Recent efforts to understand the transmission

Transcription

1 Commentary Kenneth N. Kuttner Recent efforts to unerstan the transmission of monetary policy have spawne a growing literature examining the response of financial markets to monetary policy. 1 Most of these stuies assess the likely impact of unanticipate changes in the target feeral funs rate, typically in a sample of well-efine policy events consisting of Feeral Open Market Committee (FOMC) meeting ays, plus the ays of unscheule funs rate changes. The problem is that economists o not always know the ays on which the policy actions took place, especially in the early 1990s. Before the FOMC began announcing its policy actions in February 1994, there was often some confusion in the financial markets as to whether there ha been a change in the funs rate target. This ambiguity has been largely ispelle by the FOMC s announcements, although, as Hamilton (2008) notes, there has been occasional speculation that the Fe has surreptitiously change the target rate. 2 Hamilton s (2008) paper is primarily an effort to aress the issue of unknown event ates. It eparts from the usual assumption that the ays of policy actions (or possible actions) are known 1 The first paper in this literature was Cook an Hahn (1989). Subsequent work inclues Poole an Rasche (2000), Kuttner (2001), Poole, Rasche, an Thornton (2002), Gürkaynak, Sack, an Swanson (2005), an Bernanke an Kuttner (2005). 2 So far, none of this speculation has prove to be correct. an uses instea a signal-extraction approach to etermine the market s reaction without conitioning on this information. His elegant approach allows the market s reaction to be estimate using the entire sample, not just event ays. Moreover, the approach allows for the measurement of financial markets response to evolving expectations of future Fe actions, a feature that allows him to extract information even when the Fe oes not surprise the markets. The analysis focuses on the response of term interest rates, as in Kuttner (2001), although there is no reason the same approach coul not also be applie to stock prices or exchange rates. The paper s key empirical results largely confirm those reporte elsewhere, which is goo news for those of us who have use the much simpler event-stuy approach. The response of term interest rates to changes in the funs rate is uniformly less than one for one, an the effect on longer-term interest rates is generally less than it is for short-term rates. It is interesting to note, however, that this latter tenency is less pronounce than it is in Kuttner s (2001) results. My iscussion will focus on two issues. The first point is somewhat technical, as it concerns the etails of how the noise in the feeral funs rate is moele. The secon is a more conceptual iscussion of how the interpretation of the shocks ientifie by Hamilton s proceure might iffer from those in conventional event-stuy analyses. Kenneth N. Kuttner is a professor of economics at Williams College an a research associate at the National Bureau of Economic Research. Feeral Reserve Bank of St. Louis Review, July/August 2008, 90(4), pp , The Feeral Reserve Bank of St. Louis. The views expresse in this article are those of the author(s) an o not necessarily reflect the views of the Feeral Reserve System, the Boar of Governors, or the regional Feeral Reserve Banks. Articles may be reprinte, reprouce, publishe, istribute, isplaye, an transmitte in their entirety if copyright notice, author name(s), an full citation are inclue. Abstracts, synopses, an other erivative works may be mae only with prior written permission of the Feeral Reserve Bank of St. Louis. FEDERAL RESERVE BANK OF ST. LOUIS REVIEW JULY/AUGUST

2 Figure 1 The Target an Effective Funs Rates, 1995 Percent Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1995 NOTE: Vertical lines enote settlement Wenesays. MODELING FUNDS RATE NOISE 3 This noise results from the fact that the New York Fe s control over the funs rate is not absolute: Their Traing Desk injects just enough reserves to hit the target funs rate, given its assessment of the factors affecting reserve eman an supply. However, because of unanticipate changes in eman or supply, the actual ( effective ) funs rate may iffer from the target. Unlike the more common event-stuy analysis, Hamilton s signal-extraction metho requires statistically moeling the noise process present in the aily effective feeral funs rate. 3 Intuitively, the reason for this is that calculating the likely signal present in any given funs rate change requires some estimate as to the amount of noise likely to be present on any given ay: The noisier the effective funs rate, the less likely it is that the observe change in the rate (an, by extension, the expecte rate implie by the current-month futures contract) represents a policy change. Observing that the magnitue of these eviations tens to increase over the course of a month, Hamilton moels the targeting error as an autoregressive process whose innovation variance is a linear function of the ay of the month (equations (8) an (9)). To get a sense of the magnitue of these targeting errors, his estimate parameters imply a 45-basis-point stanar eviation on the 31st ay of the month, 34 basis points on the 30th ay, an 17 basis points on the 1st ay. Although this is not an unreasonable first pass, some refinements are possible. First, because there is no reason to think that the en-of-month volatility in 31-ay months is greater than it is for 30-ay months, it woul be esirable to relax the assumption of 31-ay months an replace equation (9) with ˆ 2 Ni t e = a+ b 05. ( ) + ν ˆ, where N i is the number of ays in month i. A secon important refinement woul be to account for the settlement Wenesay effect. Especially in the early part of the sample, the Wenesays associate with the final ay of the reserve maintenance perio were often associate JULY/AUGUST 2008 FEDERAL RESERVE BANK OF ST. LOUIS REVIEW

