The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Does The Federal Reserve Affect Asset Prices? Vefa Tarhan Working Papers Series Issues in Macroeconomics Research Department Federal Reserve Bank of Chicago January 1992 (WP-92-3) FEDERAL RESERVE BANK OF CHICAGO DOES THE FEDERAL RESERVE AFFECT ASSET PRICES? by Vefa Tarhan Loyola University of Chicago Department of Finance June 1991 First Revision: Second Revision: May 1990 October 1990 I would like to thank Tim Bollerslev, Ravi Jagannathan, David Marshall, Paul A. Spindt, and Steven Strongin. The usual caveat about any error in the paper applies. Eric Klusman provided excellent research assistance. Financial support from the Federal Reserve Bank of Chicago is gratefully acknowledged. ABSTRACT The Federal Reserve is probably one of the institutions most closely monitored by investors. This indicates that investors believe the actions of the Fed have implications for asset prices. However, studies detect no empirical relation between money growth and interest rates. To this date, the trading activities of the Fed in the financial markets have not been examined to see whether the Fed has the ability to influence asset prices. Using daily data on Open Market Operations (OMOs) and asset prices, this study fills this void. One finding of the paper is that OMOs Grangercause both short and long term interest rates. Judging by the impulse response paths, the effects of monetary policy appear to be confined to the short run. Furthermore, the sign of the relationship confirms the existence of the much hypothesized liquidity effect. Additionally, daily OMOs appear to have some impact on exchange rates, but not on stock prices. This paper also investigates the impact of monetary policy on asset return volatility. The evidence indicates that OMOs have a dampening effect on volatility in some of the financial markets examined. 1 1. INTRODUCTION This paper investigates the empirical link between Open Market Operations (OMOs) and asset prices. Additionally, the connection between OMOs and the volatility of asset returns is also examined. While the primary emphasis of the paper is the impact of OMOs on both short and long term interest rates, the influence of OMOs on the stock and the foreign exchange markets are also analyzed. Whether or not the Fed has the ability to influence asset prices is of prime interest to macroeconomists that investigate the connection between monetary policy and the real economy. The issue is also important to investors who are concerned with the value of their portfolios. Recent papers by Bernanke (1990), Bernanke and Blinder (1990), Kuttner and Friedman (1991), Stock and Watson (1989), and Strongin (1991), demonstrate that interest rates are very informative about future movements of real macro variables. that the spreads, Federal and the perform very well funds rate, various spread between the In particular, they find short short term and as predictors of business interest long term cycles. If rate rates indeed there is such a link between interest rates and the real sector of the economy, the implication of these findings is that monetary policy can be used to influence output, provided the Fed has the ability to influence interest rates. However, attempts to empirically document the much hypothesized negative correlation between monetary policy and interest rates (the liquidity effect) has met with failure. In a survey paper, after updating some of 2 the empirical studies on this topic, Reichenstein the conclusion empirical that support since for at the least April existence of (1987) 1975, the reaches there liquidity is no effect. However, the failure of previous studies to document the liquidity effect may have been due to their research design. a strong case interest can be made rates is not that the the impact appropriate In particular, of money nexus for growth the on empirical examination of the liquidity effect. This paper provides strong evidence for the contention that daily Open Market Operations influence asset prices. the sign effect: rates, of the relation Injection and of is as reserves hypothesized into reserve withdrawals the by system increase Furthermore, the liquidity lowers interest interest rates. Taken together with the evidence linking interest rates to real macro economic activity, this finding supports the view that monetary policy influences the real sector of the economy. The finding that the actions of the Fed influence asset prices probably does not come as a surprise to most investors. managers in convinced houses, the of this employ activities of financial that all economists the Fed. sector major as By "Fed of the banks, economy as well watchers", employing these to In fact must as be so brokerage monitor individuals the their employers must be hoping to receive information about the "correct" interpretation of the Fed's transactions. This in return, presumably, leads to potentially profitable trading strategies, or avoidance of losses on portfolio values. 3 In addition prices, this to paper the impact empirically of the examines Fed's actions the connection daily OMOs and the volatility of asset returns. asset returns considerations. is crucial to investors for on asset between Volatility of asset pricing It is also likely that volatility has a bearing on the real sector of the economy by influencing capital budgeting and consumption decisions. There is a substantial body of papers that document the time variation of asset return distributions. the form of volatility of asset returns is well While documented by statistical conditional variance models, the sources of volatility has not been investigated as extensively.1 Monetary policy may be an important factor in asset return volatility. However, it is not clear, on theoretical grounds, whether monetary policy would dampen or magnify the volatility of asset returns. 2. MONETARY AUTHORITY AND ASSET PRICES There are some theoretical models in the literature that demonstrate that the monetary authority, by conducting open market operations, can influence interest rates. Grossman and Weiss (1983) and Rotemberg (1984) show that in a world where money is needed to execute transactions both in the goods and financial markets, a one time unanticipated sale of bonds by the central bank will raise interest rates. In Grossman and Weiss, the central intervals. bank go to As a result, supply at any one time. the bank the agents that trade with to withdraw cash at fixed they hold a small portion of the money Given this, the open market sale raises 4 interest rates, not because it changes inflationary expectations or the real rate, but because the traders do not have the ability to obtain more money, i.e. they are liquidity constrained. Lucas (1988) developed a model along the same lines. In his model, the agents that trade in goods and securities face separate liquidity constraints, but are members of the same family, bound by a household utility function. The representative household has three members, balance. an endowment of goods, and an initial cash The initial cash balances are allocated to the purchase of goods and securities at the beginning of the period. The only shock to the system in this model is in the form of an open