Central Bank Intervention and Foreign Exchange Volatility.
This paper provides additional empirical evidence on the topic of the effectiveness and the impact of Federal Reserve intervention on U.S. exchange rates. Using a daily measure of exchange rate intervention in the yen/dollar and mark/dollar exchange markets for the period January 3, 1985 to March 19, 1997, this paper finds a statistically significant impact of intervention on spot rates. A generalized auto-regressive conditional heteroskedasticity exchange rate equation is used to measure the impact of intervention on exchange rate uncertainty. This study finds that intervention is associated with a significant increase in the interday conditional variance (uncertainty) of both bilateral spot exchange rates. This supports the view of Friedman and Schwartz that exchange rate intervention serves to destabilize the foreign exchange market by introducing additional levels of exchange rate uncertainty. (JEL F31)
Of the many policies undertaken by the U.S. central bank, few have seemed to generate as much controversy as foreign exchange market intervention. Opponents argue that sterilized intervention has little lasting influence on exchange rates. Moreover, when the exchange rate market disturbance is neither domestic in origin nor monetary in nature, non-sterilized intervention conflicts with price stability.
Friedman  provides the classic argument against central bank intervention in foreign exchange markets. Later, the introduction of models that allowed for imperfect information [Brainard, 1967; Poole, 1970] led to the conclusion that exchange rate policies could be used for stabilization purposes. Boyer's  work on optimal foreign exchange market intervention helped to achieve an uneasy consensus in the theoretical literature. It was shown that optimal exchange rate policies lie between the theoretical extremes of complete exchange rate fixity and flexibility. Optimal policy responses were shown to be a function of the nature of the shocks to the economy as well as dependent on the degree of capital mobility in the economy.
In contrast, empirical work on the actual impact of foreign exchange intervention has not yielded a consensus framework or result. Studies that regressed the spot exchange rate on intervention variables have often found coefficients with ambiguous signs. (1) For example, one might interpret a negative coefficient as evidence that official sales of foreign exchange caused the dollar to depreciate (a perverse response) or that official sales prevented a steeper depreciation from occurring, a "leaning against the wind" response [Humpage, 1988; Dominguez and Frankel, 1993]. Friedman  suggests a simple way to determine the desirability of official intervention: test if intervention is profitable. Friedman claims that stabilizing intervention will make money on average since monetary authorities will sell (buy) currencies when they are on average above (below) their equilibrium values. Therefore, assessing the profitability of intervention can help determine whether such policies are on average stabilizing or destabilizing. Taylor  finds that official intervention is almost always unprofitable. These initial findings led to numerous studies on this topic, some of which find strong evidence of profitable intervention. (2) Most recently, Leahy  finds that official intervention by the Federal Reserve has consistently generated profits.
These conflicting results have led many researchers to adopt different empirical methodologies to study the impact of intervention. (3) However, these studies have done little to narrow the gap in opinion concerning intervention. Recent academic work concerning the appropriateness and effectiveness of official intervention range from Dominguez and Frankel's  generally favorable view to Schwartz'  contention that intervention is an "exercise in futility" that at best can have only a very short-run effect on exchange values and at worse serve to introduce harmful amounts of uncertainty and volatility in foreign exchange markets.
This paper provides additional empirical evidence on the topic of the effectiveness and the impact of Federal Reserve intervention on U.S. exchange rates. Using a daily measure of exchange rate intervention in the yen/dollar and mark/dollar exchange markets for the period January 3, 1985 to March 19, 1997, this paper finds a statistically significant impact of intervention on spot rates. Following Baillie and Humpage , a generalized autoregressive conditional heteroskedasticity (GARCH) exchange rate equation is used to measure the impact of intervention on exchange rate uncertainty. Researchers in finance and economics have argued that a GARCH framework provides an efficient parametric way of modeling uncertainty in high frequency econometric time series. This study finds that intervention is associated with a significant increase in the interday conditional variance (uncertainty) of both bilateral spot exchange rates. This supports the view of Friedman  and Schwartz  that exchange rate inte rvention serves to destabilize the foreign exchange market by introducing additional levels of exchange rate uncertainty.