3 Figure 2 The Target an Effective Funs Rates, 2002 Percent Percent 2.00 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2002 NOTE: The vertical lines in the top panel enote settlement Wenesays; in the bottom panel they mark the last ay of the month. with extremely large funs rate spikes, as shown in Figure 1. (The vertical lines enote settlement Wenesays.) To account for this pattern, a reasonable specification for the targeting error might be something like r ξ = 03. ( r ξ )+ 14W + 19M + eˆ 1 1 ˆ 2 e = W + 1, 422W 27M , M + ν ˆ, where W is a ummy equal to 1 on settlement Wenesays an M is a ummy equal to 1 on the last ay of the month. The other notation is the same as Hamilton s. 4 Three features of this alternative specification are particularly interesting. One is that there are 4 The parameter estimates are estimate using orinary least squares from May 17, 1989, through October 12, 2007, excluing September 2001 an December significant level effects associate with settlement Wenesays an with the last ay of the month: Errors on these ays ten to be positive. The secon is that the stanar eviation of the targeting error is 27 basis points higher on settlement Wenesays. Thir, unlike in Hamilton s specification, there is no evience of a month-en effect, except on the very last ay of the month. Other information about changes in the feeral funs market can also be brought to bear to further refine the specification. One such change is the shift to lagge reserve accounting as of July 30, Partly as a result of this change, the monthen an settlement-wenesay volatility of the funs rate, as well as the overall variance, has fallen sharply in recent years. Post 1998, the stanar eviation of the last-ay-of-month targeting error is only 22 basis points (compare with Hamilton s last-ay estimate of 45 basis points), an there is no longer any evience of a settlement-wenesay spike. That the Feeral Reserve FEDERAL RESERVE BANK OF ST. LOUIS REVIEW JULY/AUGUST

4 Bank of New York s Traing Desk has improve control over the funs rate is reaily apparent in Figure 2. (The vertical lines in the top panel enote settlement Wenesays; in the bottom panel they mark the last ay of the month.) Finally, in refining the estimates of the targeting error process, one woul want to make allowances for special circumstances affecting the feeral funs market. Hamilton alreay makes one such allowance, omitting September 2001 from the sample use for estimating the moel. December 1999 shoul be roppe for similar reasons: With the Y2K changeover approaching, the Fe flooe the market with reserves in an effort to assuage liquiity concerns. Consequently, the funs rate trae as much as 150 basis points below its target as the en of the month approache. Incluing atypical episoes, such as this one, coul overestimate the amount of noise normally present in the effective funs rate. It is important to emphasize that none of these observations unercuts in any way the sounness of Hamilton s basic approach. In particular, I can think of no reason to suspect that any misspecification in equations (8) or (9) woul necessarily bias the parameter estimates reporte in Hamilton s Table 2. Instea, it is more akin to the problem of choosing inappropriate weights in a weighte-least-squares proceure: In that case, while the parameter estimates may not be biase, the proceure is not making optimal use of the information containe in the ata. ON INTERPRETING THE HAMILTON SHOCKS The secon part of my remarks concerns the interpretation of the funs rate shocks unerlying Hamilton s proceure. By way of backgroun, it may be useful to istinguish between two ifferent regimes. In the first regime, changes in the funs rate target are equally likely on any ay but changes in the target are not announce by the FOMC. This regime plausibly correspons to the pre-1994 worl, in which policy actions were generally not isclose an a significant fraction of rate changes took place between meetings. In this regime, ay-to-ay changes in the futuresimplie rate on any particular ay woul plausibly represent the market s inference as to whether the Fe ha change its target on that ay. In the secon regime, which is more relevant post 1994, the ays of the rate changes are largely known; an even when policy actions are taken between FOMC meetings, the changes are announce, an not in response to any specific news that might have arrive on that ay. In this case, the ay-to-ay change in the futures rate on ays other than event ays (i.e., ays of rate changes or FOMC meetings) woul reflect changes in the market s expectation of the target funs rate on some future ate. Now consier the sources of news that coul affect policy expectations. One source is new macroeconomic information: higher-than-expecte employment, for example, or lower-than-expecte inflation. The other source woul be changes in the Fe s perceive preferences regaring inflation vis à vis output the presume source of monetary policy shocks, as the term is commonly use in the literature. These istinctions bear on how we shoul interpret the information containe in alternative measures of monetary policy shocks or surprises. In the secon regime, policy surprises (i.e., changes in the futures rate) occurring on event ays are more likely to be riven by the secon category of news: changes in the Fe s perceive preferences. 5 Changes occurring on ays other than event ays woul, for the most part, be associate with the arrival of economic news. In the first regime, however, ay-to-ay changes in the futures rate coul result from either source: changes in policy preferences or macro news. Thus, conitioning on known event ays allows the econometrician to istinguish between the enogenous response of policy expectations to new economic information an otherwise inexplicable policy shocks. This istinction can be critically important in assessing the financial market response to monetary policy. As shown in 5 It is also possible that the change woul be interprete as the Fe reacting to private information, although the evience for this view is weak; see Faust, Swanson, an Wright (2004) JULY/AUGUST 2008 FEDERAL RESERVE BANK OF ST. LOUIS REVIEW