market operation. This shock takes place after the allocation of the family's funds among the two purchasing activities has already been made. Furthermore, this shock is observed only by the agents that trade in the securities market. As a result, the shock in question affects neither the distribution of cash balances between financial market and goods market purchases, nor the prices in the goods market. The only response to open market operations shocks takes the form of changes in bond prices. Given that the cash raised from the sale of family endowments is not available in the current period, and that the goods market is unaware of the shock, bond prices need to change for the markets to clear. The marginal rate of substitution in this model is constant. Bond prices change due to the liquidity expectations transaction. constraint, implications of and the not due to unanticipated inflationary open market 5 The Keynesian concept of liquidity is somewhat different. In a Keynesian world, goods prices do not respond to open market transaction shocks because prices in the goods market are "sticky". However, unlike the case in the models discussed above, the real rate changes as a result of the unanticipated open market operation. The Keynesian model typically is not cast in the context of the individual consumer. However, presumably individuals hold money for purposes of executing goods and securities transactions. Faced with an unanticipated open market sale, in order for the individual to be convinced to hold more securities (thus, consume less today), he has to be offered a higher real rate. This means that, in a Keynesian world, the nominal rate changes are triggered by changes in the real rate when the monetary authority engages in a bond sale. Differences in what is meant by the liquidity effect not withstanding, above both the Keynesian model and the models discussed agree on the sign of the interest rate response to open market transactions: security purchases by the monetary authority lower interest rates, while security sales cause the rates to be higher. However, empirical studies fail to confirm the existence a negative correlation between money growth and interest rates.2 In a recent study, using data targeting operating procedure period, from the Fed funds rate Cook and Hahn (1989) show that changes in the Fed funds rate target caused changes in other interest September rates. 1979, They find changes in that the during Fed September funds rate 1974 through caused large 6 movements in the short term rates and movements in the longer term rates. small While but this significant is not direct evidence of the existence of the liquidity effect in the sense of a negative correlation between money and interest rates, it indicates that the Fed has the ability to influence interest rates. Additionally, the findings of the money supply announcements literature also implies that the Fed has the ability to influence asset prices. These studies document the reaction of interest rates to money announcement surprises. While the interest rate response is consistent both with the expected liquidity effect and possible changes in inflationary expectations, recent evidence (Strongin and Tarhan (1990), Hardouvelis (1987)) support the contention that interest rates respond due to expected liquidity considerations.3 If indeed the response of interest rates is triggered by anticipations of the Fed's reaction to the money figures (expected liquidity effect), the failure to document the actual liquidity effect may just be indicative of a problem in empirical test design. The relationship between OMOs and interest rates may prove to be a better forum in which to investigate the existence of the liquidity effect than examining the link between money and interest rates. One difference between this study and the previous studies is that, in this paper, the relationship examined is the one between open market operations and asset prices, and not the one between money and asset prices. Since the question investigated in this study is the ability of the Fed to affect interest rates, 7 the causal link between asset prices and a variable over which the Fed has direct control is the appropriate avenue of inquiry. The Fed has the ability to control the level of reserves in the system by its conduct of open market operations (OMOs) . Changes in the level of reserves triggered by OMOs translate into changes in money, via the multiplier process, as investors and financial institutions respond to the Fed's actions. intentions of the Fed more Thus, OMOs capture the accurately than the growth rate of money, which is jointly determined by the Fed, the public, and the financial institutions. Additionally, OMOs have the advantage of being events that are readily observed by market participants as soon as they are executed.4 In contrast, when investors observe the growth studies show that it is not clear exactly in money. In fact, empirical interest rates respond to money announcements even though the money growth that is being announced had already been determined ten days prior to the announcement.5 This may indicate that investors do not observe money growth when it takes place. OMOs are free of this problem. The sample period in this study 31, 1984) October 6, (October 2, 1979 - December covers the October 1979 - October 1979, the Federal Reserve 1982 announced period. changes in On its operating procedures. Prior to this date, the short term focus of monetary policy centered on maintaining the Fed funds rate within a narrow target range. operating policy from The new procedures changed the focus of targeting Fed funds, to a policy attempts to accomplish monetary policy objectives by that targeting 8 reserves. Specifically, the desk was directed to set and maintain a target path for nonborrowed reserves consistent with long term money growth Reserve regime again can targets.6 changed best be In the late Fall of 1982, its operating procedures. characterized by a set of the Federal The post-1982 procedures that targets borrowed reserves.7 Using data from the post October 1979 period is ideal for the empirical question examined in this paper. was characterized by a regime that targets If the period studied interest rates, the power of empirical tests conducted would be potentially low. To see this, assume that in fact the Fed can affect interest rates by its actions. As the Fed funds rate starts to diverge target range, the Fed will supply reserves to or drain reserves from the system in order to prevent rates from changing. from its If the Fed indeed has the ability to control interest rates, the data will show substantially more variation in OMOs than in interest rates. In fact, in the extreme, if the Fed had a point target rather than a range, and is successful in achieving its target, there will be no variation in interest rates. In such a scenario, the empirical tests will detect no causality between OMOs and interest rates, when in reality the Fed will have perfect control over interest rates.3 3. OPEN MARKET OPERATIONS The operating policy objective of the trading desk is to implement the monetary policy objectives set by the Federal Open Market 9 Committee (FOMC). The FOMC meets six to eight times a