VELOCITY AND THE VOLATILITY OF UNANTICIPATED AND ANTICIPATED MONEY SUPPLY IN THE UNITED KINGDOM

Transcription

1 INTERNATIONAL ECONOMIC JOURNAL 61 Volume 9, Number 3, Autumn 1995 VELOCITY AND THE VOLATILITY OF UNANTICIPATED AND ANTICIPATED MONEY SUPPLY IN THE UNITED KINGDOM JOHN THORNTON International Monetary Fund, and University of Wales, Bangor PHILIP MOLYNEUX* University of Wales, Bangor This paper uses the Granger causality method to test on United Kingdom data the relationship between income velocity and unanticipated and anticipated money supply and their respective volatilities over the period The results suggest that unanticipated and anticipated money supply and the volatility of these variables did not help predict income velocity and thus were not important factors in the decline in income velocity in the 1980s. [E52] 1. INTRODUCTION After increasing steadily in the 1960s and 1970s the income velocity of money in the United Kingdom (U.K.) fell sharply through much of the 1980s (see e.g. Goodhart 1986, 1989). This decline has been interpreted as undermining any relationship between changes in the money supply and subsequent changes in national income, as complicating the task of choosing an appropriate monetary aggregate to target, and as putting into question the implications for future inflation of the growth of the money supply. In the United States (U.S.) context Friedman (1983, 1984) argued that the decline in velocity could be explained by developments expanding the money supply: specifically, that the cause of the decline was more volatile money supply in the 1980s which increased the degree of perceived uncertainty and thereby increased the demand for money. Thus Friedman rejected the argument that the decline in velocity undermined the case for a monetary policy which emphasized controlling the growth of the money supply. In tests of the Friedman hypothesis on U.S. data, Hall and Noble (1987) and Fisher and Serlitis (1989) concluded that Friedman was correct in that the volatility of money caused income velocity in the Granger sense. However, later studies by Brocato and Smith (1989) and Mehra (1989) demonstrated that these results were sensitive to the time period chosen and to equation specification. In recent paper in this Journal, Arize (1993) tested Friedman s hypothesis on data for the U.K. over the period and found no significant causal relationship. Arize s findings confirmed those * The views expressed in this paper are those of the authors and should not be attributed to the institutions with which they are associated.

2 62 J. THORNTON AND P. MOLYNEUX of an earlier study using U.K. data by Thornton (1991) and are in line with Thornton (1995) which tests Friedman s hypothesis for several industrial countries. However, in his study Arize raised the issue of whether focussing on unanticipated money growth volatility might produce conclusions more in line with Friedman s hypothesis. In this note we distinguish between the unanticipated and anticipated components of several definitions of the money supply and their respective volatilities in the U.K. to investigate their effects on income velocity. 2. MODELLING CAUSALITY AND MONEY SUPPLY SHOCKS We used the following autoregressive time series model to test the hypothesis that unanticipated or anticipated money supply and their respective volatilities Grangercaused income velocity in the UK: a Vt = α 0 + α i V t-i + β i (M - Ma ) t-i + Ω i d b c Ma t-i + Φ i VOL( M - Ma ) t-i + δ i VOL( Ma ) t-i + ε t e (1) w h e r e V is income velocity, M is the money supply, (M - M a ) is the unanticipated money supply, M a is the anticipated money supply, VOL(M - M a ) is the volatility of the unanticipated money supply, VOL(M a ) is the volatility of the anticipated money supply, is the difference operator, and ε is a well behaved normally distributed error term. To implement the Granger causality test, if in equation (1) β i = 0 then past unanticipated money supply has no effect on future income velocity (i.e., unanticipated money supply does not Granger-cause V t ). In the same way: if Ω i = 0, Φ i = 0, and δ i = 0, then anticipated money supply, unanticipated money supply volatility, and anticipated money supply volatility, respectively, do not cause Grangercause V t ). Thus we test for causal relationships between anticipated and unanticipated money supply and their volatilities and income velocity by carrying out standard F- tests. The estimation procedure was conducted in three steps. The first step was to test the income velocity and money supply variability series for unit roots using the familiar augmented Dickey-Fuller test. With the series in level form the results (not published) indicated that it was not possible to reject the null hypothesis that all contained unit roots (i.e. were nonstationary). The results using first-difference data unanimously rejected the hypothesis of a unit root. In view of this, the analysis was carried out in with first-difference data. The second step was to generate series for the

3 VELOCITY AND VOLATILITY OF MONEY SUPPLY 63 unanticipated and anticipated money supply for each measure of money. This was done by regressing the money supply on lagged values of itself and a time trend, with the unanticipated money supply comprising the residuals from the equation. The procedure followed for each aggregate was to test down from a general equation containing eight quarterly lags so as to achieve a data acceptable representation of the series. Once this was done the results were used to generate the unanticipated and anticipated money supply series and their respective volatilities. In the third step, equation (1) was estimated by OLS and Granger causality tests were carried out. A common approach in Granger tests is to choose arbitrarily lag lengths for a, b, c, d, and e in equation (1). However, since results from Granger tests are sensitive to the selection of lag length we use the minimum final prediction error (FPE) criterion suggested by Akaike (1969) to determine lag length DATA AND GRANGER CAUSALITY TEST RESULTS Money supply volatility was proxied by an eight-quarter (current quarter and seven lags) standard deviation of the quarterly percentage change in each measure of m o n e y ; 2 GDP at current prices ( money GDP ) was the measure of income used. 3 Results are reported in Table 1 from tests using four monetary aggregates (M1, M3, M4, and M5) over two sample periods: the full sample period (1966QI-88QIII) and the period excluding the decline in income velocity in the 1980s (1966QI-79QIV). 4 The F-test denoted F 1 is the test of the hypothesis that in a regression of V t on lagged values of itself and unanticipated and anticipated money supply and their respective volatilities the coefficients on unanticipated money supply are zero. In the same manner, the F-tests denoted F 2, F 3, and F 4 test the hypotheses that in such a regression the coefficients of anticipated money supply are zero (hypothesis F 2 ), the coefficients of unanticipated money supply volatility are zero (F 3 ), and the coefficients of anticipated money supply volatility are zero (F 3 ). Over the sample period 1966QI-1989III the F values of income velocity are consistent with the inference that unanticipated and anticipated money supply and its volatility influence income velocity in the case of the M1 monetary aggregate. In the case of the other measures of money the hypothesis that these variables do not cause income velocity cannot be rejected at the 5 percent level of significance. The M1 result is not robust to the sample period, however. Estimates over the sub-period 1 The conclusions of the study are unchanged when arbitrary lag lengths of 2, 4, and 6 quarters are used in place of FPE determined lags. 2 The measure of monetary instability was chosen to be consistent with Friedman (1983). 3 Monetary data are from Long Runs of Monetary Data, , Bank of England, 1989; Money GDP is from Economic Trends, Annual Supplement 1989, Central Statistical Office, London. 4 The sample period was ended in 1989QIII because of changes in the way some money supply series were defined.

