. But for many households below the top brackets, liquid assets (bank deposits, savings accounts, savings bonds) are virtually the only assets held; their total serves as a good proxy for total wealth. In other cases, the remaining constituents of net worth are not easily convertible into current purchasing power—home real estate, consumers' durable goods, pension rights. Consumption functions fitted to survey data indicate a positive effect of liquid asset holdings on consumption, especially for households suffering income reverses (see Michigan, University of … 1954).
Income and wealth distribution
Aggregate consumption can be affected not only by changes in aggregate income and wealth but also by changes in the distribution of a given aggregate income or wealth. However, these distributional effects are commonly exaggerated by observers who are struck by the wide discrepancies in average propensities to consume among different economic and social groups. Redistribution will affect total consumption only to the extent that the marginal propensities of the affected groups differ. And these appear to differ much less than average propensities. Nevertheless, the prevailing evidence is that certain redistributions of income and wealth would increase saving; for example, from low or middle to top income brackets, from employees to self-employed, from urban residents to farm operators.
It is necessary to beware of another frequently heard argument, namely that reduction in taxes levied on individuals with high saving propensities, in particular those in the upper brackets of income and wealth, will by itself increase saving. This increase in disposable income will doubtless increase personal saving and probably by more than would equal tax reductions benefiting others. But it will increase consumption too. By raising disposable income relative to net national product, it raises the propensity to consume with respect to national product and leaves less room in the economy for nonconsumption expenditures.
The hypothesis that “capitalists” save while “workers” spend has a long history, and it has recently been revived in connection with certain analyses of economic growth and development. Kaldor (1955–1956) has gone so far as to build a theory of the functional distribution of income on the differences between capitalists and workers in their marginal propensities to save. Houthakker, who has conducted some pioneering studies of intercountry differences in saving propensities, finds (1957) that the saving ratio appears to be higher in countries where the property share of income is higher. This statistical result can, however, be alternatively rationalized.
This hypothesis may be more relevant to underdeveloped economies with a well-defined division of the population between high-income property owners and low-income workers than to advanced economies like the United States, where ownership of property is widely diffused and some rentiers are poor while some “workers” are rich. But if “capitalists” are identified with corporations and their saving with retained corporate earnings, the proposition has some applicability to the United States.
The importance of the distribution of income has also been stressed in the analysis of “forced saving” during inflation. If wages lag behind prices and the marginal propensity to consume from profits is less than from wages, the process of inflation suppresses some demand for consumption goods, which would occur if prices were stable. The same effect can occur without a change in income distribution if consumers are slow to adjust their money expenditures to rising prices and money incomes.
The rate of interest
Prior to Keynes the economic variable that economists usually stressed in analyzing the choice between consumption and saving was the interest rate. The classic exposition of the theoretical relation of saving decisions to the rate of interest is Fisher's (1930). Conard (1959) provides a good review of doctrine on this subject. An increase in the rate of return on saving was expected to tip the balance of choice in favor of the future. But it was recognized that for many positive savers, an increase in interest rates means an over-all increase in consumable resources, from which current as well as future consumption might benefit.
Whether because of this real ambiguity or because of the narrow range of variation of observed interest rates, econometricians have not been able to detect significant interest rate effects on saving. Perhaps these effects would be more evident if it were possible to relate net saving not merely to rates paid on savings accounts and bonds, but to the effective rates at which consumers can borrow and the rates of return on business capital, corporate equities, houses, other real estate, and consumer durables. These assets, after all, absorb the bulk of national saving; but their returns are difficult to measure. The positive correlation internationally between saving and the property share of income cited above could be interpreted to mean that saving is higher in those countries where its yield is greater.
Observed differences among households in consumption and saving behavior are, of course, attributable to a long list of differences in their circumstances, habits, and preferences. Some of these are, like income and wealth, variables whose influence is the major interest of economists. Others are, like demographic characteristics, variables that can differ widely among households even though their distribution over the population changes only very slowly. It is nonetheless important to measure their effects, if only to disentangle them from the measurement of the influence of variables more important in economic fluctuations and economic policy. Considerable statistical effort has been devoted to the “life cycle” variables—age, marital status, family size and composition—and other demographic characteristics, such as educational attainment, occupation, race, and geographical location (see, for example, Michigan, University of … 1954; Lydall 1955).
A set of variables of a different nature are the “psychological” ones—attitudes, intentions, expectations, personality attributes—which Katona and his associates seek to measure. Unlike demographic variables, the distribution of some of these psychological variables in the population may change radically in the short run, in ways that can be ascertained in our present state of knowledge only by new surveys. If household surveys are to contribute further to our understanding of the propensity to consume and make possible more powerful tests of competing theories, they will have to take a longer perspective. To measure the effects of past and expected levels of income and wealth and of retirement and bequest objectives, it is necessary to observe not only the current status of households but their lifetime histories, plans, and aspirations.
