The Critics of Keynesian Economics

XX

ARTHUR F. BURNS was born in Austria in 1904. He studied at Columbia University, took his doctor’s degree there in 1934, and became professor of economics there in 1944. He has been director of research at the National Bureau of Economic Research since 1930, and is now president of that organization. From 1953 to 1956 he was a chairman of the President’s Council of Economic Advisers. In 1946 he collaborated with the late Wesley C. Mitchell in the study Measuring Business Cycles. The following is an excerpt from the article “Economic Research and the Keynesian Thinking of Our Times” which appears in a collection of sixteen of Dr. Burns’s essays published by Princeton University Press for the National Bureau of Economic Research in 1954 under the title: The Frontiers of Economic Knowledge.

KEYNES’ THEORY OF UNDEREMPLOYMENT EQUILIBRIUM

ARTHUR F. BURNS

I have said enough to set the theme of my report, which is to relate the work of the National Bureau to the Keynesian thinking of our times. The opinion is widespread that Keynes has explained what determines the volume of employment at any given time, and that our knowledge of the causes of variations in employment is now sufficient to enable government to maintain a stable and high level of national income and employment within the framework of our traditional economic organization. If this opinion is valid, the solution of the basic problem of democratic societies is in sight, and the National Bureau would do well to reconsider its research program. Unhappily, this opinion reflects a pleasant but dangerous illusion.

The basis for the Keynesians’ confidence is Keynes’ theory of underemployment equilibrium, which attempts to show that a free enterprise economy, unless stimulated by governmental policies, may sink into a condition of permanent mass unemployment. The crux of this theory is that the volume of investment and the “propensity to consume” determine between them a unique level of income and employment. The theory can be put simply without misrepresenting its essence. Assume that business firms in the aggregate decide to add during a given period $2 billion worth of goods to their stockpiles, using this convenient term to include new plant and equipment as well as inventories. This then is the planned investment. Assume, next, that business firms do not plan to retain any part of their income;1 so that if they pay out, say, $18 billion to the public, they expect to recover $16 billion through the sale of consumer goods, the difference being paid out on account of the expected addition to their stockpiles. Assume, finally, that the “consumption function” has a certain definite shape; that if income payments are, say, $18 billion, the public will spend $17 billion on consumer goods and save $1 billion, and that one-half of every additional billion dollars of income will be devoted to consumption and one-half to savings. Under these conditions, the national income per “period” should settle at a level of $20 billion.

The reason is as follows. If income payments were $18 billion, the public would spend $17 billion on consumer goods. But the firms that made these payments expected to sell $16 billion worth to the public and to add $2 billion worth to their stockpiles; the actual expenditure of $17 billion on consumer goods would therefore exceed sellers’ expectations by $1 billion, and stimulate expansion in the consumer goods trades. On the other hand, if income payments were $22 billion, the public would spend $19 billion on consumer goods; this would fall short of sellers’ expectations by $1 billion, and set off a contraction in the output of consumer goods. In general, if income payments fell below $20 billion, the sales expectations of business firms would be exceeded; while if income payments rose above $20 billion, the expectations of business firms would be disappointed. In either case, forces would be released that would push the system in the direction of the $20 billion mark. Hence, in the given circumstances, $20 billion is the equilibrium income, and it may be concluded that the basic data—that is, the volume of investment and the consumption function—determine a national income of unique size. If we assume, now, a unique correlation between income and employment, it follows that the basic data determine also a unique volume of employment—which may turn out to be well below “full” employment.

This is the theoretical skeleton that underlies the Keynesian system. The theory implies that when unemployment exists, an increase in consumer spending out of a given income will expand employment; so too will an increase in private home investment or in exports, and so again will governmental loan expenditure, its effect on employment being in a sense similar to that of private investment expenditure. The theory implies also that the magnitude of the expansion in employment by any of these routes is a precisely calculable quantity, since the determinants of employment are alleged to have been isolated. To get more out of the theory, more specific assumptions must be made.

At this vital juncture the Keynesians differ somewhat among themselves, but two institutional assumptions dominate the thinking of the school. The first is that consumer outlay is linked fairly rigidly to national income and is unlikely to expand unless income expands; in other words, there is little reason to expect, at least in the short run, that a condition of unemployment will be corrected through a reduction in individual savings. The second assumption is that investment opportunities are limited in a “mature” economy such as our own; consequently, private investment may continue, year in and year out, at a level that falls considerably short of what the community would save if “full employment” existed. If neither an upward shift in the consumption function, nor an expansion of private investment at home, nor an increase in net exports can be confidently counted on, it follows that our lot may be persistent mass unemployment. We may escape the fate of secular stagnation, however, if the effective demand for employment is supplemented by governmental spending. Furthermore, this remedy for secular stagnation is also the remedy for business cycles, since the most that can be expected of private investment is that it may rise sufficiently to generate “full employment” during a fleeting boom.