This paper proceeds as follows. The second section discusses the recent history of Federal Reserve foreign exchange intervention in the post-Bretton Woods period. The third section provides a discussion of the data used in this study, and the fourth section outlines the statistical methodology and discusses the empirical results. The fifth section provides a brief conclusion.
A Brief History of Intervention
In March 1973, the major industrialized countries discontinued pegging their exchange rates to the dollar, thereby ushering in an era of managed floating exchange rates. Throughout the remaining decade, periodic central bank intervention in the foreign exchange market took place. Schwartz  identified the three most important episodes in this period. From October 1974 to March 1975, the Federal Reserve (in conjunction with the German and Swiss central banks) coordinated intervention to minimize exchange rate volatility and halt the depreciation of the U.S. dollar. Second, from September 1977 to December 1979, U.S. monetary authorities (in cooperation with the Japanese and German central banks) intervened to attempt to halt the depreciation of the dollar. By the end of the decade, the Federal Reserve was intervening in foreign exchange markets "virtually on a day-to-day basis" [Schwartz, 1996, p. 385].
The period 1981 to 1986 saw a dramatic change in exchange rate policy. Official intervention by U.S. monetary authorities was severely limited. This period saw a large rise in the value of the dollar from 1981 to 1985. However, in the second Reagan administration, the U.S. Treasury (led by James Baker) backed away from the previous nonintervention stance of U.S. policy. The famous Plaza accord of September 1985 committed U.S. and foreign monetary authorities to sell dollars in order to depreciate the perceived overvalued dollar.
The Louvre accord, reached on February 20, 1987, marks the next interesting episode of exchange rate intervention. This was an agreement of finance ministers of the G7 nations (U.S., United Kingdom, France, Germany, Japan, Italy, and Canada) concerning intervention. In contrast to the Plaza, the Louvre accord focused on stabilizing exchange rates around current levels. The next two years saw frequent and large amounts of exchange rate intervention by the Federal Reserve. In more recent years, official U.S. intervention has usually been limited to the Japanese yen and German mark [Schwartz, 1996 p. 387]. Furthermore, in recent years, the Federal Reserve has routinely sterilized U.S. intervention. Monetary policy makers have not allowed their intervention to interfere with the attainment of their federal funds rate target. Nevertheless, unexpected changes in monetary policies can affect exchange rates, and interventions that coincide with them will appear more successful than they otherwise would.
Since the yen and the mark are the two major currency instruments used by the Federal Reserve in its intervention policies, this study uses the spot exchange rate between these two currencies against the U.S. dollar to examine the effectiveness of the Federal Reserve's intervention policies.
The daily intervention data used in this study was provided by Owen Humpage  of the Federal Reserve Bank of Cleveland. The intervention data are official U.S. interventions against marks and yen. A positive (negative) value for the intervention term represents net purchases (sales) of the foreign currencies under investigation. The exchange rates are the mark and the yen per dollar. Table 1 presents summary statistics for these interventions.
All official exchange rate intervention is undertaken by the Federal Reserve Bank of New York as an agent of both the U.S. Treasury and the Federal Reserve Open Market Committee. Almost all U.S. intervention occurs between 9:00 a.m. and 4:00 p.m. Therefore, it is assumed that all intervention on day t occurs before the close of the New York market on day t. This time series runs from January 3, 1985 to March 19, 1997. The bilateral exchange rates for this period are afternoon closing quotations (bids) from the New York market.
