پیامدهای اقتصاد باز سیاست پولی آمریکا

We consider the open economy consequences of U.S. monetary policy, extending the identification approach of Romer and Romer (2004) and adapting it for use with asset prices. Intended policy changes are orthogonalized against the economy’s expected future path, which captures any effects from open economy variables. Estimated from a set of bilateral VARs, the dynamic responses of the exchange rate, foreign interest rate, and foreign output are consistent with recent work that identifies U.S. policy via futures market changes and a priori impulse response bounds. We compare the two approaches, finding important commonalities. We also outline some advantages of our approach.

Estimating the impact of unanticipated and exogenous monetary policy shocks on open economy variables such as exchange rates and foreign interest rates is a perennial desideratum in the field of international finance. Important contributions to the open economy empirics of monetary policy include Eichenbaum and Evans, 1995, Kim and Roubini, 2000 and Faust and Rogers, 2003, and Faust et al. (2003). The identification of monetary policy shocks from broader movements in monetary instruments poses a particular challenge in the open economy context because of simultaneity amongst asset returns invalidating many of the short-run restrictions used to identify policy innovations in structural vector autoregressions (SVARs). In a partially identified framework, Faust and Rogers (2003) show that the exclusion restrictions associated with the recursive ordering of variables posited by Eichenbaum and Evans (1995) for a set of two-country VARs can be rejected for some country pairs. Furthermore, they find that the evidence for delayed exchange rate overshooting is sensitive to relaxing such restrictions. Our strategy for addressing such simultaneity concerns is to incorporate information from outside of a VAR framework to achieve monetary policy identification.2 We can then allow the contemporaneous and dynamic responses of open economy variables in the model to be completely unrestricted. The approach builds upon the work of Romer and Romer (2004), who use a two-step procedure to identify U.S. monetary policy shocks. In the first step, they use narrative sources to determine intended changes in the U.S. target federal funds rate. In the second step, these federal funds rate changes are decomposed into two components, one that can be explained by the Federal Reserve’s Greenbook (in-house) forecasts for output growth, inflation, and unemployment, and one that cannot be explained by those forecasts. The second component, which captures interventions unrelated to economic forecasts, defines the monetary policy shock. In our implementation, we extend both parts of the identification procedure. In the first stage, we augment the Romers’ narratively identified changes in the target federal funds rate at Federal Open Market Committee (FOMC) meetings for the period 1969–1996 with announced changes in the target for the period 1997–2001.3 In the second stage, the augmented series is orthogonalized against a meeting-based information set that combines the Greenbook forecasts used by the Romers and a measure of capacity utilization constructed by the Federal Reserve.4 If the controls in the orthogonalization adequately summarize the FOMC’s information set regarding its objectives and expectations at the time of its decision, then the subsequent residuals represent the unanticipated and exogenous component of U.S. monetary policy.5 We convert the resulting identified policy series from a meeting frequency to a monthly frequency via a novel aggregation method which ensures that period average asset price responses are correctly estimated. The identified policy series is then an exogenous variable in the estimation of the dynamic effects of U.S. policy on the exchange rate and other open economy variables. To facilitate comparisons with previous work, we include our identified policy series in a set of 6 bilateral, monthly VARs, where the U.S. is Home and one of the non-U.S. G7 countries is Foreign. We find that the open economy dynamic responses to a contractionary monetary policy shock which raises the U.S. t-bill rate by 100 basis points (b.p.) are the following: 1. There is always an impact appreciation of the U.S.$ exchange rate. 2. The maximal appreciation of the exchange rate occurs between 1 and 2 months (Canada and Germany) and 20 months (France and Italy). It lies between 0.66% (Canada) and 4.38% (France). 3. There is strong, positive interest rate pass-through from the U.S. to the foreign countries, with the maximum response occurring between 3 months (Canada) and 10 months (Japan and the U.K.). It lies between 0.46 (Japan) and 1.24 (Canada). 4. Foreign output shows a mixed initial response (some positive and some negative) which uniformly becomes negative at horizons of 16 months (earlier for many). U.S. output responses are generally negative, with the maximum contractions ranging from 1.3 to 2.1 percentage points and occurring between 17 and 27 months after the initial U.S. policy shock. The U.S. price responses in the baseline model are disappointing, showing a price puzzle. We investigated the price responses in detail, finding them to depend critically on the number of lags of the policy shock included in the model. As the number of policy shock lags is increased, the price puzzle disappears while preserving many of the other impulse responses. The response of U.S. non-borrowed reserves exhibits the classic liquidity effect, where a contractionary monetary policy shock leads to a decline in reserves