year, decides on period, the a course for monetary policy. targets for the conduct During of the monetary and 1979-1982 policy were expressed in terms of the desired growth for monetary aggregates. The Desk, then, had the responsibility of converting these money growth targets into the implied target path of nonborrowed reserves for the intermeeting period. The next step was conduct the OMOs, so as to make realized for the Desk to level of nonborrowed reserves for the intermeeting period equal the weekly average of the specified target path. An algorithm describing the implementation of the monetary policy during this time period is developed in Spindt and Tarhan (1987). The first step in converting the FOMC's monetary policy objectives involves the computation of the weekly target for money. The next step is to determine the path of required reserves implied by the projected path for accounting regime that was banks' money. Under the lagged in effect during the reserve sample period, required reserves for a given week were determined on the basis of deposits they held two weeks previously. path for required reserves Given this, the consistent with the intermeeting period targets is obtained by multiplying the average reserve requirement ratio with projected money figure from two weeks earlier. the projections for excess reserves are added to the Next, required reserves path to obtain the desired path for total reserves. The target directives, path is for nonborrowed then calculated reserves by implied by subtracting the the FOMC level of 10 discount window borrowing assumed by FOMC Assumption"), from the total reserves path. Desk is to nonborrowed conduct reserves OMOs, such during the that ("Initial Borrowing The objective of the the realized intermeeting average weekly nonborrowed reserves target. level period The of equal the step in last determining the magnitude of the OMOs for a given week involves estimating the exogenous factors that will influence the level of reserves in the system. These projections, then, are netted out from the target nonborrowed reserves path to arrive at the size of OMOs to be conducted during that week. Among the exogenous factors ("technical operating factors") that the Desk needs to forecast, three are especially important.8 These factors difficulty, Reserve have the float. considerable most crucial Float arises of due size. In these to terms factors the fact of forecasting is the Federal that the Federal Reserve credits banks for the checks presented, on a shorter time schedule than the time it takes to collect on these checks. Because of unforeseen problems in communication and transportation, the level of float can show substantial variability. During the sample period float projection errors in the magnitude of 500 million 1 billion dollars per day were not uncommon. In comparison, to the average amount of reserves injection needed to sustain a 6 percent annual growth in Ml during 1981 was in the order of 50 million dollars per week.9 The second source of uncertainty about the actual level of reserves in the system involves the deposits of the U.S. 11 Treasury. The Treasury maintains accounts with various commercial banks (Tax and Loan accounts). It periodically transfers funds from these accounts to banks. its checking accounts at the Federal When this happens, the amount of reserves declines. When the Treasury writes Reserve in the system checks against its accounts at the Federal Reserve, on the other hand, reserves are injected into the system once the Fed credits the reserve balances of commercial banks. At times, there can be significant errors in projecting the reserve implications of these transactions.10 The portfolio decisions of foreign central banks constitute the third significant accommodates these their trades, exogenous reserve factor. institutions by becoming a When the Fed counter party to it in effect injects or drains reserves. Thus, the desk needs to adjust its OMOs for the expected level of foreign central bank trades with the Fed. Once the size of the 0M0 is determined, the Desk must determine how to execute the required trades. On this issue, the Desk has considerable flexibility. The alternatives available to the Desk have three broad characteristics. the choice between OMOs that will permanently affect the level transactions whose initial reserve implications will be reversed in the near (repurchase transactions), versus of reserves those future (outright One decision involves agreements and matched sale-purchases). The second decision that needs to be made is whether to execute trades in the secondary market, with dealers, or with foreign central banks. The third issue involves the type of securities to be used 12 in conducting the OMOs. Here, one choice would be between short term Treasury Treasuries securities (Treasury bills), longer term (coupon securities). Another choice would be between Treasuries and Federal Agency securities. the Desk, and The primary concern of in comparing the various alternatives, is to select a course of action such that the Fed will not have undue influence in the determination of security prices. As a result, the Fed monitors the security portfolios of dealers, and attempts to conduct OMOs only dealers with those quantities. securities Thus, that have in sufficient even when the Fed wants to influence security prices, it does not want the security prices to change because of market micro that, structure considerations. for the most part, In practice, this means Treasury bills end up as the preferred instrument of OMOs. An 0M0 could be in the form of an outright transaction. In the case of an outright purchase (sale) , the reserves in the system increase (decrease) permanently. Repurchase agreements (RPs) and matched sale-purchases (MSPs), on the other hand, change the level of reserves for a limited time period. In an RP transaction, the level of reserves increases in the current period, but this increase is drained from the system when the RP matures. transaction, in essence, is a reverse RP, and thus, A MSP the initial decline in the level of reserves is offset at its term. The sample December 31, period 1984. for this study is Open market operations October 2, 1979 to are measured as the 13 changes in the Fed's portfolio. The daily data on the Fed's portfolio was obtained from the Federal Reserve Board. 4. EMPIRICAL RESULTS: OPEN MARKET OPERATIONS AND ASSET PRICES The impact of the Fed's actions on asset prices is first examined by conducting bivariate Grange causality tests between interest rates and OMOs. These tests are later repeated for stock returns and returns on foreign exchange market variables. i. Open Market Operations and Interest Rates The hypothesis that open market operations (AFPt) do not Granger cause interest rates (rt) is examined by testing the joint hypothesis that >Si=0 in the following regression. 