موقعیت جغرافیایی، مهارت ها و یا هر دو: چه چیزی که دقت پیش بینی ناظران بانک فدرال از سیاست پولی آمریکا را توضیح می دهد؟

The paper shows that there is a substantial degree of heterogeneity in the ability of Fed watchers to forecast US monetary policy decisions. Based on a novel database for 268 individual professional forecasters since 1999, the average absolute forecast error of FOMC decisions varies 5–10 basis points between the best and worst-performers across the sample. This heterogeneity is found to be related to both the skills of analysts – such as their educational and employment backgrounds – and to geography. In particular, forecasters located in regions which experience more idiosyncratic economic conditions perform worse in anticipating monetary policy. This evidence is indicative that limited attention and heterogeneous priors are present even for anticipating important events such as monetary policy decisions.

Over the last two decades, a major evolution has taken place in the world of central banking towards a transparent conduct of monetary policy and away from a monetary policy that had often largely surprised the public. In light of this development, central banks have repeatedly stressed the importance of predictability of their decisions, which has indeed improved remarkably over time (e.g., Poole et al., 2002 and Lange et al., 2003). While much of the empirical work has focused on predictability based on the financial market consensus (Kuttner, 2001; Hamilton, 2009; Gürkaynak et al., 2007), others have documented the role of disagreement and heterogeneity among agents in forecasting monetary policy (Bauer et al., 2006 and Swanson, 2006). The latter papers suggest that the increasingly transparent monetary policy of the Federal Reserve has not only led to a better prediction by financial markets in general, but is also reflected in more synchronized forecasts of monetary policy decision. Nonetheless, these papers also document that a remarkable degree of disagreement among forecasters about future central bank actions seems to persist. The literature suggests a number of factors that may generate disagreement among economic forecasters. Hong and Stein (2007), surveying the discussion on financial market forecasting, stress asymmetries in information availability and information processing. An example relating to information availability is the notion of “gradual information flow”, where the arrival of information is staggered across agents. Examples relating to information processing include limited attention (where agents neglect or overweight information because of limits in their information processing capabilities) and heterogeneous prior beliefs (where agents receive the same information, yet interpret it differently).1 Differences in information processing seem of particular importance in the area of monetary policy forecasts, where, as a rule, the relevant macroeconomic information governing central bank decisions, as well as all relevant central bank communication is available to all forecasters.2 The present paper focuses on the heterogeneity among forecasters of monetary policy decisions by the U.S. Federal Reserve, and its determinants. The paper is motivated by the fact, as we will show, that such heterogeneity is surprisingly high, even at a very short forecasting horizon. To understand this heterogeneity, we concentrate on forecaster asymmetries in information processing related to skills and geography. Given an abundance of potentially relevant data, a major challenge in forecasting monetary policy decisions is to make an appropriate selection of information and apply proper weights. Limited attention as well as heterogeneous priors can therefore easily generate disagreement. Both mechanisms suggest that skills have an important role to play, as better skilled forecasters devote the appropriate attention to the relevant signals, or have priors which more closely reflect the actual FOMC behavior. Another important implication is that geographical location matters. This is because local information is salient, which, in the presence of limited attention might bias information processing and distract forecasters’ attention from other signals. In addition, geographical location could influence priors, for instance, because the salience of local information shapes the analytical framework of forecasters or analysts with certain given skill sets cluster in particular localities. Our paper is also related to the literature on information and geography, 3 including Berger et al. (2009), who find evidence that geographic factors play a role for the predictability of ECB monetary policy decisions. The present paper is substantially wider in scope. 4 The paper uses a novel dataset of 268 professional forecasters – covering many major investment banks, commercial banks and forecasting institutions – who are located across 98 cities in 15 countries, for FOMC decisions between February 1999 and September 2005. The dataset is very rich, containing not only each forecaster’s survey expectations for FOMC decisions, but also information about the individual’s forecasts of the macroeconomic releases for other variables, such as inflation and economic activity. Moreover, the data includes information related to analysts’ skills, e.g. the type of institution, his or her position within that institution, employment record and educational background. We combine this dataset with information about the economic conditions specific to the region in which each individual is located. As a key stylized fact, the degree of heterogeneity in the forecast performance across individuals is large: after grouping forecasters by performance over the full sample period, the absolute forecast error by the group of the 10% of the worst forecasters is 5 basis points (b.p.) higher than that of the best decile of analysts, when measured across all FOMC meetings. This difference rises to 10 b.p. when analyzing only those FOMC meetings that had some degree of heterogeneity across forecasters. This is of the same order of magnitude we have found for the heterogeneity of forecasts of ECB monetary policy decisions (Berger et al., 2009) and given the frequency of forecasters’ participation cannot be the result of pure chance. This level of heterogeneity is non-negligible from a financial market point of view, as it suggests that forecasters could be wrong by 25 basis points every fifth FOMC meeting or nearly twice a year. Such a performance can lead to sizable financial losses, especially if we assume that the forecasters’ institutions have taken corresponding positions in financial markets (after all, the mere fact that financial institutions typically devote substantial resources to their Fed-watching activities suggests that there are possibly large returns to be gained from accurate forecasts of monetary policy). Interestingly, the observed differences in forecasting ability are mirrored in financial market data—a finding that should be of relevance to monetary policy-makers. We show that the larger the observed heterogeneity of monetary policy expectations, the higher is financial market volatility. This suggests that a thorough analysis of the impact of monetary decisions on the formation of interest rate expectations and, ultimately, financial markets will deepen the Federal Reserve’s understanding of monetary policy transmission through the interest rate channel, including possible divergences in policy effectiveness across regions. Moreover, reducing market volatility will help improve the signaling quality of market prices and, thus, the effectiveness efficient allocation of capital. Next, we find that a significant part of the heterogeneity in forecasting accuracy is systematic. That is, there is compelling empirical evidence that skills and geography play a significant and substantial role. As to geography, we find that a number of locational factors systematically influence the ability of forecasters to anticipate US monetary policy. For instance, forecasters located in New York City or in other financial centers, either in the USA or abroad, as well as those located in Washington DC, i.e. in immediate proximity of the Board of Governors of the Federal Reserve, perform better on average. Moreover, we find that forecasters take a local perspective in the sense that regional economic developments shape their forecasting ability for US monetary policy. We take this as evidence that salience of information is an important factor in the forecasting process. As to skills, there are a number of factors that affect forecasters’ performance. For instance, analysts who work for investment banks do better than those in other financial and non-financial companies. Second, it is intriguing that analysts who have the position of Economist in their institution do better than forecasters with higher-ranking titles, in particular executives. Our interpretation is that executives are less specialized and can devote less time and resources to following the Fed. The results seem to support the limited attention hypothesis. Third, professional experience and education matter for forecast accuracy. We find that analysts who previously worked for the Federal Reserve’s Board of Governors perform better, as do analysts with a Master’s degree. A related result of the empirical analysis is that forecasters who do well in predicting monetary policy also do well in anticipating other economic variables. The paper is structured as follows. Section 2 discusses in detail the data for the monetary policy forecasts. Section 3 starts by outlining our hypotheses before presenting the empirical results. Section 4 concludes.

