1. Introduction. (C) The difference between the impact and the cumulative effects of a change in government spending or taxation.

Transcription

1 Comment on Long Run Effects of Fiscal and Monetary Policy on Aggregate Demand, by James Tobin and Willem Buiter* by Milton Friedman In Monetarism, edited by Jerome L. Stein, pp Amsterdam and New York: North- Holland, El Sevier 1. Introduction Three different issues are involved in James Tobin s continuing skirmish with the allegedly characteristic monetarist proposition that pure fiscal policy does not matter for aggregate real demand, nominal income, and the price level. (A) The theoretical conditions under which this assertion would be rigorously and precisely correct for both impact and long-run effects. (B) The empirical importance of the effects of pure fiscal policy arising from the failure of these conditions to be satisfied perfectly; i.e., whether, as a practical matter, sufficiently correct results will be obtained by treating them as if they were perfectly satisfied. (C) The difference between the impact and the cumulative effects of a change in government spending or taxation. Tobin s and Buiter s paper is directed mostly to analyzing the third issue for a specific model (or pair of models). I shall leave detailed criticism of that analysis to the other discussants of the paper, on grounds both of their greater competence for this task, and my tentative conclusion that these models do not cast much light on whatever real issues may divide economists labeled monetarists from economists labeled Keynesian. The models may be useful, and the results enlightening, for other purposes. On that question, I express no judgment. I shall restrict my comments therefore primarily to the first two issues, which means, to the initial more general pages of the Tobin Buiter paper, where the authors use my views as a springboard (or straw man?) for their analysis. I shall then add some general comments on what seem to me the real issues that need resolution. One man s characteristic monetarist proposition is not another s. In my reply to Tobin s earlier comments on my two J.P.E. articles on a monetary framework, I explicitly refused to accept as characteristically monetarist a slightly different variant of the proposition that Tobin and Buiter now describe in these terms. 1 I suspect that our difference is a byproduct of my Marshallian approach to theory, which leads me to regard issue (B) as fundamental, versus Tobin s Walrasian approach, which makes it difficult for him to recognize issue (B) as a meaningful issue at all. 2. The Purely Theoretical Issue Whether that be correct or not, it is desirable to get the purely theoretical issue straight. Tobin and Buiter state that a necessary condition for the proposition [that pure fiscal policy does not matter for aggregate real demand, nominal income, and the price level] is zero elasticity of 1

2 demand for money with respect to interest rates. This is clearly wrong, even on the level of the most elementary IS LM model. 2 A horizontal IS curve is a sufficient condition for the stated result, whatever the shape of the LM curve. And a horizontal IS curve has as much claim to consideration on theoretical and empirical grounds as a vertical LM curve (or a horizontal LM curve of liquidity trap fame). Frank Knight consistently treated an infinitely elastic demand for new investment, which implies a horizontal IS curve, as a key element of his capital theory. One historical observation that is no less surprising than it is well documented is the relative constancy of the real interest rate over long periods of time, which could be regarded as reflecting a roughly horizontal IS curve for given price expectations. On a purely theoretical level, it is interesting to note that much practical Keynesianism has treated the LM curve as if it were horizontal, the IS curve as if it were vertical; whereas much practical monetarism has treated the LM curve as it were vertical, the IS curve as if it were horizontal. In this sense, the two can be regarded as polar opposites. On this same level, it is interesting that in one of their longer-term models intended to demonstrate the necessity of a vertical LM curve for the long-run ineffectiveness of fiscal policy, Tobin and Buiter introduce the counterpart of a vertical IS curve (their GT curve in Model I). 3. The Practical Issue Tobin and Buiter explicitly reject issue (B) as of any significance by noting that monetarists usually soften their assertions with qualifying adjectives and adverbs minor, almost, etc. and then adding hedges of this order really do not alter the monetarist message for theory and policy. From my Marshallian point of view, such hedges are highly relevant to both the theory that we find useful and the policy measures that we recommend. I continue to believe that the fundamental differences among us are empirical, not theoretical; that what we really differ on are precisely what effects are and are not minor, not on whether an effect is precisely zero. Despite their rejection of issue (B), Tobin and Buiter devote much attention to my words certain to be temporary and likely to be minor. They treat these words from a 1967 Newsweek column of mine as referring to a rise in government expenditures or reduction in taxes, whereas they were actually written in reference to an increase in taxes designed to halt inflation. That is a trivial point, involving only a change of sign, except that it has implications for the initial conditions implicit in the analysis. A more important point is that Tobin and Buiter state that I used these words to refer to fiscal effects but do not specify what fiscal effects I was referring to. In the original column, these words referred to the net decrease in spending from these sources, and these sources are (1) that some of the funds not borrowed by the Federal government may be added to idle balances rather than spent or loaned, and (2) it takes time for borrowers and lenders to adjust to reduced government borrowing [see Gordon (1974, pp )]. Is this net decrease in spending all the fiscal effects that there are? For example, this statement does not say that lower interest rates are temporary though on a fuller analysis they might be. And later, on the same page of my J.P.E. reply, interpreting my Newsweek statement, I assert, once the new velocity is reached there is no further downward pressure. Is the lowered velocity not a fiscal effect? 2