Model and Empirical Results
A large body of literature in the international finance arena suggests that exchange rates follow a Martingale process with heteroskedastic errors. Economists often quantify exchange rate volatility in terms of the second moment of the exchange rate process. Hung  uses the standard deviation and Bonser-Neal and Tanner  consider implied volatility as defined by option pricing. Baillie and Bollerslev , Almekinders and Eijffinger , Dominguez , and Baillie and Humpage  have applied GARCH methods to model the heteroskedastic error terms in exchange rate equations. This study takes this latter tack and employs the following exchange rate model: (4)
[increment of [S.sub.t]] = [alpha] + [beta][int.sub.t] + [[epsilon].sub.t], (1)
[[epsilon].sub.t]\[[OMEGA].sub.t-1] [sim] D(0, h), (2)
[h.sub.t] = [[omega].sub.0] + [SIGMA][gamma][[epsilon].sup.2.sub.t-i] + [lambda][int.sub.t], (3)
where [increment of[S.sub.t]] is the log change in the exchange rate at time t; [[OMEGA].sub.t-1] is the information set in t-1; and [int.sub.t] is the intervention variable at time t. (5) Variables [int.sub.t] and [[OMEGA].sub.t-1] are statistically independent since the frequency of the data renders the possibility that they are both influenced by an excluded variable improbable. In addition, the possibility is also statistically untestable due to the lack of data at the frequency needed for this analysis. Equation (2) shows that the distribution of the error term is conditional on the information set,[[OMEGA].sub.t-1], available at time t - 1 . The conditional density, D, which can be either normal or student t, can often accommodate the extreme kurtosis that high-frequency exchange rate data frequently exhibit. The distribution approaches normal as parameter v approaches 30 [Baillie and Bollershev, 1989]. Equation (3) models the conditional variance as an MA(q) process and a function of intervention at time t. T he model allows us to estimate the effects of intervention on both the conditional mean and variance of the exchange rate process.
Table 2 presents ordinary least squares (OLS) results for equations in which the changes in the yen/dollar and mark/dollar exchange rates are regressed against the corresponding intervention terms. The signs on the intervention terms suggest that official intervention sales of dollars against either currency are associated with dollar depreciation. ARCH tests (for lags 1, 4, and 8) reveal a strongly significant and persistent pattern for the conditional variance of both equations.
Table 3 presents the results obtained from estimating ARCH equations for both bilateral exchange rates, which allows the intervention terms to affect both the conditional mean and variance of the series. (6) The conditional variance provides an excellent proxy for near term exchange rate volatility. Wald test statistics indicate that the conditional variance is stationary for all four regressions reported in Table 3. In addition, the signs on the intervention terms in the mean equations suggest that official sales of dollars against either currency are associated with dollar depreciation. The coefficients on the intervention terms in the conditional variance equations reveal that official intervention leads to an increase in exchange rate uncertainty. (7) This can be due to the fact that official intervention provides a signal to market participants about concern over the stability of the market. In addition, it adds a degree of additional (policy) uncertainty concerning the intensity and persistence of the i ntervention policies. This indicates that far from stabilizing exchange markets, official intervention resulted in an increased amount of instability. Ljung-Box Q statistics fail to reject the null hypothesis of no residual correlation in the estimated ARCH regressions. These findings provide further empirical support for the theoretical arguments concerning the perils of exchange rate intervention found in Friedman  and Schwartz . (8)
This paper assessed the impact of official U.S. exchange rate intervention against the yen and mark. Using daily data on intervention, covering the period January 3, 1985 to March 19, 1997, it is found that the volume of official intervention is associated with a significant increase in exchange rate uncertainty. This finding supports the view of those who argue that exchange rate intervention serves to disrupt exchange rate markets.
(*.) Ohio University--U.S.A.
(1.) The Federal Reserve reveals the timing and direction of their official policies to foreign exchange market participants. As is consistent with many other studies, our intervention variable is a simple dummy that takes on the value of 1 for a particular intervention (one dummy variable for sales and one for purchases of a particular currency).
(2.) See Edison  for a survey of the profitability studies and all recent work on intervention.
(3.) Almekinders and Eijffinger  provides useful surveys of these studies.
(4.) Since the periodicity of the data used in this model is daily, we are precluded from incorporating the possible influence of other market fundamentals such as output, interest rates, and inflation.
(5.) Both the Dickey-Fuller and Phillips-Perron unit root tests show that the exchange rate series are not stationary. So this study uses the first differences of the log of the exchange rates.
(6.) The conditional variance is modeled as an ARCH(8) process. Similar results are obtained if a GARCH(1,1) model is used instead.
(7.) We interpret the positive correlation between official intervention and the conditional variances as evidence concerning the impact of intervention. We argue that it suggests a causal relationship because decisions to intervene are made before the start of each trading day. Therefore, intervention could not be caused by the increased volatility since data on the contemporaneous amount of volatility would not be observed until sometime after the trading day had begun.
