Money supply announcements and foreign exchange futures prices for five countries.
Since the Federal Reserve began targeting money supply growth in October 1979, the effects of Fed money supply announcements on financial markets have been intensively examined. Studies have considered the impacts on financial assets including yields on short-term and long-term Treasury securities, forward interest rates, exchange rates, stock prices, and commodity prices. These studies suggest that money supply announcements have had significant effects on all these financial assets. The effects, however, have not been consistent across time, assets, or countries.
This study adds to the money supply announcement literature, focusing on foreign exchange spot and currency future prices. Mussa  provides a justification for this approach. He demonstrates that exchange rates move in accord with the efficient markets theory of asset price determination. In addition, he finds most exchange rate changes are unexpected. Thus, exchange rate changes generally should be related to "news" since unexpected changes must be due to new information. How exchange rates respond to news depends on how that information alters expectations. Those expectations, however, depend on economic agents' historical experience and on policymakers' anticipated reactions. Thus, the effects may vary across countries and across policymaking regimes.
We believe this study makes four contributions to the money announcement literature. First, we develop a model of money announcements where expected announcements may be important even when markets are efficient. While we consider money announcements in particular, the model suggests a general result. Even expected announcements may affect financial market variables simply by confirming prior expectations. Second, we extend and update the prior literature by examining whether recent money supply announcements have an impact on exchange rates and whether recent impacts differ from those found through 1985. The results suggest the impact of money announcements may have diminished marginally. Third, we extend the analysis to exchange rate currency futures. Comparing the reactions of spot and futures rates yields further evidence on the "policy anticipation" versus the "expected inflation" effects. And fourth, most prior studies of unanticipated money growth do not distinguish between positive and negative surprises. If positive and negative surprises influence market prices asymmetrically, however, prior statistical tests may not correctly identify the response of financial markets to unanticipated money growth. Markets also may react differently to good news versus bad based on asymmetric loss functions. Thus, we test for asymmetry of money announcement effects. Recent studies by Cover , Sichel , and Beaudry and Koop  demonstrate the potential importance of macroeconomic asymmetry. The results below strongly suggest asymmetric response.
Futures markets have received much attention in finance. In contrast, the linkages between futures prices and other markets have received little attention. The internationalization of world financial markets and the international coordination of monetary policy, however, suggests that the linkages between the money supply, expectations, and foreign exchange futures prices are a crucial concern to traders in foreign exchange markets and to foreign exchange operating sections of central banks. Thus, this study has important implications for monetary policy as well as for trading in foreign exchange markets. In particular, does the unanticipated component of money supply growth convey information about future exchange rates? Money growth larger than expected may cause expected appreciation of the dollar through interest rate channels. Alternately, unanticipated money growth could generate expected inflation leading to expected depreciation of the dollar.
Section II discusses alternative hypotheses to explain the relationship between unexpected money shocks and financial asset prices. It also briefly reviews the related empirical literature. Section III states our theoretical model. Section IV discusses the data and empirical methodology employed. In Section V we present statistical results on the relationship between U.S. money surprises and currency future and spot prices for Canada, Japan, Germany, Switzerland, and the United Kingdom. Section VI offers a summary and some conclusions.
II. Alternative Hypotheses on the Effects of Unanticipated Money Announcements
The widely documented positive correlation between interest rates and unexpected money growth is consistent with alternate interpretations. Unanticipated money growth may affect financial markets because of an expected reaction - or inaction - by the Federal Reserve. Alternately, markets may respond based on information inferred about future economic activity from the money surprise. Three hypotheses attempt to explain the effect of money supply announcements: the "expected inflation" hypothesis, the "policy anticipation" hypothesis, and the "real activity" hypothesis.(1)
The expected inflation hypothesis assumes financial markets do not believe the Fed is committed to a given money growth policy. Simply stated, Fed policy is not credible. The Fed is expected to accommodate a positive monetary surprise, at least in part, by increasing money supply growth thus raising the inflation premium in nominal interest rates. Increased U.S. inflation relative to other countries also leads financial markets to expect the dollar to depreciate in both spot and futures markets.
According to the policy anticipation hypothesis, market participants believe the Fed's attempt to offset a positive money surprise will raise some future real interest rates, assuming short-run price rigidity. Interest rates rise today anticipating their future increase. Market participants also expect an increase in the spread between U.S. and foreign real interest rates, leading to a capital inflow and an immediate appreciation of the U.S. dollar. Future real interest rates beyond some point, however, are not expected to change. Thus, exchange rate futures may increase for shorter contracts but remain unchanged for longer ones. The policy anticipation effect assumes Fed policy is credible. Market participants believe that faced with a demand shock the Fed will pursue monetary restraint to meet its policy objectives.
The expected inflation and policy anticipation effects both predict that nominal interest rates will rise with unanticipated money growth. The expected inflation effect also predicts depreciation of the dollar with higher inflation while the policy anticipation effect predicts appreciation of the dollar due to increases in real interest rates.
A third hypothesis, the real activity effect, is based on the information that monetary surprises contain about future output and money demand. If future output is a significant factor explaining variations in money demand, then unanticipated money growth may reflect an increase in money demand that indicates higher than expected future output. If the Fed does not accommodate increased money demand, real interest rates will rise. Thus, interest rates change based on the information that the announcement conveys about future output and money demand. This third hypothesis also implies the dollar appreciates with an unanticipated money increase. Increased expected real output will increase the demand for dollars, cause higher real interest rates, and lead to a capital inflow. Spot exchange rates and all futures rates should increase.
