شوک های سیاست پولی روزانه و خرید و فروش های خانه های جدید
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|26451||2008||20 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Monetary Economics, Volume 55, Issue 7, October 2008, Pages 1171–1190
The conventional notion of a monetary policy shock as a surprise change in the fed funds rate is misspecified. The primary news for market participants is not what the Fed just did, but is instead new information about the Fed's future intentions. Revisions in these anticipations show up instantaneously in long-term mortgage rates. Home sales do not respond until much later. This paper attributes this delay—and hence much of the hump-shaped response of economic activity to monetary policy—to cross-sectional heterogeneity in search times. This framework allows one in principle to measure policy impacts at the daily frequency.
How can the effects of monetary policy on the economy be measured? One popular approach (e.g., Christiano et al., 1999) is based on a structural VAR. Let ymym denote a vector of variables observed for month mm of which the average fed funds rate over the month, rmrm, is one element. Consider a linear forecast of ym+sym+s based on lagged values of yy (denoted Ωm-1Ωm-1) and some subset of the current values of yy (denoted ΛmΛm). How would news about the value of rmrm, denoted umum, cause the forecast to change? The standard impulse-response function is simply a graph of the answer to this question1: equation(1) View the MathML source∂E^(ym+s|um,Λm,Ωm-1)∂um. Turn MathJax on Much of the discussion in the literature concerns which elements of ymym to include in the contemporaneous information set ΛmΛm. However, this choice often proves of limited consequence. Fig. 1 displays impulse-response functions for a fairly standard VAR including industrial production, the CPI, commodity prices, the fed funds rate, and M2.2 One sees the same broad hump-shaped response, with an increase in the fed funds rate being followed after a substantial delay by a slowdown in industrial production, regardless of the specification of ΛmΛm.Although the choice of ΛmΛm makes little difference for the answer to this forecasting question, the specification of the lagged information set Ωm-1Ωm-1 is quite significant, as stressed by Rudebusch (1998) and Brissimis and Magginas (2006). The top panel of Fig. 2 plots the errors umum associated with a monthly VAR (in which the fed funds rate is ordered last) between 2003:01 and 2006:06. The second panel plots the difference between rmrm and the forecast implied by the 1-month fed funds futures contract on the last day of month m-1m-1. Over this period, changes in the fed funds rate that would be characterized by the VAR as monetary policy shocks were in fact almost perfectly anticipated by market participants.This is not to say that there were no surprises in monetary policy over this period. However, any surprises were not about what the Fed just did, but instead reflected new information about what the Fed was going to do in the future. The bottom panel plots revisions during each month in the anticipation of what the fed funds rate was going to be 2 months after the indicated month. For example, what the VAR classifies as a surprisingly high fed funds rate in July 2004 actually showed up as news to markets in a much more modest adjustment in the July fed funds contract price between April 30 and May 31. This paper presents evidence that, for purposes of determining long-term mortgage rates or new home sales, only unanticipated monetary policy changes matter, that is, it is the bottom panels rather than the top panel inFig. 2 that will affect the economy. Before proposing an alternative to the forecasting question posed by (1), let me clarify what it is that I believe we are trying to estimate. The primary input the Fed needs from empirical researchers is an answer to questions like the following: We’re trying to decide between a funds rate of 5 or 5.25. How would the predicted path for ym+sym+s be different under the two choices? This question is potentially related to the impulse-response function in (1), in that both represent questions about a conditional forecast. However, interpreting an object like (1) as telling us the answer to the policy question of interest faces two challenges. First, for purposes of the policy question, it is clear that the information set we would like to use is all the information available to the Fed prior to the decision. That suggests that the bottom panels of Fig. 2 are more promising measures than the top, and indeed ideally we would want to use forecasts from the day before rather than a month before the Fed's decision. The second challenge is whether the conditions that caused surprise movements in rmrm in the sample are comparable to those that would govern the outcome of the policy question currently contemplated. For example, if historically the Fed raised rmrm in response to new inflation fears that month, to what extent is the change in ym+sym+s a result of the inflation itself, and to what extent does it result from the choice made by the Fed? A number of papers have sought to resolve these problems by looking at the change