6 6 Art = ao + E a . Art - i + E i =1 P i AFI>t - \ + e t (i) l =1 The interest rates used cover the maturity spectrum from 1 day (the Fed funds rate) to 30 years (the 30 year Treasury bond rate). In addition to these two rates other interest rates examined are the Treasury bills of 3, 6 and 12 month maturities and Treasury bond rates of 1, 2, 3, 5, 10 and 20 year maturities. The Granger causality test results obtained from (1) for changes portfolio in interest rates and the (AFPt) are presented in Table I.11 change estimating in the Fed's The results indicate 14 that the null hypothesis that OMOs do not Granger-cause interest rates is rejected . These results show that OMOs affect both short and long term interest rates. While these results strongly indicate that the actions of the Fed influence interest rates, what needs to be determined is whether or not the sign of the relationship is as predicted by the liquidity effect. To see whether increases in the size of the Fed's portfolio (transactions that inject reserves into the system) result in standard lower interest deviation examined. shock rates, impulse response in orthogonalized OMO paths to innovations a are Cumulative response of interest changes as of 1, 6, and 12 days after the OMO shock are reported in Table II. These results appear reasonable on various grounds: predicted by negative, indicating injections. the liquidity Second, that A innovations one rates as expected, more than long term rates, maturity. effect, the fall sign in in First, as all response cases to is reserve short term rates are affected and the impact of OMOs decline with standard deviation shock in orthogonalized OMO generate a 7.5 basis point reduction in the Fed funds rate in the first period. The bill rates decline by around 1.2 basis points, and the 30 year rate declines by 0.8 basis points.12 Cook and Hahn (1989) find that Fed funds rate changes constant effect across the 3 months to 12 months horizon. have a This is confirmed by the results displayed in Table II. However, there are two differences: First, in their case the magnitude of the T-bill response is roughly one half of the change in the Funds rate while 15 here it is around 16 percent. is constant across the Second, while the initial response Treasury bill maturity spectrum, the cumulative response after 6 and 12 days varies with the maturity of Treasury bills. After 6 days the cumulative Fed funds rate response is 13 basis points, and the bill rate responses are in the range of 20 to 3 0 percent of this magnitude. Cumulative responses after 12 days are similar in relative magnitude to the response after 6 days. One notable exception is that at the longest end of the maturity spectrum (20 and 30 year rates) , the cumulative response is close in magnitude to the initial response. The statistical significance of the impulse responses can be seen in Figures 1 to 4, which display the impulse response paths of selective interest rates (in basis points) to a one standard deviation shock in OMOs over a 20 day horizon (the response paths are bounded by 95% confidence intervals). The interest rates in question are the Fed funds rate, the 3 months bill rate, Treasury bond rates of 2, and 20 year maturities. and the These graphs show that the impact of monetary policy on interest rates is felt very quickly. The adjustment of interest rates to monetary shocks is close to completion within 10 days. Combined with the results of Table II, these figures indicate that the adjustment of interest rates to monetary policy is fast, and also that monetary policy have a lasting impact on interest rates (interest rates stabilize at lower levels following a reserves injection). Table II also displays the response of OMOs to own shocks. The initial period shock is around 1.9 billion dollars. The initial 16 positive response of OMOs is reversed later on. As a result, the cumulative response after 12 days is in the range of 850 to 870 million dollars. In order to test whether interest rates Granger cause OMOs, the following equation is estimated: 6 A F P t = Yo + £ 6 Yi A F P t_5 + £ i “1 6 i A r t _, + u t (2) i “1 The results obtained from estimating (2) are also displayed in Table I. that The null hypothesis of no causality (joint hypothesis <Sj = 0) cannot be rejected for any of the interest rates. first glance, includes Tarhan the this is surprising. October (1987) report 1979 After all, to October empirical 1982 the sample period period. results that At Spindt supports contention that during this time period the target the and Fed's of monetary policy was nonborrowed reserves and not interest rates. However, the presence of Granger causality running from interest rates to OMOs does not necessarily imply that the Fed was following leaning-against-the-wind policy with respect to interest rates. a In fact, upon examination of the 0M0 impulse response to innovations in interest rates, it becomes apparent that both the sign and size of the responses are incompatible with an interest rate targeting procedure. The response of OMOs displayed in Table III. to innovations in interest rates is Comparing the impulse responses to 0M0 17 innovations (Table II) with impulse responses innovations, two things stand out. to interest rate First, looking at the initial period responses, the size of the reserve injection triggered by a one standard deviation shock in interest rates is very small in relation to the size of the shock: For example, a 63 basis increase in the Fed funds rate results in a reserve increase in the amount of only reserve 304 million increases dollars. in the Whereas, magnitude of as 1.9 discussed billion before, dollars required to move the Fed funds rate by 7.5 basis points initial period. increases initially case of in the Perhaps more importantly, while an interest rate leads to a reserve injection, withdrawn from the system in the subsequent periods. in the is a Fed funds rate shock, even reserves are For example, though reserves increase initially by 304 million dollars, the cumulative response after 12 days is a reserve decrease of 25.6 million dollars. pattern holds true for interest rate shocks of all maturities. This In all cases small initial increases in reserves in the initial period are reversed in the subsequent periods such that the net response to interest rate increases is a small withdrawal of reserves from the system. decline, The important point here is not that reserves actually but that they do not change by a meaningful amount in response to interest rate increases. Thus, while there is evidence of Granger causality running from interest rates to open market operations, the size and the sign of the cumulative 0M0 response is not indicative rates. of a central bank policy that targets interest 18 Figures dollars) to 5 and a standard interest rates 6 show the estimated deviation shock response in two of OMOs (in representative (the Fed funds rate and the 5 year Treasury rate respectively). The cumulative response of interest rates to own shocks indicate that the initial movement in interest rates persist. By the end of 12 days, initial interest rate shocks in the magnitude of 63 to 12 basis points produce a net increase of 40 to 14 basis in interest rates, but a small decline in the volume of reserves in the system (26 to 135 million dollars). Figures 7 and 8 enable us to compare the response of interest rates and OMOs to own shocks versus shocks in the "other" variable. Figure 7 graphs how the Fed fund rate reacts to own shocks compared with OMO shocks. Since interest rates differ, the unit of measurement for OMOs and the response