[See alsoIncome and employment theory.]
Ackley, Gardner 1951 The Wealth-Saving Relationship. Journal of Political Economy 59:154–161.
Arena, John J. 1963 The Wealth Effect and Consumption: A Statistical Inquiry. Yale Economic Essays 3:251–303.
Bodkin, Ronald 1959 Windfall Income and Consumption. American Economic Review 49:602—614.
Brady, Dorothy S.; and Friedman, Rose D. 1947 Savings and the Income Distribution. Volume 10, pages 247–265 in Conference on Research in Income and Wealth, Studies in Income and Wealth. New York: National Bureau of Economic Research.
Burns, Arthur F. 1954 The Frontiers of Economic Knowledge: Essays. Princeton Univ. Press.
Conard, Joseph W. 1959 An Introduction to the Theory of Interest. Berkeley: Univ. of California Press.
Cornfield, Jerome; Evans, W. Duane; and Hoffenberg, Marvin 1947 Full Employment Patterns, 1950. Monthly Labor Review 64:163–190, 420–432.
Duesenberry, James S. 1949 Income, Saving and the Theory of Consumer Behavior. Harvard Economic Studies, Vol. 87. Cambridge, Mass.: Harvard Univ. Press.
Farrell, Michael J. 1959 The New Theories of the Consumption Function. Economic Journal 69:678–696.
Fisher, Irving (1930) 1961 The Theory of Interest. New York: Kelley. → Revision of the author's The Rate of Interest, 1907.
Friedman, Milton 1957 A Theory of the Consumption Function. National Bureau of Economic Research, General Series, No. 63. Princeton Univ. Press.
Friedman, Milton; and Meiselman, David 1963 The Relative Stability of Monetary Velocity and the Investment Multiplier in the United States, 1897–1958. Pages 165–268 in Stabilization Policies, by E. Cary Brown et al. Englewood Cliffs, N.J.: Prentice-Hall.
Houthakker, Hendrik S. 1957 An International Comparison of Household Expenditure Patterns, Commemorating the Centenary of Engel's Law. Econometrica 25:532–551.
Kahn, Richard F. 1931 The Relation of Home Investment to Unemployment. Economic Journal 41:173–198.
Kaldor, Nicholas 1955–1956 Alternative Theories of Distribution. Review of Economic Studies 23:83–100.
Katona, George 1960 The Powerful Consumer: Psychological Studies of the American Economy. New York: McGraw-Hill.
Katona, George; and MUELLER, EVA 1953 Consumer Attitudes and Demand, 1950–1952. A Survey Research Center Publication. Ann Arbor: Univ. of Michigan.
Katona, George; and Mueller, Eva 1956 Consumer Expectations 1953–1956. Survey Research Center Publication No. 16. Ann Arbor: Univ. of Michigan Press.
Keynes, John Maynard 1936 The General Theory of Employment, Interest and Money. London: Macmillan. → A paperback edition was published in 1965 by Harcourt.
Keynes, John Maynard 1940 How to Pay for the War: A Radical Plan for the Chancellor of the Exchequer. New York: Harcourt; London: Macmillan.
Koopmans, Tjalling 1942 The Dynamics of Inflation. Review of Economics and Statistics 24:53–65.
Kuznets, Simon 1946 National Product Since 1869. New York: National Bureau of Economic Research.
Kuznets, Simon 1953 Shares of Upper Income Groups in Income and Savings. Publication No. 55. New York: National Bureau of Economic Research.
Lintner, John 1953 The Determinants of Corporate Savings. Pages 230–258 in Walter W. Heller, Francis M. Boddy, and Carl L. Nelson (editors), Savings in the Modern Economy. Minneapolis: Univ. of Minnesota Press. → Includes four pages of comments.
Livingston, S. Morris 1945 Forecasting Postwar Demand. Econometrica 13:15–24.
Lubell, Harold 1947 Effects of Redistribution of Income on Consumers' Expenditures. American Economic Review 37:157–170, 930.
Lydall, Harold 1955 The Life Cycle in Income, Saving, and Asset Ownership. Econometrica 23:131–150.
Mack, Ruth P. 1948 The Direction of Change in Income and the Consumption Function. Review of Economics and Statistics 30:239–258.
Michigan, University of, Survey Research Center 1954 Contributions of Survey Methods to Economics. Edited by Lawrence R. Klein. New York: Columbia Univ. Press.