(8.) Equations (2) and (4) of Table 3 allow the conditional standard deviation to enter the mean equation for the exchange rate models. The results of these ARCH(8)-M models provide similar conclusions concerning the effect of intervention than did the previous ARCH models.
Almekinders, Geert J.; Eijffinger, Sylvester C. W. "Empirical Evidence on Foreign Exchange Market Intervention: Where Do We Stand?," Weltwirtschaftliches Archiv, 127, December 1991, pp. 645-77.
Baillie, Richard T.; Bollerslev, Tim. "The Message in Daily Exchange Rates: A Conditional Variance Tale," Journal of Business and Economic Statistics, 7, July 1989, pp. 297-305.
Baillie, Richard; Humpage, Owen F. "Post-Louvre Intervention: Did Central Banks Stabilize the Dollar?," working paper, Federal Reserve Bank of Cleveland, October 1994.
Bonser-Neal, Catherine; Tanner, Glenn. "Central Bank Interventions and the Volatility of Foreign Exchange Rates: Evidence from the Options Market," Journal of international Money and Finance, 15, December 1996, pp. 853-78.
Boyer, Russell S. "Optimal Foreign Exchange Market Intervention," Journal of Political Economy, 86, 1978, pp. 1045-55.
Brainard, William C. "Uncertainty and the Effectiveness of Policy," American Economic Revenue Papers and Proceedings, 57, May 1967, pp. 411-25.
Dominguez, Kathryn Mary. "The Informational Role of Official Foreign Exchange Intervention Operations: The Signaling Hypothesis," Exchange Rate Efficiency and the Behavior of International Asset Markets, New York, NY: Garland Publishing, 1992.
Dominguez, Kathryn Mary; Frankel, Jeffrey A. Does Foreign Exchange Intervention Work?, Washington, DC: Institute for International Economics, 1993.
Edison, Hali J. "The Effectiveness of Central Bank Intervention: A Survey of the Literature After 1982," Special Papers in Economics, 18, July 1993.
Friedman, Milton. Essays in Positive Economics, Chicago, IL: University of Chicago Press, 1953.
Humpage, Owen F. "Interventions and the Dollar's Decline," Economic Review, 24, 1988, pp. 2-17.
___. "U.S. Intervention: Assessing the Probability of Success," working paper, Federal Reserve Bank of Cleveland, January 1997.
Hung, Juann H. "Assessing the Effect of Sterilized U.S. Foreign Exchange Intervention: A Noise Trading Perspective," unpublished manuscript, Federal Reserve Bank of New York, January 1992.
Leahy, Michael. "The Profitability of U.S. Intervention in Foreign Exchange Markets," Journal of International Money and Finance, 14, 1995, pp. 823-44.
Poole, William. "Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model," Journal of Political Economy, 84, May 1970, pp. 197-216.
Schwartz, Anna J. "U.S. Foreign Exchange Market Intervention Since 1962," Scottish Journal of Political Economy, 43, September 1996, pp. 379-97.
Taylor, Dean. "Official Intervention in the Foreign Exchange Market, or Bet Against the Central Bank," Journal of Political Economy, 90, April 1982, pp. 256-68.
