اثرات سیاست پولی بر تولید چیست؟ نتایج حاصل از یک روش شناسایی اگنوستیک
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|25555||2005||39 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Monetary Economics, Volume 52, Issue 2, March 2005, Pages 381–419
This paper proposes to estimate the effects of monetary policy shocks by a new agnostic method, imposing sign restrictions on the impulse responses of prices, nonborrowed reserves and the federal funds rate in response to a monetary policy shock. No restrictions are imposed on the response of real GDP to answer the key question in the title. I find that “contractionary” monetary policy shocks have no clear effect on real GDP, even though prices move only gradually in response to a monetary policy shock. Neutrality of monetary policy shocks is not inconsistent with the data.
What are the effects of monetary policy on output? This key question has been the focus of a substantial body of the literature. And the answer seems easy. The “Volcker recessions” at the beginning of the 1980s have shown just how deep a recession a sudden tightening of monetary policy can produce. Alternatively, look at Fig. 1, which juxtaposes movements in the federal funds rate from 1965 to 1996 with growth rates in real GDP, flipped upside-down for easier comparison. In particular, for the first half of that sample, it is striking, how rises in the federal funds rate are followed by falls in output (visible as rises in the dotted line, due to the upside-down flipping). The case is closed.Or is it? Eyeball econometrics such as Fig. 1 or case studies like the Volcker recessions can be deceptive: many things are going on simultaneously in the economy, and one may want to be careful to consider just a single cause-and-effect story. If the answer really is so obvious, it should emerge equally clearly from an analysis of multiple time series, which allows for additional channels of interaction and other explanations, at least in principle. Thus, many researchers have followed the lead of Sims, 1972, Sims, 1980 and Sims, 1986 and proceeded to analyze the key question in the title with the aid of vector autoregressions. Rapid progress has been made in the last 10 years. Bernanke and Blinder (1992) shifted the focus on the federal funds rate. The ‘price puzzle’, raised by Sims (1992), and other anomalies, led to the inclusions of e.g. nonborrowed reserves, total reserves as well as a commodity price index in VAR studies, see e.g. Eichenbaum (1992), Strongin (1995), Christiano and Eichenbaum, 1992a and Christiano and Eichenbaum, 1992b, Leeper and Gordon (1992), Gordon and Leeper (1994), Christiano et al., 1996, Christiano et al., 1997 and Christiano et al., 1999 and Kim (1999). Recently, Bernanke and Mihov, 1998a and Bernanke and Mihov, 1998b have reconciled a number of these approaches in a unifying framework, and Leeper et al. (1996) have summarized the current state of the literature, while adding new directions on their own. Additional excellent surveys are in Canova (1995), Christiano et al. (1999) and Bagliano and Favero (1998). There seems to be a growing agreement that this literature has reached a healthy state, and has provided a list of facts, which now theorists ought to explain, see e.g. Christiano et al., 1996, Christiano et al., 1997 and Christiano et al., 1999 or Leeper and Sims (1994). The key step in applying VAR methodology to the question at hand is in identifying the monetary policy shock. While this is usually done by appealing to certain informational orderings about the arrival of shocks, there also is a more informal side to the identification search: researchers like the results to look “reasonable”. According to conventional wisdom, monetary contractions should raise the federal funds rate, lower prices and reduce real output. If a particular identification scheme does not accomplish this, then the observed responses are called a puzzle, while successful identification needs to deliver results matching the conventional wisdom. The “facts” that are obtained this way are thus necessarily influenced by a priori theorizing. There is a danger that the literature just gets out what has been stuck in, albeit more polished and with numbers attached. Without being explicit about this a priori theorizing, it is hard to distinguish between assumptions and conclusions. This circularity is well recognized in the literature, has already been clearly pointed out by Cochrane (1994), and has been dealt with in a variety of ways. Leeper et al. (1996) explicitly appeal to the reasonableness of impulse responses as an “informal” identification criterion. Gali (1992) directly asks whether the “IS-LM model fit[s] the postwar U.S. data” rather than indirectly presuming that this is the only model worth fitting. Cochrane (1994) and Rotemberg (1994) argue that economic theory is crucially important for identifying monetary policy shocks: a VAR