paths are scaled by dividing the response of each variable by the square root of its residual variance. Thus, the responses are measured in terms of fractions of standard deviations. Figure 8 displays the reaction of OMOs to own shocks compared with the reaction to a shock in a selective interest rate (6 months T-bill rate). What emerges from figures 7 and 8 is that for both OMOs and interest rates, the reaction to own shocks is more pronounced than their response to innovations in the "other" variable.13 ii.Qpen Market Operations and Other Financial Asset Returns Open Market Operations could influence the prices of other financial assets by influencing interest rates. The change in 19 interest rates, in return, may change asset prices if it means that the discount rate used in valuing asset cash flow streams change with interest rates. Alternatively, the actions of the Fed may change asset prices by influencing the market risk premia. This also will lead investors to revise their required rates of return. A third possibility is that OMOs may have an impact returns by influencing the real sector of the economy. on asset In addition to interest rates, this paper investigates the influence of OMOs on stock returns, Eurocurrency deposit rates (Eurodollar, Euroyen, and Eurodeutchmark deposits of one month maturity), and spot exchange rates (the U.S. dollar-DM rate and the U.S. dollar-Japanese Yen rate). Table IV summarizes results Granger causality tests conducted. obtained portfolio (VWRET), are employed: and equally the battery of The link between stock returns and OMOs are examined in a bivariate model. stock return measures from Two alternative daily Value weighted weighted CRSP CRSP market market return portfolio (EWRET) . Monetary policy may affect stock prices via the three channels discussed. For example, OMOs indicating a policy of monetary ease may boost stock prices by lowering interest rates and leading investors to revise downwards the discount rates they use in valuing expected equity cash flows. The policy of monetary ease may also reduce investor uncertainty and lower the risk premia, once again increasing stock prices as a result of lower required rates of returns on equity. To the extent lower interest rates are associated with higher expected output in the real sector, stock 20 prices may go up as a result of an increase in expected corporate profits. Monetary policy could have a significant impact on share prices since it can lower equity capitalization rates and increase in expected equity cash flows simultaneously. Investors mention all of these channels in arguing the importance of the actions of the Fed for stock market returns. However, in spite of these arguments, as an empirical matter, at least for the sample period used in this study, there is no evidence that the Fed influences share prices. obtained from Table two IV also vector exchange market variables. The variables in the displays the autoregressions significant of OMOs and results foreign There are four variables in each VAR. first model are OMOs, Eurodollar and Euro DM deposit rates one month maturity (differences), and the spot Dollar-DM exchange rate (DMs per dollar in log differences). The second VAR includes the same variables, but the Japanese yen. is substituted in place of the DM. One result causality between is that OMOs there is evidence and one month for bidirectional Eurodollar deposit rates. Given, the relationship between OMOs and domestic interest rates, this result is not surprising. Another significant result is that the two spot exchange rates and OMOs are related. Japanese yen The relationship is bidirectional for the (JY) , but in the case of the DM, OMOs but is not caused by OMOs. JY exchange rates Granger it Granger causes The fact that both the DM and the cause OMOs may indicate that stabilization of the value of the dollar is one of the goals of 21 monetary policy. stabilization However, attempts are it is also confined to possible the that foreign the exchange intervention activities of the Fed, and it may be that OMOs appear to be related to the exchange rates due to the sterilization of the intervention operations. phase In fact impulse responses (not reported here) indicate that a shock in the form of an increase in the value of the dollar vis-a-vis both the DM and the JY trigger an Open Market Sale. policy of This will be consistent with an intervention purchasing dollars when the value of the dollar increases, and offsetting the increase in the money supply caused by the intervention by means of open market sales.14 Table IV also between exchange indicates the presence of Granger causality rates and Euro deposit rates. This result is consistent with the Interest Rate Parity relationship. 5 5. EMPIRICAL RESULTS: OMOs AND THE VOLATILITY OF ASSET PRICES Time conditional established volatility empirical phenomena. Bollerslev (1989),and Connolly volatility of rates. analyzed in Stambaugh exchange Baillie of and (1987), Schwert For and returns example, Taylor Volatility DeGennaro (1989) asset (1990), a Baillie (1990) in is stock French well and examine returns Schwert and Poterba and Summers the is and (1987). Econometric models that document time conditional volatility of financial assets (ARCH, GARCH), do not typically investigate what economic variables could account for observed volatility. policy may be an important factor in asset return Monetary volatility. 22 First, stabilization of interest rates may be one of the targets of monetary policy. financial If so, the Fed may respond to volatility in the markets., volatility. Second, OMOs may On theoretical grounds, affect interest rate it is not clear whether the activities of the Fed would increase or dampen the volatility in financial markets. There information (for flows is evidence example, in in the stock market that the form of earnings announcements) tend to increase volatility of asset returns. It is conceivable that the actions of the Fed may produce similar results as investors try to extract the policy content of OMOs. other hand, if stabilization is a goal On the of monetary policy, the Fed's actions may signal to the market that it intends to reduce volatility. This, in return may reduce volatility if the Fed has credibility. The general point is that the connection between monetary policy and volatility of asset prices needs to be examined empirically. The effect of OMOs on the conditional mean and variance of financial assets is examined by: Art = b0 + b^MO^ o* = a0 + a1et.12 + + Where DT is Open Market Operations. et a2(abs ( O M O ^ ) ) (3) (4) The conditional mean (3) and the conditional variance (4) equations are jointly estimated using the Berndt, Hall, Hall and Hausman maximum likelihood procedure. 