Liquidity preference as behavior Towards risk One of the basic functional relationships in the Keynesian model of the economy is the liquidity preference schedule, an inverse relationship between the demand for cash balances and the rate of interest. This aggregative function must be derived from some assumptions regarding the behavior of the decision-making units of the economy, and those assumptions are the concern of this paper. Nearly two decades of drawing downward loping liquidity preference curves in textbooks and on classroom blackboards should not blind us to the basic implausibility of the behavior they describe. Why should anyone hold the non-interest bearing obligations of the government instead of its interest bearing obligations? The apparent irrationality of holding cash is the same, moreover, whether the interest rate is 6%, 3%or i of 1%o. What needs to be explained is not only the existence of a demand for cash when its yield is less than the yield on alternative assets but an inverse relationship between the aggregate demand for cash and the size of this differential in Two kinds of reasons for holding cash are usually distinguished: transactions reasons and investment reasons I Transactions balances: size and composition. No economic unit--firm or house- hold or government-enjoys perfect synchronization between the seasonal patterns of its flow of receipts and its flow of expenditures. The discrepancies give rise to balances which accumulate temporarily, and are used up later in the year when expenditures catch up. Or, to put the same phenomenon the other way, the discrepancies give rise to the need for balances to meet seasonal excesses of expenditures over receipts. These balances re transactions balances. The aggregate requirement of the economy for such balances depends on the institutional arrangements that determine the degree of synchronization between individual receipts and expenditures. Given these institutions, the need for transactions balances is roughly proportionate to the aggregate volume of transactions The obvious importance of these institutional determinants of the demand for tra ctions balances has led to the general opinion that other possible determinants, including 1 I am grateful to Challis Hall, Arthur Okun, Walter Salant, and Leroy Wehrle for helpful earlier drafts of this paper. in a world involving no transaction friction and no uncer for a spread between the yield on any two assets, and hence demand ba it clear that li
Liquidity Preference as Behavior Towards Risk' One of the basic functional relationships in the Keynesian model of the economy is the liquidity preference schedule, an inverse relationship between the demand for cash balances and the rate of interest. This aggregative function must be derived from some assumptions regarding the behavior of the decision-making units of the economy, and those assumptions are the concern of this paper. Nearly two decades of drawing downwardsloping liquidity preference curves in textbooks and on classroom blackboardshould not blind us to the basic implausibility of the behavior they describe. Why should anyone hold the non-interest bearing obligations of the government instead of its interest bearing obligations ? The apparent irrationality of holding cash is the same, moreover, whether the interest rate is 6 %, 3 % or i of 1 %. What needs to be explained is not only the existence of a demand for cash when its yield is less than the yield on alternative assets but an inverse relationship between the aggregate demand for cash and the size of this differential in yields.2 1. Transactions balances and investment balances. Two kinds of reasons for holding cash are usually distinguished: transactions reasons and investment reasons. 1.1 Transactions balances: size and composition. No economic unit-firm or household or government-enjoys perfect synchronization between the seasonal patterns of its flow of receipts and its flow of expenditures. The discrepancies give rise to balances which accumulate temporarily, and are used up later in the year when expenditures catch up. Or, to put the same phenomenon the other way, the discrepancies give rise to the need for balances to meet seasonal excesses of expenditures over receipts. These balances are transactions balances. The aggregate requirement of the economy for such balances depends on the institutional arrangements that determine the degree of synchronization between individual receipts and expenditures. Given these institutions, the need for transactions balances is roughly proportionate to the aggregate volume of transactions. The obvious importance of these institutional determinants of the demand for transactions balances has led to the general opinion that other possible determinants, including 1 I am grateful to Challis Hall, Arthur Okun, Walter Salant, and Leroy Wehrle for helpful comments on earlier drafts of this paper. 2" .... in a world involving no transaction friction and no uncertainty, there would be no reason for a spread between the yield on any two assets, and hence there would be no difference in the yield on money and on securities . . . in such a world securities themselves would circulate as money and be acceptable in transactions; demand bank deposits would bear interest, just as they often did in this country in the period of the twenties." Paul A. Samuelson, Foundations of Economic Analysis (Cambridge: Harvard University Press, 1947), p. 123. The section, pp. 122-124, from which the passage is quoted m s it clear that liquidity preference must be regarded as an explanation of the existence and level not of the interest rate but of the differential between the yield on money and the yields on other assets. 65
