Liquidity Preference

views updated

Liquidity Preference

The demand for money

Liquidity preference, monetary theory, and monetary management


“Liquidity preference” is a term that was coined by John Maynard Keynes in The General Theory of Employment, Interest and Money to denote the functional relation between the quantity of money demanded and the variables determining it (1936, p. 166). He also used this term, or such variants of it as “liquidity preference function” and “liquidity function,” to denote more narrowly the relation between the quantity of money demanded and the rate of interest (see, for example, p. 168). Since the General Theory the term “liquidity preference” has come to be used to refer to the hypothesis or theory that the aggregate quantity of money demanded by the economy will, ceteris paribus, tend to be smaller the higher the rate of interest.

Keynes’s analysis of the systematic and intimate relation between the demand for money and interest rates and its implications is generally acknowledged to be one of his major contributions to economics. It is one of the two main pillars on which the edifice of the General Theory rests, the other being the hypothesis that in a contemporary monetary economy, money prices and especially money wages tend to be rigid in the downward direction (see “Liquidity preference, monetary theory and monetary management,” below).

The demand for money

Pre-Keynesian theories

Information about the history of theories of the demand for money may be found elsewhere [see especiallyMoney, articles On Quantity Theoryand Velocity Of Circulation; see also Marget 1938 and Patinkin 1956, pp. 373–472. It will suffice to recall here that although monetary theorists had long recognized that money is a “store of value” as well as a “medium of exchange,” prevailing theories of the demand for money before the General Theory tended to stress the role of money as a medium of exchange and the “transaction demand.” The two major, broadly accepted formulations before the General Theory were that of Irving Fisher and that of the Cambridge school. Fisher (1911) started from the now well-known identity called Fisher’s equation of exchange: MV = PT, where M is the quantity of money in circulation, T is the volume of transactions, P is the price level, and V is the “transaction velocity of circulation.” This equation is also frequently restated as MVu = PX = Y, where X is “real income,” Y is money income, and Vu is the “income velocity of circulation.” From these identities Fisher derived his theory by hypothesizing (1) that at a given point of time V can be taken as constant (or at least as largely independent of M) and (2) that V tends to change, at best, very slowly over time, being largely determined by institutional and technological factors with a high degree of inertia. The major factors of this kind include the frequency of receipts and disbursements (intimately related in turn to the so-called income period, which is the length of the interval between the dates at which various types of income, such as wages, salaries, and dividends, are typically paid), the degree of synchronization of receipts and expenditures, prevailing financial arrangements, the rapidity of transportation, and so on. [seeMoney, article on Velocity Of Circulation.]

By contrast, the so-called Cambridge school tried to put the explanation of the demand for money into the more familiar format of value theory, i.e., in terms of a demand-for-money equation, Mậ = kY, an exogenously given supply of money, M, and a clearing-of-market equation, Mậ = M, implying M = kY (see, e.g., Pigou 1917; Marshall 1923). By comparing this equation with Fisher’s equation above, one can readily see that k = l/Vu, i.e., that k is the reciprocal of the velocity of circulation. Indeed, in analyzing the determinants of k and the reasons for its hypothesized stability, the Cambridge school tended to stress largely the same forces on which Fisher’s theory rests.

The Fisher and Cambridge models are generally regarded as providing the definitive basis for the so-called “quantity theory of money,” a view of very old standing according to which the price level, P, tends to be directly proportional to M. In order for this relationship to follow logically from these models, not only must M not affect V (or k), as those models imply, but also one must suppose that money is “neutral” in the wider sense that it does not affect any of the “real” variables of the system—inputs, outputs, and relative prices, including interest rates. Under this assumption, which, as shown below (see “The significance of liquidity theory under wage flexibility”), might provide a reasonable approximation under conditions of perfect wage and price flexibility, real income, X, may be taken as fixed at the “full employment” level, say ‾X. From either the Fisher or the Cambridge equation it then follows that

that is, the price level is proportional to the quantity of money, M.

It should be acknowledged that some of the writers in the Cambridge tradition did at times suggest that the demand for money might depend on wealth and that they did make some occasional references to the possible influence of interest rates (see, for example, Pigou [1917] 1951, p. 166; Lavington 1921, p. 30; Marshall 1923, chapter 4; for still earlier references, see Eshag 1963, pp. 13–14). But they failed to explore systematically the effect of interest rates on the demand for money and the implications of this effect. This failure is even more conspicuous in Fisher. He makes no mention of interest rates in his list of factors affecting velocity, and although he makes fleeting mention of the “waste of interest” involved in holding money (1911, p. 152), one finds no reference to this passage in the index under the rubric “interest rates.”

Two authors who anticipated Keynes in giving adequate recognition to the role of interest rates are Walras, in 1899, and Schlesinger, in 1914 (see Patinkin 1956, notes C and D), but their contributions were largely overlooked at the time. The most significant pre-Keynesian analysis of liquidity preference is generally acknowledged to be that of Hicks (1935), which, however, preceded the General Theory by but one year and was partly inspired by Keynes’s earlier work, A Treatise on Money, published in 1930. This contribution to monetary theory, which in some respects has turned out to be even more influential for further developments than that of Keynes, will be touched upon below.

Keynes’s theory

In chapters 13 and 15 of the General Theory, Keynes distinguished three “motives” for holding money. The first, the “transaction motive” —sometimes broken down into an income motive and a business motive—corresponds quite closely to the motives stressed by Fisher and the Cambridge school. Like his predecessors, Keynes did not regard transaction balances as being significantly affected by interest rates. The second motive is the “precautionary motive.” Under this heading Keynes included balances not earmarked for some definite expenditure in the near future but held instead to “provide for contingencies requiring sudden expenditure and for unforeseen opportunities of advantageous purchases” (1936, p. 196). But why should these balances be kept in the form of idle cash instead of being invested in some kind of readily marketable securities, to be converted into cash if and when the contingency arises? The reason is that the market value of a debt instrument (or “bond”), if it is liquidated before its maturity, is uncertain, even if there is absolutely no risk of default. It depends on the market rate of interest prevailing at the future time of liquidation for loans having a duration equal to the remaining life of the bond: the higher this rate, the lower the market value. This uncertainty about the realization value of a bond would not by itself make bonds inferior to cash as a store of ready purchasing power if the sum of the uncertain liquidation value and the cash interest earned could be counted on to exceed the amount initially invested. However, there can be no such assurance, since between the times of purchase and liquidation interest rates could rise sufficiently to produce a capital loss in excess of the interest earned. Keynes suggested in particular that the likelihood of a net loss would be larger the smaller the yield of the bond originally acquired. This is because the smaller the yield, the smaller the rise in the rate of interest (in absolute as well as in percentage terms) that will produce a capital loss sufficient to wipe out the accrued interest earned. Furthermore, Keynes suggested that if the current rate is low by historical standards, it will usually be regarded as more likely to rise than to fall. He concluded that the lower the current rate, r, the stronger the incentive to hold precautionary reserves in the form of cash instead of securities. Therefore the (real) demand to hold money for precautionary reasons will tend to be inversely related to r. At the same time, somewhat surprisingly, Keynes did not appear to regard precautionary balances as very sensitive to r. Accordingly, much of the time he lumped together the demand for transaction and for precautionary reasons and regarded the sum, which he labeled Ml 5 as primarily controlled by—or a function of—current income. Thus, in his notation M1 = L1 (Y), where the function Ll denotes the demand for money resulting from the transaction and precautionary motives.