analysis of these shocks only has a chance to be convincing, if the results look plausible to begin with. Christiano et al. (1999) propose to throw out all impulse responses inconsistent with some given set of theories, some of which are at odds with the conventional wisdom. Joint estimation of a theoretical model and a VAR is done in e.g. Altig et al. (2002). Priors for a VAR from an explicitly formulated theory are constructed in Del Negro and Schorfheide (2003). In sum, the answer to the key question—here, the impact of monetary policy shocks on GDP—is often already substantially narrowed down by a priori theorizing, be it implicit or explicit. What is therefore desirable as a complement to the existing literature is some way to make the a priori theorizing explicit (and use as little of it as possible), while at the same time leaving the question of interest open. This paper proposes to push this idea all the way, and to identify the effects of monetary policy shocks by directly imposing sign restrictions on the impulse responses. More specifically, I will assume that a “contractionary” monetary policy shock does not lead to increases in prices, increase in nonborrowed reserves, or decreases in the federal funds rate for a certain period following a shock. While theories with different implications can fairly easily be constructed, these assumptions may enjoy broad support and in any case are usually tacitly assumed in most of the VAR literature. In the approach here, they are brought out into the open and can therefore be subject to debate. Crucially, I impose no restrictions on the response of real GDP. Thus, the central question in the title is left agnostically open by design of the identification procedure: the data will decide. I call the procedure “agnostic” for this reason. One can think about the procedure as identifying all shocks which are consistent with these fairly weak a priori restrictions, and that the literature (insofar it delivers impulse responses also obeying the sign restrictions) uses further a priori identifying restrictions to only select a subset of these shocks. This will not be a free lunch, nor should one expect it to be. When imposing the sign restrictions, one needs to take a stand on for how long these restrictions ought to hold after a shock. Furthermore, one needs to take a stand on whether a strong response in the opposite direction is more desirable than a weak one. I will try out a variety of choices and look at the answers. Section 2 introduces the method with most of the technicalities postponed to Appendices A and B. Section 3 shows some results, based on the data set provided by Bernanke and Mihov, 1998a and Bernanke and Mihov, 1998b, extended until the end of 2003. Section 4 concludes. My approach is asymmetric in that I am agnostic about the response of output but not of some other variables. This is intentional: the response of output is the focus of this investigation. Nonetheless, it is interesting to also report findings about the other variables, keeping in mind that they are tainted by a priori sign restrictions. I find the following: 1. “Contractionary” monetary policy shocks have an ambiguous effect on real GDP. With View the MathML source23 probability, a typical shock will move real GDP by up to ±0.2±0.2 percent, consistent with the conventional view, but also consistent with e.g. monetary neutrality. Indeed, the usual label “contractionary” may thus be misleading, if output is moved up. Monetary policy shocks account for probably less than 25% of the variance for the 1-year or more ahead forecast revision of real output, and may easily account for less than 2% at any given horizon. 2. The GDP price deflator falls only slowly following a contractionary monetary policy shock. The commodity price index falls more quickly. 3. I also find, that monetary policy shocks account for only a small fraction of the forecast error variance in the federal funds rate, except at horizons shorter than half a year, as well as for prices. While these observations confirm some of the results found in the empirical VAR literature so far, there are also some potentially important differences in particular with respect to my key question: “contractionary” monetary policy shocks do not necessarily seem to have contractionary effects on real GDP. Our conclusion from these results: one should feel less comfortable with the conventional view and the current consensus of the VAR literature than has been the case so far. The new method introduced here complements the work by Blanchard and Quah (1989), Lippi and Reichlin, 1994a and Lippi and Reichlin, 1994b and in particular by Dwyer (1997), Faust (1998), Gambetti (1999), Canova and Pina (1999) and Canova and de Nicolo (2002): these authors also impose restrictions on the impulse responses to particular shocks. Like Faust, Dwyer and Canova–de Nicolo, my aim is to make explicit restrictions which