Modigliani, Franco 1949 Fluctuations in the Saving-Income Ratio: A Problem in Economic Forecasting. Volume 11, pages 371–443 in Conference on Research in Income and Wealth, Studies in Income and Wealth. New York: National Bureau of Economic Research.
Modigliani, Franco; and Ando, Albert 1963 The Life Cycle Hypothesis of Saving: Aggregate Implications and Tests. American Economic Review 53:55–84.
Modigliani, Franco; and Brumberg, Richard 1954 Utility Analysis and the Consumption Function: An Interpretation of Cross Section Data. Pages 388–436 in Kenneth K. Kurihara (editor), Post Keynesian Economics. New Brunswick, N.J.: Rutgers Univ. Press.
Pigou, A. C. 1947 Economic Progress in a Stable Environment. Economica New Series 14:180–188.
Rosett, Richard N. 1958 Working Wives: An Economic Study. Pages 51–99 in Thomas F. Dernburg et al., Studies in Household Economic Behavior. Yale Studies in Economics, Vol. 9. New Haven: Yale Univ. Press.
Ruggles, Richard; and Ruggles, Nancy (1949 ) 1956 National Income Accounts and Income Analysis. 2d ed. New York: McGraw-Hill. → First published as Introduction to National Income and Income Analysis.
Samuelson, Paul A. 1943 Full Employment After the War. Pages 27–53 in Seymour E. Harris (editor), Postwar Economic Problems. New York and London: McGraw-Hill.
Smithies, Arthur 1943 The Behavior of Money National Income Under Inflationary Conditions. Quarterly Journal of Economics 57:113–128.
Smithies, Arthur 1945 Forecasting Postwar Demand. Econometrica 13:1–14.
Tobin, James 1951 Relative Income, Absolute Income, and Saving. Pages 135–156 in Money, Trade, and Economic Growth: Essays in Honor of John Williams. New York: Macmillan.
Watts, Harold W. 1958 Long-run Income Expectations and Consumer Saving. Pages 101–144 in Thomas F. Dernburg et al., Studies in Household Economic Behavior. Yale Studies in Economics, Vol. 9. New Haven: Yale Univ. Press.
The classical economists were concerned with the economic categories of consumption, production, and exchange. One description of the classical Say’s Law is that it states that production and consumption are identical. From the perspective either of underconsumption or oversaving, one category is perceived as a limit of the other. In the hands of John Maynard Keynes, however, consumption became a function, relating aggregate consumption, C, mainly to aggregate disposable income, Y, defined as income less taxes and transfer payments. This equation has been called the absolute income hypothesis (AIH). Like a good gardener, Keynes weeded out many variables that could influence consumption, settling on disposable income as the most important one. Keynes wrote the implicit form of this relation as C w = X (Y w ). In another context, he held that C = φ 1(W, Y ), where the additional variable W represents the state of the news, a term that changes with long-term expectation.
The relation of consumption to income follows a psychological law stating that consumption increases with income but not by the same proportion. This law makes the consumption function a behavioral relationship that can be juxtaposed with data as opposed to a structural or identity equation. A testable linear form of the AIH is obtained by expanding the implicit relationship between consumption and income by the Taylor series, ignoring the nonlinear terms. The intercept, a, of the line captures autonomous consumption, that is, any consumption that is not induced by income. The slope of the line, 0 < dCl dY = b < 1, is the marginal propensity to consume (MPC)—on average, the amount of an additional dollar that is consumed in a community. The Keynesian model uses the MPC to estimate a multiplier that predicts how a change in investment will boost GDP. By examining data for the United States, Keynes estimated the MPC to be between 60 and 70 percent.
The Keynesian consumption function sparked a new research program. For example, significant works by Milton Friedman on the permanent income hypothesis (PIH) and Franco Modigliani on the life cycle hypothesis (LCH) were in large part the catalysts for their Nobel Prizes in 1976 and 1985, respectively. The PIH-LCH models reconciled anomalies in the prediction of the AIH during the post–World War II period. The major anomaly was that the average propensity to consume was over 90 percent, whereas short-run MPC was between 60 and 70 percent. Modigliani and James Duesenberry reconciled the differences by postulating a previous peak income in the consumption function. Whereas Modigliani’s work evolved into his LCH hypothesis, Duesenberry’s relative income hypothesis (RIH) remained an example of external effects on consumption in line with Veblenesque norms; defined in terms such as emulation, convention, and molding, it holds that a person’s consumption depends on the level and types of other people’s consumption. RIH also holds that consumers would want to maintain a consumption pattern established by the highest income they had previously received. During a downswing phase of a business cycle, consumers experience a reduction in income but are hesitant to adjust their spending behavior away from what they had established during a previous peak period. Consumers prefer to draw down their savings or borrow in order to maintain their previous peak consumption. Only when they recover their previous peak level income will their consumption behavior change. The change is an unusual one, a sort of quantum leap or “ratchet” upward, perhaps due to pent-up demand during the downswing. This upward effect amounts to a reconciliation of the short- and long-run MPCs.