held and a rise in domestic interest rates. We discuss all of these responses in more detail later. There are of course other approaches to estimating the open economy effects of U.S. monetary policy which incorporate information from outside a VAR framework in order to achieve identification. We contextualize our findings by comparing them to those of Faust et al. (2003), a notable recent example which is a natural comparator for our study.6 They estimate the high frequency (intra-day or daily) responses of spot exchange rates, Home and Foreign spot interest rates, and three and six month Home and Foreign forward interest rates to U.S. monetary policy shocks. Following Kuttner (2001), the policy shocks are measured from changes in the daily closing price of current-month federal funds futures contracts around FOMC meetings. The maintained assumptions for the validity of their approach are that: (1) market expectations (as captured by the federal funds futures price just prior to the FOMC meeting) account for U.S. policy that is either anticipated and/or endogenous to other macroeconomic shocks; (2) there is no relevant information arrival over the interval for which federal funds futures price changes and asset returns are measured, other than the FOMC announcement; and (3) term premia in the forward interest rate contracts are time-invariant. Faust et al. combine restrictions derived from the high frequency regressions with a priori bounds on monetary policy’s effects in order to partially identify impulse response functions in two bilateral open economy VARs, where Home is the U.S. and Foreign is either Germany or the U.K. (see Faust (1998) for a full description of the partial identification methodology). Under our identification scheme, we find that the within-month response of the US$/GRM exchange rate to a monetary policy shock that raises the U.S. t-bill rate by 100 b.p. is an appreciation of 1.70%. The response of the US$/UK£ rate is a 1.30% appreciation. Under Faust et al. (2003)’s identification scheme, the impact estimates are 2.44% and 1.41% respectively, indicating a strong agreement under the two identifications. The maximum exchange rate responses that we estimate occur after 2 months in the case of the US$/GRM exchange rate and 6 months in the case of the US$/UK£ rate. However, the uncertainty surrounding these estimates is considerable. Similar to Faust et al., delays of several years in the maximum exchange rate response cannot be ruled out. The dynamic responses of foreign interest rates are comparable across the two identifications, while the foreign output effects that we estimate are close to the maximum effects associated with the partial identification. We argue that the robustness of the results for Germany and the United Kingdom across the two monetary policy identifications is an important finding. It casts light on the validity of the different identifying assumptions which underlie the two approaches. In Section 4, we compare the U.S. monetary policy shocks derived from the federal funds futures price (whose effects on asset returns are a key input to Faust et al.’s partial identification) with those derived from the narrative evidence, the Greenbook forecasts, and the capacity utilization index. Although the policy shock magnitudes differ, there is a strong positive correlation between the two identified series. This suggests that the federal funds futures price and the identifying information we employ represent comparable instruments for eliminating anticipated and endogenous movements in U.S. monetary policy. In addition to validating many of the results in Faust et al. (2003), there are comparative advantages to our approach that we highlight. First, the U.S. monetary policy identification approach that we follow is readily implementable, since it only requires the construction of a single time series for policy shocks which may then be included in a standard reduced-form VAR model or another impulse response model (e.g., local projections). Moreover, estimation can be undertaken for a large sample of countries, whereas Faust et al.’s approach requires liquid markets in a wide range of financial instruments, restricting its application to models featuring either the U.S. and Germany or the U.S. and the U.K. Second, we do not impose any a priori bounds on what a reasonable impulse response should be. Under the assumptions of our identification approach, we can allow the estimated responses to be unrestricted. Consequently, any estimated contemporaneous or dynamic effects represent the actual correlations of the endogenous variables with our identified U.S. monetary policy shock measure. Another contribution of the paper is an exploration of the robustness of the U.S. domestic results presented in Romer and Romer (2004). They show that their identification scheme which combines narrative and Greenbook information, eliminates the price puzzle associated with many other identifications and increases the magnitude and speed of output’s response to monetary policy. In addition to extending the identification approach used by Romer and Romer, we apply a different method for aggregating the shocks from a meeting frequency to a monthly frequency that we argue is more appropriate, given that macroeconomic data are measured on a period average as opposed to an end-of-period basis. It is particularly relevant for the correct estimation of asset price responses. We also estimate models that control for a broader range of macroeconomic variables than do those used by Romer and Romer, and employ lag structures more typical of the literature. Our results indicate that the