23 The sign of the estimate for the a2 coefficient would whether OMOs magnify or dampen volatility. of OMOs variable simple ARCH is deleted from specification. When the absolute value (4) , the model When the indicate simple collapses to a ARCH model was estimated for the assets under consideration the estimates for a1 proved to be very highly significant, ARCH effects. indicating the presence of If the results obtained from (3) and (4) show that a1 becomes insignificant, this would indicate that ARCH effects on asset returns can be explained by OMOs. It is also possible that ARCH effects are not related to policy. Table V and VI display the results obtained from estimation of the conditional consideration. mean and variance equations for assets under The estimate for b, is negative and significant for all the interest rates considered, confirming again, the presence of the liquidity effect Fed's portfolio Estimates of assets considered, since it indicates that increases in the (reserve injections) lower a1 continue to be highly interest significant rates. for all indicating that ARCH effects cannot completely be explained by OMOs. The estimate for a2 is negative and highly significant in the Fed funds equation. Apparently, the Fed has the ability to reduce volatility in this market. The estimate for this coefficient is also significant in the 1 month Eurodollar, and the 30 year Treasury rates. The same coefficient is marginally significant in the 6 month T-bill rate, and the 5 year and 20 year Treasury rates. In all these cases the sign continues to be 24 negative, indicating the actions of the Fed have a stabilizing effect in these markets. An interesting result that emerges from Table VI is that while OMOs do not appear to have any impact on the level of stock prices, they significantly reduce the volatility of stock returns since a2 is negative and significant for both stock return measures. In sum, it appears that when OMOs turn out to be a significant explanatory variable in the conditional volatility of asset returns, in all but one case (spot yen/dollar exchange rate), the effect is in the direction of reduced volatility. 7. CONCLUSIONS Even though the Federal Reserve is one of the most closely monitored institutions by investors, the impact of OMOs on asset prices has not been investigated empirically. Using data from a period during which the Fed did not peg interest rates, this study examines this issue in the context of Granger-causality tests. The major findings of the study are the following: 1) OMOs Granger-cause interest rates across the maturity spectrum. sign of the relationship confirms hypothesized liquidity effect. the existence of 2) The the much 3) As expected, the magnitude of the response of the interest rates decay with the term to maturity. 4) Interest rate response to monetary policy appears to be very fast. 5) While there is Granger causality running from interest rates to OMOs, this does not appear to be indicative of interest rate targeting. 6) Monetary policy appears to effect some asset 25 returns in the foreign exchange market but not the stock market. 7) When monetary policy influences the conditional volatility of asset returns, it is the in the actions direction of the of Fed reducing have a volatility, indicating that stabilizing influence. 8) Finally, combined with the evidence that show that there is a link between interest rates and the real sector of the economy, this paper's finding that the OMOs affect interest rates indicate that monetary policy influences the real sector. ENDNOTES 1. One paper that examines the influence of policy on volatility is Connolly and Taylor (1990). They investigate the connection between spot Japanese Yen exchange rate volatility and central bank intervention. Another paper, Lastrapes (1989), looks at the volatility of exchange rates under different monetary regimes. Other papers examined the connection between trading volume and volatility. 2. The earlier tests on this topic were conducted by Gibson (1970), Cagan (1972) and Cagan and Gandolfi (1969). These studies provide results supporting the liquidity effect. However, later tests on this topic reached a different conclusion. For example, Mishkin (1981) and (1982) finds no empirical support for the liquidity effect that has the correct sign in the case of both the short and long term rates. More recent studies also do not detect the existence of a negative correlation between money growth and interest rates. See for example, Melvin (1983), Makin (1983), and Wilcox (1983). In this survey paper,-Reichenstein (1987) concludes that since at least April 1987 there is no empirical support for the much hypothesized liquidity effect. 3. Strongin and Tarhan (1990) examine the reaction of interest rates to money announcements. Their model discriminates between the two hypotheses by directly taking into account investor expectations regarding the Federal Reserve's monetary stance. Their results strongly support the expected liquidity hypothesis. Hardouvelis (1987), on the other hand, investigates the reaction of interest rates to reserves announcements. His results show that during the May 1980 - October 1982 period real interest rates reacted to unanticipated nonborrowed reserves announcements. To the extent reserves announcements provide information about future money supply changes, the interest rate reaction indicates the presence of the liquidity effect. 4. Actually they the Fed is about from the primary which of the bids are observed even before they are executed. When to execute an OMO, it asks for bids and offers government securities dealers. Then it decides and offers to accept.* 5 5. The figures regarding the level of the monetary base are released ten days prior to the announcement. Thus, the information being revealed by the money supply announcement is information regarding the value of the multiplier. The observed reaction of interest rates to the announced money figure in reality captures the response of interest rates to the fact that the multiplier implied by the money figure was different than expected. 6. Spindt and Tarhan (1987) examine the October 1979 - October 1982 period, to determine whether or not the Fed indeed changed its operating procedures. Some critics of the Fed claim that the new policy still amounted to a procedure that pegged interest rates. If during this period the Fed was targeting nonborrowed reserves, it should be the case that innovations in money Granger cause innovations in borrowed reserves. An empirical finding that indicates a causal link between money and nonborrowed reserves, on the other hand, would be indicative of a policy that targets the Fed funds rate. Their results show the existence of causality running from money to borrowed reserves. Thus, it appears that the focus of monetary policy during this time period was what the Fed claimed it to be. 7. Tests conducted in this paper were repeated for two sub sample periods: October 1979 - October 1982 and October 1982 - December 1984. However, the results for the sub sample periods were not materially different from the full sample period results. Thus, only the full sample period results are reported here. 8. For a detailed description of the daily activities of the Federal Reserve trading desk during the sample period, see Meek (1981), and Melton (1985). 9. See Meek (1981), Chapter 7. 10. Even though the Treasury accounts normally are exogenous to the activities of the Desk, at times, these balances can become a tool of reserve management to the Fed. For example, when the Desk wants to drain reserves from the system, if it feels it is not receiving 'good' bids from the dealers for security sales, it may request that the Treasury transfer funds from its Tax and Loans accounts to its accounts at the Fed. This will have the same reserve drainage implication as the sale of securities from its portfolio. 