TABLE 1 U.S. Intervention: Basic Statistics Against Marks Against Yen Sales Purchases Sales Mean 182.78 -145.17 190.56 Median 100.00 -100.00 150.00 Standard Deviation 192.99 -124.08 179.71 Minimum 15.00 -797.00 3.00 Maximum 850.00 -10.00 800.00 Number of Days Intervened 86 138 82 Against Yen Purchases Mean -136.34 Median -100.00 Standard Deviation 106.94 Minimum -555.00 Maximum -3.00 Number of Days Intervened 108 Notes: Figures are in millions of dollars TABLE 2 OLS Model and ARCH Tests Mark/Dollar Exchange Rate Yen/Dollar Exchange Rate OLS Constant -0.000497 (0.000264) (**) -0.041850 (0.017780) (*) Intervention -8.91E-06 (4.55E-06) (**) -0.000875 (0.000332) (*) ARCH Tests ARCH(1) 119.3802 (*) 66.53352 (*) ARCH(4) 232.5774 (*) 105.0125 (*) ARCH(8) 381.7880 (*) 111.6118 (*) Notes: (*) and (**) denote significance at the 1 and 5 percent levels, respectively. Standard errors are in parentheses. The Lagrange n x [R.sup.2], where n is the number of observations and [R.sup. 2] is the coefficient of determination for the estimated (1). This test statistic is distributed as [X.sup.2], with degrees of freedom equal to the number of the lag terms in the model. The critical values at the 10 percent level with 1, 4, and 8 degrees of freedom are 2.706, 7.779, and 13.362, respectively. The dependent variable is long change in closing mark/dollar to yen/dollar exchange rate. TABLE 3 Estimation of Exchange Rate Variance as a GARCH Process Mark Dollar (1) (2) Conditional Mean [square root of (h)] -0.1111481 (0.065286) (***) Constant -0.000259 0.001006 (0.000208) (*) (0.000802) Intervention -1.29E-05 -1.27E-05 (3.17E-06) (*) (3.11E-06) (*) Conditional Variance Constant 6.34E-05 6.33E-05 (3.24E-06) (*) (3.25E-06) (*) ARCH(1) 0.112376 0.116226 (0.017216) (*) (0.017644) (*) ARCH(2) 0.073258 0.068332 (0.015878) (*) (0.015896) ARCH(3) 0.063280 0.064296 (0.015445) (*) (0.015543) (*) ARCH(4) 0.104270 0.108548 (0.016190) (*) (0.017203) (*) ARCH(5) 0.099229 0.094104 (0.019353) (*) (0.018926) (*) ARCH(6) 0.076142 0.076046 (0.016150) (*) (0.016549) (*) ARCH(7) 0.062226 0.034596 (0.016006) (*) (0.015987) (*) ARCH(8) 0.093475 0.092714 (0.015887) (*) (0.015917) (*) Intervention 7.91E-08 7.92E-08 (2.74E-08) (*) (2.70E-08) (*) Diagnostics Log Likelihood 9,035.262 9,036.242 Q(15) 13.899 13.420 Yen Dollar (3) (4) Conditional Mean [square root of (h)] -0.172687 (0.079755) (**) Constant -0.011412 0.134964 (0.0163032) (0.071454) (**) Intervention -0.000876 -0.000852 (0.000215) (*) (0.000216) (*) Conditional Variance Constant 0.432471 0.430951 (0.14906) (*) (0.014974) (*) ARCH(1) 0.228858 0.227890 (0.010359) (*) (0.010506) (*) ARCH(2) 0.056690 0.054127 (0.014427) (*) (0.013904) (*) ARCH(3) 0.081250 0.082597 (0.013874) (*) (0.013740) (*) ARCH(4) 0.012402 0.013612 (0.011650) (0.011803) ARCH(5) 0.087601 0.082390 (0.015523) (*) (0.015042) (*) ARCH(6) 0.056345 0.059947 (0.014777) (*) (0.014543) (*) ARCH(7) -0.004357 -0.002406 (0.009499) (*) (0.009659) (*) ARCH(8) 0.062008 0.062947 (0.012369) (*) (0.012341) (*) Intervention 0.000815 0.000844 (0.000143) (*) (0.000145) (*) Diagnostics Log Likelihood -4,077.826 -4,075.728 Q(15) 13.412 11.827 Note: (*), (**), and (***) denote significance at the 1, 5, and 10 percent levels, respectively. There was 3,073 observations. The Ljung-Box statistic is distributed as [X.sup.2] with 15 degrees of freedom, Q(15). The critical values are 22.307, 24.996, and 30.578 at the 10, 5, and 1 percent levels, respectively.
|Printer friendly Cite/link Email Feedback|
|Author:||Doroodian, K.; Caporale, Tony|
|Publication:||International Advances in Economic Research|
|Article Type:||Statistical Data Included|
|Date:||Nov 1, 2001|
|Previous Article:||Unemployment Insurance, Moral Hazard, and Economic Growth.|
|Next Article:||Intraindustry Trade in Intermediate Goods.|