The third and remaining source of demand for money is the speculative motive, a rather complex mechanism that Keynes had partly anticipated in A Treatise on Money (1930). In essence, speculative balances are balances held in cash rather than invested in (long-term) bonds, not just because of the risk that interest rates might rise but rather because of a definite expectation that the price of long-term bonds is likely to fall, and at a rate that more than offsets the interest earned by holding them. A person entertaining such an expectation would prefer to hold cash yielding nothing rather than invest it in what he regards as overpriced long-term bonds that would yield him a negative return. Since the price of long-term bonds varies inversely with long-term interest rates, we may equally well characterize speculative balances as those held by persons who regard the current long-term rate as untenably low and about to rise sufficiently rapidly.

The real significance of the speculative motive is that it may significantly impair, or even thwart altogether, efforts of the central bank to reduce long-term interest rates to the extent necessary to maintain investment at the level consistent with full utilization of resources (see “Liquidity preference, monetary theory, and monetary management,” below). Normally, the central bank can expect to enforce lower long-term interest rates, or higher prices of long-term bonds, by buying such bonds with newly created money. Suppose, However, that a large portion of the market holds definite views about the minimum maintainable level of the long-term rate and hence the maximum maintainable level of bond prices. If, then, the bank attempts to bid up the price of bonds to that maximum or beyond, it will find the public prepared to dump a large portion of its long-term bond holding. The bank will therefore have very little success in lowering the long-term rate, even though it is prepared to acquire a large volume of bonds and to expand the money supply correspondingly. What happens in this situation is that the increase in the money supply is absorbed, not by an increased transaction demand, but by an offsetting increase in the speculative demand, with a resulting fall in the velocity of circulation. In other words, the expansion in M, instead of achieving the desired expansion in income, Y, that would occur if Vu, the velocity of circulation, remained constant, tends to generate an offsetting change in Vu, with little effect on Y. A situation of this type has come to be known in the Keynesian literature as a “liquidity trap.”

Keynes denoted speculative balances by M,1 and wrote the demand function for such balances as M2 = L2 (r), where L2 is a decreasing function of r (1936, p. 199). This formulation—that M2 increases as r falls—is somewhat misleading, since presumably M2 should depend not on r as such but only on r in relation to the prevailing market expectations about the maintainable rate, say re. Nor can re be supposed to stay constant through time or to be uniquely related to r itself. Keynes’s formulation might be defended as a useful “short run” approximation: at a given point in time, re can be taken as a constant or, at least, as changing more slowly than r. Hence, a fall in r would necessarily imply a fall relative to re and thus a rise in M2(ibid., pp. 201-202). Under this interpretation, however, one should be aware that L2 may be subject to significant shifts through time as a result of shifts in market expectations.

The sum of the transaction and precautionary demand, Ml 5 and the speculative demand, M2, is the total demand for money proposed by Keynes: M = M1 + M2 = L1 (Y) + L2 (r) (ibid., p. 199). The Keynesian literature has tended to de-emphasize the sharp distinction between the three motives for holding money and to write the demand for money in the more general form M = L(r, Y). There has also been a tendency to minimize the role of interest expectations, re, and to associate the liquidity trap with a low absolute level of the interest rate.

The implied relation between M and r for a

given value of Y, say Y0, is shown in Figure 1 by the curve labeled L0 L0. (The choice of coordinates is dictated by the economists’ peculiar convention, popularized by Marshall, of representing demand curves with the quantity demanded measured on the abscissa and the price on the ordinate.) The quantity of money demanded increases continuously as r falls, until, for some sufficiently low value, rm, the liquidity trap is reached and the curve becomes horizontal (the demand becomes infinitely elastic). Alternatively, the demand curve might be drawn to approach the level rm asymptotically. Just how low rm may be depends some-what on “institutional” factors and on whether r is understood to be the long-term or the short-term rate. But we can, with complete generality, place a lower bound on rm: in a monetary economy, rm can never be more negative than the (marginal) cost of storing money. In particular, when money is an intangible, the cost of storing it (at least in the form of bank deposits) is essentially zero, and therefore rm cannot be (significantly) negative. Indeed, a negative r can be regarded as a premium paid by the lender to the borrower for carrying money over; for example, a short-term rate of —2 per cent per period means that the lender is willing to pay $100 to receive only $98 at the end of the period. If the cost of storing is less than 2 per cent, everybody would wish to borrow indefinitely large amounts, since by merely holding the money one would earn the excess of 2 per cent over storage costs. This implies in particular that with a zero (marginal) storage cost, at a negative rate of interest the demand for money must become indefinitely large—or, equivalently, that no matter how large the quantity of money, r can never be negative. Hence, the demand curve must tend to approach a horizontal asymptote, r = rm (or possibly reach it from above for some finite M and become discontinuous). Furthermore, rm cannot be lower than zero (quite generally, it cannot be more negative than the marginal cost of storing money), although it may well be higher, as in Figure 1.

The curve labeled L1L1 illustrates the effect on the demand for money of increasing Y, say from Y0 to 2Y0 in Figure 1. Clearly, the demand for money must then be greater at any given rate r; that is, LL must shift to the right. The relation between L0L0 and L1L1 becomes especially simple if the demand function L(r,Y) takes a more specialized form, which was suggested, for example, by Pigou (1917) and tested by Latane (1954; 1960) and which has been gaining favor in recent writings—namely, M = k(r)Y = Y/V(r). This formulation provides an obvious bridge between Keynes’s original formulation and the received Fisher and Cambridge models. It implies that for a given r the fraction k (or the velocity of circulation, V) will be constant but that k will tend to fall (or V to rise) as the rate of interest rises. In terms of Figure 1, it implies that L1L1 is obtainable from L0L0 by multiplying by 2 the abscissa value of L0L0 corresponding to any given r. More generally, it implies that the LL curve corresponding to any given Y is simply the graph of k(r), up to a proportionality factor, Y. Similarly, the graph of V(r),

the velocity of circulation as a function of the interest rate, is the graph of the reciprocal of L0L0 up to a proportionality factor, 1/Y0. The general shape of the graph of V(r) is shown by the vv curve of Figure 2.

Post-Keynesian developments

As indicated earlier, post-Keynesian developments of liquidity theory were inspired not only by Keynes’s General Theory but at least as much by two germinal ideas advanced by Hicks (1935). Hicks’s first suggestion was that the major reason why transactors hold money balances having little or no yield when they could invest them in a large number of income-yielding assets, some at least not significantly less safe than money, is to be found in the costs and the “bother” of the transactions necessary to move from money into earning assets and back to money (p. 19).

The portfolio approach. Hicks’s second suggestion was that the theory of the demand for money must be developed out of a more general theory of the allocation of wealth among various assets. This theory, Hicks suggested, should be analogous to the standard theory of consumers’ choice, except that the object of choice, instead of being consumption flows, would be the various stocks appearing on the asset and liability side of the balance sheet, and prices would be replaced by expected yields. He saw this substitution as presenting a real challenge, since yields in contrast to prices would have to be recognized as uncertain, and this uncertainty in turn would have important implications for the nature of choices.

Both ideas have been extensively pursued with the help of the emergence of the theory of choice under uncertainty [seeDecision Making, article on Economic Aspects]. At present the major differences of view between monetary theorists (and they are not very major) seem related to the relative importance assigned to each of Hicks’s two ideas.