peak effect of policy on output occurs after 24 months, matching the Romers’ estimates, but that the size of the peak is smaller, reflecting both the inclusion of additional controls and differences in the shock aggregation method. As mentioned earlier, evidence of a price puzzle reappears in our results, and is protracted in some cases. The main reason appears to be the inclusion of only 6 monthly policy shock lags in the models that we estimate, compared to the 48 lags in the models estimated by Romer and Romer. The inclusion of additional lags in the models that we estimate eliminates any price puzzle and leads to large and statistically significant price reductions at the four year horizon. The paper is structured as follows. In Section 2, we sketch out the nature of the identification problem in an open economy context and explain how narrative evidence, the Greenbook forecasts and the capacity utilization index are used to isolate unanticipated and exogenous U.S. monetary policy shocks. In Section 3, we present our main empirical results for the exchange rate and foreign interest rates, foreign output and U.S. output and prices, and compare them to those in Faust et al. (2003). We also consider the robustness of our findings to excluding data associated with special events in the sample, eliminating units roots from the empirical models and simulating impulse response functions using the method of local projections. In order to probe the reasons for common features across our results and those of Faust et al., in Section 4, we compare federal funds future-identified shocks and our identified shocks. Section 5 concludes the paper with a summary of our main arguments and results, and a discussion of possible future research directions.

Monetary policy’s open economy consequences are difficult to estimate due to the simultaneity inherent in the relationships between asset prices, such as interest rates and exchange rates. Moreover, the influence of relevant omitted or unobserved variables (such as inflation expectations) may complicate the interpretation of any estimated results. In this paper, we have extended the U.S. monetary policy identification strategy pioneered by Romer and Romer (2004) to overcome such obstacles to the estimation of the open economy consequences of U.S. monetary policy. The approach leverages narrative evidence, the Federal Reserve’s real-time information (as captured by the Greenbook forecasts), and a proxy for latent policymaker perceptions of the economy’s cyclical position (the capacity utilization index), to achieve identification. If these variables are predetermined, and adequately summarize the Federal Reserve’s information set regarding its objectives and expectations, then the approach is able to recover the unanticipated and exogenous component of U.S. monetary policy. In particular, we require that open economy variables influence the Federal Reserve’s monetary policy decisions only through their effects on expectations of the Federal Reserve’s objectives (such as output growth, inflation, unemployment, and the output gap). We included the identified U.S. monetary policy shock as an exogenous variable in a set of 6 bilateral monthly VARs, where the foreign country is one of the non-U.S. G7 members. After a contractionary U.S. monetary policy shock, we found that the U.S.$ exchange rate always appreciates, achieving a maximum appreciation within 1–20 months. Specifically, the results indicated that a one percentage point rise in the U.S. t-bill rate (from the identified shock) induces a contemporaneous U.S.$ appreciation which exceeds one percentage point. The U.S.$ appreciation continues for some months, creating large UIP deviations over a 6 month horizon. There is strong, positive interest rate pass-through from the U.S. to the foreign countries, with the maximum lying between 0.46 and 1.24 and occurring within 3–10 months. Foreign output responses are uniformly negative at horizons of 16 months, possibly indicating that the expenditure-reducing effects of the U.S. policy contraction dominate any expenditure-switching effects. We also considered the robustness of Romer and Romer (2004)’s estimated effects of monetary policy on U.S. domestic variables. The response shapes were generally similar. However, the estimated response magnitudes were smaller, likely reflecting the greater number of conditioning variables in our estimated models. We also found that the U.S. price response was sensitive to the number of exogenous policy lags included in the model. Our results are consistent with those reported in recent work by Faust et al. (2003) for the responses of Germany and the U.K. They partially identify U.S. monetary policy’s effects via a combination of restrictions derived from high frequency regressions and a priori impulse response bounds. We compared our identified shocks with U.S. monetary policy shocks derived from the federal funds futures market, which is an important component of Faust et al.’s partial identification approach. There appears to be considerable overlap between the identifying information leveraged across the two identifications schemes, leading to the similarity of results. However, as evidenced by the larger set of countries we consider, our approach is more readily implementable. Moreover, it does not require that a priori bounds on the impulse responses be established, allowing the estimated responses to be unrestricted. Accordingly, applications of our identification approach should be relatively straight-forward, enabling the open economy consequences of U.S. monetary policy to be explored for a wide variety of countries.