Friedman and Kuttner - 1992 - Money, Income, Prices, and Interest Rates.pdf

Money, Income, Prices, and Interest Rates Author(s): Benjamin M. Friedman and Kenneth N. Kuttner Source: The American Economic Review , Jun., 1992, Vol. 82, No. 3 (Jun., 1992), pp. 472492 Published by: American Economic Association Stable URL: https://www.jstor.org/stable/2117316 JSTOR is a not-for-...

Money, Income, Prices, and Interest Rates Author(s): Benjamin M. Friedman and Kenneth N. Kuttner Source: The American Economic Review , Jun., 1992, Vol. 82, No. 3 (Jun., 1992), pp. 472492 Published by: American Economic Association Stable URL: https://www.jstor.org/stable/2117316 JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms is collaborating with JSTOR to digitize, preserve and extend access to The American Economic Review This content downloaded from 92.202.12.4 on Thu, 02 May 2024 05:32:07 +00:00 All use subject to https://about.jstor.org/terms Money, Income, Prices, and Interest Rates By BENJAMIN M. FRIEDMAN AND KENNETH N. KUTTNER* Including data from the 1980's sharply weakens the postwar time-series evidence indicating significant relationships between money (however defined) and nominal income or between money and either real income or prices separately. Focusing on data from 1970 onward destroys this evidence altogether. Evidence indicating cointegration of real income and real money balances, with due allowance for the effect of interest rates, also deteriorates when the sample extends through the 1980's. A positive finding is that the spread between the commercial paper rate and the Treasury bill rate consistently contains highly significant information about future movements in real income. (JEL E52) Economists have long understood that the quantity of money, or its growth rate, can play a useful role in the monetary policy process only to the extent that fluctuations in money over time regularly and reliably correspond to fluctuations in income, prices, or whatever other aspects of economic activity the central bank seeks to influence. The same is true, of course, for any other financial quantity (e.g., nonmoney assets, measures of credit) or, for that matter, interest rates and any other financial prices. Especially in the case of money, a rich literature developed over many years has investigated in some detail the requirements that the relationships connecting money to income or prices must satisfy in order to warrant focusing monetary policy on money in any of several specific ways (see e.g., the literature surveyed in Friedman ). An equally rich empirical literature has repeatedly sought to establish whether these requirements have actually been satisfied * Department of Economics, Harvard University, Cambridge, MA 02138, and Federal Reserve Bank of Chicago, respectively. The authors are grateful to James Stock and Mark Watson for helpful discussions, to two anonymous referees for helpful comments on an earlier draft, and to the National Science Foundation, the General Electric Foundation, and the Harvard Program for Financial Research for research support. The views presented here are those of the authors; they do not necessarily reflect the official position of the Fed- eral Reserve Bank of Chicago. at specific times and in specific places (see e.g., the literature surveyed by Bennett T. McCallum ). Different ways of conceptually basing the monetary policy process on money place different empirical requirements on the relationships between money and the economic variables that are of ultimate policy concern. These relationships in turn depend on such basic dimensions of economic behavior as the nature of the economy's aggregate supply process, the degree of price flexibility, the interest and wealth elasticities of aggregate demand, and importantly, the public's money-demand behavior and the banks' money-supply behavior. For example, under familiar circumstances using money as an "intermediate target" (i.e., determining the money growth rate most likely ex ante to be consistent with the central bank's macroeconomic policy objectives and then conducting monetary policy operations during some time interval as if achieving money growth along that path were itself the policy objective) is closer to being optimal as the demand for money is closer to being both nonstochastic and interestinelastic. Looser relationships suffice for it to be optimal to use money as an "information variable," that is, adjusting policy operations during some interval in response to actual money growth that departs from the ex ante path and perhaps also in response to analogous departures for other variables, but in any case not in a way designed 472 This content downloaded from 92.202.12.4 on Thu, 02 May 2024 05:32:07 +00:00 All use subject to https://about.jstor.org/terms VOL.82 NO. 3 FRIEDMANAND KU7TNER: MONEY, INCOME, PRICES, AND INTERESTRRATES 473 necessarily to restore money growth to the ex ante path.1 However, what is essential to either of these ways of proceeding-and to others besides-is that there be at least some reliably exploitable connection between money and either income or prices, so that observed departures of money from some ex ante path bear a systematic implication for income or prices in the future. Otherwise, as a variable that the central bank cannot set directly as a policy instrument, money has no role in the policy process. From an information-variable perspective, there is no point to the central bank's reacting to fluctuations in money if those fluctuations bear no implication for subsequent movements in income or prices. From an intermediate-target perspective, there is even less point to making policy as if controlling money were stochastically equivalent to controlling income and prices if in fact there is no relation