3 My point here is not to defend my analysis as correct or my Newsweek statements as unambiguous. In view of Tobin s and Buiter s interpretation of them, the statements are clearly not unambiguous, and no doubt if I were writing the passage today, I would write it differently. 3 My point is only that I was quite explicit, certainly in my J.P.E. exegesis, in rejecting the view that the fiscal effects are precisely zero, so that Tobin s and Buiter s examination of my statement must be regarded as dealing with issue (B), not issue (A), despite their disclaimer. Even more puzzling is Tobin s and Buiter s further statement: by labeling this effect on income not only minor but temporary he seems to be saying that non-monetary financing of the accompanying budget deficits moves LM to the left, cancelling the expansionary shift of IS (my italics). As it happens, there is no need for Tobin and Buiter to conjecture what I meant since I was painfully explicit though, of course, possibly wrong. Let me quote (with apologies for repeating myself): Why certain to be temporary? Because the leftward shift in the IS curve is a once-for-all shift. Put in monetarist terms, the lowered interest rate resulting from the federal government s absorbing a smaller share of annual savings will reduce velocity; the transition to the lower velocity reduces spending for a given money stock, but once the new velocity is reached there is no further downward pressure. Why likely to be minor? Because the monetarist view is that saving and investment have to be interpreted much more broadly than the Keynesians tend to interpret it. Hence even a fairly substantial tax increase will produce only a minor shift in the IS curve. 9 [Footnote 9 in J.P.E.: Note that what is at issue is not primarily the elasticity of saving or investment, though monetarists would probably set that elasticity higher than neo-keynesians, but rather the absolute magnitude affected (the position of the relevant investment and savings curves).] 4 Let me make two comments here. First, clearly judgments about the slope of the IS curve and not merely of the LM curve enter into my belief that, for a given money supply, the effect of higher taxes on inflation is likely to be minor. This is related to the theoretical point made above with reference to issue (A). Second, my analysis up to this point was on a superficial level, since it did not take into account the effect of the accumulation of government debt and decumulation of private debt. However, I did comment on this point as well, in connection with a deficit financed by borrowing, the opposite of the case considered above, and again Tobin and Buiter do not refer to this comment. Let me therefore repeat it: As Tobin says, is a rain of Treasury bills of no consequence for the price level, while a rain of currency inflates prices proportionately? The answer is that the evidences of government debt are largely in place of evidences of private debt people hold Treasury bills instead of bills issued by, for example, U.S. Steel. The total nominal volume of debt grows by less and I believe much less than the size of the deficit. Moreover, even this growth is offset by two other factors: the increase expected in future tax liabilities accompanying the growth of the government debt, which Tobin refers to; and the reduction in the physical volume of assets created because of lowered private productive investment. On the other hand, the dollar bills are a net addition to the total nominal volume of assets. This analysis is, of course, greatly oversimplified. The effects of changes in asset structure are far more complex. Yet, I believe that even these brief remarks indicate that they are of a different order than the effect of financing deficits by creating money. 10 [Footnote 10 in J.P.E.: 3