Among those who have pursued the portfolio, or wealth, approach, the formulation of Friedman (1956), developed in numerous writings, has been particularly influential [seeMoney, article on Quantity Theory]. Friedman views the demand for money as being determined by wealth (broadly understood as the present value of expected net future receipts from all sources), by the distribution of wealth between human and nonhuman (i.e., marketable) wealth, by the expected yield of all major types of assets that are alternatives to money as ways of holding wealth, and by the “utility attached to the services rendered by money” relative to other assets—namely, bonds, equities, and physical commodities. By combining this theory with his suggestion (1957) for ways of approximating wealth (or, more precisely, “permanent income,” which is, however, essentially proportional to wealth as defined above) Friedman has endeavored to cast his theory in testable form and actually to test it (1959). He has concluded that his model fits the facts well, in that the demand for money increases with wealth and more than in proportion, although he can find little evidence that interest rates in fact play a significant role.

Other authors have been more concerned with developing and refining theoretical aspects of the Keynes-Hicks approach (see the very useful survey provided by Johnson 1962). Among their attempts, especially worth noting are the recent contribution of Turvey (1960) and the elegant formulation of the theory of choice between money and bonds of various maturities developed by Tobin (1958) along the lines of the modern theory of portfolio selection.

Transaction costs and the Neo-Fisherian approach. The portfolio approach suffers from one major inadequacy. As long as there exist any interest-bearing obligations that are issued by credit-worthy borrowers and are of sufficiently short maturity—for example, redeemable on demand or on very short notice—it is impossible to explain why any portion of the portfolio should be held in the form of money, yielding less or nothing at all— except by explicit analysis of the role of transaction costs.

Even Keynes’s stricture that, for sufficiently low interest rates, money may dominate bonds because of the uncertainty of the realization value cannot apply to short maturities or demand loans. These instruments dominate money in every possible dimension: they are equally safe, they yield an income, and they can be converted into the medium of exchange, if and when it is needed, at face value. Why, then, should anyone hold money, except for the very instant he receives a payment or is about to make one? The necessary and sufficient condition, as Hicks rightly pointed out, is that out-of-pocket costs and the effort required in moving from cash to bonds and back to cash exceed the yield. At first sight these transaction costs may appear too trivial to account for any substantial holding of cash, let alone for the observed cash holdings. (Aggregate cash holdings of U.S. consumers at the end of 1963 were estimated to represent slightly less than two months’ income.) But this casual impression is misleading. It is well known, for instance, from the theory of optimum inventory holdings that transaction costs do account for a substantial portion of inventories held by business (which in the United States amount to some three months’ sales). This parallel between business inventories and cash holdings is not fortuitous, for in many respects the holding of a stock of cash by transactors is closely analogous to the holding of a stock of goods by business. In fact, Allais (1947, chapter 8a) and Baumol (1952) pioneered in showing that the holding of cash balances could be analyzed by a straightforward application of the so-called lot-size formula of inventory theory: in order to avoid incurring too frequently the costs involved in transforming securities into cash, it pays to secure cash in a bulk or “lot” that will take care of expenditure requirements for a certain length of time, even though interest will be forgone on the amount withdrawn. Similarly, if a transactor is receiving money in a more or less continuous trickle, it will pay to accumulate a “lot” before investing it. The size of the lot, and hence the average cash balance held relative to the rate of outpayments (or receipts), which corresponds to the Cambridge k or to the reciprocal of the velocity of circulation, will be positively associated with the size of transaction costs and inversely associated with the rate of interest. Tobin (1956) refined and improved on this analysis, applying it more specifically to the consumer receiving his income in bulk at income-payment dates and spending it gradually over the income period.

Although these contributions are to be regarded as illustrative rather than as aimed at deriving an exact demand equation for money, they do point up one very fundamental principle. The amount that can be earned by investing an amount of cash, m, that will not be needed to meet expenditures for some span of time, t, in a security yielding r per cent per year, is approximately m(tr — c), where c is the brokerage fee, if any, per dollar of investment. The investment will not be worthwhile unless this product exceeds the lump-sum cost of the two-way transaction, including both the out-of-pocket and the bother costs. To illustrate the order of magnitudes involved, suppose that a person earns $12,000 a year, paid monthly; he then receives $1,000 once a month. Suppose he spends these receipts at an even rate. He might then consider keeping half the sum for current expenditure and investing the remaining half, or $500, which he will not need until the first half is exhausted—that is, for half a month. Suppose the yield of a 15-day security, net of commissions, is 3 per cent per year; then all he stands to earn from the transaction is $500 x .03/24, or a mere 62.5 cents. If the two-way transaction cost and bother exceeds this, he will invest none of the monthly receipts and thus will end up holding, on this account, an average cash balance of $500, or 1/24 of his (annual) income. Note that if he were paid twice as frequently—that is, $500 every two weeks—it would a fortiori not pay him to bother, and he would be holding an average cash balance of $250, or 1/48 of his income.

The conclusion to be drawn from these illustrations can be summarized as follows: In a money-using economy, transactors are paid in money and in turn must pay in money; lack of synchronization between receipts and payments gives rise to pools of money that will not be needed for some length of time. Given the rate of return and the cost and effort of transactions, it will not pay to invest such pools unless the product of their size and the length of the “idle” time exceeds some critical threshold level. Thus, the basic reason for holding idle cash balances is not that they provide a useful service but simply that it does not pay to shed them. Obviously, given the rate of interest, the extent to which it does not pay to shed idle money, and thus the average cash balance held, will depend on such institutional-technological factors as (a) transaction costs—the higher the cost, the smaller the incentive to shed; (b) the size and nature of the transactor’s business —large transactors may be confronted with pools so large that it pays to shed them even for very short periods, and they may also have an incentive to set themselves up so as to minimize marginal transaction costs; and (c) the frequency of income payment and settlement dates—the greater the frequency, the smaller the average cash balance. But these are, by and large, precisely the factors emphasized by Fisher in explaining the determinants of the velocity of circulation. The new element is the recognition that given all these factors, the average cash balance demanded will tend to fall with the rate of interest, which provides the incentive to shed.

How does the Keynesian liquidity trap fit into this model? The first point to be noted is that Keynes’s theory of the speculative demand suffers from his excessive concentration on long-term bonds as the alternative to cash, to the neglect of short-term instruments. The proposition that people will flee from long-term bonds when the price of those bonds is deemed to be untenably high seems valid enough, but the obvious abode for the funds accruing from moving out of long-term bonds should be short-term ones, not cash. However, a massive endeavor to move from long-term into short-term instruments will unavoidably depress short-term rates, perhaps to such an extent that for many investors the investment will no longer be worth the effort. Thus, they may eventually end up holding cash, but because of the low level of short-term rates, not directly in response to the low level of long-term rates. In short, the central bank’s endeavor to depress long-term rates by buying bonds and increasing the money supply can always be counted on to depress short-term rates. However, it may not be very successful in depressing long-term rates to the desired extent, except insofar as a persistent low level of the short-term rate may eventually persuade the public that the long-term rate is really not unreasonably low. A good example of such a development is provided by the United States in the late 1930s. Because of a sizable monetary expansion after 1932, by 1939–1940 the short-term rate on government bills had been driven down very nearly to zero (below 2/10 of 1 per cent), but the long-term rate on high-grade bonds was still hovering around 3 per cent (down from about 4.7 per cent in 1929). In this sense the Keynesian liquidity trap must still be acknowledged as a possible serious hindrance to the effectiveness of monetary policy. And in any event, the proposition that no market rate—long or short—can ever be negative retains its validity.