An extensive empirical literature has examined these three hypotheses.(2) Prior studies generally have found that anticipated money growth announcements have little or no effect on capital market prices. Money surprises, however, have significant impacts on short-term and long-term interest rates. While the evidence has been mixed, most studies support the policy anticipation hypothesis. These studies also find that the impacts increased following the Federal Reserve's October 1979 change to a nonborrowed reserves operating procedure and decreased following the Fed's switch to a borrowed reserves procedure period in October 1982.(3) Most studies conclude that money surprises had a significant effect during the October 1979 - October 1982 period. Uncertainty remains on the economic and statistical significance of money surprises outside this interval.
The money announcement literature has focused on the response of financial variables over very short intervals. Our model considers a time interval short enough that a money announcement is made, efficient financial markets react to the announcement and no other information becomes available. We express the supply of foreign currencies algebraically as:
[Mathematical Expression Omitted]
where [e.sub.t] is the exchange rate, [I.sub.t] a vector of available information, and [X.sub.t] a vector of other factors influencing supply. Similarly, demand can be expressed as:
[Mathematical Expression Omitted]
where [Z.sub.t] is a vector of other factors influencing demand. To the above equations we add an equilibrium condition:
[Mathematical Expression Omitted]
that can be used to yield a reduced form expression for the exchange rate:
[e.sub.t] = [f.sub.3]([I.sub.t], [X.sub.t], [Z.sub.t]). (4)
We make three modifications to equation (4). First, since the variables included in X and Z change infrequently (like GDP) we exclude these vectors. Second, we decompose the information set into prior money growth, exchange rates and money announcements. We also include expectations of future money growth to incorporate a view on future monetary policy. And third, for analytical simplicity we assume the function is linear. Equation (4) becomes:
[Mathematical Expression Omitted]
[M.sup.a] [is equivalent to] previously announced money stock (ignoring data revisions), and
[Mathematical Expression Omitted]
i is an index referring to one of the [k.sub.1] periods within a week. j is a weekly index where j [is equivalent to] INTEGER((i - [k.sub.1])/[k.sub.1]). The money announcement is made once per week, [k.sub.2] periods ago with the announcement j weeks ago made at period t -j[k.sub.1] - [k.sub.2]. The next announcement, expected to be [M.sup.e], will be announced in [k.sub.1] - [k.sub.2] periods. Expectations about money growth are assumed to be formed [k.sub.3] periods ago. While expectations may be revised, we assume those revisions are based solely on information contained in the exchange rate in periods since expectations were stated, i.e., [e.sub.t-1], e.sub.t-2], ..., [e.sub.t-[k.sub.3]]. The efficient markets hypothesis predicts that news will be quickly embodied in financial market variables and thus will be reflected in the exchange rate.
Assuming a week has forty periods, [k.sub.1] = 40. If the money announcement is made in period 32 (Thursday's close) and expectations are calculated, say, in period 9 (Tuesday's open), and if the current time is 36 (mid-Friday), then [k.sub.2] = 4 and [k.sub.3] = 31. Previous money announcements were made 4, 44, 84, . . periods ago, and future announcements will occur in 36, 76, 116, ... periods.
Equation (5) assumes that financial market participants respond the same way each week as well as within each week. Since the latter assumption may depend on institutional arrangements, we only compare the function at the same time interval from one week to the next. Differencing equation (5), when the interval from t to t + 1 includes the money announcement, yields:
[Mathematical Expression Omitted].
This equation is a more general expression for the impact of money announcements than employed in previous studies.(4) In particular, for reasons explained in detail below, fully anticipated events have a channel of influence on financial market variables even with the assumption of efficient markets.
Equation (6) should be contrasted with the specification traditionally used to estimate the impact of money announcements on financial market variables:
[Mathematical Expression Omitted].
Equation (7) imposes five restrictions on equation (6):
(a) [summation of] ([[Alpha].sub.i+1] [M.sub.t+1-i] - [[Alpha].sub.i] [M.sub.t-i]) where i = 1 to H = 0
(b) [summation of] ([[Beta].sub.i+1] [e.sub.t+1-i] - [[Beta].sub.i] [e.sub.t-i]) where i = 1 to J = 0
(c) [Mathematical Expression Omitted]
(d) [Mathematical Expression Omitted]
(e) [[Gamma].sub.1] = [[Psi].sub.1] = [[Rho].sub.1].
Restriction (a) implies that prior money supply changes have no aggregate impact on the exchange rate during the announcement period.(5) One potential justification is that the liquidity effect and inflation premium effect are just offsetting. Alternately, Brown and Santoni's  estimate of [[Alpha].sub.i], less than 15 basis points per month, suggests pragmatically that restriction (a) holds. The daily [[Alpha].sub.i] coefficients would be expected to average less than .8 basis points. Even with a liquidity effect, however, the first restriction could hold if money growth does not change appreciably from day-to-day, a not unlikely occurrence, and the [Alpha].sub.i]'s are approximately equal. Nevertheless, at times during a month the liquidity effect or inflation premium effect could dominate. Thus, a priori, [[Alpha].sub.i]'s signs are ambiguous.
Constraint (b) requires prior exchange rate changes to have no impact on the exchange rate change during the money announcement period. Under restrictive assumptions about information availability, we may interpret this requirement to impose informational efficiency. Given only one observation per week on the expected money announcement, we use Roley's [25; 26] and Hein's  hypothesis. That is, prior exchange rates reflect information becoming available between the survey of expectations and the money announcement.(6)
An extreme view of constraint (c) consistent with prior money announcement studies restricts all money announcements expect the current to have no impact on the change in the exchange rate, or [[Gamma].sub.i] = 0 for i greater than one. More generally, constraint (c) states that prior money announcements provide no information to financial market participants about the probable future course of monetary policy and thus probable future exchange rates. Any one week's announcement may be heavily discounted given the significant random noise in the announcement series. Nevertheless, at the margin this week's announcement may alter the perceived informational content of prior announcements.