REVIEW OF ECONOMIC STUDIES interest rates, are negligible. This may be true of the size of transactions balances, but the composition of transactions balances is another matter. Cash is by no means the only asset in which transactions balances may be held. Many transactors have large enough balances so that holding part of them in earning assets, rather than in cash, is a relevant possibility. Even though these holdings are always for short periods, the interest earnings may be worth the cost and inconvenience of the financial transactions involved Elsewhere 2 I have shown that, for such transactors, the proportion of cash in transactions balances varies inversely with the rate of interest; consequently this source of interest asticity in the demand for cash will not be further discussed here. 1.2 Investment balances and portfolio decisions. In contrast to transactions balances the investment balances of an economic unit are those that will survive all the expected easonal excesses of cumulative expenditures over cumulative receipts during the year ahead. They are balances which will not have to be turned into cash within the year. Consequently the cost of financial transactions--converting other assets into cash and vi ersa-does not operate to encourage the holding of investment balances in cash. cash is to have any part in the composition of investment balances, it must be because of expectat fears of loss on other assets. It is here in what Keynes called the specu- tive motives of investors, that the explanation of liquidity preference and of the interest- elasticity of the demand for cash has been sought The alternatives to cash considered, both in this paper and in prior discussions of the ubject, in examining the speculative motive for holding cash are assets that differ from ash only in having a variable market yield. They are obligations to pay stated cash amounts at future dates, with no risk of default. They are, like cash, subject to changes in real value due to fluctuations in the price level. In a broader perspective, all these assets, including cash, are merely minor variants of the same species, a species monetary assets--marketable, fixed in money value, free of default risk. The differences of members of this species from each other are negligible compared to their differences from the vast variety of other assets in which wealth may be invested corporate stocks, real estate unincorporated business and professional practice, etc. The theory of liquidity preference does not concern the choices investors make between the whole species of monetary assets, on the one hand, and other broad classes of assets, on the other. 4 Those choices are the concern of other branches of economic theory, in particular theories of investment and of consumption. Liquidity preference theory takes as given the choices determining how much wealth is to be invested in monetary assets and concerns itself with he allocation of these amounts among cash and alternative monetary assets I The tradi the velocity of money has, however, probably exaggerated the true that such institutions as the degree of response to money rates. For example, there is a good deal of in the prompt ty with which bills are rendered and settled. 3 Costs of financial transactions have the effect of dete changes if the investor were starting over agaitimal even when a different composition n如二配中 4 For an attempt by the author to apply to this wider choice some of the same
66 REVIEW OF ECONOMIC STUDIES interest rates, are negligible.' This may be true of the size of transactions balances, but the composition of transactions balances is another matter. Cash is by no means the only asset in which transactions balances may be held. Many transactors have large enough balances so that holding part of them in earning assets, rather than in cash, is a relevant possibility. Even though these holdings are always for short periods, the interest earmngs may be worth the cost and inconvenience of the financial transactions involved. Elsewhere2 Ihave shown that, for such transactors, the proportion of cash in transactions balances varies inversely with the rate of interest ; consequently this source of interestelasticity in the demand for cash will not be further discussed here. 1.2 Investment balances and portfolio decisions. In contrast to transactions balances, the investment balances of an economic unit are those that will survive all the expected seasonal excesses of cumulative expenditures over cumulative receipts during the year ahead. They are balances which will not have to be turned into cash within the year. Consequently the cost of financial transactions-converting other assets into cash and vice versa-does not operate to encourage the holding of investment balances in cash.3 If cash is to have any part in the composition of investment balances, it must be because of expectations or fears of loss on other assets. It is here, in what Keynes called the speculative motives of investors, that the explanation of liquidity preference and of the interestelasticity of the demand for cash has been sought. The alternatives to cash considered, both in this paper and in prior discussions of the subject, in examining the speculative motive for holding cash are assets that differ from cash only in having a variable market yield. They are obligations to pay stated cash amounts at future dates, with no risk of default. They are, hke cash, subject to changes in real value due to fluctuations in the price level. In a broader perspective, all these assets, including cash, are merely minor variants of the same species, a species we may call monetary assets-marketable, fixed in money value, free of default risk. The differences of members of this species from each other are negligible compared to their differences from the vast variety of other assets in which wealth may be invested: corporate stocks, real estate, unincorporated business and professional practice, etc. The theory of liquidity preference does not concem the choices investors make between the whole species of monetary assets, on the one hand, and other broad classes of assets, on the other.4 Those choices are the concern of other branches of economic theory, in particular theories of investment and of consumption. Liquidity preference theory takes as given the choices determining how much wealth is to be invested in monetary assets and concerns itself with the allocation of these amounts among cash and alternative monetary assets. " The traditional theory of the velocity of money has however, probably exaggerated the invariance of the institutions determining the extent of lack of syronizaton between individual receipts and expenditur. It is no doubt true that such institutions as the degree of vertical integration of production and the periodicity of wage, salary, dividend, and tax payments are slow to change. But other relevant arrangements can be adjusted in response to money rates. For example, there is a good deal of flexibility in the promptness and regularity with which bills are rendered and settled. * "; The Interest Elasticity of the Transactions Demand for Cash," Review of Economics and Statistics, vol. 38 (August 1956), pp. 241-247. ' Costs of financial transactions have the effect of deterring changes from the existing portfolio, whatever its composition; they may thus operate agaimst the holding of cash as easily as for it. Because ofthese costs, the status quo may be optimal even when a different composition of assets would be preferred if the investor were starting over again. 4 For an attempt by the author to apply to this wider choice some of the same theoretical tools that are here used to analyze choices among the narrow class of monetary assets, see" A Dynamic Aggregative Model ", Journal of Political Economy, vol. 63 (April 1955), pp. 103-115