The theory that emerges from the preceding discussion emphasizes the flow of transactions (and therefore income rather than wealth) and interest rates, especially the short rate and the rate on savings deposits, as the main arguments of the demand function for money. It further suggests that the parameters of this function are largely determined by the forces emphasized by Fisher and should therefore tend to change at best slowly through time. Because the model represents an obvious blend of the motives emphasized by Fisher and by Keynes and Hicks, we have referred to it as the Neo-Fisherian approach (although this terminology is not in general use).

Although the contrast between the “portfolio” approach and the “transaction” approach has deliberately been emphasized here, it is well to recognize that the difference between these two models is minor—largely a matter of relative emphasis— both in principle and in terms of practical implications. In particular, these models concur in the conclusion that the demand for money should be “homogeneous of first degree in current prices”— that is, that a change in the price level, other things being equal, should give rise to a proportional change in the demand for money while leaving unaffected “real demand” (demand measured in terms of purchasing power over commodities).

Empirical verification. Since the appearance of the General Theory, considerable effort has been devoted to assessing empirically the responsiveness of the demand for money to variations in interest rates and more generally to estimating demand functions for money and testing their stability (see, for the United States, Johnson 1962, pp. 354–357).

These investigations have tended to confirm that the demand for money is positively and closely associated with income or wealth or both and that a change in the price level tends to result in a proportional change in demand. They have also overwhelmingly tended to confirm that this demand is significantly responsive to changes in interest rates in the direction hypothesized by Keynes. The only significant exception in this regard is Friedman’s results, cited in the section “The portfolio approach,” above. His contrary conclusions, however, have been criticized for being very much dependent on the specific definition of money he uses (which includes means of payment and some, but not all, savings deposits), on the specific period chosen for his tests, and on his statistical techniques. They have also been criticized because his model, although it apparently fits the period from the second half of the last century to the late 1940s quite well, is not able to account for the very significant rise in velocity that has occurred since the beginning of the 1950s, concomitantly with the marked rise in interest rates. In particular, Meltzer (1963) and Brunner and Meltzer (1964), who otherwise fully sympathize with Friedman’s basic theory, have found marked and significant interest-rate effects, whether one uses as additional variables income, or permanent income, or a measure of nonhuman wealth. The major novelty in their results is the strong showing of the nonhuman wealth variable as compared with current income, although these results contrast with those reported by other investigators using a different measure of wealth (e.g., Bronfenbrenner & Mayer 1960). On the whole, it seems fair to say that at the moment the evidence is not adequate for the fine discrimination between the wealth and the neo-Fisherian formulations of the demand for money.

Liquidity preference, monetary theory, and monetary management

The Keynesian revolution

As suggested earlier, the two major analytical contributions of the Gen- eral Theory are the hypotheses of liquidity preference and of wage rigidity. The systematic analysis of the implications of these two highly fruitful hypotheses and their interaction was made more powerful and incisive by a third novelty, which is primarily methodological. This is the development of “aggregative analysis,” or what has since come to be known as macroeconomic analysis. Economists had long before been used to analyzing economic variables as reflecting the interaction of simultaneous relations, and the notion of equilibrium was used precisely to denote the value of the variables simultaneously satisfying all the relevant relations. However, before the General Theory this method of analysis was generally applied in so-called “partial equilibrium analysis,” that is, the study of some portion of the economy—say, the market for a particular commodity or a group of interrelated commodities. The method had also been applied with some success, largely by Walras, to the economy as a whole in “general equilibrium analysis,” which formally recognizes the interactions of all possible markets, treating the economy as a very large scale closed system of simultaneous equations. The novelty of aggregative analysis consists in lumping together a large number of commodities having common characteristics for the problem at hand and treating the aggregate as a single commodity. This approach makes possible the approximation of the whole economy with a small system of simultaneous relations, and, by permitting closer scrutiny and understanding of the interactions, it has proved to be highly fruitful.

Analysts of Keynes’s work have correctly pointed out that none of these basic ingredients of the General Theory—liquidity preference, wage rigidity, or the aggregative approach—was entirely new. We have documented this point above with respect to liquidity preference. The novelty consisted in the masterly way in which the ingredients were blended, which enabled Keynes to provide an analytical explanation of the phenomenon of unemployment and its possible persistence in an advanced capitalistic economy and to shed new light on the role and limitations of monetary and fiscal policy in controlling the level of employment and prices. It is this achievement, and its enormous impact on economics, that has since come to be known as the Keynesian revolution.

The rest of this section, relying largely on aggregative analysis, endeavors to sketch out the role of liquidity preference, first under the classical assumptions of perfect wage and price flexibility and then in combination with the empirically far more relevant hypothesis of downward wage rigidity. Our focus is primarily on the significance of liquidity preference as seen today, some thirty years after the appearance of the General Theory, rather than on summarizing or criticizing Keynes’s original formulation. Accordingly, in what follows, the post-Keynesian elaborations are freely drawn upon.

The basic model

In Keynes’s General Theory and, more particularly, in later endeavors by other authors to formalize its message (e.g., Hicks 1937; Lange 1938; Modigliani 1944; 1963; Patinkin 1956), the whole economy is reduced (explicitly or implicitly) to four aggregates: aggregate output, X; labor, N; money, M; and bonds, B. For each of these aggregate commodities there is a “market” characterized by supply conditions, demand conditions, and the “clearing-of-market” or equilibrium requirement that demand must equal supply. Demand and supply, in turn, are controlled by three prices or terms of trade between each commodity and money: P, the price of output (the “price level”); W, the price of labor (the “wage rate”); and 1 + r, the number of dollars obtainable next period by lending a dollar today, where r is the rate of interest. To understand the mechanism determining the level of economic activity in a given short interval and the role of liquidity preference, we must examine the structure of the four markets and their interaction.

The demand for output in the commodity market, usually referred to in the literature as “aggregate demand” and denoted here by Xd, is a central construct of Keynesian analysis. It has given rise to a voluminous literature, both theoretical and empirical, which can be summarized here very briefly, since it is covered in other articles [see in particularIncome And Employment Theory; Consumption Function; Investment, article on The Aggregate Investment Function]. TWO sources of demand are distinguished: current consumption, C, and investment demand, I—i.e., demand for current output destined to increase the stock of productive capital. Thus, Xd = C + I. Theoretical considerations, and the empirical evidence, suggest that consumption in turn is primarily controlled by (a) the level of real income that, disregarding for the moment the fiscal activity of the government sector, can be equated with aggregate output, X; (b) net real private wealth, A; and possibly (c) the rate of interest, r. This can be formalized by means of the “consumption function,” C = (X, A, r). Investment demand can be taken to be positively associated with aggregate output and negatively associated with the rate of interest and the pre-existing stock of capital, K0; thus, 1 = & (r0, X, K0). Finally, net private real wealth, A, the sum of all privately held assets minus private debt, can be expressed as A = K0 + G/P; that is, it consists of the stock of capital plus the money value of the outstanding government debt, G, deflated by the price level, P, to express it in terms of purchasing power over output.