Constraint (d) is similar to constraint (c). It implies that the money announcement may alter expectations of future money growth. Unlike the other constraints, this last restriction often has been mentioned in the money announcement literature. The expected liquidity effect and the inflation premium effect both are predicated on the expected future policy response to a current surprise. However, constraint (d) cannot be tested directly with available survey data since data is available for [Mathematical Expression Omitted] but not for [Mathematical Expression Omitted].
Constraint (e) focuses on the value of [[Gamma].sub.1] versus [[Psi].sub.1] and suggests the extent that financial market participants are forward looking versus backward looking. The conventional assumption that [[Gamma].sub.1] = [[Psi].sub.1] implies that financial market participants place equal weights on the announced and expected money growth as determinants of the current exchange rate. That is, the net impact of past inflation premium and liquidity effects is the same as future expected liquidity and inflation premium effects. While this assumption may be true, it certainly is not a necessary condition for the efficient markets hypothesis to hold. Non-risk-neutral agents also may respond differently to actual versus expected values.
To begin estimation we impose four restrictions on equation (6). First, for consistency with prior studies, we impose constraint (e) to obtain a term for the unexpected component of the money announcement [Mathematical Expression Omitted]. This restriction is central to testing the standard hypothesis that only unexpected monetary changes have any impact on financial variables. Second, we assume that four weeks is sufficient to capture lugged impacts. Longer lags were not significant, and this restriction does not appear to be binding. Third, since information on revisions of future expectations is unavailable, we must drop those terms from the estimated equations. Thus, we impose constraint (d) based on data availability. And fourth, only a single lagged exchange rate term is included, consistent with Roley's  interpretation. This variable, [[Delta].sub.[e.sub.TTH]], captures new information acquired by the market after the survey of expectations has been conducted. Other exchange rate terms were insignificant. Thus, this restriction also does not appear to be binding.
We make two additional changes to equation (6) to make our results comparable with prior studies . We include prior expected money in the estimated equation rather than prior money announcements. Since actual money growth already is included in the equation, the resulting equations effectively will be the same.(7) We also convert the subscripts to a weekly basis. Rearranging terms and adding a constant (for comparability with prior equations) implies equation (6) can be written as:
[Mathematical Expression Omitted]
where [Mathematical Expression Omitted] indicates the expectation in week t of the money growth to be announced during that week.
IV. Data and Methodology
We employ weekly money supply announcements, daily spot exchange rates (at market close) and daily foreign exchange future prices (also at market close) for the U.S. dollar relative to the currencies of Canada, Japan, Germany, Switzerland, and the United Kingdom.
Future currency prices are obtained from the Chicago Mercantile Exchange International Monetary Market data base.(8) Future contracts for the five countries examined exist for the months of January, March, April, June, July, September, October and December. Delivery (expiration) of these contracts is on the third Wednesday of the contract month. We limit our analysis to the four widely traded contracts - March, June, September, and December. From the set of all trading days for these four contracts between 1981 and 1990, we extract a smaller data set containing only the last three months prior to the expiration month. Splicing data from these months gives us the contract having the highest trading volume.(9) For example, for the March contract, we use trading prices for the months of December, January, and February. We then identify the dates that correspond to weekly money announcements and employ the price data that correspond to these dates. The resulting data set contains 482 trading days for Canada, 481 days for Germany, 409 days for Japan, 481 days for Switzerland, and 411 trading days for the United Kingdom. Our first observation is the money announcement date December 1, 1980 (using the March, 1981 contract). The last observation is the money announcement date August 30, 1990 (using the September, 1990 contract).
The unexpected money supply is generated using the median of Money Market Services Inc.'s weekly survey of money growth expectations.(10) The unexpected value is the actual change in M1 minus the expected change.(11) For the actual change we employ the initially announced change in the narrowly defined monetary aggregate reported in the H.6 release of the Federal Reserve.(12) The announced change is the difference between the preliminary estimate of the average MI money stock for the seven days ending Wednesday of the previous week and the revision of the prior week's preliminary estimate of the money supply.(13)
Before estimating equation (8), we examine whether all series are difference stationary, as required for inference given that equation . We use the augmented Dickey-Fuller test (ADF) and present the results in Table I. Without exception, we can reject the null hypothesis of a unit root for all series.(14) All series appear to be stationary in differences, as specified in equation (8).
Based on equation (8), we attempt to determine the reaction of spot and futures exchange rates to weekly money announcements. Consistent with our discussion above on the liquidity effect, no lagged money term ever is significant. In addition, the lagged change in the exchange rate is insignificant as are expectations prior to the current week. These terms are deleted from the equations presented here, and equation (8) reduces to:
[Mathematical Expression Omitted].
We also present results imposing the standard constraint in this literature that [[Alpha].sub.1] equals zero.
Most prior discussion in the money announcement literature has focused on the coefficient [[Alpha].sub.2], whose sign is unknown a priori. The expected inflation hypothesis predicts a positive value for [[Alpha].sub.2] while the policy anticipation and real activity hypotheses predict negative values. Given prior findings, we expect [[Alpha].sub.2] to be negative before October 5, 1982 because the Federal Reserve is thought to have offset prior money shocks during this period.(15) The post-October 5, 1982 period is more difficult to categorize. Clearly, less emphasis was placed on controlling M 1 after October 1982. The degree that financial markets believed the Fed's behavior changed, however, is less apparent.
[TABULAR DATA OMITTED]
V. Empirical Results
We begin by breaking the data at the end of February 1985. This break corresponds to the removal of daily price limits on currency futures.(16) Later we consider a number of different subperiods, although efforts to identify a clear break point were unsuccessful. We also initially examine the distributions of changes in future and spot exchange rates and changes in unexpected money growth for the pre- and post-February 1985 subperiods. The most noteworthy aspect of these distributions is that all three series are highly leptokurtic. The leptokurtic nature of exchange rates has been documented previously. The non-normal distribution of the dependent variable requires that all tests be undertaken using nonparametric techniques. Thus, we estimate equation (9) using standard OLS as well as OLS applied to the data in ranks.