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK Why should any investment balances be held in cash, in preference to other monetary assets? We shall distinguish two possible sources of liquidity preference, while recognizing that they are not mutually exclusive. The first is inelasticity of expectations of future nterest rates. The second is uncertainty about the future of interest rates. These two sources of liquidity preference will be examined in turn. 2. Inelasticity of interest rate expectations 2.1 Some simplifying assumptions. To simplify the problem, assume that there is only one monetary asset other than cash, namely consols. The current yield of consols in perpetuity. The yield of cash is assumed to be zero; however, this is not essential, as it is the current and expected differentials of consols over cash that matter. An investor with a given total balance must decide what proportion of this balance to hold in cash A1, and what proportion in consols, A2. This decision is assumed to fix the portfolio for year 2.2 Fixed expectations of future rate. At the end of the year, the investor expects the rate on consols to be re. This expectation is assumed, for the present, to be held with certainty and to be independent of the current rate r. The investor may therefore expect with certainty that every dollar invested in consols today will earn over the year ahead not only the interest Sr, but also a capital gain or loss g For this investor, the division of his balance into proportions A, of cash and A2 of consols is a simple all-or-nothing choice. If the current rate is such that r+g is greater than zero, then he will put everything in consols. But if r+g is less than zero, he will put everything in cash. These conditions can be expressed in terms of a critical level of the current rate re, where (22) At current rates above re, everything goes into consols; but for r less than re, everythin goes into cash 2.3 Sticky and certain interest rate expectations. So far the investor's expected interest rate re has been assumed to be completely independent of the current rate r. This assum- ption can be modified so long as some independence of the expected rate from the current rate s maintained. In Figure 2.1, for example, re is shown as a function of r, namely p(r) has only one intersection with the 45 line, and at this intersection its slope( +o) is less than one. If these conditions are met, the intersection determines a critical rate re such that if r exceeds re the investor holds no cash, while if r is less than re he holds no consols As noted above, it is the costs of financial transactions that impart inertia to portfolio Every reconsideration of the portfolio involves the investor in ex money. The frequency with which it is worth while to review inecisioatsione ma dn a ging ie isaac ia ran nsactions. Thus the releval od of time, a portfolio is never so irrevocably frozen that there are n conceivable events during the period which would induce the he fact that composition due to, the costs of transactions and of decision makino g account of the inertia in portfolio sed in this paper is, therefore, not a wholly satisfactory way of
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK 67 Why should any investment balances be held in cash, in preference to other monetary assets ? We shall distinguish two possible sources of liquidity preference, while recognizing that they are not mutually exclusive. The first is inelasticity of expectations of future interest rates. The second is uncertainty about the future of interest rates. These two sources of liquidity preference will be examined in turn. 2. Inelasticity of interest rate expectations. 2.1 Some simplifying assumptions. To simplify the problem, assume that there is only one monetary asset other than cash, namely consols. The current yield of consols is r per " year ". $1 invested in consols today will purchase an income of $r per " year " in perpetuity. The yield of cash is assumed to be zero ; however, this is not essential, as it is the current and expected differentials of consols over cash that matter. An investor with a given total balance must decide what proportion of this balance to hold in cash, A1, and what proportion in consols, A2. This decision is assumed to fix the portfolio for a full " year ".1 2.2 Fixed expectations offuture rate. At the end of the year, the investor expects the rate on consols to be re. This expectation is assumed, for the present, to be held with certainty and to be independent of the current rate r. The investor may therefore expect with certainty that every dollar invested in consols today will earn over the year ahead not only the interest $r, but also a capital gain or loss g (2.1) g r 1 re For this investor, the division of his balance into proportions A1 of cash and A2 of consols is a simple all-or-nothing choice. If the current rate is such tfhat r + g is greater than zero, then he will put everything in consols. But if r + g is loss than zero, he will put everything in cash. These conditions can be expressed in terms of a critical level of the current rate rc, where: (2.2) rc - _re 1 + re At current rates above rc, everything goes into consols ; but for r less than rc, everything goes into cash. 2.3 Sticky and certain interest rate expectations. So far the investor's expected interestrate re has been assumed to be completely independent of the current rate r. This assumption can be modified so long as some independence of the expected rate from the current rate is maintained. In Figure 2.1, for example, re is shown as a function of r, namely p(r). Correspondingly 1 is a function of r. As shown in the figure, this function 1 +l has only one intersection with the 450 line, and at this intersection its slope (1 is less than one. If these conditions are met, the intersection determines a critical rate rc such that if r exceeds rc the investor holds no cash, while if r is less than rc he holds no consols. 2 As noted above, it is the costs of financial transactions that impart inertia to portfolio composition. Every reconsideration of the portfolio involves the investor in expenditure of time and effort as well as of money. The frequency with which it is worth while to review the portfolio will obviously vary with the investor and will depend on the size of his portfolio and on his situation with respect to costs of obtaining information and engaging in financial: transactions. Thus the relevant " year " ahead for which portfolio decisions are made is not the same for all investors. Moreover, even if a decision is made with a view to fixing a portfolio for a given period of time, a portfolio is never so irrevocably frozen that there are no conceivable events during the period which would induce the investor to reconsider. The fact that this possibility is always open must influence the investor's decision. The fiction of a fixed investment period used in this paper is, therefore, not a wholly satisfactbry way of taking account of the inertia in portfolio composition due to the costs of transactions and of decision making