The four equations given above can be conveniently reduced to a single one by first substituting for A in the consumption function and then substituting this function and the investment function into the definition of aggregate demand:

Next, we observe that in equilibrium, aggregate demand Xd must equal aggregate supply, or

We use this property to replace X with Xd in the right hand side of (1). The resulting equation contains Xd on both sides of the equality. We can, however, “solve” the equation explicitly for Xd, obtaining finally an expression of the form

which will be referred to hereafter as the aggregate demand relation. Note that aggregate demand, Xd, may be expected to be negatively associated with r. This is because an increase in r will reduce investment demand directly, and this reduction, in turn, will reduce aggregate demand even further by means of its depressing effect on consumption demand, which depends on total output — this is the so-called multiplier effect [seeConsumption Function]. Insofar as investment demand itself depends on output, Xd may in fact not decrease continuously as r rises, but this complication will be ignored here.

To complete the description of the output market, we also need an “aggregate supply function.” Aggregate supply, X, may be expected to be positively associated with (a) the price, P, at which firms can sell their output relative to the wage rate, W, they must pay, or P/W, and (fr) with the preexisting capital stock, K0 (on the convenient approximation that the increment in the stock of capital resulting from current investment will not become productive until the next period); thus,

In the labor market, the aggregate demand for labor, N, can be inferred from the so-called aggregate production function, relating output, X, to the input of labor and the stock of capital, K0. This function implies that N can be expressed in terms of X and K0 , say, N = F(X, K0). It is, however, more convenient to replace X in this equation with the right-hand side of (3), thus obtaining the “labor demand” equation

The description of the supply side of the labor market is a somewhat more complex task, for it is here that we must formalize the Keynesian notion of “downward wage rigidity.” In its broadest sense, this term connotes the absence of “wage flexibility,” of a state of affairs in which money wages fall promptly whenever the supply of labor exceeds the demand for it and keep falling as long as the excess supply persists. In a narrower definition, it means that the current money wage will not be bid below some floor level, W0 (reflecting the past history of the system), no matter how large the excess supply of labor — though it can be freely bid up in response to excess demand for labor. For present purposes, we shall rely on this narrower version, which we label “absolute” rigidity, because it is more readily formalized. However, the conclusion of the analysis below would not change qualitatively if the wage rate had some tendency to fall for sufficiently large unemployment and falling prices, as long as the reaction was sluggish and unsystematic. There can be little doubt that wage rigidity in this sense is, and has been for some time, a feature of free market economies.

To formalize the hypothesis of absolute wage rigidity we need to introduce the notion of a “potential supply of labor function,” say, ȓ (W/P), which gives the level of employment “desired,” or labor force available, at any given real wage, W/P. Now, let E denote the actual level of employment. Then absolute wage rigidity can be expressed as follows:

Line (a) of (5) states in essence that if at the rigid wage W0 the demand for labor falls short of the potential supply, then the actual wage rate will coincide with W0. Employment, being equal to the demand for labor as stated by (6), will then fall short of the potential supply, and the difference will represent the so-called involuntary unemployment. If, however, at W0 the demand exceeds the potential supply, then line (b) of (5) becomes applicable: the floor level loses its relevance, and the actual wage will have to rise enough to equate the demand with the potential supply. This formulation of wage rigidity has the advantage that it can encompass wage flexibility as a limiting case, in which we assign to W0 a value so small that the relevant portion of (5) will necessarily be line (b).

In the money market, the demand, Md, can be expressed as Md = L(PX, PA, r), where L is, of course, the liquidity preference function that (in recognition of the two major points of view summarized in the section “Post-Keynesian developments” of liquidity preference) is written as a function of both money income (PX) and wealth (PA). By expressing A in terms of its components, K0 and G, and using the property that a change in the price level should tend to give rise to a proportional change in the demand for money, the preceding equation can be rewritten as

Md = PL (X, K0 + G/P, r),

However, for the purpose of the graphical analysis that is developed below, we shall frequently find it convenient to rely on the specialized version Md = PX/V(r), where V(r), it will be recalled, denotes the velocity of circulation as a function of the rate of interest. As to the supply side, unless otherwise specified, it will be assumed that money is created by the banking system in the process of purchasing debt instruments (bonds) issued either by the private sector or by the government, and that the total supply of money, M, is exogenously determined through central bank policy. Since in equilibrium we must have Md = M, the description of the money market can be reduced to a single equation obtained by replacing Md with M in the above equations:


Equations (1) to (7) involve seven endogenous variables: Xd, X, N, E, P, W, r. They therefore form a closed system whose solution describes the shortrun equilibrium of the economy. This solution also depends, of course, on the parameters of the various equations, on initial conditions, such as K0 and G, and on policy variables, of which in the present case there is but one, the money supply, M. The demand and supply for the remaining commodity, namely, bonds, B, are not explicitly displayed in the system because, by a well-known principle called Walras’s law, it can shown that the demand and supply for one commodity out of the set of all commodities are necessarily equal when all other markets are “cleared” (that is, when demand equals supply); we choose the bond market as the redundant one (Modigliani 1963).

With the help of this system we can now focus on the role of monetary forces, in particular that of liquidity preference, in the determination of equilibrium, beginning with the classical assumption of wage flexibility.

Wage flexibility

As already noted, under the assumption of wage flexibility the labor supply conditions are fully described by line (b) of equation (5). But this equation, together with that for labor demand, equation (4), and the equilibrium condition (6), turns out to involve only three variables: N, E, and P/W (or its reciprocal, the real wage). They therefore form a closed subsystem which can be solved independently of the rest. This solution yields the equilibrium real wage, W∇/P (where “∇” denotes an equilibrium value), and employment, ∇N (which is also “full employment” since it coincides with the labor supply). From (3) we can then infer the equilibrium or full-employment level of output, ∇X.

At this point it becomes useful to distinguish two possible cases, the one in which there is no national debt and the one in which there is.

No national debt. Referring back to the aggregate demand relation (1’), we observe that if G is zero, then the third argument of the aggregate demand function is necessarily zero, no matter what value P may take: in other words, aggregate demand does not depend on the price level. (This important implication, it should be noted, depends critically on the approximation implicit in the formulation of the consumption function, that aggregate consumption, C, depends only on aggregate net wealth, not on its distribution between households. Changes in the price level will of course affect the demand of individual consumers by causing redistributions of wealth between creditors and debtors, but our aggregative assumption implies that such redistributions will affect only the distribution of consumption between households, without affecting the total.) Since K0 is a given initial condition, it can be seen that the right-hand side of (1’) contains only one variable, r. It follows that from (1’) we can infer the equilibrium value of r, ≷r, which makes the aggregate demand, Xd, equal to the aggregate supply, ∧X. Next, substituting ∧r and ∧X into the money-market equation

(7), we can determine the price level ΄P that equates the demand for money with the given supply (provided such a value of P exists; see below). Finally, from ΄P and the equilibrium real wage, W∧/P, we can infer the equilibrium money wage, ∧W.

The nature of the solution can be clarified by means of Figure 3, in which X is measured on the ordinate and r on the abscissa. The horizontal line labeled SS, cutting the ordinate at ∧X, represents the aggregate supply consistent with full employment —that is, with the clearing of the labor market. It is a horizontal line because, as should be apparent from the derivation above, the value of ∧X does not explicitly depend on r. The curve labeled DD is the graph of the aggregate demand relation (1’), which is shown falling from left to right for the reasons stated earlier. Equilibrium in the commodity (and labor) market is thus represented by the point of intersection of the demand and supply curves, namely, point a, with coordinates (∧X,∧f).