in expectations of the fed funds rate on the day of a Fed policy change or announcement itself, supposing that on such days, the answer to the forecasting question might isolate the effect of policy alone. Such studies include Kuttner (2001), Cochrane and Piazzesi (2002), Faust et al. (2004), Gürkaynak et al., 2005a and Gürkaynak et al., 2005b, and Andersson et al. (2006), among others. All of the above papers simply assume that the conditional forecasting question has a different answer on these days relative to others. To my knowledge, mine is the first paper to test this assumption, and I find that over the period 1988–2006, it appears not to be the case. That finding opens up a vastly bigger and richer data set than previous researchers have used for purposes of calculating how revisions of forecasts of the fed funds rate are associated with revisions of the forecasts of other macro-variables of interest. This paper also differs from most previous studies in making primary use of daily innovations in the 1- and 2-month-ahead futures contracts rather than the spot-month contract. As suggested in the second and third panels of Fig. 2, the most important news about the Fed in recent years has been information about what it is going to do rather than information about what it just did. Poole and Rasche (2000), Gürkaynak (2005), and Gürkaynak et al. (2007) looked at the comovements between near-month contracts (rather than spot-month contracts) and asset prices on policy days. But the current paper again appears to be the first to link changes in these contracts directly to subsequent changes in a measure of real economic activity. The way in which the paper does so is also methodologically novel, proposing a new method for combining data observed at different frequencies based on parametric restrictions inspired by the observed cross-sectional heterogeneity in search times. The plan of the paper is as follows. Section 2 reviews evidence on the time-series properties of daily changes in near-term fed funds futures prices, and concludes that these changes primarily result from daily changes in a rational anticipation of what the Fed is going to do next. Section 3 documents that weekly mortgage rates follow a near martingale, and relates its innovations to daily changes in fed funds futures. This relation appears to be invariant with respect to which day of the week one uses, whether one uses only those changes associated with policy announcement days, days of particular macroeconomic news releases, or the level of time aggregation up to a month. Section 4 investigates the forecasting relation between interest rates and the level of new home sales, documenting that there is a very long, sustained lag. Some of the sales for a given month depend on mortgage rate changes that occurred during the previous month, while sales of other homes within that same month appear to be responding to mortgage rates up to 6 months earlier. The paper attributes this lag in part to heterogeneity across households. The mean lag of the time-series relation turns out to match closely the mean lag of the cross-sectional distribution across different households in the time spent searching before buying a home. Taken together, the evidence supports the following interpretation of the way in which monetary policy affects the economy. Current mortgage rates reflect a rational anticipation of everything the Fed may do in the future. If the Fed wants to change mortgage rates, it has to do something other than what the market expected. Any new information about what the Fed is going to do shows up essentially instantly in mortgage rates, but due to heterogeneity across households in information-processing and search times, shows up only gradually over time in new home sales. The biggest effect on home sales is observed 15 weeks after the change in policy is first perceived by futures markets.
نتیجه گیری انگلیسی
The current mortgage rate reflects a rational anticipation of all future Fed policy actions. In order to change the mortgage rate, the Fed must do something other than what the market anticipated, and any change in Fed policy shows up in mortgage rates as soon as the market anticipates it. An unanticipated 10-basis-point increase in the level of the term structure of near-term expected fed funds rates raises the mortgage rate by 5 basis points. An unanticipated 10-basis-point increase in the slope raises the mortgage rate by 13 basis points. The consequences of such changes do not have their peak effect on new home sales until 15 weeks after mortgage rates go up. This delay might be attributed to heterogeneity across households in the time required to learn about changes in mortgage rates and to buy a new home. These dynamic relations, which have been directly estimated in detail here using daily and weekly time-series data, are claimed to account for some of the long lags found in more traditional analysis using time-aggregated monthly data. The framework also enables us to summarize on a daily or even minute-by-minute basis, if desired, the cumulative consequences of recent innovations in Fed policy or hypothetical future scenario as of any particular historical moment