REVIEW OF ECONOMIC STUDIES (r) 点 CHto a csamt FIGURE 2.1 Stickiness in the Relation between Expected and Current Interest Rate. 2.4 Diferences of opinion and the aggregate demand for cash. According to this model, the relationship of the individual's investment demand for cash to the current rate of interest would be the discontinuous step function shown by the heavy vertical lines LMNW in Figure 2.2 How then do we get the familiar Keynesian liquidity preference function, a smooth, continuous inverse relationship between the demand for cash and the rate of interest? For the economy as a whole, such a relationship can be derived from individual behaviour of the sort depicted in Figure 2.2 by assuming that individual investors diffei in their criticalratesre. Suchan aggregate relationship is shown in Figure 2.3 FIGURE 2.2 Individual Demand for Cash Assuming Certain but Inelastic Interest Rate Expectations
68 REVIEW OF ECONOMIC STUDIES *~~~~~~~~~~~~r .b(a 1+1 1+ FiGuRE 2.1 Stickiness in the Relation between Expected and Current Interest Rate. 2.4 Differences of opinion and the aggregate demand for cash. According to this model, the relationship of the individual's investment demand for cash to the current rate of interest would be the discontinuous step function shown by the heavy vertical lines LMNW in Figure 2.2 How then do we get the familiar Keynesian liquidity preference function, a smooth, continuous inverse relationship between the demand for cash and the rate of interest ? For the economy as a whole, such a relationship can be derived from individual behaviour of the sort depicted in Figure 2.2 by assuming that inaividual investors diffei in their criticalrates r. Suchan aggregate relationship is shown in Figure 2.3. L \A M~~~~~~~~~~~~~~~~ \ r _ IiGuRt 2.2 Inldividual Demand for Cash Assumiing Certain but Inelastic Interest Rate Expectations
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK 一一一 FIGURE 2.3 Aggregate Demand for Cash Assuming Differences Among Individuals in Interest Rate Expectations. At actual rates above the maximum of individual critical rates the aggregate demand for cash is zero, while at rates below the minimum critical rate it is equal to the total invest ment balances for the whole economy. Between these two extremes the demand for cash Figure 2. 3. The demand for cash at r is the total of investment balances controlled by investors whose critical rates rc exceed r. Strictly speaking, the curve is a step function; out, if the number of investors is large, it can be approximated by a smooth curve. Its shape depends on the distribution of dollars of investment balances by the critical rate of the investor controlling them the shape of the curve in figure 2. 3 follows from a uni modal distribution 2.5 Capital gains or losses and open market operations. In the foregoing the size of investment balances has been taken as independent of the current rate on r. This is not the case if there are already consols outstanding. Their value will depend inversely on the current rate of interest. Depending on the relation of the current rate to the previously fixed coupon on consols, owners of consols will receive capital gains or losses. Thus the investment balances of an individual owner of consols would not be constant at w but would depend on r in a manner illustrated by the curve ABC in Figure 2.2.1 Similarly, the investment balances for the whole economy would follow a curve like ABC in Figure 2.3, instead of being constant at 2w. The demand for cash would be described by LMBC in both figures Co gly the demand for consols at any interest rate would be described by the horizontal distance between LMBC and ABC. The value of consols goes to infinity as the rate of interest approaches zero for this reason, the curve BC may never reach the horizontal axis. The size of investment balances would be bounded if the monetary assets other than cash consisted of bonds with definite maturities rather than consols According to this theory, a curve like LMBC depicts the terms on which a central bank can engage in open-market operations, given the claims for future payments outstanding in the form of bonds or consols. The curve tells what the quantity of cash must be in order for the central bank to establish a particular interest rate. However, the curve will in holdings of assets other reflect at least in part these cap
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK 69 L I I minr FIGURE 2.3 Aggregate Demand for Cash Assuming Differences Among Individuals in Interest Rate Expectations. At actual rates above the maximum of individual critical rates the aggregate demand for cash is zero, while at rates below the minimum critical rate it is equal to the total investment balances for the whole economy. Between these two extremes the demand for cash varies inversely with the rate of interest r. Such a relationship is shown as LMNEW in Figure 2.3. The demand for cash at r is the total of investment balances controlled by investors whose critical rates r, exceed r. Strictly speaking, the curve is a step function; but, if the number of investors is large, it can be approximated by a smooth curve. Its shape depends on the distribution of dollars of investment balances by the critical rate of the investor controlling them; the shape of the curve in Figure 2.3 follows from a unimodal distribution. 