There remains to be shown the role of money and the money market in the determination of equilibrium. For this we refer back to equation (7’) and note that for a given value of M/P this equation expresses a relation between the two variables X and r. It is therefore amenable to graphical representation in our figure. Indeed, the shape of this graph can be readily inferred by solving (7’) for X to obtain X = V(r)(M/P). Given M/P, the graph of this equation is simply that of V(r), already shown in Figure 2, except for a proportionality factor and for the fact that the axes are interchanged, r now being measured on the abscissa instead of the ordinate. The result is a curve such as M’M’, which represents the locus of (7’) for an arbitrarily chosen value of the “real” money supply, (M/P)’. It rises from left to right because, as r increases and transactors are induced to economize on their cash holdings, the velocity of circulation increases, and thus a given “real” money supply is capable of financing a larger and larger volume of transactions, X.

It should be readily apparent that the curve corresponding to some other value of M/P, say, a value ra times larger, can be obtained from the M’M’ curve by multiplying by the factor m the ordinate of the M’M’ curve corresponding to any given value of r. It follows that provided ∧r (the abscissa of point a) is to the right of rm there will be some unique value of M/P, say, M∧/P, such that the corresponding MM curve will go through the point of intersection, a, of the other two curves. This unique curve is represented by ∧M∧ in the figure. Thus, with a real money supply M∧/P, the money market as well as the commodity and labor markets will all be simultaneously cleared with the output ∧X and the rate of interest ∧r. But this in turn means that, given the actual money supply M, the price level must tend to the equilibrium level ∧P, such that M/∧P = (M∧/P), or ∧P = M/(M∧/P). Similarly, ∧W = ∧P(W∧/P), where (W∧/P) is the full-employment equilibrium real wage. A higher value of W and P would make the money supply inadequate to transact the full-employment income, unless the rate of interest were higher than r. But a higher r would reduce the aggregate demand below the full-employment supply. This in turn would cause unemployment which, with flexible wages, would lead to a fall of W and hence of P to the equilibrium levels ∧W and ∧P; the converse would be true for values of P and W below the equilibrium levels.

There are three main implications of this analysis to which attention must be called:

(a) Provided ∧R > rm, the only economic effect of M is to determine the price level ∧P; furthermore, it is apparent from the derivation of the last paragraph that ∧P is proportional to M, so that, in this sense, the quantity theory of money holds.

(b) The equilibrium value of P corresponding to a given M depends, not only on full-employment output ∧X, which controls the position of SS, and on slowly changing institutional factors determining the shape of V(r), but also on r As can be seen from Figure 3, the larger ∧r is, the smaller will be the equilibrium real money supply. But ∧r, for a given ∧X, is in turn associated with the position of the aggregate demand relation, DD. A rise in the aggregate demand relation—reflecting an increase in consumption or investment demand or both at each level of income and of the interest rate—will result in an upward shift of the DD curve, and this in turn will move to the right the point of intersection, a, of aggregate demand and supply, increasing its r coordinate. Such a shift is illustrated by the curve D’D’ intersecting SS at a’. If in the face of such a shift the central bank does not force an appropriate contraction in M, excess demand will arise in the commodity and labor markets that will force wages and prices up. This will reduce the real money supply, lowering the MM curve, until a value of P is reached such that MM coincides with M’M’. If the price rise is to be avoided, the central bank must enforce an appropriate reduction in M (in the same proportion in which prices would rise otherwise). We deduce that once liquidity preference is recognized, if the monetary authority is concerned with maintaining the stability of the price level over time—as it must be if a monetary economy is to work smoothly—it must actively manage the money supply, enforcing a (relatively) larger money supply, and a smaller value of r, when demand tends to be slack and a relatively smaller supply, and higher r, when demand tends to be more active.

(c) Suppose, however, that in some period demand is slack and the DD curve is so depressed that it intersects SS at a value of r smaller than rm, as illustrated by D“D” intersecting SS at b in Figure 3. It is then apparent that there can be no possible value of M/P such that the corresponding MM curve will go through b, since regardless of the value of M/P, every MM curve must lie entirely to the right of rm. In this situation, sometimes referred to as “the Keynesian case” or “the special Keynesian case,” the economic system will not have any equilibrium solution (a set of prices and interest rates that can simultaneously equate all demands and supplies). If prices and wages are flexible, they will both tend to fall indefinitely under the pressure of excess supplies. But this fall, which under normal conditions would re-establish equilibrium by shifting MM up, can now never prove sufficient. By the same token, monetary policy also breaks down: there is no feasible expansion of the money supply sufficient to eliminate the excess supply of goods and labor.

Thus, from liquidity preference Keynes was able to derive the important and novel result that under certain conditions an economy using a token money may simply break down, having no maintainable position of equilibrium (except through government fiscal policy or wage rigidity, which will be discussed below).

Positive national debt. The government debt, G, may consist of interest-bearing instruments (government bonds) or government fiat money or both, circulating along with or instead of the money created by the banking system. In any case, if G is positive, it is apparent from equation (1’) that aggregate demand depends not only on r but also on P. In terms of Figure 3, equation (1’) must now be represented by a family of curves, one for each value of P. For the sake of illustration, suppose that the curve DD in the figure corresponds to the received price level, P0. It can readily be established that to a different value of P, say, P1 < P0 , there will correspond a new DD curve higher and to the right, such as D’D’. This is because a fall in P will increase the real value of the government debt held by the public and hence the real net worth of the private sector. This in turn will tend to increase consumption demand, and hence total demand, for any given r. Conversely, a rise in P will shift DD downward and to the left.

This dependence of aggregate demand on P when G is not zero has come to be known in the literature as the “real-balance effect,” and also as the “Pigou effect” because Pigou called attention to it in a very influential work (1947). However, the point had been made earlier by others, in particular by Scitovsky (1940). The main implication of the real-balance effect is that even with flexible wages the system will in general have a position of full-employment equilibrium. In other words, it rules out the possibility of the “Keynesian case” discussed above. To illustrate this point, suppose that corresponding to the received price, P0, the aggregate demand function had the position D“D” in the figure, which could not possibly intersect an MM curve on SS. Since the position of DD now depends on P, as P falls under the pressure of excess supply the DD curve will keep shifting to the right at the same time that MM shifts upward. Except under very special ad hoc assumptions, MM and DD will eventually intersect on SS at some point to the right of rm.

This demonstration that, provided G is positive, a system with flexible wages will possess a position of full-employment equilibrium, contrary to Keynes’s conclusion, has been seized upon by some of Keynes’s critics as disposing of one of his most significant and novel results. They have concluded that underemployment equilibrium can arise only from wage-price rigidities. This view must be regarded as unwarranted, mainly for the following reasons: (a) Keynes’s conclusion stands when G = 0. (b) Even when G < 0, the conclusion that a full-employment solution would exist is valid only under the assumption, implicit in the model, that falling prices do not generate perverse expectations of further falls, which would reduce demand. Furthermore, it ignores the likelihood that a violent deflation, which might be necessary to produce a sufficient increase in the real value of the national debt, would severely disrupt a monetary economy by producing wholesale debtors’ insolvency. In view of these considerations, Pigou’s demonstration has little practical relevance, as Pigou himself acknowledged ([1947] 1951, p. 251). Even if full employment could be re-established by sufficient deflation of prices and wages, it would be preferable to avoid this outcome by relying on the kind of fiscal policy devices, discussed in the next section, that one would have to fall back on when G = 0. To look at the matter in a slightly different light, wage and price rigidity, instead of hindering the working of a monetary economy, may provide it with a degree of price stability that in the long run contributes to its smooth working, even though this rigidity makes the task of successful monetary management more challenging.