4 64 J. THORNTON AND P. MOLYNEUX Table 1.Tests of Granger Causality from Unanticipated and Anticipated Money Supply and Its Volatility to Income Velocity Hypothesis test Causal inference F1 F2 F3 F4 a. 1966QI-88QIII M1 > M1 velocity * * * * M3 > M3 velocity M4 > M4 velocity M5 > M5 velocity b. 1966QI-79QIV M1 > M1 velocity M3 > M3 velocity M4 > M4 velocity M5 > M5 velocity Notes: An * indicates significance at the 5 percent level (rejection of the hypothesis of noncausality). Lag lengths datermined on the basis of the FPE criterion suggested by Akaike (1969). 1966QI-1979QIV indicate that the hypothesis that these variables do not Grangercause velocity cannot be rejected for all the money aggregates. In Table 2 we examine M1 income velocity more closely. The table presents the results from estimates of equation (1) over the full sample period but including a zeroone dummy variable that takes the value of 1 in 1980QI-1988QIII and zero otherwise. This dummy allows for a different trend in M1 velocity over the 1980s. The causality result changes completely from the full sample period result in Table 1. The hypothesis that unanticipated and anticipated money supply and its volatility do not cause M1 velocity can no longer be rejected. Table 2.Granger Causality Test Results for M1 Including Zero-One Dummy, 1966QI-88QIII Hypothesis test Causal inference Lags F1 F2 F3 F4 M1 > M1 velocity FPE

5 VELOCITY AND VOLATILITY OF MONEY SUPPLY CONCLUSIONS The purpose of this paper was to test empirically the relationship between income velocity and the unanticipated and anticipated components of the money supply and the volatility of these series in the U.K. Our results are broadly consistent with those of Thornton (1991, 1995) and Arize (1993). They suggest that unanticipated and anticipated money supply and its volatility do not help predict income velocity in the U.K. and thus they were not important factors in the decline in income velocity in the 1980s. Accordingly, we find no support in the U.K. context for Friedman s hypothesis. Alternatively, Goodhart (1986, 1989) has argued that the decline in velocity in the U.K. reflected the impact of financial innovation on the demand for money with an enlarged money stock being held willingly at the prevailing pattern of interest rates. To the extent that financial innovation was the explanation, the decline in income velocity was beyond the control of the U.K. monetary authorities and, as such, a money supply growth rule was unlikely to have succeeded in the period. REFERENCES Akaike, Hirotugu, Fitting Autoregressive Models for Prediction, Annals of the Institute of Statistical Mathematics, 1969, Arize, Augustine C., Money Growth Volatility and Income Velocity in the United Kingdom, International Economic Journal, Autumn 1993, Brocato, Joe., and Kenneth L. Smith, Velocity and the Variability of Monetary Growth: Evidence from Granger-Causality Tests, Journal of Money, Credit and Banking, May 1989, Fisher, Douglas., and Apostolos Serlitis, Velocity and the Growth of Money in the United States, , Journal of Macroeconomics, Summer 1989, Friedman, Milton, Monetary Variability: The United States and Japan, Journal of Money, Credit and Banking, August 1983, Friedman, Milton, Lessons from the Monetary Policy Experiment, The American Economic Review Papers and Proceedings, May 1984, Goodhart, Charles E.A., Financial Innovations and Monetary Control, O x f o r d Review of Economic Policy, Winter 1986, Goodhart, Charles E.A., Money, Information and Uncertainty, 2nd Edn., London: Macmillan, Hall, Thomas E., and Nicholas R. Noble, Velocity and the Variability of Money Growth: Evidence from Granger Causality Tests, Journal of Money, Credit and Banking, February 1987, Mehra, Yash P., Velocity and the Variability of Money Growth: Evidence from Granger Causality Tests, Journal of Money, Credit and Banking, May 1989, Thornton, John, Velocity and the Variability of Monetary Growth in the United Kingdom, Applied Economics, April 1991,

6 66 J. THORNTON AND P. MOLYNEUX Thornton, John, Friedman s Money Supply Volatility Hypothesis: Some International Evidence, Journal of Money, Credit and Banking, February 1995, Mailing Address: Dr John Thornton, The International Monetary Fund, Room , th Street, N.W., Washington, D.C , U. S. A. Mailing Address: Dr. Philip Molyneux, School of Accounting, Banking and Economics, University of Wales, Bangor, Gwyned LL57 2DG, UNITED KINGDOM