Table II reports results of estimating equation (9) while imposing the restriction a [[Alpha].sub.1] = 0. It presents the results of equations estimated using both changes and ranks (of changes), for the spot and futures exchange rates, for both subperiods, and for all countries.(17) Our results in Table II are comparable to prior studies. Table III presents similar results without the restriction [[Alpha].sub.1] = 0.(18)
We find that the spot price responds to unanticipated money announcements only for the early period (before March 1985) for Canada, Germany, and Switzerland. The estimates of [[Alpha].sub.2] in [TABULAR DATA OMITTED] [TABULAR DATA OMITTED] the spot rate equations, both in changes and ranks, are negative for all countries but are significant only for Canada, Germany, and Switzerland. This result indicates that spot exchange rates respond negatively to a positive money surprise, resulting in the appreciation of the U.S. dollar relative to the Swiss franc, Canadian dollar, and German deutsche mark prior to 1985. The consistency of the equations estimated in changes and in ranks suggests that the results are robust.
Our results for the early period are consistent with the findings of Cornell , Engel and Frankel  and Hardouvelis  for spot exchange rates.(19) Our results also support the interpretation that the market believed that the Fed would quickly move to offset money shocks. This evidence is consistent with the policy anticipations effect but inconsistent with the expected inflation effect.
In the early period we also find that some exchange rates respond while some do not. This result suggests, for example, that the exchange rate between the mark and the yen must change with a U.S. money surprise. That the yen and the pound do not change suggests that the central banks in Japan and the U.K. may have followed Fed actions while the other central banks pursued independent monetary policy. The closeness of the German and Swiss results is not surprising. The difference between the German and U.K. results suggests that the U.K. has been a more reluctant participant in the EMS.
Differences between countries should not be surprising, having been previously reported by Hardouvelis , for example. These differences can be explained by examining the differences in the monetary and exchange rate policies of the home countries. They are less easily explained when restricting the analysis only to U.S. factors. An examination of other countries' monetary policies and announcements, however, is beyond the scope of this paper.
During the post-February 1985 period we find no significant impact of monetary surprises on the spot exchange rate for any country. The estimates of [[Alpha].sub.2] do not even have a consistent sign. These results support neither the policy anticipations effect nor the expected inflation effect in isolation. It is possible, however, these two effects were approximately offsetting during this period.
Given the dramatic differences apparent in the spot exchange rate results, is there a significant change from the early to the late period? At the 5 percent significance level, a Chow test rejects the null hypothesis of no change in regimes for the German spot rate equations (F(2, 477) = 3.10 with a significance level of .046 for the equation in changes; F(2, 477) = 4.29 with a significance level of .014 for the rank equation) and for the Canadian spot rate equations (F(2, 478) = 4.43 with a significance level of .012 for the equation in changes; F(2, 478) = 4.77 with a significance level of .009 for the rank equation).(20) We find little evidence of a significant regime change for the other three countries.(21)
Turning to the future currency price results, money surprises influence only the German futures price in the first period. Even the deutsche mark futures reaction is only about 60 percent of the spot rate change. The more limited futures reactions strongly support the policy anticipations hypothesis and is inconsistent with both the expected inflation and real activity effects. Change in the futures price should represent the expected reaction of exchange rates approximately three months ahead. Based on the German results, it appears that exchange market participants expect the Fed to offset approximately 40 percent of any unanticipated weekly money growth change within the next three months. Based on the insignificant Canadian and Swiss results, market participants expect the Fed to offset most changes within the next three months.
We also constructed a series of futures rates approximately 4 to 6 months ahead using the methodology described above to create a future rate series 1 to 3 months ahead. The results using the 4 to 6 months ahead futures data uniformly indicate no impact of unanticipated money. This result further strengthens the case for the policy anticipations hypothesis and suggests market participants expect unanticipated changes will be entirely offset within six months.
The results for the first period also suggest a potential concern with interest rate parity (IRP). Three prior pieces of evidence are important. First, Edison and Pauls  present results they interpret as not providing strong support for IRP. Second, prior studies including Hardouvelis  indicate U.S. interest rates respond more than foreign interest rates to unanticipated U.S. money growth. And third, Hafer and Sheehan  among others demonstrate that short term U.S. rates respond more than long term U.S. rates to unanticipated money growth. These prior results suggest that spot exchange rates or very near term futures contracts should respond more to unanticipated U.S. money growth than should longer term futures contracts. The results presented here are consistent with prior studies, albeit one might question whether they are consistent with IRP. Developing a model that fully incorporates IRP in the money announcements literature is beyond the scope of this paper but appears a potentially fruitful topic for further research.
In the second period, currency future prices never respond to unanticipated money growth, either in the change or rank equations. This result is consistent with the spot rate results. The insignificance of both spot and futures prices provides no support for either the policy anticipations or the expected inflation effect.
We also examine whether the period prior to October 1982, when the Fed used a reserve aggregate operating target, differs from the post-October 1982 period (from October 6, 1982 through February, 1985) a period when the Federal Reserve appeared to return to an interest rate operating procedure. Insignificant Chow tests for all countries indicate that the response of the exchange rate to unanticipated money announcements did not change at this date. This result should not be unanticipated given prior findings, including Hafer and Sheehan  who also did not find a break at October 1982.(22)
To further examine the robustness of the results, we consider the sensitivity of the conclusions to breaks in the data. Given Hafer and Sheehan's conclusions, it is not clear that we can uniquely identify a break point (or multiple breaks) in the data. Given the low explanatory power of the equations, Quandt tests to determine where the most likely break point lies likely will be ineffective. The likelihood function probably is very flat suggesting a Quandt test would not be revealing. Instead, we break the data at points corresponding to the introduction of each new contract, beginning with the October 1982 contract and ending with the December 1987 contract.