Downward wage rigidity

The working of the system when the level of the rigid wage W0 is sufficiently high to be at least potentially effective can be illustrated by Figure 4, which is a simple variant of Figure 3. For this purpose it is convenient to

introduce a new symbol to denote money income, Y, which is related to other variables of the system by the identity

Also, for the sake of exposition we deal formally with the case G = 0, with some occasional reference to the (rather minor) modifications called for if this restriction is discarded.

We recall that with G = 0 the right-hand side of (1’) contains only the variable r. From (1’),

(2), and (3) we can then derive a relation between Y/W and r. Here Y/W is income measured in what Keynes called wage units (that is, income measured in terms of labor as a numéraire). We first solve equation (3) for P/W in terms of X and write the solution as

a “Marshallian” short-run supply function indicating the price—in terms of the cost of labor—needed to call forth a given supply, X. Next, using (1’) and (2), we can express X as a function of r. It follows that Y/W can itself be expressed as a function of r—say, . This equation is an obvious variant of the aggregate-demand relation (1’), shown as DD in Figure 3, except that output is expressed in wage units. Accordingly, its graph, shown by the yy curve of Figure 4, bears a close relation to that of DD in Figure 3, from which it differs only by the factor P/W. In particular, yy must fall from left to right if DD does, since P/W is an increasing function of X. The horizontal line ss again represents “full-employment output” in wage units, (P∧/W)∧X, where P∧/W and ∧X can be inferred from the solution of the system under flexible wages. The portion of yy above ss has been dashed to indicate that it can never be “effective,” since real income there exceeds the full-employment level.

The curve rising from left to right and labeled ∧M/W0 is again derived from the money market equation (7’), on the assumption that the given money supply is M. First solve (7’) for PX, obtaining PX = V(r)M. Next replace M with ∧M and divide both sides by W0. This yields . Its graph must look like that of MM in Figure 3, for it is again the graph of V(r) up to a proportionality factor M/W0 (instead of M/P, as in Figure 3). It shows the level of income (in wage units) that can be transacted at each level of r, given the money supply in wage units.

If the yy curve and the money-market curve intersect in their effective range—below or on ss— as in Figure 4, then the coordinates of their point of intersection, labeled a, show the equilibrium value of income, ­Y/W, and of the rate of interest, ­r. If this intersection does not fall on ss, then the equilibrium is one of less than full employment. It is a position of equilibrium despite the presence of unemployment because, under wage rigidity, the excess supply of labor does not lead to any further adjustment (at least in the short run). By contrast, if wages were flexible, the excess supply would bid down W which, with ­M given, would raise ­M/W, shifting the MM curve upward and moving its point of intersection with yy upward and to the left until it coincided with the full-employment point, a’ in the figure. (If G is assumed to be positive, the fall in W will also tend to shift yy upward, moving a’ to the right.)

Even under wage rigidity, output and employment could be increased by increasing the money supply, which, with W given at W0, would raise M/W and hence the MM curve. In fact, provided that a’ is to the right of rm, there is an ideal money supply, $, that produces an MM curve that intersects yy at a’. Alternatively, the goal of optimal monetary policy might be visualized as that of en-forcing the rate of interest that would generate an aggregate demand equal to full-employment output, supplying whatever quantity of money is needed to enforce that rate. In terms of Figure 4, the rate of interest called for is, of course, ∧ (which is the r coordinate of a’), and the corresponding quantity of money is again ∧M.

This analysis should help to show how the interaction of liquidity preference and wage rigidity makes the task of economic stabilization through monetary policy a highly complex and difficult one. In the absence of wage-price rigidities the concern of monetary policy would be reduced to the maintenance of price stability. And in the absence of liquidity preference the velocity of circulation could be counted upon to be sufficiently stable to make this task a relatively easy one. In a stationary economy it would call essentially for a stable money supply, whereas in an expanding economy it would call for a money supply that keeps pace with the growth of full-employment output, a growth that also appears to be characterized by a fair degree of stability.

But under wage rigidity, monetary policy has the double task of trying to achieve both price stability and full employment. Furthermore, because liquidity preference causes the velocity of circulation to vary with interest rates, the money supply needed to reach these goals will vary, relative to full-employment output, with variations in aggregate demand conditions. In terms of Figure 4, a rise in consumption or investment demand relative to income will shift the yy curve to the right; a corresponding fall will shift it to the left. These shifts have to be countered by contrary adjustments of the money supply relative to the level of full-employment output. Furthermore, failure to adjust the money supply properly will tend to have asymmetrical consequences. An excessive money supply will still give rise to increases in prices that could have been avoided and that are largely irreversible. But too small a money supply will result in an insufficient aggregate demand that, aside from deflationary effects on the price level, will result in the waste and social scourge of unemployment.

Note also that the central bank’s control over the price level is at best partial and largely unidirectional. The price level is anchored to the wage rate, which monetary policy can readily push up by being too expansive but which it can hardly hope to force down, except possibly through the painful and wasteful route of prolonged and wide-spread unemployment. Furthermore, if the minimum money wage—the W0 of equation (5) — tends to be pushed up even before full employment is reached, whether through powerful unions or through partial bottlenecks or both, and if the rise tends to exceed the rate of increase of productivity, then monetary (as well as fiscal) policy will be faced with the unsavory choice between “creeping inflation” and chronic unemployment. Whether this dilemma is in fact a serious and real one revolves around the issue of the determinants of the overall level of money wages, an issue that the Keynesian analysis has opened up but that is still far from settled. [SeeInflation And Deflation; see also Phillips 1958.]

One other implication of the Keynesian framework, which can be only touched upon in this survey dealing primarily with monetary aspects, is that fiscal policy provides an alternative approach to the control of aggregate demand for economic stabilization [see Fiscal Policy]. Fiscal policy can be accommodated in our macroeconomic model by adding government expenditure on goods and services as a component of aggregate demand in equation (1’), making consumption (and possibly investment) depend on taxes as well as on income produced, and adding an equation relating tax collection to income and tax rates. Without attempting to pursue this line here we may indicate that, in terms of Figure 4, fiscal policy—defined as policy concerned with the level of government expenditure and taxation—will affect the position and shape of the aggregate demand relation yy. An increase in expenditure will shift it upward and to the right; an increase in tax rates will shift it in the opposite direction. Thus, given a position of less than full employment equilibrium such as a in Figure 4, output and employment could be raised toward or up to the full-employment level by increased government expenditure, tax reductions, or both, which would shift the yy curve to the right.

The possibility of affecting equilibrium output and employment through fiscal tools becomes of critical importance in the special “Keynesian case,” in which the aggregate demand is so depressed that the yy curve intersects ss to the left of the minimum achievable interest rate, rm. In this case (illustrated by the curve y’y’ in Figure 4) full employment, as we have seen, is beyond the reach of monetary policy, for no money curve can have points to the left of rm. Fiscal policy is then the only effective tool of stabilization policy, at least until the yy curve has been shifted rightward enough to cut ss to the right of rm.

Beyond this point—and, more generally, whenever the intersection of yy and ss is to the right of rm—either fiscal or monetary tools can be used in the pursuit of full employment and price stability. Of course, both tools can be used simultaneously and in coordinated fashion. This should be clear from the fact that, in terms of Figure 4, fiscal policy acts basically on yy whereas monetary policy acts basically on the money curve. (The graphical apparatus of Figure 4 was chosen to illustrate the working of the system partly because of its convenience in isolating the modus operandi of monetary and fiscal policy.)