Of late this theory has been refined and elaborated, so that “deficit financing” need no longer be the key instrument for coping with unemployment, and I shall refer to one of these refinements at a later point. But the practical significance of the modifications of the theory is problematical, and in any event the theory as I have sketched it still dominates the thinking of the Keynesians when they look beyond the transition from war to peace. The similarity of this theory to the Ricardian model is unmistakable. The most important proposition in Ricardian economics is that the production function in agriculture has a certain shape, that is, the marginal product diminishes as the input of labor increases. The most important proposition in Keynesian economics is that the consumption function has a certain shape, that is, consumer outlay increases with national income but by less than the increment of income. The Ricardians treated the production function as fixed, and deduced the effects on income distribution of an increase or decrease in population, or of a tax or bounty on the production of corn. The Keynesians treat the consumption function as fixed, and deduce the effects on the size of the national income of an increase or decrease in private investment, or of an increase or decrease in governmental loan expenditure. The Ricardians believed that population was the key dynamic variable, and they drew a gloomy picture of the course of events if that exuberant variable was not counteracted. The Keynesians believe that investment is the key dynamic variable, and they draw a gloomy picture of the course of events if that timid variable is not fortified by governmental loan expenditure. To be sure, the Ricardians recognized that the production function in agriculture was subject to change, and they frequently inserted qualifications to their main conclusions. The Keynesians likewise recognize that the consumption function is not absolutely rigid, and they frequently insert qualifications to their main conclusions. But I have formed the definite impression that the Keynesians—except when they discuss changes in personal taxation—attach even less importance to their qualifications than did the Ricardians; all of which may merely reflect the fact that the Ricardians were concerned largely with secular changes, while the Keynesians are mainly concerned, despite their anxiety over secular stagnation, with comparatively short-run changes.

There is, of course, nothing unscientific about Ricardianism as such. But ceteris paribus is a slippery tool, and may lead to serious error if the premises accepted for purposes of reasoning are contrary to fact, or if the impounded data are correlated in experience with factors that the theorist allows to vary, or if the very process of adjustment induces changes in the impounded data. Let us go back to the theoretical skeleton of the Keynesian system and examine it more carefully. Suppose that the volume of intended investment is $2 billion, income payments $20 billion, and consumers’ outlay at this level of income $18 billion. On the basis of these data, the economic system is alleged to be in equilibrium. But the equilibrium is aggregative, and this is a mere arithmetic fiction. Business firms do not have a common pocketbook. True, they receive in the aggregate precisely the sum they had expected, but that need not mean that even a single firm receives precisely what it had expected. Since windfall profits and losses are virtually bound to be dispersed through the system, each firm will adjust to its own sales experience, and within a firm the adjustment will vary from one product to another. Under the circumstances the intended investment cannot—quite apart from “autonomous” changes—very well remain at $2 billion, and the propensity to consume is also likely to change. Our data therefore do not determine a unique size of national income; what they rather determine is a movement away from a unique figure. Of course, we cannot tell the direction or magnitude of the movement, but that is because the basic data on which the Keynesian analysis rests are not sufficiently detailed for the purpose.

I have imagined that Keynes’ aggregative equilibrium is realized from the start. But suppose that this does not happen; suppose that, in the initial period, the intended investment is $2 billion, income payments $16 billion, and that savings at this level of income are zero. Will income now gravitate towards the $20 billion mark, as ihe theory claims it should? There is little reason to expect this will happen. In the first place, windfall profits will be unevenly distributed, and the adjustment of individual firms to their widely varying sales experiences will induce a change in the aggregate of their intended investment. In the second place, unemployed resources will exercise some pressure on the prices of the factors of production, and here and there tend to stimulate investment. In the third place, if an expansion in the output of consumer goods does get under way, it will induce additions to inventories for purely technical reasons; further, the change in the business outlook is apt to stimulate the formation of new firms, and to induce existing firms to embark on investment undertakings of a type that have no close relation to recent sales experience. In the fourth place, as income expands, its distribution is practically certain to be modified; this will affect the propensity to consume, as will also the emergence of capital gains, the willingness of consumers to increase purchases on credit, and the difficulty faced by consumers in adjusting many of their expenditures to increasing incomes in the short run. These reactions, and I have listed only the more obvious ones, are essential parts of the adjustment mechanism of a free enterprise economy. Under their impact the data with which we started—namely, the amount of intended investment and the consumption function—are bound to change, perhaps slightly, perhaps enormously. It is wrong, therefore, to conclude that these data imply or determine, even in the sense of a rough approximation, a unique level at which the income and employment of a nation will tend to settle. In strict logic, the data determine, if anything, some complex cumulative movement, not a movement towards some fixed position.

If this analysis is sound, the imposing schemes for governmental action that are being bottomed on Keynes’ equilibrium theory must be viewed with skepticism. It does not follow, of course, that these schemes could not be convincingly defended on other grounds. But it does follow that the Keynesians lack a clear analytic foundation for judging how a given fiscal policy will affect the size of the national income or the volume of employment. Fiscal policy is now the fashion among economists, and three fiscal paths to “full employment” have recently been delineated. The first is to increase expenditure but not taxes. The second is to increase taxes as much as expenditure. The third is to reduce taxes but leave expenditure unchanged. The first of these methods—that is, loan expenditure—avoids, we are told, the excessively large expenditures of the second method, and the excessive deficits of the third. This is a highly suggestive conclusion, and may have much to recommend it on practical grounds. But to accept it as an approximation to scientific truth we must be willing to make assumptions of the following type: (1) the consumption function is so shaped that the dollar volume of savings increases as income increases, (2) the consumption function is practically invariant except in response to personal taxation, (3) an increase in taxes will lower the consumption function considerably but by less than the addition to taxes, (4) a reduction in taxes will raise the consumption function but by considerably less than the tax reduction, (5) the planned savings of business enterprises are correlated simply and uniquely with income payments, (6) monopolistic practices of business firms can safely be neglected, (7) private investment will not be influenced appreciably by the character of the fiscal policy pursued by government. Although assumptions such as these may be extremely helpful at a stage in our thinking about an exceedingly complicated problem, it seems plain that the inferences to which they lead cannot be regarded as a scientific guide to governmental policies.

1 This assumption is not essential to the Keynesian system; I make it here in order to simplify the exposition. The figures used throughout are merely illustrative. Further, the exposition is restricted to the proximate determinants of employment in Keynes’ system; this simplification does not affect the argument that follows.