2.5 Capital gains or losses and open market operations. In the foregoing analysis the size of investment balances has been taken as independent of the current rate on consols r. This is not the case if there are already consols outstanding. Their value will depend inversely on the current rate of interest. Depending on the relation of the current rate to the previously fixed coupon on consols, owners of consols will receive capital gains or losses. Thus the investment balances of an individual owner of consols would not be constant at W but would depend on r in a manner illustrated by the curve ABC in Figure 2.2.1 Similarly, the investment balances for the whole etonomy would follow a curve like ABC in Figure 2.3, instead of being constant at SW. The demand for cash would then be described by LMBC in both figures. Correspondingly the demand for. consols at any interest rate would be described by the horizontal distance between LMBC and ABC. The value of consols goes to infinity as the rate of interest approaches zero; for this reason, the curve BC may never reach the honrzontal axis. The size of investment balances would be bounded if the monetary assets other than cash consisted of bonds with definite maturities rather than consols. According to this theory, a curve like LMBC depicts the terms on which a central bank can engage in open-market operations, given the claims for, future payments outstanding in the form of bonds or consols. The curve tells what the quantity of cash must be in order for the central bank to establish a particular interest rate. However, the curve will 1 The size of their investment balances, held in cash and consols may not vary by the full amount of these changes in wealth ; some part of the changes may be reflected in holdings of assets other than monetary assets. But presumably the size of investment balances will reflect at least in part these capital gains and losses
REVIEW OF ECONOMIC STUDIES be shifted by open market operations themselves, since they will change the volume of outstanding bonds or consols. For example, to establish the rate at or below min res the central bank would have to buy all outstanding bonds or consols. The size of the community s investment balances would then be independent of the rate of interest it would be represented by a vertical line through, or to the right of, B, rather than the curve ABC. Thus the new relation between cash and interest would be a curve lying above LMB, of the same general contour as LMN2w 2.6 Keynesian theory and its critics. I believe the theory of liquidity preference I have just presented is essentially the original Keynesian explanation. The general Theory suggests a number of possible theoretical explanations, supported and enriched by the experience and insight of the author. But the explanation to which Keynes gave the greatest cmphasis is the notion of a"normal "long-term rate, to which investors expect the rate of interest. to return. When he refers to uncertainty in the market, he appears to mean disagreement among investors concerning the future of the rate rather than subjective doubt in the mind of an individual investor. Thus Kaldors correction of Keynes is more verbal than substantive when he says, "It is . not so much the uncertaint concerning future interest rates as the inelasticity of interest expectations which is responsible for Mr. Keynes''liquidity preference function, Keynes'use of this explanation of liquidity preference as a part of his theory of under employment equilibrium was the target of important criticism by Leontief and Fellner Leontief argued that liquidity preference must necessarily be zero in equilibrium, regardless of the rate of interest. Divergence between the current and expected interest rate is bound to vanish as investors learn from experience; no matter how low an interest rate may be it can be accepted as"normal" if it persists long enough. This criticism was a part of Leontief,'s general methodological criticism of Keynes, that unemployment was not a feature of equilibrium, subject to analysis by tools of static theory, but a phenomenon of disequilibrium requiring analysis by dynamic theory. 3 Fellner makes a similar criticism of the logical appropriateness of Keynes explanation of liquidity preference for the purposes of his theory of underemployment equilibrium. Why, he asks, are interest rates the only variables to which inelastic expectations attach? Why don,'t wealth owners and others egard pre-depression price levels as" normal"levels to which prices will return If they did consumption and investment demand would respond to reductions in money wages and prices, no matter how strong and how elastic the liquidity preference of These criticisms raise the question whether it is possible to dispense with the assumption of stickiness in interest rate expectations without losing the implication that Keynesian The General Theory of Emp d Money(New York: Harcourt pp.168-172and201-20 of Economic Studies, vol. 7( editor The Ne W. Fellner, Monetary Policies and Full Employment(Berkeley: University of California Press, 1946)
70 REVIEW OF ECONOMIC STUDIES be shifted by open market operations themselves, since they will change the volume of outstanding bonds or consols. For example, to establish the rate at or below min rc, the central bank would have to buy all outstanding bonds or consols. The size of the community's investment balances would then be independent of the rate of interest ; it would be represented by a vertical line through, or to the right of, B, rather than the curve ABC. Thus the new relation between cash and interest would be a curve lying above LMB, of the same general contour as LMNE W. 2.6 Keynesian theory and its critics. I believe the theory of liquidity preference I have just presented is essentially the original Keynesian explanation. The General Theory suggests a number of possible theoretical explanations, supported and enriched by the experience and insight of the author. But the explanation to which Keynes gave the greatest emphasis is the notion of a " normal" long-term rate, to which investors expect the rate of interest, to return. When he refers to uncertainty in the. market, he appears to mean disagreement among investors concerning the future of the rate rather than subjective doubt in the mind of an individual investor.' Thus Kaldor's correction of Keynes is more verbal than substantive when he says, " It is . .. not so much the uncertainty concerning future interest rates as the inelasticity of interest expectations which is responsible for Mr. Keynes' ' liquidity preference function,' 912 Keynes' use of this explanation of liquidity preference as a part of his theory of underemployment equilibrium was the target of important criticism by Leontief and Fellner. Leontief argued that liquidity preference must necessarily