There is a substantial literature concerned with the analysis of the relative advantages and shortcomings of monetary and fiscal policy in terms of such criteria as reliability, response delays, ease of implementation, and reversibility (for monetary policy, see Johnson 1962, pp. 365-377; for fiscal policy see, e.g., Reiser 1964, part 5), effects on long-run economic growth (e.g., Smith 1957; Modigliani 1961), and, more recently, differential impact on the balance of payments (e.g., Mundell 1962). Because of the complexity of the problem it is not surprising that there have been substantial differences in points of view between economists favoring the use of one tool, or of some specific mix, and those favoring others. These differences can be traced in part to differences in the subjective valuation of different goals. But in part they revolve around disagreement about the empirical importance of the “Keynesian case” in which monetary policy becomes powerless to maintain or reestablish full employment, either because it is ineffective in reducing interest rates any further (at least in the short run) or because the achievable reduction in interest rates is insufficient to induce the required expansion in investment and aggregate demand.

Liquidity preference, that is, the proposition that the demand for money is systematically and significantly affected by interest rates, has proved to be a major, lasting contribution to economic analysis, well supported by empirical evidence. From an analytical point of view its great significance lies in the implication that under certain conditions— the “special Keynesian case” —even an economy with flexible wages and prices might not possess a stable full-employment equilibrium. But beyond this fundamental theoretical contribution, the dramatic impact of the General Theory on economic theory and policy can be traced to its insightful analysis of the role of liquidity preference in a world of widespread wage and price rigidities. This analysis has led to a new understanding and fundamental reappraisal of the role of money and of the tasks and limitations of monetary and fiscal policy.

With downward wage rigidity (and even ignoring international trade) money cannot be regarded, even in first approximation, as “neutral,” a mere veil having no effect on the economy other than the determination of the price level, except possibly when the money supply is excessive. Under conditions of less than full employment due to lack of demand, and barring the special Keynesian case, monetary policy plays a crucial role in the determination of income and employment. In the special Keynesian case, on the other hand, monetary policy breaks down, since it is incapable, at least in the short run, of affecting either output or prices.

Franco Modigliani


Allais, Maurice 1947 Économic & intérét: Présentation nouvelle des problémes fondamentaux relatifs au ròle économique du taux de I’intérét et de leurs solutions. 2 vols. Paris: Librairie des Publications Officielles.

Baumol, William J. 1952 The Transactions Demand for Cash: An Inventory Theoretic Approach. Quarterly Journal of Economics 66:545–556.

Bronfenbrenner, Martin; and Holzman, Franklyn D. 1963 Survey of Inflation Theory. American Economic Review 53:593–661.

Bronfenbrenner, Martin; and Mayer, Thomas 1960 Liquidity Functions in the American Economy. Econometrica 28:810–834.

Brunner, K.; and Meltzer, Allan H. 1964 Some Further Investigations of Demand and Supply Functions for Money. Journal of Finance 19:240–283.

Eshag, Eprime 1963 From Marshall to Keynes: An Essay on the Monetary Theory of the Cambridge School. Oxford: Blackwell.

Fisher, Irving (1911) 1920 The Purchasing Power of Money: Its Determination and Relation to Credit, Interest and Crises. New ed., rev. New York: Macmillan.

Friedman, Milton (editor) 1956 Studies in the Quantity Theory of Money. Univ. of Chicago Press. → See especially pages 5-21, “The Quantity Theory of Money—A Restatement,” by Friedman.

Friedman, Milton 1957 A Theory of the Consumption Function. National Bureau of Economic Research, General Series, No. 63. Princeton Univ. Press.

Friedman, Milton 1959 The Demand for Money: Some Theoretical and Empirical Results. Journal of Political Economy 67:327–351.

Hicks, John R. (1935) 1951 A Suggestion for Simplifying the Theory of Money. Pages 13-32 in American Economic Association, Readings in Monetary Theory. Philadelphia: Blakiston.

Hicks, John R. 1937 Mr. Keynes and the “Classics”: A Suggested Interpretation. Econometrica 5:147–159.

Johnson, H. G. 1962 Monetary Theory and Policy. American Economic Review 52:335–384.

Reiser, Norman F. 1964 Macroeconomics, Fiscal Policy, and Economic Growth. New York: Wiley.

Keynes, John Maynard (1930) 1958–1960 A Treatise on Money. 2 vols. London: Macmillan. → Volume 1: The Pure Theory of Money. Volume 2: The Applied Theory of Money.

Keynes, John Maynard 1936 The General Theory of Employment, Interest and Money. London: Macmillan. → A paperback edition was published in 1965 by Harcourt.

Lange, Oskar 1938 The Rate of Interest and the Optimum Propensity to Consume. Economica New Series 5:12-32.

Latane, Henry A. 1954 Cash Balances and the Interest Rate: A Pragmatic Approach. Review of Economics and Statistics 36:456–460.

Latane, Henry A. 1960 Income Velocity and Interest Rates: A Pragmatic Approach. Review of Economics and Statistics 42:445–449.

Lavington, Frederick 1921 The English Capital Market. London: Methuen.

Marget, Arthur W. 1938 The Theory of Prices: A Re-examination of the Central Problems of Monetary Theory. Vol. 1. Englewood Cliffs, N.J.: Prentice-Hall.

Marshall, Alfred 1923 Money, Credit and Commerce. London: Macmillan.

Meltzer, Allan H. 1963 The Demand for Money: Th Evidence From the Time Series, Journal of Political Economy 71:219–246.

Modigliani, Franco (1944) 1951 Liquidity Preference and the Theory of Interest and Money. Pages 186–239 in American Economic Association, Readings in Monetary Theory. New York: Blakiston. → First published in Volume 12 of Econometrica.

Modigliani, Franco 1961 Long Run Implications of Alternative Fiscal Policies and the Burden of the National Debt. Economic Journal 71:730–755.

Modigliani, Franco 1963 The Monetary Mechanism and Its Interaction With Real Phenomena. Review of Economics and Statistics 45 (Supplement): 79-107.

Mundell, Robert A. 1962 The Appropriate Use of Monetary and Fiscal Policy for Internal and External Stability. International Monetary Fund, Staff Papers 9:70-77.

Patinkin, Don (1956) 1965 Money, Interest, and Prices: An Integration of Monetary and Value Theory. 2d ed. New York: Harper.

Phillips, A. W. 1958 The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom: 1861–1957. Economica New Series 25:283–299.

Pigou, A. C. (1917) 1951 The Value of Money. Pages 162–183 in American Economic Association, Readings in Monetary Theory. Philadelphia: Blakiston.

Pigou, A. C. (1947) 1951 Economic Progress in a Stable Environment. Pages 241–251 in American Eco nomic Association, Readings in Monetary Theory. Philadelphia: Blakiston. → First published in Volume 14 of Economica New Series.

Pigou, A. C. 1950 Keynes1’s General Theory: A Retrospective View. London: Macmillan.

Scitovsky, Tibor 1940 Capital Accumulation, Employment and Price Rigidity. Review of Economic Studies 8:69-88.

Smith, Warren L. 1957 Monetary-Fiscal Policy and Economic Growth. Quarterly Journal of Economics 71:36-55.

Tobin, James 1956 The Interest-elasticity of Transactions Demand for Cash. Review of Economics and Statistics 38:241–247.

Tobin, James 1958 Liquidity Preference as Behavior Towards Risk. Review of Economic Studies 25, no. 2: 65-86.

Turvey, Ralph (1960) 1961 Interest Rates and Asset Prices. New York: Macmillan.