4 In terms of real rather than nominal magnitudes, the difference can be interpreted as reflecting the effect of using different tax structures to finance government expenditures.] 5 Once again, the issue between us is largely empirical rather than theoretical. Treasury bills could be a sufficiently better substitute for money than U.S. Steel bills to raise appreciably desired velocity, or for that matter, conversely. I was expressing the empirical judgment that this was not so plus the theoretical argument that changes in Treasury bills could not be considered separately from the related changes in the whole structure of assets. On the particular issue Tobin and Buiter raise, if I am asked whether changes in government spending and taxation which do not affect the money supply can in principle have effects on velocity, total money income, interest rates, real income, etc., my answer is unambiguously, Of course. But I will hasten to add, so can changes in the relative price of oil produced by an international cartel or in the relative price of food produced by crop failures or in the relative price of computer services produced by technological change and so on and on. I would add further that in all these cases, I believe that the major effect is on the allocation of resources, or the distribution of income, and that the effects in each case on aggregate income and the price level are minor compared with the effects of a comparable change in the quantity of money (i.e., a change in the quantity of money over a given period equal in dollar amount to the extra or deficient government spending or taxation or the extra cost of oil). In addition, I would not deny that the cumulative effect of the continuous stream of such changes accounts for a sizable fraction of short-term fluctuations in nominal income. (As Franco Modigliani recalled, my estimate has generally been that about 50 percent of the variance of nominal income over quarterly or annual periods can be attributed to fluctuations of velocity. This is an upper estimate of the fraction attributable to the effect of non-monetary forces, since some of these fluctuations in velocity may be the delayed effect of earlier monetary changes not fully accounted for in my calculations.) 4. The Ambiguity of Fiscal Measures As something of a digression from the main theme, let me note the ambiguity that attaches to fiscal measures and in particular to the concept government spending. In econometric work, government spending is usually equated to national income account spending on goods and services. This is a defective measure of the relevant theoretical concept. Suppose the government had imposed a special excise tax on automobiles and had used the proceeds to pay for the antipollution equipment on automobiles that it now requires each of us to pay for separately. That would have been economically equivalent to the present requirement, yet it would have shown up in the national income accounts as government rather than private spending. Suppose current relief recipients were relabeled government civil servants assigned to child care duties. That would raise both recorded government spending and recorded total income. In general, I believe much too sharp a distinction has been drawn between transfers and expenditures on goods and services. As a final example, what of off-budget governmental guaranteed financing by F.H.A. insurance of mortgages, or through Fannie May or Freddie Mac? Should the implicit interest subsidy be regarded as part of government expenditures? Or total capital expenditures? Or what? I believe these problems are not merely intellectual puzzles but of very great empirical importance for econometric studies using government spending as a variable. 4

5 I have long been impressed with the empirical ambiguities in the application of the Keynesian apparatus most of the controversy surrounding the paper that I wrote some years ago with Meiselman centered on how to define autonomous expenditures, almost none on how to define money. The similar growing ambiguity about what government expenditures means reinforces my skepticism about the value of using these concepts to analyze the determinants of aggregate income and the price level, or the division of changes in aggregate income between changes in real income and in prices. 5. The Fundamental Differences Among Us I continue to believe, to return to Tobin s and Buiter s paper and the general subject, that the fundamental differences among us are empirical, not theoretical. But this does not mean that we all use the same theory. It means that our differences do not arise from analytical mistakes in manipulating theory. No one can use in practice a fully comprehensive theory. We must all work with simplified theories. Different empirical presumptions make different simplifications appropriate and thereby make different theoretical structures (different analytical filing cases) most useful. I must plead guilty to having contributed to confusion on this score. In the attempt to communicate, I have tried for example to present monetarist analysis in IS LM terms, even though recognizing that this was a cumbrous theoretical structure for this purpose, that the quantity theory structure was much more convenient. In concluding, let me comment briefly in very general terms on what I believe to be the major areas in which there are important empirical differences of judgment that need to be resolved. Like Thomas Mayer, I do not believe that these give rise to a dichotomy between monetarists and non-monetarists, that people who share the same views in one area necessarily do so in other areas. (1) Money versus credit. For the monetarist/non-monetarist dichotomy, I suspect that the simplest litmus test would be the conditioned reflex to the question, What is the price of money? The monetarist will answer, The inverse of the price level ; the non-monetarist (Keynesian or central banker) will answer, the interest rate. The key difference is whether the stress is on money viewed as an asset with special characteristics, or on credit and credit markets, which leads to the analysis of monetary policy and monetary change operating through organized money, i.e., credit, markets, rather than through actual and desired cash balances. Though not so obvious, the answer given also affects attitudes toward prices: whether their adjustment is regarded as an integral part of the economic process analyzed, or as an institutional datum to which the rest of the system will adjust. (2) Build up or down. A second dichotomy, not at all identical with the first, is between the view that the best approach to analyzing aggregate change is from the top down that is, analyzing the behavior of aggregates and then perhaps supplementing that analysis with an examination of the determinants of how the aggregate is divided among components or from the bottom up that is, analyzing directly each component and constructing the aggregate as a sum of components. This is essentially the little versus big model issue. It clearly cuts across the money-credit issue. For example, Keynes, like me, was clearly in the top-down category. And someone who believed that the quantity of money was critical could perfectly well believe that it was necessary to look at its influence, along with other forces, in disaggregated form. 5