11. Consistent estimates of the covariance matrix of estimated coefficients was obtained by using Robusterrors procedure of RATS version 3.0. 12. The decline in the size of the coefficients is not monotonic with maturity. In fact, the decline in the 1 year Treasury rate is actually larger than the decline in the 1 year T-bill rate. Strongin and Tarhan (1990) find a similar response to money announcement shocks. Their explanation of this is that the l year Treasury bond, because it is a coupon paying security, has a shorter effective maturity than a 1 year discount security like the Treasury bill.3 1 13. Figures 7 and 8 are representative of the relative impulse responses for all interest rates examined. The finding that own innovations ara a bigger source of variation in both the OMOs and interest rates is confirmed by variance decompositions. Except for the Fed funds rate, own innovations in interest rates account for over 90 percent of the forecast errors in interest rates (20 day horizon). Similarly, over 90 percent of the OMO variation is OMO specific. In the case of the Fed funds rate, OMO innovations account for 24 percent of the variation and the Fed fund innovations account for the remaining 76 percent. These results are not sensitive to the order of othogonalization. 14. To understand the Granger causality running from OMOs to the Japanese exchange rate, impulse response to a one standard deviation shock in OMOs was examined. Impulse responses indicate that an increase in the U.S. money supply lowers the value of the dollar. While this could also be intervention related, it is also possible that the decline in the value of the dollar is due to lower domestic interest rates (via the interest rate parity) or due to higher expected U.S. inflation rate (via purchasing power parity). REFERENCES 1. Baillie, Richard T. and Tim Bollerslev. "The Message in Daily Exchange Rates: A Conditional Variance Tale," Journal of Business and Economic Statistics. 7 (1989), 297-305. 2. Baillie, Richard T. and Ramon P. DeGennaro. "Stock Returns and Volatility," Journal of Financial and Quantitative Analysis. 25 (1990), 203-214. 3. Bernanke, Ben and Alan Blinder. "The Federal Funds Rate and the Channels of Monetary Transmission," NBER Working Paper no. 3487, October 1990. 4. Bernanke, Ben. "On the Predictive Power of Interest Rates and Interest Rate Spreads," New England Economic R e view. Federal Reserve Bank of Boston, November/December 1990, 51-68. 5. Cagan, P.D. "The Channels of Monetary Effects," National Bureau of Economic Research, New York, 1972. 6. Cagan, P.D. and A. Gandolfi. "The Lag in Monetary Policy as Implied by the Time Pattern of Monetary Effects on Interest Rates," American Economic Review. (May 1969), 277-84. 7. Connolly, Robert A. and William M. Taylor. "The Impact of Intervention on Exchange Rate Volatility," Unpublished Manuscript, Graduate School of Management, University of California, Irvine (1990). 8. Cook, Timothy and Thomas Hahn. "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1970's," Journal of Monetary Economics. 24 (1989), 331-51. 9. French, Kennneth R . , G. William Schwert and Robert F. Stambaugh. "Expected Stock Returns and Volatility," Journal of Financial Economics. (19), 1987, 3-30. 10. Friedman, Benjamin M. and Kenneth N. Kuttner. "Why is the Paper-Bill Spread Such a Good Predictor of Real Economic Activity?", NBER Conference on New Research on Business Cycles, Indicators and Forecasting, Forthcoming. 11. Hardouvelis, Gikas A. "Reserve Announcements and Interest Rates: Does Monetary Policy Matter?" Journal of Finance. June 1987, 407-422. 12. Gibson, William E. "The Lag in the Effect of Monetary Policy on Income and Interest Rates," Quarterly Journal of Economics (1970), 288-300. 13. Grossman, Stanford and Laurence Weiss. "A Transactions-Based Model of the Monetary Transmission Mechanism," American Economic Review. 73 (1983), 871-880. 14. Lastrapes, William D. "Weekly Exchange Rate Volatility and U.S. Monetary Policy Regimes: An Application of the ARCH Model," Journal of Money Credit and Banking. 21, (1989), 66-77. 15. Lucas, Robert E. "Liquidity and Interest Rates," University of Chicago working paper, September 1988. 16. Makin, J. H. "Real Interest, Money Surprises, Anticipated Inflation and Fiscal Deficits," The Review of Economics and Statistics. August 1983, 374-384. 17. Meek, Paul. U.S. Monetary Policy and Financial Markets, Federal Reserve Bank of New York, 1982. 18. Melvin, Michael. "The Vanishing Liquidity Effect of Money on Interest: Analysis and Implications for Policy," Economic Inquiry. (1983), 188-202. 19. Melton, William C. Inside the-Fed: Making Monetary Policy. Homewood, Illinois. Dow Jones-Irwin, 1985. 20. Mishkin, Frederick S. "Monetary Policy and Long-Term Interest Rates: An Efficient Markets Approach," Journal of Monetary Economics. (1981), 29-55. 21. __________ "Monetary Policy and Short-term Interest Rates: An Efficient Markets-Rational Expectations Approach," Journal of Monetary Economics. (1982), 63-72. 22. Poterba, James and Lawrence Summers. "The Persistence of Volatility and Stock Market Fluctuations," American Economic Review. 76, (1986), 1142-1151. 23. Reichenstein, W. "The Impact of Money on Short-Term Interest Rates," Economic Inquiry. (1987), 67-82. 24. Rotemberg, Julio J. "A Monetary Equilibrium Model with Transactactions Costs," Journal of Political Economy. (1984), 40-58. 25. Schwert, G. William. "Why Does Stock Market Volatility Change Over Time," Journal of Finance. 44, (1989), 1115-1153.* 6 2 26. Spindt, Paul A. and Vefa Tarhan. "The Federal Reserve's New Operating Procedures: A Post Mortem." Journal of Monetary Economics. (1987), 107-123. 27. Stock, James and Mark Watson. "New Indexes of Coincident and Leading Economic Indicators," NBER Macroeconomics Annual. Oliver J. Blanchard and Stanley Fisher, eds. Cambridge, MIT Press, (1989), 351-394. 28. Strongin, Steven and Vefa Tarhan. "Money Supply Announcements and the Market's Perception of Federal Reserve Policy," Journal of Money, Credit and Banking. May 1990, 135-153. 