between them. It is for just this reason that the events of the 1980's have proved so subversive to what had almost come to be standard ways of thinking about monetary policy, not just in the United States but in many other countries as well. While there was never any lack of debate about the strength or weakness of the empirical relationships connecting money to income and prices, and therefore about the appropriate role of money in the monetary policy process, before the 1980's there was widespread agreement that fluc- cluding the Federal Reserve System in the United States, have altered or abandoned the ways in which they had previously relied on money to make policy. The object of this paper is to show how the passage of time-in particular, the experience since 1980-has altered familiar empirical relationships previously taken to support a central role for money in the monetary policy process. Section I reports results for a variety of significance tests and forecast error variance decompositions, in each case seeking to establish whether fluctuations in money or interest rates are useful for predicting subsequent fluctuations in income or prices. Here the consistent finding is that the positive results for the monetary aggregates in this context that are fa- miliar from earlier time periods do not hold up when the sample is extended to include data from the 1980's. A further consistent result is that the difference between the commercial paper rate and the Treasury bill rate does contain incremental information about real income (but not prices). Section II focuses in more detail on this predictive power of the paper-bill spread. Section III shifts the paper's focus to longrun relationships, reporting results for tests of the cointegration of movements of money and income. Here, as in Section I, the relationships that would have to hold in order to warrant using money as the central focus of monetary policy disappear when the analysis includes data from the 1980's. Sectuations in money did contain at least potion IV briefly concludes by drawing together the implications of these findings for tentially useful information abput future income and price movements. In the 1980's, monetary policy. however, the empirical basis underlying that I. Autoregression Tests and agreement disappeared. Empirical investigation based on sample periods that include Variance Decompositions the 1980's simply does not lead to results Ever since the early work of Christopher corroborating what were commonly ac- A. Sims (1972), empirical consideration of whether money (or any other aggregate) can usefully play a role in the monetary policy process has appropriately focused not just on whether fluctuations of money help preIJohn H. Kareken et al. (1973) first formally intro- dict future fluctuations of income (or prices, duced the information-variable concept into the analyetc.), but on whether they help predict fusis of monetary policy. See Friedman (1975) for a ture fluctuations of income that are not formal analysis of the intermediate-target-variable proalready predictable on the basis of fluctuacedure. cepted as facts of economic behavior not so many years earlier. It is not surprising that in this environment some central banks, in- This content downloaded from 92.202.12.4 on Thu, 02 May 2024 05:32:07 +00:00 All use subject to https://about.jstor.org/terms 474 THE AMERICAN ECONOMIC REVIEW JUNE 1992 TABLE 1- F STATISTICS FOR FINANCIAL VARIABLES IN NoM Variable 1960:2-1979:3 1960:2-1990:4 1970:3-1990:4 A. Three-Variable System (Nominal Income, Fiscal Variable, Financia A ln(base) 3.68** 1.85 0.82 A ln(M1) 7.23** 3.75** 2.27t A ln(M2) 5.12** 4.49** 1.85 A ln(credit) 5.54** 1.55 0.21 Arp 1.60 5.55** 4.39** ArB 1.81 4.82** 3.96** rp- rB 4.43** 4.68** 3.16* B. Two-Variable System (Nominal Inco A ln(base) 4.07** 2.19t 0.93 A ln(M1) 7.39** 3.75** 2.14t A ln(M2) 5.19** 5.15** 2.37t A ln(credit) 4.96** 1.68 0.39 Arp 1.91 5.84** 4.36** ArB 2.15t 5.05** 3.83** rp-rB 5.34** 5.38** 3.35* Notes: Estimated regressions use four dollars, and the fiscal variable is mi tStatistically significant at the P < 0 *Statistically P < 0.01 level. tions of income itself or other readily observable variables. Especially in the context of the information-variable approach to monetary policy, this interpretation of Sims-type autoregression tests is precisely what is at issue, and the much-debated question of whether such tests constitute valid tests of "causality" is beside the point. As long as movements in money do contain information about future movements in in- significant at the P zero in autoregressions of the form 4 4 (1) i=l i=l A\ 4 where y, m, and g (all in natural logarithms) are, respectively, nominal income, the financial variable indicated, and mid- come beyond what is already contained in movements in income itself, monetary polexpansion federal expenditures; a, the P3i, icy can exploit that information by respondthe yi, and the 8i are coefficients to be ing to observed money growth regardless of estimated; and u is a disturbance term.2 whether the information it contains reflects true causation, reverse causation based on anticipations, or mutual causation by some 2The monetary base and the two "M's" are based independent but unobserved influence. on the conventional Federal Reserve Board definitions. The first four lines in the upper portion Credit is the outstanding indebtedness of domestic nonfinancial borrowers. Income is gross national prodof Table 1 present F statistics for tests, uct. (Substituting domestic absorption for gross naacross different time periods, of the null tional product does not systematically change the rehypothesis that all of the coefficients on the sults reported here or in most of the tables introduced lagged growth of either the