The results for the impact of unanticipated changes on spot or futures exchange rates suggest that the choice of a break point is not critical either in terms of significance of coefficient sign. Breaking the data at any point between the end of August 1982 (after the September 1982 contract) through the end of August 1987 (after the September 1987 contract) implies that spot and futures exchange rates respond in the early period but not in the later period. We find a drop in significance of the early period coefficients at the beginning of 1987 that suggests unanticipated money growth has minimal impact on exchange rates after late 1986. These results are consistent with the policy anticipation hypothesis holding from 1980 through late 1986. Alternately, the Fed pursued a policy of credible monetary restraint throughout the 1980-86 period. We cannot comment on the credibility of the Fed after late 1986, however, because the spot rate also is not significantly different from zero. Thus, there is no initial impact that market participants expect to be offset shortly. The Fed, however, did discontinue providing a monthly target range for M1 at this time.
Table III reports the results from estimating the unconstrained equation (9) including the expected component of the money announcement. The coefficients on unexpected money in the spot rate equations remain negative for all countries and now are significant at the 10 percent level for all countries except Japan. Thus, the discussion of Table III focuses on Canada, Germany, Switzerland, and the United Kingdom.
The Table III money surprise coefficients in the spot rate equations are close to but slightly larger than their Table II values. The change and rank equations again virtually always are in accord. (The single substantive difference is the significance of the expected term in the Canadian spot rate equation.) The negative coefficients again support the policy anticipation hypothesis.
The expected money announcement also is significant in the spot rate equations. The expected terms, however, have positive coefficients. Thus, an expected increase in the money announcement leads to depreciation of the dollar. This positive coefficient provides no evidence on the policy anticipations versus expected inflation hypotheses, however.
Our findings also indicate that the spot rate does not respond either to expected or to unexpected money announcements post-1985. While these results may indicate offsetting policy anticipations and expected inflation effects, the insignificance of both the expected and unexpected terms suggests a more likely explanation. Given that monthly M1 targets were discontinued in mid 1986, the weekly M1 announcements may be ignored by financial market participants.(23)
For the futures equations, the coefficients on unexpected money generally are negative in the early period. The Canadian rank equation is the sole exception. The German and Swiss results are significant while the Canadian and U.K. results are not. Excluding Japan, all estimated coefficients on money surprises fall from the spot to the futures equations, albeit the magnitudes of the drops differ substantially. In Germany and Switzerland, the decline is on the order of 25 percent while Canada has virtually a 100 percent decline. These results again generally support the policy anticipations hypothesis. Current higher than anticipated money growth is expected to be largely offset by slower future money growth.
The expected money announcement also has a significant impact on currency future prices. [TABULAR DATA OMITTED] Higher anticipated money growth leads to higher future prices. After early 1985, however, neither unexpected nor expected money has any impact on currency futures.
Several prior studies have found that expected money changes have significant impacts [4; 5; 16; 25; 32]. One explanation for the statistical significance of expected money is that a simple form of informationally efficient markets may be too restrictive. A more general explanation, consistent with the above model, is that money supply announcements are not viewed by financial market participants as independent events but, over some period of time, have a cumulative impact on the information set. Money announcements may alter expectations of future monetary policy and may change or clarify the interpretation of previous announcements. For example, current announcements may contain information concerning whether prior money shocks should be considered permanent or temporary or confirming prior expectations about monetary policy.
We consider one final issue concerning the impact of money announcements on exchange rates. Prior studies of financial market responses to unanticipated money growth generally do not distinguish between positive and negative surprises. If positive and negative surprises have asymmetric effects on market prices, however, the statistical tests may not identify the response of financial markets to unanticipated money growth. Consider an extreme example. If positive and negative shocks both lead to a decrease in the exchange rate, then the estimated coefficient on a single unanticipated money growth variable may approximately equal zero. Then standard tests would find no impact of the announcements.
Cover  employed Barro's  general test procedure to determine whether quarterly anticipated and unanticipated money growth rates have real economic impacts. Cover also allowed for differences between positive and negative money surprises. His results indicate that negative money surprises decreased real output while positive surprises had no impact. These results suggests that financial markets may react more strongly to "bad" news, viewed as an unexpected negative innovation, than to "good" news. A negative money surprise usually is associated with the fear of at least temporarily higher real interest rates, a decrease in foreign price competitiveness, and an economic slowdown. Zellner  also hypothesizes that market participant's reactions to news may differ based on asymmetric loss functions.
To test for asymmetry, we break unanticipated and anticipated money growth rates into positive and negative components and re-estimate equation (9). Table IV presents the change results only since the rank equations yield the same implications. We consider the results for the first subperiod only since all changes in the second subperiod are insignificant. The results strongly support asymmetry. The significance of unanticipated money growth in the spot exchange rate equations found in Tables II and III appears to be due solely to negative money changes or "bad" news. An unanticipated decrease in money growth has a significant impact on the spot exchange rate at the 10 percent level for all countries except Japan. An unanticipated increase always has an insignificant impact. This result is consistent with Cover's conclusions.
For the futures equations, all coefficients on negative money shocks are smaller than the corresponding coefficients in the spot rate equations, but only the German coefficient is significant. This result is consistent with the policy anticipations hypothesis and inconsistent with the expected inflation hypothesis and the real activity hypothesis.