are often used implicitly. But there are also important differences. I do not impose a particular shape of the impulse response as in Lippi and Reichlin (1994a) or Dwyer (1997) or impose a zero impulse response at infinity as in Blanchard and Quah (1989). Instead, I am content with restrictions on the sign at a few periods following the shock, making for substantial differences between their approach and ours. The intention here is to be minimalistic and to impose not (much) more than the sign restrictions themselves, as they can be reasonably agreed upon across many economists. Faust (1998) also only imposes sign restrictions to restrict monetary policy shocks. His focus is a different one. Faust examines the fragility of the consensus conclusion, that monetary shocks account for only a small fraction of GDP fluctuations, see Cochrane (1994), while this paper aims at estimating that response. Furthermore, Faust only imposed sign restrictions on impact. In my discussion (Uhlig, 1998), I have shown how his approach can be extended, when one wishes to impose the sign restrictions for several periods following the shock. The method by Canova and de Nicolo (2002) and its application in Canova and Pina (1999) identifies monetary disturbances by imposing sign restrictions on the cross-correlations of variables in response to shocks, adding restrictions until the maximum number of shocks is uniquely identified. The identification here proceeds differently by using impulse responses rather than cross-correlations, by using other criteria used to select among orthogonal decompositions satisfying the restrictions, and by not imposing increasingly stringent restrictions to eliminate candidate orthogonalizations. I do not aim at a complete decomposition of the one-step ahead prediction error into all its components due to underlying structural shocks, but rather concentrate on identifying only one such shock, namely the shock to monetary policy. Similarly, Bernanke and Mihov, 1998a and Bernanke and Mihov, 1998b or Christiano et al. (1999) only identify a single shock or a subset of shocks. They impose considerably more structure than I do here. Again, the aim is to be minimalistic, and to use as little a priori reasoning about other shocks as possible in order to identify the effects of monetary policy shocks. The identification of additional shocks can help in principle, as orthogonality between the shocks provides an additional restriction for identifying the monetary policy shock, and there may be those who argue that it is even necessary. The method can fairly easily be extended in this direction; if necessary, see Mountford and Uhlig (2002) for an example.
نتیجه گیری انگلیسی
This paper proposed a new agnostic method to estimate the effects of monetary policy, imposing sign restrictions on the impulse responses of prices, nonborrowed reserves and the federal funds rate in response to a monetary policy shock. No restrictions are imposed on the response of real GDP. It turned out that 1. “Contractionary” monetary policy shocks have an ambiguous effect on real GDP, moving it up or down by up to ±0.2%±0.2% with a probability of 2/3. Monetary policy shocks account for probably less than 25% of the k-step ahead prediction error variance of real output, and may easily account for less than 3%. 2. The GDP price deflator falls only slowly following a contractionary monetary policy shock. The commodity price index falls more quickly. 3. Monetary policy shocks account for only a small fraction of the forecast error variance in the federal funds rate, except at horizons shorter than half a year. They account for about one quarter of the variation in prices at longer horizons. In sum, even though the general price level moves very gradually for a period of about a year, monetary policy shocks have ambiguous real effects and may actually be neutral. These observations largely confirm the results found in the empirical VAR literature so far, except for the ambiguity regarding the effect on output. This exception is, of course, a rather important difference. “Contractionary” monetary policy shocks do not necessarily seem to have contractionary effects on real GDP. One should therefore feel less comfortable with the conventional view and the current consensus of the VAR literature than has been the case so far. The key identifying assumption explaining the difference between my results and the results of, say, a conventional Cholesky decomposition appears to be that I do not restrict the on-impact response of real GDP to be zero. The paper agrees with a number of other publications in the literature, that variations in monetary policy account only for a small fraction of the variation in any of these variables. Good monetary policy should be predictable policy, and should not rock the boat. From that perspective, monetary policy in the U.S. during this time span has been successful indeed.