6 (3) Stocks and flows. Here I believe the key issue is whether one considers stock flow relations as more stable and predictable in the short run than flow flow relations (in the long run all stocks and flows must be in equilibrium). Again this cuts across the other categories, since different groups may regard different capital flow ratios or flow flow ratios as important. For example, monetarists regard the money income ratio as of special importance; some development theorists, the capital output ratio. Keynesians regard the consumption income ratio as critical. Sidney Weintraub, the wages income ratio. (4) Substitution versus wealth effects. The Pigou effect, of which James Tobin was one of the independent discoverers and certainly one of the first to see its theoretical significance [see his contribution to Seymour Harris (1947)], is of great importance on a purely analytical level, as one way to eliminate an apparent flaw in the price system, i.e., to show that in principle there are enough equations to solve for all the unknowns, including the level of employment. However, I have never myself thought that wealth effects of changes in the quantity of money, or of price changes which altered the real quantity of money, were of any empirical importance for short-run economic fluctuations. I have always believed that substitution effects were the important way in which changes in money exerted influence. That is why, for example, I have never attributed great empirical importance to the inside outside money distinction. Here is again a dichotomy which, as Thomas Mayer emphasized, cuts across the others. Recent years have seen a convergence of professional judgments in each of these areas, particularly the first and third. The papers included in this volume most notably to my mind, Stein s and the empirical and theoretical work they stimulate, will, I trust, accelerate this trend. Bibliography Friedman, M., 1966, "Interest Rates and the Demand for Money," Journal of Law and Economics 9, Oct. Reprinted in: 1969, Optimum Quantity of Money and Other Essays (Aldine, Chicago, Ill.) ch. 7. Gordon, R. J., ed., 1974, Milton Friedman s Monetary Framework (University of Chicago Press, Chicago, Ill.). Harris, S. E., ed., 1947, The New Economics (Kelley, Clifton, N. J.). Notes * The paper that is the subject of this comment is printed in Monetarism, edited by Jerome L. Stein, pp Amsterdam and New York: North-Holland, These articles, comments by the critics, and my reply are all now available in Robert J. Gordon (1974). On the present point, see pp. 78, I have examined this issue on a broader level in my Interest Rates and the Demand for Money [Friedman (1966)]. 3 In particular, I would distinguish more sharply (as Stanley Fischer suggested in his verbal comments) between the once-for-all effect of shifting to a new velocity, which would produce a once-for-all decline in the price level, and the absence of any continuing effect on inflation there-after, even though the higher level of taxes were maintained, which would mean that the braking effect on the economy which the column was analyzing would be temporary. 6

7 4 Gordon (1974, pp ). 5 Gordon (1974, pp. 1 2). Reprinted from Monetarism, edited by Jerome L. Stein, Comment on Long Run Effects of Fiscal and Monetary Policy on Aggregate Demand, by James Tobin and Willem Buiter, pp El Sevier /18/13 7