Robert Frank expanded and articulated the RIH paradigm in relation to the question of how current relative consumption will dominate future relative consumption. Parents who prefer to buy a house in a good school district now may be negatively affecting their future consumption after retirement. Conversely, spending now on a suit for a job interview may have a positive impact on future income. The experience of low relative consumption now may also set expectations for a low relative consumption in the future. Juliet Schor’s “new consumerism” is also based on lifestyles and norms, and posits that consumers elevate their consumption to unsustainable levels that lead to mounting debts and bankruptcies, as well as longer working hours.
The LCH-PIH hypotheses advance consumption theory by introducing wealth or assets as well as income into the consumption function. In this scenario, consumers draw on their lifetime income and assets to smooth their consumption expenditures over their life cycle. We can speak of permanent income —changes in which have more significant effects on consumption than temporary or transitory changes in income. The two hypotheses have one major difference: Friedman made the income stream infinite, whereas Modigliani made the income stream finite. For instance, Modigliani estimated the consumption function as: C = .766Y + .073A. Here the short run MPC is 0.77, and if assets, A, are approximately five times income, while labor income is approximately 80 percent of income, then a long-run MPC of 0.98 = 0.8(.766Y) + 5(.073Y) is reached. Essentially, the presence of wealth in the AIH causes it to drift upward.
With the introduction of these hypotheses, the stage was set for a paradigmatic shift in the consumption function. Friedman estimated the PIH through a distributed lag model, which John Muth showed to be optimal under rational expectation assumptions. Because Friedman left the definition of income vague, Muth proceeded to measure permanent income by an exponentially moving average equal to the conditional expected value under rational expectations. Robert Lucas expanded the rational expectation concept by shifting the meaning of the consumption function from one relating consumption and income, to one relating permanent income and observed income. Robert Hall rescued the consumption function from that line of attack by postulating that only surprising events could be responsible for unexpected results. The model he specified maximizes the expected value of lifetime utility subject to an unchanging real interest rate. He presumed that consumers would make the ratios of marginal utilities for present and future consumption equal to the ratio of their prices. Hall tested his consumption function in the form C t + 1 = λC t + error t , a random walk model. Clive Granger referred to the consumption theory as “manna from heaven to macro-econometricians. It was easily stated and understood, and appears to be easy to test” (Granger 1999, pp. 42–43). In practical parlance, consumers will tend to adjust their individual consumption so that it will not differ from an expected level. This fact reinforces the underlying principle that consumers tend to smooth out spending over time, and that this practice relates to some uncertainty about income. Hall’s model rendered the lagged income effect insignificant on consumption. If consumers have a quadratic utility function, then they will want to consume at the level where their future income will equal its mean value.
Marjorie Flavin, a student of Hall’s, made two findings that furthered the development of the consumption function. One finding is that future consumption is sensitive to the previous level of consumption and can show a strong variation. Another finding is that the surprise element does not cause much variation in future consumption. John Campbell and Gregory Mankiw had the idea of combining these two findings in a convex way. This means that a proportion of the variation will be captured. Following Hall’s model, the surprise element varies by a certain proportion and thus income will explain the less than proportional expected consumption. The combination of Hall’s model with the LCH consumption equation resulted in a simple test of a change in consumption based on a change in disposable income. Since Hall’s work, research on the consumption function has been escalating.
SEE ALSO Absolute Income Hypothesis; Adaptive Expectations; Class, Leisure; Class, Rentier; Conspicuous Consumption; Consumerism; Consumption; Consumption Tax; Economics, Keynesian; Expectations, Rational; Life-Cycle Hypothesis; Macroeconomics; Permanent Income Hypothesis; Propensity to Consume, Marginal; Propensity to Save, Marginal; Relative Income Hypothesis; Underconsumption
Campbell, John Y., and N. Gregory Mankiw. 1989. Consumption, Income, and Interest: Reinterpreting the Time Series Evidence. NBER Macroeconomics Annual 4: 185–216.
Flavin, Marjorie A. 1981. The Adjustment of Consumption to Changing Expectation about Future Income. Journal of Political Economy 89 (5): 974–1009.
Frank, Robert H. 1999. Luxury Fever: Why Money Fails to Satisfy in an Era of Excess. New York: Free Press.
Frank, Robert H. 2005. Americans Save So Little, but What Can Be Done to Change That? New York Times, March 17.