The impact of anticipated changes is entirely different. Coefficients on positive expected money growth are uniformly positive in both the spot and futures equations while coefficients on negative expected money growth generally are negative. Higher expected money growth leads to a significant increase in the German and United Kingdom spot rates and to a significant increase in the futures rate for all countries but Canada at the 10 percent significance level. Lower expected money growth leads to a statistically significant decrease only in the forward rates for Germany, Switzerland and the United Kingdom. In all countries, however, for both the spot and the futures rates the magnitude of the estimated coefficient is larger for negative expected changes than for positive.(24) In general, the significance of expected money changes appears to be largely driven by expected increases in money growth.
Our results suggest expected money growth leads to a depreciation of the dollar. This finding could be a function of higher expected inflation, consistent with futures rates reacting more strongly than spot rates. Alternately, it could be due to higher expected real economic activity, also consistent with futures rates reacting more strongly than spot rates. Finally, the asymmetric response could be due to asymmetric behavior by central bankers. For example, our results are consistent with policy makers trying to prevent their currencies from depreciating but not intervening to stop their currency from appreciating.
VI. Summary and Conclusion
This study examines the impact of US money supply announcements on foreign exchange future and spot prices for Canada, Germany, Japan, Switzerland, and the United Kingdom. We begin by developing a theoretical model that allows both anticipated and unanticipated announcements to influence financial markets even under efficient markets. The empirical results extend prior studies by considering the impacts of unanticipated money announcements through most of 1990. Our results suggest that unanticipated money growth did not have an impact on either spot or futures exchange rates beginning in late 1986 or early 1987. The data do not allow us to pin-point when surprises ceased to have an impact. The results prior to 1985 also indicate that money surprises have a significant negative impact on Canadian, German, and Swiss currencies. Spot exchange rates decrease with a positive money surprise, resulting in the appreciation of the U.S. dollar relative to the Canadian dollar, German deutsche mark, and Swiss franc. Currency future prices also are affected but the responses are smaller than those for the spot rates. Furthermore, all results appear to be relatively robust.
Our results generally support the policy anticipations hypothesis. Unanticipated money growth leads to increases in spot exchange rates. Those increases, however, are expected to be quickly offset since the futures response is much smaller. This result implies that the Federal Reserve had credible monetary policy in terms of money growth targets through 1986. After 1986 the results suggest that market participants did not believe the Fed's monetary targets since unanticipated money growth had no impact even on spot exchange rates.
One additional conclusion is noteworthy. We find that only negative unanticipated surprises lead to exchange rate changes. We interpret this to mean that only bad news impacts exchange rates. This result is consistent with Cover's  conclusion that only unanticipated negative monetary shocks have an impact on real output. This result also is consistent with the view that central banks respond asymmetrically to positive and negative surprises.
1. Sheehan  has shown that these hypotheses are not mutually exclusive. The policy anticipation effect and the expected inflation effect may be complements or substitutes depending on financial market participants' perspectives on Fed goals. Sheehan contends that there is no reason to believe that these hypotheses operate in isolation. For example, findings that suggest money surprises have no impact on stock prices appear inconsistent with the real activity hypothesis. However, they may indicate only that the money demand effect by itself cannot explain all the impacts of money surprises.
2. A thorough review is beyond the scope of this paper. For a survey of the empirical literature prior to 1985, see Sheehan . For post-1985 literature dealing with the effects of money supply announcements on interest rates, see Hafer and Sheehan  and the references therein. For recent literature examining the impact of money announcements on exchange rates, see Cunningham and Cunningham .
3. Thornton [30; 31] adopts a less traditional approach in testing these competing hypotheses. He investigates the consistency of the response to unanticipated money growth across the bond, stock and foreign exchange markets over the period January 5, 1978 to January 26, 1984. While he finds weak evidence favoring the policy anticipations hypothesis, his results indicate no hypothesis adequately explains the stylized facts. In particular, Thornton contends that only a small proportion of unanticipated money changes are associated with movements in Treasury bill rates regardless of Fed operating procedure. The frequency and magnitude of foreign exchange market responses increased, however, when the Federal Reserve returned to interest rate targeting. These results support Hakkio and Pearce  who found that exchange rates did not respond to money surprises before October 1979, but after, the dollar appreciated in response to a positive money surprise.
4. If we difference the equation during a period other than the money announcement period, the third and fourth terms drop out of the equation. Further assuming that expectations do not change yields interest rate equations of the form typically estimated to measure the liquidity effect [7; 22].
5. Strictly speaking, restriction (a) may be phrased in terms of money stock changes only if [[Alpha].sub.i-1] = [[Alpha].sub.i].
6. It is conceptually appealing to assume that expectations concerning future money announcements change in that period when the money stock is announced. The reader may wish to think intuitively of two revisions, one occurring at the money announcement and one at some other time of the week.
7. Using the announcements does not alter the results. This finding should not be surprising given the conclusion of prior studies that expectations are an unbiased estimator of the announced series. Furthermore, if expected money announcements and actual surprises are orthogonal, then the estimate of [[Rho].sub.1] would be identical with or without the expected money in the regression. Belongia and Sheehan  find the coefficient on unexpected money is unaffected by the inclusion of expected money. Hakkio and Pearce's  results suggest that the expected money change has no statistically significant effect on bilateral rates.
8. Currency futures also are traded on the Philadelphia Option Exchange. That volume, however, is small relative to that of the Chicago Mercantile Exchange.
9. We also considered data spliced for the prior three months (four to six months previously) and briefly discuss those results below. Estimating equations considering only one contract are not particularly interesting for two reasons. First, for much of the contract life little trading occurs. And second, the length to maturity of the contract changes thus clouding the interpretation of the estimated coefficients.
10. On September 29, 1977 Money Market Services Inc. began to survey money market participants weekly for their forecasts of economic announcements to be made that week. Starting with the implementation of Friday money announcements, February 8, 1980, the survey was collected on the Tuesday before the announcement. Previously, survey data were collected on Tuesdays and Thursday. Since February 8, 1980, surveys have been conducted only on Tuesdays. Pearce and Roley  present a more detailed discussion of this survey as well as evidence indicating that these forecasts are unbiased and efficient.