be zero in equilibrium, regardless of the rate of interest. Divergence between the current and expected interest rate is bound to vanish as investors learn from experience ; no matter how low an interest rate may be, it can be accepted as " normal" if it persists long enough. This criticism was a part of Leontief's general methodological criticism of Keynes, that unemployment was not a feature of equilibrium, subject to analysis by tools of static theory, but a phenomenon of disequilibrium requiring analysis by dynamic theory.3 Fellner makes a similar criticism of the logical appropriateness of Keynes' explanation of liquidity preference for the purposes of his theory of underemployment equilibrium. Why, he asks, are interest rates the only variables to which inelastic expectations attach ? Why don't wealth owners and others regard pre-depression price l.evels as "normal" levels to which prices will return ? If they did, consumption and investment demand would respond to reductions in money wages and prices, no matter how strong and how- elastic the liquidity preference of investorS_4 These criticisms raise the question whether it is possible to dispense with the assumption of stickiness in interest rate expectations without losing the implication that Keynesian 1*J. M. Keynes, The General -Theory of Employment, Interest, and Money (New York: Harcourt Brace, 1936), Chapters 13 and 15, especially pp. 168-172 and 201-203. One quotation from p. 172 will illustrate the point : " It is interesting that the stability-of the system and its sensitiveness to changes in the quantity of money should be so dependent on the existence of a variety of opinion about what is uncertain. Best of all that we should know the future. But if not, then, if we are to control the activity of the economic system by changing the quantity of money, it is important that opinions should differ." 2 N. Kaldor, "Speculation and Economic Stability," Review of Economic Studies, vol. 7 (1939), p. 15. 3 W. Leontief, " Postulates: Keynes' General Theory and the Classicists ", Chapter XIX in S. Harris, editor, The New Economics (New York-: Knopf, 1947), pp. 232-242. Section 6, pp. 238-239, contains the specific criticism of Keynes' liquidity preference theory. 4 W. Fellner, Monetdry Policies and Full Employment (Berkeley: University of California Press, 1946), p. 149
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK theory drew from it. Can the inverse relationship of demand for cash to the rate of interest be based on a different set of assumptions about the behaviour of individual investors This question is the subject of the next part of the paper 3. Uncertainty, risk aversion, and liquidity preference 1 The locus of opportunity for risk and expected return. Suppose that an investor is not certain of the future rate of interest on consols: investment in consols then involves risk of capital gain or loss. The higher the proportion of his investment balance that he holds in consols, the more risk the investor assumes. At the same time, increasing the proportion in consols also increases his expected return. In the upper half of Figure 3.1 the vertical axis represents expected return and the horizontal axis risk. a line such as OCi pictures the fact that the investor can expect more return if he assumes more risk. In the lower half of Figure 3. 1, the left-hand vertical axis measures the proportion invested in consols. A line like OB shows risk as proportional to the share of the total balance held in consols The concepts of expected return and risk must be given more precisio The individual investor of the previous section was assumed to have, for any current rate of interest, a definite expectation of the capital gain or loss g( defined in expression (2. 1 )above) he would obtain by investing one dollar in consols. Now he will be assumed instead to be uncertain about g but to base his actions on his estimate of its probability distribution. This probability distribution, it will be assumed, has an expected value of zero and is independent of the level of r, the current rate on consols. Thus the investor considers a doubling of the rate just as likely when rate is 5% as when it is 2%, and a te just as likely when it is 1% as when a portfolio consists of a proportion A, of cash and Ag of consols, where A, and A2 initial investment balance in dollars. Negative values of A, and Ag are excluded by definition; only the government and the banking system can issue cash and government consols. The return on a portfolio R is (3.1) R=A2(+g) Since g is a random variable with expected value zero, the expected return on the portfolio E(R)
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK 71 theory drew from it. Can the inverse relationship of demand for cash to the rate of interest be based on a different set of assumptions about the behaviour of individual investors ? This question is the subject of the next part of the paper. 3. Uncertainty, risk aversion, and liquidity preference. 3.1 The locus of opportuinity for risk and expected return. Suppose that an investor is not certain of the future rate of interest on consols ; investment in consols then involves a risk of capital gain or loss. The higher the proportion of his investment balance that he holds in consols, the more risk the investor assumes. At the same time, increasing the proportion in consols also increases his expected return. In the upper half of Figure 3.1, the vertical axis represents expected return and the horizontal axis risk. A line such as OC1 pictures the fact that the investor can expect more return if he assumes more risk. In the lower half of Figure 3.1, the left-hand vertical axis measures the proportion invested in consols. A line like OB shows risk as proportional to the share of the total balance held in consols. The concepts of expected return and risk must be given more precisioThe individual investor of the previous section was assumed to have, for any current rate of interest, a definite expectation of the capital gain or loss g (defined in expression (2.1) above) he would obtain by investing one dollar in consols. Now he will be assumed instead to be uncertain about g but to base his actions on his estimate of its probability distribution. This probability distribution, it will be assumed, has an expected value of zero and is independent of the level of r, the current rate on consols. Thus the investor considers a doubling of the rate just as likely when rate is 5 % as when it is 2%, and a halving of the rate just as likely when it is 1 % as when it is 6%. A portfolio consists of a proportion A1 of cash and A2 of consols, where A1 and A2 add up to 1. We shall assume that A1 and A2 do not depend on the absolute size of the initial investment balance in dollars. Negative values of A1 and A2 are excluded by definition; only the government and the banking system can issue cash and government consols. The return on a portfolio R is: (3.1) R 4=A2(r+g) 0 < A2 < 1 Since g is a random variable with expected value zero, the expected return on the portfolio is: (3.2) E(R) = =R A2 r