5 Bernanke an Kuttner (2005), the stock market s reaction to unanticipate funs rate changes is effectively zero when those changes occurre on the same ay as an employment report, a pattern that was common in the early 1990s. Any analysis that faile to make this istinction coul provie a misleaing answer to the primary question of interest to policymakers: how the market will react to an unexpecte change in the fun rate target. Within Hamilton s framework, it woul be easy to make this istinction. In fact, it suggests an interesting test of the null hypothesis that the response to (calenar-ajuste) changes in the futures rate is the same on event ays as it is on non-event ays. The alternative, of course, woul be that the reaction iffers in a systematic way. Given the richness of the ataset, it woul be possible to go further an istinguish between event ays an the ays of specific economic news releases (e.g., inflation, employment, gross omestic prouct). This assumes that the relevant ays are known, of course but after 1994, this is not such a ba assumption. Econometrically, the only moification to Hamilton s proceure woul be to allow the relevant ummy variables to interact with the slope coefficient in equation (26). CONCLUSION None of these points takes away from the bottom line: The paper is a classic Hamilton timeseries tour e force. It aresses an important question using elegant econometrics, an it incorporates a etaile knowlege of the market for feeral funs. Using more of that knowlege to refine the targeting-error specification woul enhance an alreay fine paper, as woul further efforts to unerstan what the shocks in the moel really represent. REFERENCES Bernanke, Ben S. an Kuttner, Kenneth N. What Explains the Stock Market s Reaction to Feeral Reserve Policy? Journal of Finance, June 2005, 60(3), pp Cook, Timothy an Hahn, Thomas. The Effect of Changes in the Feeral Funs Rate Target on Market Interest Rates in the 1970s. Journal of Monetary Economics, November 1989, 24(3), pp Faust, Jon; Swanson, Eric T. an Wright, Jonathan H. Do Feeral Reserve Policy Surprises Reveal Private Information About the Economy? Contributions to Macroeconomics, 2004, 4(1), pp Gürkaynak, Rafet S.; Sack, Brian P. an Swanson, Eric T. Do Actions Speak Louer Than Wors? The Response of Asset Prices to Monetary Policy Actions an Statements. International Journal of Central Banking, June 2005, 1(1), pp Hamilton, James D. Assessing Monetary Policy Effects Using Daily Feeral Funs Futures Contracts. Feeral Reserve Bank of St. Louis Review, July/August 2008, 90(4), pp Kuttner, Kenneth N. Monetary Policy Surprises an Interest Rates: Evience from the Fe Funs Futures Market. Journal of Monetary Economics, June 2001, 47(3), pp Poole, William an Rasche, Robert H. Perfecting the Market s Knowlege of Monetary Policy. Journal of Financial Services Research, December 2000, 18(2/3), pp Poole, William; Rasche, Robert H. an Thornton, Daniel L. Market Anticipations of Monetary Policy Actions. Feeral Reserve Bank of St. Louis Review, July/August 2002, 84(4), pp FEDERAL RESERVE BANK OF ST. LOUIS REVIEW JULY/AUGUST

6 4 0 4 JULY/AUGUST 2008 FEDERAL RESERVE BANK OF ST. LOUIS REVIEW