Frank, Robert H. 2005. The Mysterious Disappearance of James Duesenberry. New York Times, June 9.
Friedman, Milton. 1957. A Theory of the Consumption Function. Trenton, NJ: Princeton University Press.
Granger, Clive W. J. 1999. Empirical Modeling in Economics: Specification and Evaluation. London: Cambridge University Press.
Hall, Robert E. 1978. Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence. Journal of Political Economy 86 (6): 971–987.
Hall, Robert E. 1989. Consumption. In Modern Business Cycle Theory, ed. Robert J. Barro, 153–177. Cambridge, MA: Harvard University Press.
Keynes, John Maynard. 1936. The General Theory of Employment, Interest, and Money. London: Macmillan. Reprint, New York: St. Martin’s Press, 1970.
Lucas, Robert E., Jr. 1976. Econometric Policy Evaluation: A Critique. In Carnegie-Rochester Conference Series on Public Policy, vol. 1, eds. Karl Brunner and Allan H. Meltzer, 19–46. Amsterdam: North-Holland.
Modigliani, Franco.  1980. Fluctuations in the Saving-Income Ratio: A Problem in Economic Forecasting. In The Life Cycle Hypothesis of Saving, vol. 2 of The Collected Papers of Franco Modigliani, eds. Andrew Abel and Simon Johnson, 4–40. Cambridge, MA: MIT Press.
Modigliani, Franco, and Richard Brumberg.  1980. Utility Analysis and the Consumption Function: An Interpretation of Cross-Section Data. In The Life Cycle Hypothesis of Saving, vol. 2 of The Collected Papers of Franco Modigliani, eds. Andrew Abel and Simon Johnson, 79–127. Cambridge, MA: MIT Press.
Muth, John F. 1960. Optimal Properties of Exponentially Weighted Forecasts. Journal of the American Statistical Association 55 (290): 299–306.
Romer, David. 1996. Advanced Macroeconomics. New York: McGraw-Hill.
Schor, Juliet B. 1998. The Overspent American: Upscaling, Downshifting, and the New Consumer. New York: Basic Books.
The notion of aggregate demand formally made its appearance in John Maynard Keynes’s (1883-1946) General Theory in 1936 and, in its numerous guises, quickly rose to become a vital concept in economists’ tool kit of analytical devices. Despite pleas by some economists, notably new classical economists, to reject the aggregate demand/supply framework because of lack of rigorous microeconomic foundations (see, among others, Barro 1994), the aggregate demand function has retained a central but highly debated role in macroeco-nomic analysis.
Though he regarded it as his major analytical innovation (King 1994, p. 5), Keynes defined his aggregate demand function in a way that would be unfamiliar to most economists nowadays. This is because the aggregate demand function was conceived as a subjective aggregate relation linking entrepreneurs’ offers of employment to the anticipated overall market demand (or expected proceeds) for their firms’ output. Keynes wrote: “Let D be the proceeds which entrepreneurs expect to receive from the employment of N men, the relationship between D and N being written D = f(N ), which can be called the aggregate demand function ” (Keynes 1936, p. 25). Given entrepreneurial perceptions of firms’ investment plans, and expected flow of household consumption arising from hypothetical employment offers, an aggregate functional relation could be delineated in a two-dimensional D-N space: “The aggregate demand function relates various hypothetical quantities of employment to the proceeds their outputs are expected to yield” (Keynes 1936, p. 55).
There is a positive relationship between aggregate income and employment because increased employment offers will bring forth higher expected proceeds from household consumption. Indeed, the greater the share of spending out of each additional dollar of income—that is, the higher the marginal propensity to consume—the higher the level of additional income associated with increased employment (Asimakopulos 1991, p. 45).
When depicted in D-N space with an aggregate supply function (the latter resting on a standard Marshallian microfoundation and representing the desired proceeds that would just make it worth the while of entrepreneurs to employ N workers), short-period equilibrium is achieved at the intersection of the aggregate demand and supply curves, dubbed the point of effective demand. On this basis, Keynes rejected classical-type theories founded on the Say’s Law principle (that “supply creates its own demand”) by arguing that the latter doctrine did not assume an independent aggregate demand function that could conceivably result in an equilibrium point at less than full employment.
While the development of his aggregate demand concept was of major theoretical and policy significance, particularly in its support of activist taxation, spending, and monetary policies of aggregate demand management, there were obvious problems with Keynes’s original formulation. For instance, unless the business sector is conceived as one large firm that can envision the impact of its employment decision on its own expected proceeds, how exactly could a multitude of uncoordinated decisions by competitive firms be collectively anticipated by entrepreneurs and represented in an aggregate demand relation? As a result of such theoretical conundrums, the concept was to undergo tremendous transformations during the post–World War II (1939-1945) period as economists sought conceptually less challengeable theoretical constructs.