monetary base, below.) All data except interest rates are seasonally Ml, M2, or credit (that is, all of the /3) are adjusted. This content downloaded from 92.202.12.4 on Thu, 02 May 2024 05:32:07 +00:00 All use subject to https://about.jstor.org/terms VOL.82 NO. 3 FRIEDMAN AND KUTTNER: MONEY INCOME, PRICES, AND INTEREST RATES 475 The lower portion of the table shows F statistics based on analogous equations excluding the government-spending variable. The results reported are based on standard quarterly data for three sample periods: 1960:2-1979:3, 1960:2-1990:4, and 1970:31990:4. For the sample spanning 1960:2-1979:3 (i.e., from the earliest time for which the Federal Reserve provides data corresponding to its current definitions of the monetary aggregates until the introduction of its new monetary policy procedures in October 1979), the F statistics in the first column of Table 1 show that the base, Ml, M2, and credit each contained information about future income movements that was statistically significant at the 0.01 level, regardless of whether the analysis includes the fiscal variable. Extending the sample to include data through year-end 1990 eliminates this significance for the base and for credit. Also, beginning the sample in 1970:3 (i.e., after the removal of the Regulation Q ceiling on interest paid on large certificates of deposit and the beginning of the [sporadic] targeting of money growth for purposes of mone- replaced by, successively, the interest rate on prime 4-6 month commercial paper, the 90-day Treasury bill rate, and the difference between these two interest rates.4 Introducing an interest rate in relationships like those under study here, either in place of or in addition to a financial aggregate, has also been familiar since the work of Sims (1980). Sims (1980) and Friedman (1983) both used the commercial paper rate for this purpose. Robert B. Litterman and Lawrence Weiss (1985), Martin Eichenbaum and Kenneth J. Singleton (1986), and James H. Stock and Mark W. Watson (1989a) all used the Treasury bill rate. These relationships too have changed with the passage of time, although in the case of the two interest rates the changes have been at least partly in the direction of stronger, rather than weaker, ability to predict income. For the sample ending in 1979, neither the commercial paper rate nor the Treasury bill rate contained statistically significant information about future fluctua- tions in income. By contrast, the spread between the two rates contained information about income that was significant at the 0.01 level. For both the 1960-1990 and the tary policy) leaves only Ml significant among 1970-1990 samples, however, the informathese three aggregates, even at the 0.10 tion contained in each of the three interest level in the presence of the fiscal variable, rate variables is significant at least at the and leaves only Ml and M2 significant 0.05 level. (again, only at the 0.10 level) without the Tables 2 and 3 present analogous results fiscal variable.3 based on equations in which the variable For purposes of comparison, Table 1 also whose movements are to be explained is, in presents F statistics for analogous autoreturn, real income and then the price level, gressions in which variable /Akm in (1) is and the equation includes lagged values of both real income and the price level as regressors.5 As is consistent with much of 3Restricting the sample to the period since the Federal Reserve's change in operating procedures (i.e., 1979:4-1990:4), renders all four aggregates not significant, even at the 0.10 level. Including a dummy variable for the 1980's in the regressions based on the longer samples (a suggestion made by an anonymous referee) makes only a small difference for the results reported above. In the 1970-1990 equations for MI, for example, adding a dummy variable equal to 1 beginning in 1979:4 results in F statistics of 2.45 when the fiscal variable is included and 2.26 when it is excluded; both are again significant only at the 0.10 level. The F statistics in the corresponding M2 equations are 1.29 and 1.62, respectively. Neither of these values is significant at the 0.10 level. 4We are grateful to Bennett McCallum for suggesting the inclusion of interest rates in this analysis for purposes of comparison. Both interest rates are en- tered in the regressions in difference form, while the spread is entered in level form, because univariate stationarity tests indicate that the two interest rates (like income and the financial aggregates) are each difference-stationary, while the spread is stationary in levels. 5Real income is gross national product in 1982 dollars. The price level is the corresponding deflator. This content downloaded from 92.202.12.4 on Thu, 02 May 2024 05:32:07 +00:00 All use subject to https://about.jstor.org/terms 476 THE AMERICAN ECONOMIC REVIEW JUNE 1992 TABLE 2- F STATISTICS FOR FINANCIAL VARIABLES IN REAL-INCOME EQUATIONS Variable 1960:2-1979:3 1960:2-1990:4 1970:3-1990:4 A. Four-Variable System (Real Income, Price Index, Fiscal Variable, Financial Variable): A A ln(base) 1.22 0.77 ln(M1) 2.71* 2.85* A ln(M2) A ln(credit) 3.63** 4.32** 2.34t 1.02 0.52 1.81 1.81 0.16 Arp 1.55 5.61** 3.94** ArB 1.81 4.52** 3.44* rp - rB 7.10** 7.16** 4.68** B. Three-Variable System (Real Income, Price Index, Financial Variable): A ln(base) 1.27 0.91 0.98 A ln(M1) 2.59* 2.65* 1.77 A ln(M2) 3.78** 4.66** 2.19t A ln(credit) Arp ArB 1.95 rp-rB 2.21 1.97 1.21 5.80** 4.76** 8.12** 0.34 4.14** 3.62** 7.70** 5.32** Notes: Estimated regressions use four lags of each variable. Income is GNP in 1982 dollars, the price index is the implicit GNP deflator, and the fiscal variable is mid-expansion federal expenditures. tStatistically significant at the P < 0.1 level. *Statistically significant at the P