Deaves  and Deaves, Melino, and Pesando  adjust the survey measure to "eliminate its demonstrably downward bias." In particular, they remove the bias by using fitted values from a regression of money on the survey measure (and a constant) for expected money, with the residual from the regression representing the unanticipated money measure. Roley  performs a similar, but, slightly different procedure. The evidence that the survey is biased, however, is mixed at best and few researchers make such an adjustment. We do not make such an adjustment, believing that results which depend on such an adjustment are unlikely to lead to useful implications.
11. Actual and expected money below are both measured in difference form. Unexpected money, however, is expressed throughout as [Mathematical Expression Omitted] but by construction is also in difference form.
12. From February 1980 through November 1981 the M1B definition of money, not adjusted for NOW accounts, is used. The current definition of money is used after this point. This choice corresponds to the measure of M1 that the survey participants were asked to forecast. Other narrowly defined monetary aggregates announced by the Fed at different times during the sample include "shift-adjusted M1B" (which corrects for inflows to NOW accounts) and M1A. One may argue that the Fed has downgraded the importance of M1, especially since 1986, and it would be more appropriate to focus on M2. Nevertheless, survey participants are asked to forecast M1, suggesting that at least some financial market participants still pay attention to M1.
13. Prior to February 8, 1980 money supply announcements were made on Thursdays. From February 8, 1980 through February 10, 1984 those announcements were made on Fridays while after that date they again were made on Thursdays. Throughout this period those announcements were made after virtually all U.S. financial markets had closed. Prior to November 29, 1982, money announcements were made at 4:10 p.m. EST. From December 6, 1982 through February 10, 1984 the announcements were made at 4:15 p.m. Since then they have been made at 4:30 p.m.
14. The tests for money announcements all refer to U.S. announcements. The small differences in the reported test statistics across countries stem from small differences in the samples across countries. These sample differences generally are based on minor difference in the data availability of the futures exchange rates.
15. Hardouvelis  reports results supporting this result as well as findings indicating [[Alpha].sub.2] was positive before October 6, 1979, supporting the belief that the federal funds operating procedure employed during that period meant the Fed did not quickly counteract positive money shocks.
16. Prior to February 25, 1985, daily price limits were in effect for all currency futures traded on the Chicago Mercantile Exchange. For example, the Deutsche mark was limited to a maximum 100 point change, and this limit was hit twice (March 27, 1981 and October 2, 1981).
17. Tables I and II include changes equal to or exceeding the limit price change effective in the first period. We have re-estimated the equations with those observations excluded and find no appreciable difference in the results. We also excluded observations of money supply announcements, exchange rate changes, and currency future prices which were more than two standard deviations from the mean, with no qualitative change in the results. Our results appear to be relatively unaffected by the deletion of a few observations. This finding is not inconsistent with the careful analysis of possible outliers and influential observations advocated by Beggs  and Belsley, Kuh, and Welsch . It also is not inconsistent with the conclusions of Hafer and Sheehan  who find a substantial response almost throughout their sample. It is, however, inconsistent with the contention of Thornton  and Deaves, Melino, and Pesando  who argue that when a few observations are deleted, the statistical significance of unanticipated money growth disappears.
18. Some question exists about whether heteroskedasticity exists in the data. We use the White test for heteroskedasticity but find no evidence for any country in either period for spot or currency future prices.
19. Hakkio and Pearce  broaden the investigation to consider the response of exchange rates to U.S. announcements of inflation, industrial production, and the unemployment rate in addition to weekly money supply announcements. Their results show that only money surprises significantly affect exchange rates. Hakkio and Peatee examine the effect of money surprises on seven bilateral spot exchange rates for three subperiods. Their results indicate little difference across countries but large differences across time. For September 1977-October 1979 they find no significant response to money surprises. They find an appreciation of the dollar from money surprises for later periods - October 1979-October 1982 and October 1982-March 1984, with [[Alpha].sub.2] larger in the latter period.
20. We could not reject the null hypothesis of no regime change in 1985 for the German futures equation in ranks (F(2, 477) = 2.05 with a significance level of. 130) nor could we reject the null hypothesis of a break at October 1982 (F(2, 203) = 1.14 with a significance level of .323), This result is one of the rare occurrences when the change and the rank results conflict.
21. For Switzerland, a Chow test rejects the null hypothesis of stability at the 10 percent level for the rank spot rate equation (F(2, 478) = 2.68 with a significance level of .069 versus F(2,478) = 1.33 with a significance level of .267 for the equation in changes). In addition, we must note that all Chow tests presented here have relatively low power given the minimal explanatory ability of the equations.
22. In contrast, many studies including Roley  and Huizinga and Leiderman  do find a significant change in the response of U.S. interest rates to money surprises following the Fed regime change in October 1982. The only significant Chow test was found for Canada with respect to equation (2), for the spot price, estimated in ranks. The null hypothesis of no regime change in 1982 could be rejected (F(2,201) = 3.04 with a significance level of .050).
23. We also conducted stability tests for the equations presented in Table III. Generally, the null hypothesis of stability was rejected. Again, however, we could not unambiguously pinpoint the location of the change in regimes.
24. Them are substantially fewer negative expected changes than positive. Splitting the sample into halves, the larger versus the smaller expected changes, yields qualitatively similar results and the magnitude of the coefficient for the smaller expected changes decreases somewhat. Our analysis of the tails of the distribution, however, strongly indicate that our results are not being driven by a few outliers.
1. Barro, Robert J., "Unanticipated Money, Output, and the Price Level in the United States." Journal of Political Economy, August 1978, 549-80.