29. Strongin, Steven. "Macroeconomic Models and the Term Structure of Interest Rates," Federal Reserve Bank of Chicago, working paper, 1990.0 3 30. Wilcox, J.A. "Why Real Rates Were So Low in the 1970's," American Economic Review. March 1983, 44-53. TABLE I Granger Causality Tests of Open Market Operations (OMOt) and Interest Rates (rt) 0M0t Does Not Grangercause rt Chi-Square Statistic (Marginal Significance level) Interest Rate Fed Funds 3 mos. T-bill 6 mos. T-bill 12 mos. T-bill 1 yr Treasury 2 yr Treasury 3 yr Treasury 5 yr Treasury 10 yr Treasury 20 yr Treasury 30 yr Treasury 29.45 10.56 14.44 16.83 19.16 22.13 22.60 15.35 17.55 21.27 15.06 (0.000) (0.103) (0.025) (0.010) (0.004) (0.017) (0.001) (0.017) (0.007) (0.002) (0.020) rt Does not Grangercause 0M0t Chi-square Statistic (Marginal S igni f icance level) 31.10 31.16 44.42 40.99 40.16 34.50 34.43 29.15 24.69 25.69 23.09 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Notes: Each equation contains a constant term, 6 lags of the forecasted variable, and 6 lags of the variable that is suspected to be Granger-causing the forecasted variable. The first marginal significance levels are for omitting 6 lags of the open market operations variable from the unrestricted OLS prediction equation for the interest rate in question. The second marginal significance level is for omitting 6 lags of the interest rate variable from the unrestricted OLS prediction equation for open market operations. Interest rates are in first differences. Open Market Operations (0M0t) are measured as the first difference of the Fed's total portfolio of securities. The data is daily. December 31, 1984. The sample period is October 2, 1979 to TABLE II Impulse Responses to a One Standard Deviation Shock in 0M0 Innovations Interest Rate Fed funds 3 Mos T-bill 6 Mos T-bill 12 Mos T-bill 1 yr Treasury 2 yr Treasury 3 yr Treasury 5 yr Treasury 10 yr Treasury 20 yr Treasury 30 yr Treasury Interest Rate Response in the Period after the Shock (Basis Points) -7.49 -1.12 -1.21 -1.20 -1.61 -1.29 -1.23 -1.00 -0.92 -0.94 -0.78 Cumulative Interest Rate Response (Basis Points) After 6 davs 12 davs -13.00 -4.21 -3.38 -2.76 -3.72 -2.94 -2.01 -1.84 -1.60 -1.24 -1.18 -11.29 -3.63 -2.79 -2.15 -2.94 -2.29 -1.51 -1.35 -1.13 -0.76 -0.78 Cumulative OMO Response to own shock 12 days After (Million Dollars) 878.9 873.1 857.1 858.8 956.1 858.2 852.3 858.8 858.5 854.7 855.5 NOTE: The magnitude of the initial OMO shock is in the range $1910 million. The response coefficients are ; s ignificant for interest rates. Confidence intervals for the response paths selected interest rates are shown in Figures 1-4. TABLE III Impulse Responses to One Standard Deviation Shock in Interest Rate Innovations Interest Rate Shock (Basis Points) Fed-Funds (63BP) 3 Mos. T-bill (22.3) 6 Mos. T-bill (19.9) 12 Mos. T-bill (16.9) 1 yr. Treasury(20.5) 2 yr. Treasury(17.3) 3 yr. Treasury(16.1) 5 yr. Treasury(14.8) 10 yr. Treasury(13.3) 20 yr. Treasury(12.6) 30 yr. Treasury(12.2) OMO Response in the Period after the Shock (Million Dollars) 304.8 153.5 222.8 217.0 218.8 212.4 212.1 205.7 175.6 163.9 157.1 Cumulative Response 12 days after (Million Dollars) -25.9 -112.6 -98.8 -102.1 -99.0 -109.6 -129.0 -103.0 -125.6 -132.7 -135.3 Cumulative Interest Rate Response to own Shocks after 12 days (Basis Points) 40.3 28.8 24.6 21.1 25.6 22.6 20.3 18.6 16.3 14.5 14.1 Notes: The numbers in parentheses in the first column are the first period interest rate responses to own shocks. 0M0 responses are significant at the 5 percent level in the case of all interest rates. The interest rate response to own shocks is also significant at the same level of significance. Confidence intervals of the response of OMOs to selective interest rates are shown in Figures 5-6. TABLE IV Granger Causality Tests of Open Market Operations and Stock Market and Foreign Exchange Market variables The First Variable does The Second Variable does not Granger-Cause not Granger-Cause the Second Variable the First Variable Chi-Scruare Statistic Relationship examined Chi-Souare Statistic OMOs, Stock OMOs, Stock Equally Weighted Returns Value Weighted Returns OMOs, Spot Deutche Mark (DM) OMOs, 1 m o s . Eurodollar Int. Rate OMOs, 1 m o s . Euro DM Int. Rate 1 mos. Eurodollar, 1 mos Euro DM Int. Rate Spot DM, 1 mos. Euro DM Int. Rate Spot DM, 1 mos. Euro Dollars Int. Rate 2.66 (0.850) 2.26 (0.894) (0.961) 2.67 (0.849) 8.35 (0.214) 36.38 (0.000) 18.84 (0.004) 43.97 (0.000) 12.39 (0.054) 7.77 (0.256) 9.18 (0.163) 10.57 (0.102) 27.61 (0.000) 10.87 (0.092) 8.74 (0.189) 1.47 48.62 (0.000) OMOs, Spot Japanese Yen (JY) 13.21 (0.039) OMOs, 1 m o s . 17.81 (0.007) Euro Dollar Int. Rate OMOs, 1 mos. Euro Yen 1.63 (0.950) 1 mos Euro Dollar 1 mos Euro Yen Int. Rate 6.39 (0.381) Spot JY, 1 mos Euro Yen Int. Rate 9.19 (0.163) Spot JY, 1 mos 36.20 (0.000) Euro Dollar Int. Rate 17.90 (0.006) 42.47 (0.000) 10.81 (0.094) 10.63 (0.100) 5.19 (0.519) 6.19 (0.402) Notes: The numbers in parentheses are marginal significance levels. The daily stock returns are the equally and value weighted CRSP portfolio returns. The Granger Causality tests reported in rows 3 to 8 are obtained from vector autoregressions with 6 daily lags of OMOs, spot DM exchange rate (DMs per dollar) , the interest rate on one month Eurodollar deposits, and the interest rate on one month Euro DM deposits. The Granger Causality tests reported in rows 9 to 14 are obtained from vector autoregressions with 6 daily lags of OMOs, spot Japanese Yen exchange rate (JYs per dollar) , the interest rate on one month Euro JY deposits. The exchange rates are in log differences, interest rates are in first differences. See Table I for additional notes. TABLE V Open Market Operations and ARCH Effects on Interest Rate Variances Art CTe2 = = b»0 a o Interest rate fe o Fed Funds -0.02 (0.02) 3 mos T-bill 0.29 (0.60) 6 mos T-bill 0.20 (10.37) 12 mos T-bill 0.22 (10.47) 1 yr T-bond 0.23 (0.41) 2 yr T-bond 0.40 (0.82) 3 yr T-bond 0.48 (1.05) 5 yr T-bond 0.40 (0.98) 0.52 10 yr T-bond (1.40) 20 yr T-bond 0.40 (1.1) 0.30 30 yr T-bond (0.87) Notes: + + b iOMOt-l aiet-i2 fei -0.003 (14.27) -0.0004 (13.07) -0.0005 (2.19) -0.0007 (3.04) -0.0008 (2.99) -0.0007 (2.92) -0.0007 (3.6) -0.0006 (3.37) -0.005 (3.41) -0.0005 (3.33) -0.0004 (2.90) + + < a2 (abs (OMOt.1)) 2502 (27.46) 328.9 (22.19) 364.3 (24.05) 253.6 (23.71) 371.04 (23.89) 268.-7 (26.47) 215.9 (24.7) 191.48 (24.79) 147.7 (20.03) 147.8 (20.5) 145.3 (22.03) Locr L ai —2 0.66 -0.25 -5970.39 (13.2) (5.76) 0.55 -0.0009 -4680.25 (12.50) (0.77) 0.16 -0.011 -4566.85 (5.30) (1.60) 0.14 0.054 -4358.6 (4.47) (0.75) 0.12 0.084 -4604.4 (4.18) (1.08) 0.14 0.0003 -4382.6 (5.41) (0.06) 0.205 -0.0003 -4269.1 (7.45) (0.07) 0.19 -0.005 -4158.2 (5.83) (1.52) 0.23 -0.004 -4013.1 (6.54) (1.27) 0.13 -0.004 -3959.5 (4.60) (1.65) 0.08 -0.006 -3909.6 (2.28) (3.31) Interest rates (rt) are in first differences. The line in parentheses under the coefficient values gives t-statistics see tables I and III for additional notes. TABLE VI Open Market Operations and ARCH effects on Exchange Rate, Euro Interest Rates and Stock Return variances. Art = = a0 Variable Examined 1 mos Euro$ 1 mos EuroDM 1 mos EuroJY Spot DM Spot JY Equally Wtd. Stock Ret. Value Wtd. Stock Ret. + bo b° -0.008 (0.97) -0.005 (2.62) -0.002 (0.54) 1.69 (1.23) 3.61 (3.22) 0.45 (1.85) 0.09 (5.06) + b 1OMOt. -1 alet-l2 fei0.000009 (2.80) 0.000007 (8.41) 0.000002 (0.82) -0.0004 (0.60) -0.0002 (0.42) 0.00006 (0.56) 0.00001 (1.52) + + et a2(abs(OMOt.1)) a©— 0.14 (32.10) 0.009 (36.02) 0.03 (59.6) 2239.3 (18.1) 1403.7 (19.5) 79.25 (23.3) 0.39 (26.2) Si- a, Locr L 0.95 (13.06) 1.16 (24.22) 0.87 (18.59) 0.20 (6.04) 0.16 (4.58) 0.07 (2.96) 0.32 (7.17) -0.00001 282.64 (10.55) -0.0000 1959.5 (0.41) -0.000003 1323.3 (7.12) -0.0004 -5744.0 (0.008) 0.074 -5473.6 (2.69) -0.005 -3478.8 (3.78) -0.00003 -132.2 (4.56) Notes: Interest rates are in first differences exchange rates are in log differences. The line in parentheses under the coefficient values gives t-statistics. See Tables I and III for additional notes. Figure 1 EFFE C T O F D T O N DFF Figure 2 E F F E C T O F D T O N D 0 3 Figure 3 E F F E C T O F D T O N D 2 4 Figure 4 E F F E C T O F D T O N D 2 4 0 Figure 5 EFFECT O F DFF ON DT Figure 6 E F F E C T O F D 6 0 O N D T Figure 7 PLOT OF RESPONSES TO DT o 2 Figure 8 PLOT OF RESPONSES TO DOS