Even among fundamentalist Keynesians of the early postwar years, such as Sidney Weintraub (1914-1983) and Paul Davidson, the aggregate demand function, D, came to be treated no longer as an expected proceeds curve as perceived by entrepreneurs, but simply as a representation of the intended spending on the part of economic agents (consumers, firms, and governments) associated with hypothetical levels of total employment. Indeed, in the hands of numerous early postwar Keynesians such as Paul Samuelson, Keynes’s original association between sales proceeds and employment was to be transformed into a relation between aggregate intended expenditures of economic agents and the level of real income or output, as depicted in the framework of the popular 45-degree diagrams found in many introductory textbooks (Dutt 2002, p. 329).
Because of its implicit assumption of fixed price, the 45-degree aggregate expenditure relation slowly succumbed to alternative formulations of the aggregate demand function as economists struggled to incorporate the effect of changes in prices within a competing analytical framework. This resulted in redefining a downward-sloping aggregate demand function within aggregate price-output space seemingly comparable to its traditional Marshallian microeconomic counterpart. However, to ensure a negative slope, this latter incarnation of the aggregate demand function had to rely on somewhat more questionable assumptions than its previous upward-sloping Keynesian aggregate expenditure relation in the context of 45-degree diagrams. This is because, as prices rise, it is assumed that the purchasing power of household wealth and cash balances declines and thereby household spending (aggregate demand) also declines. These so-called wealth effects and real balance effects assume that currency held by households plus reserves held by banks exceed the value of bank deposits. In fact, however, bank deposits greatly exceed the value of bank reserves plus currency held by households. Hence, the relevance of real balance effects has been seriously questioned. This is why modern macroeconomic textbooks have slowly been abandoning this form of aggregate demand analysis (in price-output space) and relying simply on a dynamic relation that links inflation to an economy-wide capacity utilization rate—a variant of the Phillips Curve. Unfortunately, the latter is a far cry from Keynes’s unique formulation of the aggregate demand function that related aggregate expected proceeds to the level of employment.
SEE ALSO Aggregate Supply; Economics, Keynesian; Economics, New Classical; Keynes, John Maynard; Lucas, Robert E.; Macroeconomics; Phillips Curve; Propensity to Consume, Marginal; Survey of Income and Program Participation
Asimakopulos, A. 1991. Keynes’s General Theory and Accumulation. Cambridge, U.K.: Cambridge University Press.
Barro, Robert J. 1994. The Aggregate-Supply/Aggregate-Demand Model. Eastern Economic Journal 20 (1): 1-6.
Dutt, Amitava Krishna. 2002. Aggregate Demand-Aggregate Supply Analysis: A History. History of Political Economy 34 (2): 321-363.
Keynes, John Maynard. 1936. The General Theory of Employment, Interest, and Money. London: Macmillan.
King, John E. 1994. Aggregate Supply and Demand Analysis since Keynes: A Partial History. Journal of Post Keynesian Economics 17 (1): 3-31.
What It Means
Aggregate demand is a term used in macroeconomics (the study of the economy as a whole, as opposed to microeconomics, which concerns the individual parts of an economy). It refers to the demand of all buyers in an economy for what are called “final products” (in other words, the willingness and ability of these buyers to purchase the products). Final products are not all the products in the economy. Rather, they are those products, such as bicycles or furniture cleaning, that are sold to their final users (consumers, businesses, and governments). A product, such as plastic, that is purchased by a business to produce another good, such as a comb, is not a final product, but the comb purchased by a consumer is.
The demand for final products is also influenced by price. Because aggregate demand refers to all buyers and all final products in an economy, what influences aggregate demand is not the price of a particular good or service but the average price of all final products in the economy. As buyers have only so much money, the higher the average price for final products, the lower will be the demand (the willingness and ability to purchase the products). The opposite is also true. The lower the average price, the higher will be the demand, as buyers can afford to purchase more.
The definition of aggregate demand, then, is the amount of final products demanded by all buyers in an economy at a given average price. Its complimentary concept, aggregate supply, is the amount of final products the economy’s sellers are willing to produce at a given average price. These macroeconomic concepts of aggregate supply and aggregate demand are based on the microeconomic concepts of supply and demand. Supply is the quantity of a good or service that a business is willing to sell at a given price, and demand is the amount of a particular good or service that consumers are willing to buy at a given price.
