No abstract

Using real-time data that reflects information available to monetary authorities at the time they are formulating policy, we find that estimated Taylor rules based on revised and real-time data differ more for Germany than for the U.S., Taylor rules using real-time data suggest differences between U.S. and German monetary policies, and Taylor rules for the U.S. using inflation forecasts are nearly identical to those using lagged inflation rates. Evidence of out-of-sample predictability for the dollar/mark nominal exchange rate with forecasts based on Taylor rule fundamentals is only found with real-time data and does not increase if inflation forecasts are used.JEL classification: C2, E5, F3

We examine rationality, forecasting accuracy, and economic value of the survey-based exchange rate forecasts for 10 developed and 23 developing countries at the 3-, 12-, and 24-month horizons. Using the data from two surveys for the period from 2004 to 2012, we find strong evidence that the forecasts for developing countries are biased at all forecast horizons. For developed countries, forecasts are strongly biased at the 3-month horizon, the bias decreases at the 12-month horizon, and increases again at the 24-month horizon. Based on the magnitude of the forecast errors and the direction of change, long-term forecasts are more accurate than short-term forecasts. Economic evaluation of the forecasts indicates that the forecasters are successful at generating positive economic profits, and economic gains of the forecasts for developed countries improve with the forecast horizon.

No abstract

This paper evaluates out-of-sample exchange rate predictability of Taylor rule models, where the central bank sets the interest rate in response to inflation and either the output or the unemployment gap, for the euro/dollar exchange rate with real-time data before, during, and after the financial crisis of [2008][2009]. While all Taylor rule specifications outperform the random walk with forecasts ending between 2007:Q1 and 2008:Q2, only the specification with both estimated coefficients and the unemployment gap consistently outperforms the random walk from 2007:Q1 through 2012:Q1. Several Taylor rule models that are augmented with credit spreads or financial condition indexes outperform the original Taylor rule models. The performance of the Taylor rule models is superior to the interest rate differentials, monetary, and purchasing power parity models. IntroductionThe past few years have seen a resurgence of academic interest in out-of-sample exchange rate predictability. Gourinchas and Rey (2007), using an external balance model, Engel, Mark, and West (2008), using monetary, Purchasing Power Parity (PPP), and Taylor rule models, and Molodtsova and Papell (2009), using a variety of Taylor rule models, all report successful results for their models vis-à-vis the random walk null. There has even been the first revisionist response. Rogoff and Stavrakeva (2008) criticize the three above-mentioned papers for their reliance on the Clark and West (2006) statistic, arguing that it is not a minimum mean squared forecast error statistic.An important problem with these papers is that none of them use real-time data that was available to market participants. 1 Unless real-time data is used, the "forecasts" incorporate information that was not available to market participants, and the results cannot be interpreted as successful out-of-sample forecasting. Faust, Rogers, and Wright (2003) initiated research on out-of-sample exchange rate forecasting with real-time data. Molodtsova, Nikolsko-Rzhevskyy, and Papell (2008) use real-time data to estimate Taylor rules for Germany and the U.S. and forecast the deutsche mark/dollar exchange rate outof-sample for 1989:Q1 -1998:Q4. Molodtsova, Nikolsko-Rzhevskyy, and Papell (2011), henceforth MNP (2011), use real-time data to show that inflation and either the output gap or unemployment, variables which normally enter central banks' Taylor rules, can provide evidence of out-of-sample predictability for the U.S. Dollar/Euro exchange rate from 1999 to 2007. Adrian, Etula, and Shin (2011) show that the growth of U.S. dollar-denominated banking sector liabilities forecasts appreciations of the U.S. dollar from 1997 to 2007, but their results break down in 2008 and 2009. Molodtsova and Papell (2009) conduct out-of-sample exchange rate forecasting with Taylor rule fundamentals, using the variables, including inflation rates and output gaps, which normally comprise Taylor rules. Engel, Mark, and West (2008) propose an alternative methodology for Taylor rule out-ofsample exchange rate forecas...