REVIEW OF ECONOMIC STUDIES The risk attached to a portfolio is to be measured by the standard deviation of R The standard deviation is a measure of the dispersion of possible returns around the mean value uR. A high standard deviation means, speaking roughly, high probability of large deviations from uR, both positive and negative. A low standard deviation means low probability of large deviations from uR; in the extreme case a zero standard deviation would indicate certainty of receiving the return Hr. Thus a high-or portfolio offers the investor the chance of large capital gains at the price of equivalent chances of large capital losses. A low-oR portfolio protects the investor from capital loss, and likewise gives him little prospect of unusual gains. Although it is intuitively clear that the risk of a portfolio is to be identified with the dispersion of possible returns, the standard deviation is neither the sole measure of dispersion nor the obviously most relevant measure. The case for the standard deviation will be further discussed in section 3. 3 below The standard deviation of R depends on the standard deviation of g, g, and on the amount invested in consols (3.3) Thus the proportion the investor holds in consols Ag determines both his expected return FR and his risk oR. The terms on which the investor can obtain greater expected return at the expense of assuming more risk can be derived from(3. 2) and(3. 3) (34) 0≤aR≤ Such an opportunity locus is shown as line OC(for r= ri) in Figure 3. 1. The slope of the line is. For a higher interest rate ra, the opportunity locus would be oc 2, and for ra, a still higher rate, it would be OCa. The relationship(3. 3 )between risk and invest- ment in consols is shown as line OB in the lower half of the Figure. Cash holding A 1-Aa) can also be read off the diagram on the right-hand vertical axis. .2 Loci of indifference between combinations of risk and expected return. The investor is assumed to have preferences between expected return uR and lisk or that can be repre- sented by a field of indifference curves. The investor is indifferent between all pairs (uR, or) that lie on a curve such as I, in Figure 3. 1. Points on I2 are preferred to those or 11: for given risk, an investor always prefers a greater to a smaller expectation of return Conceivably, for some investors, risk-lovers, these indifference curves have negative slopes Such individuals are willing to accept lower expected return in order to have the chance of unusually high capital gains afforded by high values of oR. Risk-averters, on the other
REVIEW OF ECONOMIC STUDIES The risk attached to a portfolio is to be measured by the standard deviation-of R, aR. The standard deviation is a measure of the dispersion of possible returns around the mean value VLR. A high standard deviation means, speaking roughly, high probability of large deviations from (lR, both positive and negative. A low standard deviation means low probability of large deviations from VtR; in the extreme case, a zero standard deviation would indicate certainty of receiving the return gR. Thus a high-aR portfolio offers the investor the chance of large capital gains at the price of equivalent chances of large capital losses. A low-gR portfolio protects the investor from capital loss, and likewise gives him little prospect of unusual gains. Although it is intuitively clear that the risk of a portfolio is to be identified with the dispersion of possible returns, the standard deviation is neither the sole measure of dispersion nor the obviously most relevant measure. The case for the standard deviation will be further discussed in section 3.3 below. The standard deviation of R depends on the standard deviation of g, ag, and on the amount invested in consols: (3.3) aR = A2 0 < As2 < 1. Thus the proportion the investor holds in consols A2 determines both his expected return ILR and his risk aR. The terms on which the investor can obtain greater expected return at the expense of assuming more risk can be derived from (3.2) and (3.3): r (3.4) zR -= a R 0 < OR < a, Such an opportunity locus is shown as line OC1 (for r = rl) in Figure 3.1. The slope of the line is -s-. For a highei interest rate r2, the opportunity locus would be OC 2; and as for r3, a still higher rate, it would be OC3. The relationship (3.3) between risk and investment in consols is shown as line OB in the lower half of the Figure. Cash holding A1(= 1 - A2) can also be read off the diagram on the right-hand vertical axis. 3.2 Loci of indifference between combinations of risk and expected return. The investor is assumed to have preferences between expected return V[R and iisk aR that can be represented by a field of indifference curves. The investor is indifferent between all pairs ([LR, cR) that lie on a curve such as I1 in Figure 3.1. Points on I2 are preferred to those on I ; for given risk, an irnvestor always prefers a greater to a smaller expectation of return. Conceivably, for some investors, risk-lovers, these indifference curves have negative slopes. Such individuals are willing to accept lower expected return in order to have the chance of unusually high capital gains afforded by high values of aR. Risk-averters, on the other 72
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK ↓n!!l 1 Ag 3 RE3.1 Portfolio Selection at Various Interest Rates and Before and After Taxation
LIQUIDITY PREFERENCE AS BEHAVIOR TOWARDS RISK C3 / 'I C2 Cl1 I q I I I I I I I I I I I I I I I i I I I I I I I I ^ ^ ^ ^ A2(r) A1 0 FIGURE 3.1 Portfolio Selection at Various Interest Rates and Before and After Taxation. t 0o 73 , I I I I I I I I I I t A I I I I .. . . . .. .... . I 6 0