2. Beaudry, Paul and Gary Koop, "Do Recessions Permanently Change Output?" Journal of Monetary Economics, April 1993, 149-63.
3. Beggs, John J., "Diagnostic Testing in Applied Econometrics." The Economic Record, June 1988, 81-101.
4. Belongia, Michael T. and Fredric Kolb, "Risk Aversion and Weekly Money: Does the Market Expect the Fed to Offset Large Increases in M1?" Economic Letters, 1984, 327-30.
5. ----- and Richard G. Sheehan, "The Informational Efficiency of Weekly Money Announcements: An Econometric Critique." Journal of Business and Economic Statistics, July 1987, 351-356.
6. Belsley, David A., Edward Kuh, and Robert E. Walsh. Regression Diagnostics: Identifying Influential Data and Source of Collinearity. New York: John Wiley, 1980.
7. Brown, W. W. and Gary J. Santoni, "Monetary Growth and the Timing of Interest Rate Movements" Federal Reserve Bank of St. Louis Review, August/September 1983, 16-25.
8. Cornell, Bradford, "Money Supply Announcements, Interest Rates, and Foreign Exchange." Journal of International Money and Finance, August 1982, 201-208.
9. Cover, James Peery, "Asymmetric Effects of Positive and Negative Money Supply Shocks." Quarterly Journal of Economics. November 1992, 1261-82.
10. Cunningham, Thomas J. and Rosemary T. Cunningham, "Another Source of Money Supply Announcement Surprises." Journal of Macroeconomics, Spring 1991, 279-289.
11. Deaves, Richard, "The Impact of Unanticipated U.S. Weekly Money on the Path of the Dollar." Quarterly Review of Economics and Business, Summer 1991, 94-106.
12. -----, Angelo Melino, and James E. Pesando, "The Response of Interest Rates to the Federal Reserve's Weekly Money Announcements: The 'Puzzle' of Anticipated Money." Journal of Monetary Economics. May 1987, 393-404.
13. Edison, Hali J. and B. Dianne Pauls, "A Re-assessment of the Relationship Between Real Exchange Rates and Real Interest Rates: 1974-1990." Journal of Monetary Economics, April 1993, 165-87.
14. Engel, Charles and Jeffery A. Frankel, "Why Interest Rates React to Money Announcements: An Explanation from the Foreign Exchange Market." Journal of Monetary Economics, January 1984, 31-39.
15. Fuller, W. A. Introduction to Statistical Time Series. New York: John Wiley, 1976.
16. Gavin, William T. and Nicholas V. Karamouzis. "Monetary Policy and Real Interest Rates: New Evidence from the Money Stock Announcements." Federal Reserve Bank of Kansas City Working Paper 8406.
17. Hafer, R. W. and Richard G. Sheehan, "The Response of Interest Rates to Unexpected Weekly Money: Are Policy Changes Important?" Southern Economic Journal, January 1990, 577-593.
18. Hakkio, Craig, S. and Douglas K. Pearce, "The Reaction of Exchange Rates to Economic News." Economic Inquiry, October 1985, 621-36.
19. Hardouvelis, Gikas A., "Market Perception of Federal Reserve Policy and the Weekly Monetary Announcements." Journal of Monetary Economics, September 1984, 225-40.
20. Hein, Scott E., "The Response of Short-Term Interest Rates to Weekly Money Announcements: A Comment," Journal of Money, Credit, and Banking, May 1985, 264-71.
21. Huizinga, John and Leonardo Leiderman, "The Signalling Role of Base and Money Announcements and Their Effects on Interest Rates." Journal of Monetary Economics, December 1987, 439-462.
22. Leeper, Eric M. and David B. Gordon, "In Search of the Liquidity Effect." Journal of Monetary Economics, June 1992, 341-69.
23. Mussa, Michael L., "Empirical Regularities in the Behavior of Exchange Rates and Theories of the Foreign Exchange Market." Carnegie-Rochester Series on Public Policy, Autumn, 1979, 9-57.
24. Pearce, Douglas K. and V. Vance Roley, "Stock Prices and Economic News." Journal of Business, January 1985, 49-67.
25. Roley, V. Vance, "The Response of Short-Term Interest Rates to Weekly Money Supply Announcements: A Note." Journal of Money, Credit and Banking, August 1983, 344-54.
26. -----, "The Response of Short-Term Interest Rates to Weekly Money Announcements: A Reply." Journal of Money, Credit and Banking, May 1985, 271-73.
27. -----, "The Effects of Money Announcements Under Alternative Monetary Control Procedures." Journal of Money Credit and Banking, August 1987, 292-307.
28. Sheehan, Richard G., "Weekly Money Announcements: New Information and Its Effects." Federal Reserve Bank of St. Louis Review, August/September 1985, 25-34.
29. Sichel, Daniel E., "Business Cycle Asymmetry: A Deeper Look." Economic Inquiry, April 1993, 224-36.
30. Thornton, Daniel L., "The Effects of Unanticipated Money on the Money and Foreign Exchange Markets." Journal of International Money and Finance, December 1989, 573-587.
31. -----, "Why do Market Interest Rates Respond to Money Supply Announcements?" Journal of International Financial Markets and Money, 1991, 33-60.
32. Urich, Thomas J. and Paul Wachtel, "The Effect of Inflation and Money Supply Announcements on Interest Rates." Journal of Finance, September 1984, 1177-88.
33. Zellner, Arnold, "Biased Predictors, Rationality and the Evaluation of Forecasts." Economics Letters, 1986, 45-48.
|Printer friendly Cite/link Email Feedback|
|Author:||Wohar, Mark E.|
|Publication:||Southern Economic Journal|
|Date:||Jan 1, 1995|
|Previous Article:||Asymmetric nominal flexibility and economic fluctuations.|
|Next Article:||Legal versus illegal U.S. immigration and source country characteristics.|