Aggregate demand is commonly represented using a diagram called an aggregate demand curve (AD curve). The AD curve shows how buyers respond to varying price levels. With price level on the vertical axis and output, or products demanded, on the horizontal axis, the quantity of goods buyers demand forms a downward-sloping curve, moving from right to left, indicating that as the price level falls, buyers want increasing quantities of products. The AD curve, together with the aggregate supply, or AS, curve, is a useful tool for understanding and predicting the behavior of the economy under a variety of different conditions.
When Did It Begin
The British economist John Maynard Keynes first developed the concept of aggregate demand in his landmark book The General Theory of Employment, Interest, and Money (1936). Whereas economists previously had analyzed the workings of individual buyers and sellers and the places or systems (called markets) that bring them together, Keynes showed the importance of theorizing about the economy from the top down. By observing and theorizing about the behavior of the total, or aggregate, activity of all those buyers, sellers, and markets, Keynes demonstrated, governments could better manage their national economies. As a result of Keynes’s work, economics has since been split into two basic subfields: microeconomics and macroeconomics.
Specifically, Keynes outlined the various components that together add up to aggregate demand: consumer demand for household goods and services, business demand for productive equipment and other investments, the demand of government (at the national, state, and local levels) for goods and services, and foreign demand for domestic products minus domestic demand for foreign products (exports minus imports).
More Detailed Information
The components of aggregate demand are the same today as when Keynes first outlined the concept. Taken together, these components represent all of the possible purchases of final products (products sold to their final buyer, rather than products sold in raw or intermediate forms, to be made into other products) in any economy.
Consumer demand for household goods and services makes up one key portion of aggregate demand. This category takes into account all of the purchases that people make for personal, as opposed to business, reasons. Examples include purchases of food, clothing, haircuts, electronics, cars, and movie tickets.
Business spending is another primary component of aggregate demand. Businesses commonly make investments in their equipment and other resources in the hope of increasing profitability. Any finished products companies buy (such as pencils, hardhats, computers, and trucks) therefore contribute to the overall level of aggregate demand. Raw materials and labor costs do not count toward aggregate demand, since they are not considered finished products.
Government spending on finished products is the third main component of aggregate demand. Governments at the local, state, and national level buy a wide variety of goods and services. Examples include buildings to house schools, machinery for building roads, computers for government workers, and missiles for the armed forces. Not all types of government spending contribute to the overall level of aggregate demand. Direct payments of money to the elderly, in the form of Social Security, or to the poor, in the form of welfare, are not directly included in aggregate demand since they do not represent additional output of products. Likewise, the salaries of government workers are not counted toward aggregate demand.
Finally, domestic products exported and sold to foreign individuals, businesses, and governments contribute to the overall level of aggregate demand, at least as long as their quantities outweigh the quantities of goods produced by foreign companies and sold domestically. In other words, the total quantity of imports minus the total quantity of exports is added to the other three components of aggregate demand. Sometimes a country spends more money on imports than exports; in this case the difference would be subtracted from the overall aggregate demand total.
The AD curve indicates the level of aggregate demand at different price levels. At low price levels aggregate demand is high, and vice versa. The resulting curve moves downward from left to right, with price level on the vertical axis and output (goods or services demanded) on the horizontal. There are three explanations for this.
First, there is what is known as the wealth effect. When prices fall consumers’ money is suddenly worth more than before. This additional purchasing power (additional wealth) leads to increased demand for economic output.
The second reason for the shape of the AD curve is the interest rate effect, which explains what happens to interest rates at different prices. Interest rates are fees charged to those who borrow money from banks or other lenders. When prices for products are low, consumers and businesses need less money to make transactions, and they put additional money into bank accounts. In response, banks and other lenders lower their interest rates, and lending activity picks up, allowing individuals and businesses to make large purchases. Thus, demand is stimulated.
Finally, the net exports effect—which deals with the relationship between imports, exports, and prices—helps explain why the AD curve slopes downward. When domestic prices are low, domestic-made goods tend to be a better value than imported goods. Thus, domestic consumers and businesses buy domestic products. Likewise, low prices at home result in low prices for the foreign buyers of exported goods, so export sales increase. Both of these developments result in high levels of aggregate demand at low price levels, which takes the form, when diagrammed, of a downward sloping AD curve.
The AD curve can shift from left to right at any given price level (that is, the whole downward-sloping line is moved over). This represents a change in aggregate demand that is independent of the price level. If the level of aggregate demand increases independently of the price level, the downward-sloping curve would shift to the right. If the level of aggregate demand decreases for some reason other than the price level, the downward-sloping curve would shift to the left. Numerous factors can cause such a shift to occur.