REVIEW OF ECONOMIC STUDIE hand, will not be satisfied to accept more risk unless they can also expect greater expected return. Their indifference curves will be positively sloped. Two kinds of risk-averters need to be distinguished. The first type, who may be called diversifiers for reasons that will become clear below, have indifference curves that are concave upward, like those in Figure 3. 1. The second type, who may be called plungers, have indifference curves that are upward sloping, but either linear or convex upward 3. 3 Indifference curves as loci of constant expected utility of wealth. The reader whe is willing to accept the indifference fields that have just been introduced into the analysis may skip to section 3. 4 without losing the main thread of the argument. But these in- difference curves need some explanation and defence. Indifference curves between uR and or do not necessarily exist. It is a simplification to assume that the investor chooses mong the alternative probability distributions of R available to him on the basis of only two parameters of those distributions. Even if this simplification is accepted, the mean and standard deviation may not be the pair of parameters that concern the investor. 3. 3. 1 One justification for the use of indifference curves between uR and be that the investor evaluates the future of consols only in terms of some two-parameter family of probability distributions of g. For example, the investor might think in terms of a range of equally likely gains or losses, centered on zero. Or he might think in terms that can be approximated by a normal distribution. Whatever two-parameter family is assumed--uniform, normal, or some other-the whole probability distribution is deter mined as soon as the mean and standard deviation are specified. Hence the investors choice among probability distributions can be analyzed by HR oR indifference curves; any other pair of independent parameters could serve equally well If the investor's probability distributions are assumed to belong to some two-parameter family, the shape of his indifference curves can be inferred from the general characteristics percentage growth in the investment balance by the end of the period. This way of formu lating the utility function makes the investor's indifference map, and therefore his choices of proportions of cash and consols, independent of the absolute amount of his initial balance On certain postulates, it can be shown that an individuals choice among probability distributions can be described as the maximization of the expected value of a utility function The ranking of probability distributions with respect to the expected value of utility will of a constant or by multiplication by a positive constant. Consequenty the addition not be changed if the scale on which utility is measured is altered either we are free to choose arbitrarily the zero and unit of measurement of the utility function U(R)as follows U(0)=0;U(-1) J and morgens pp.1530,pp.617632 The er
74 REVIEW OF ECONOMIC STUDIES hand, will not be satisfied to accept more risk unless they can also expect greater expected return. Their indifference curves will be positively sloped. Two kinds of risk-avertexs need to be distinguished. The first type, who may be called diversifiers for reasons that will become clear below, have indifference curves that are concave upward, like those in Figure 3.1. The second type, who may be called plungers, have indifference curves that are upward sloping, but either linear or convex upward. 3.3 Indifference curves as loci of constant expected utility of wealth. The reader who is willing to accept the indifference fields that have just been introduced into the analysis may skip to section 3.4 without losing the main thread of the argument. But these indifference curves need some explanation and defence. Indifference curves between fR and UR do not necessarily exist. It is a simplification to assume that the investor chooses among the alternative probability distributions of R available to him on the basis of only two parameters of those distributions. Even if this simplification is accepted, the mean and standard deviation may not be the pair of parameters that concern the investor. 3.3.1 One justification for the use of indifference curves between ULR and aR would be that the investor evaluates the future of consols only in terms of some two-parameter family of probability distributions of g. For example, the investor might think in terms of a range of equally likely gains or losses, centered on zero. Or he might think in terms that can be approximated by a normal distribution. Whatever two-parameter family is assumed-uniform, normal, or some other-the whole probability distribution is determined as soon as the mean and standard deviation are specified. Hence the investor's choice among probability distributions can be analyzed by 9R-aR indifference curves any other pair of independent parameters could serve equally well. If the investor's probability distributions are assumed to belong to some two-parameter family, the shape of his indifference curves can be inferred from the general characteristics of his utility-of-return function. This function will be assumed to relate utility to R, the percentage growth in the investment balance by the end of the period. This way of formulating the utility function makes the investor's indifference map, and therefore his choices of proportions of cash and consols, independent of the absolute amount of his initial balance. On certain postulates, it can be shown that an individual's choice among probability distributions can be described as the maximization of the expected value of a utility function.' The ranking of probability distributions with respect to the expected value of utility will not be changed if the scale on which utility is measured is altered either by the addition of a constant or by multiplication by a positive constant. Consequently we are free to choose arbitrarily the zero and unit of measurement of the utility fuinction U (R) as follows U(O) = O; U(-1)=-1. 1 See Von Neumann, J. and Morgenstern, O., Theory of Games and Economic Behavior, 3rd Edition (Princeton: Princeton University Press, 1953), pp. 15-30, pp. 617-632 ; Herstein, I. N. and Milnor, J., " An Axiomatic Approach to Measurable Utility ", Econometrica, vol. 23 (April 1953), pp. 291-297; Marschak, J., " Rational Behavior, Uncertain Prospects, and Measurable Utility ", Econometrica, vol. 18 (April 1950), pp. 111-141 ; Friedman, M. and Savage, L. J., " The Utility Analysis of Choices Involving Risk ", Journal of Political Economy, vol. 56 (August 1948), pp. 279-304, and " The Expected Utility Hypothesis and the Measurability of Utility ", Journal of Political Economy, vol.-60 (December 1952), pp. 463-474. For a treatment which also provides an axiomatic basis for the subjective probability estimates here assumed, see Savage, L. J., The Foundations of Statistics (New York: Wiley, 1954)