Changes in the expectations of consumers and businesses can cause the AD curve to shift. Consumers base their buying decisions not only on the amount of money they have in the present but on how much money they expect to have in the future. Similarly, companies make decisions about investing in new equipment based on how good they expect business to be in the future. If consumers and businesses are optimistic, aggregate demand can rise even though there has been no change in the price level. This would be represented by a rightward shift of the AD curve. Pessimism on the part of consumers and businesses would likely have the opposite result.
Changes in wealth not related to prices can also cause the AD curve to shift. For example, many consumers in the 1990s were owners of shares of stock (portions of company ownership that can be bought and sold and that gain value when companies prosper). In the mid-to-late 1990s, the stock market boomed, and shares of stock gained value dramatically. In such an environment, with prices for consumer goods remaining steady, an influx of wealth derived from the stock market resulted in increased aggregate demand, which shifted the AD curve rightward. Stock markets sometimes crash without warning, as well, suddenly making shareholders poorer than they were a day or a week earlier. This could cause the AD curve to shift to the left even if the price level for products is unchanged.
Another common cause of shifts in the AD curve is a change in businesses’ capital (equipment) needs. If, for example, new computer technologies make it necessary for businesses across the economy to purchase all new office equipment, the AD curve might shift to the right as a result of increased business spending. On the other hand, it is also possible for businesses to fulfill their equipment needs fully; once they have done so, they have no need to continue buying equipment. If this happens across the economy, a leftward shift of the AD curve could result.
John Maynard Keynes was responsible not only for founding macroeconomics but also for setting the stage for government management of the economy. Keynes showed that governments could effectively manage aggregate demand (shift a country’s AD curve to the right or the left) by manipulating tax rates and levels of government spending. Since government spending on final products is itself a component of aggregate demand, changes to this form of spending directly shifts the AD curve. Changes to other types of government spending, such as direct payments to the poor or the elderly, shift the AD curve indirectly by making people more or less wealthy (and thus more or less likely to buy products). Similarly, increasing or decreasing taxes is an indirect way of inducing a shift in the AD curve, since increased taxes result in decreased wealth, and vice versa. The United States and other governments successfully recovered from the Great Depression by implementing Keynes’s ideas, and for the next several decades the U.S. government attempted to manage aggregate demand through fiscal policy (changes to tax and spending programs).
By the 1970s many economists had come to believe that managing aggregate demand through fiscal policy was messier and less effective than using monetary policy (changes in the quantity of money circulating in the economy) to influence the economy. When the government increases the size of the money supply, lenders have more money available to make loans, and business investment and consumer spending increase. This shifts the AD curve to the right. Decreasing the quantity of money in circulation sets off a chain reaction in the opposite direction, shifting the AD curve to the left.
Today central banks (those government entities responsible for setting monetary policy, such as the Federal Reserve System in the United States) take the lead in managing the economy. Though some people see this as a repudiation of Keynesian-style management of aggregate demand, others see it as a refinement of and addition to Keynes’s ideas about aggregate demand.
Aggregate demand is the total amount of goods and services that U.S. consumers and businesses are willing to buy at specific price levels. As prices for goods and services fall, consumers, businesses, and government agencies tend to buy more. In addition to the consumption of goods and services aggregate demand includes the money consumers and firms invest in government expenditures and net exports (that is, exports minus imports). When aggregate demand increases aggregate supply generally rises to keep up with it. Aggregate supply is the total output or production of goods and services.
Aggregate demand increases when consumers spend more or save less, when businesses believe the profitability of their investments will increase, or when the government spends more or lowers taxes. Aggregate demand will also rise when foreign consumers or businesses increase their purchases of U.S. products, when U.S. consumers buy fewer imports and more U.S. products, and when the money supply is increased. Because each of these factors can change fairly quickly, aggregate demand is more unpredictable than aggregate supply.
Aggregate demand can also be more easily shaped by government policy than aggregate supply can. British economist John Maynard Keynes (1883–1946) popularized the view that the best way to increase aggregate demand is to raise government spending or cut taxes. On the other hand so-called monetarists like Milton Friedman (b. 1912) argue that aggregate demand is best stimulated by lowering interest rates or loosen the supply of money circulating in the economy. Keynes believed that the Great Depression was caused by the federal government's failure to come to the rescue of an inherently unstable U.S. economy. Friedman argued that the Depression would never have occurred if the government had not sharply tightened the money supply in the late 1920s and early 1930s.
Between 1945 and 1990 the U.S. Federal Reserve never allowed the U.S. money supply to shrink as dramatically as it had just before and during the Great Depression. During this forty-five year period there were no major depressions. Those who followed Keynes argued that fiscal policy rather than money supply was the best way to pump up aggregate demand. But when the administration of President Ronald Reagan (1981–89) sharply lowered income tax rates in the early 1980s, aggregate demand remained largely unaffected.