The monetary policy of the Federal Reserve has become increasingly predictable over time, also given its remarkable progress towards more transparency. This process has not only led to fewer monetary policy actions that have surprised the public, it has also synchronized the views of individual Fed watchers. However, disagreement among market participants remains, and is still sizable. Based on a novel dataset of 268 individual professional forecasters located across 98 cities in 15 countries, we found that the degree of heterogeneity in the forecast performance across individuals is large: the average absolute forecasts error by the group of the 10% of the worst forecasters is 5 b.p. higher than that of the best decile of analysts (10 b.p. if we focus on FOMC meetings where not all forecasters agreed). This heterogeneity has repercussions for trading behavior, by significantly increasing financial market volatility, as well as influencing the monetary policy transmission process and the market allocation of funds. Both geography and skill influence the heterogeneity of forecasts. The paper has found that monetary policy expectations exhibit a significant and systematic regional pattern in the United States, in that regional economic developments shape their forecasting ability about monetary policy. In particular, forecasters make larger errors the more economic developments in their home region differ from their average. As to the role of skills, the result show that forecasters that are good in forecasting inflation also perform well in predicting monetary policy decisions. Moreover, analysts who work for investment banks or specialized forecast institutions, have a graduate degree or have an employment history with the Federal Reserve’s Board of Governors all conduct better forecasts. Among other things, the performance of investment bankers and other forecasting specialists is likely to reflect strong incentives to forecast accurately, as well as resource advantages. What do these findings imply for policy? First of all, it should be stressed that not all heterogeneity in expectations is necessarily undesirable from a policy perspective, in particular if such differences are the result of different degrees of investment in information gathering by analysts’ institutions. Moreover, differential expectations about policy decisions may at times also provide useful information to policy-makers. Therefore, the primary nature of the analysis of the paper is a positive one, i.e. to document the magnitude and understand the determinants of the heterogeneity in monetary policy expectations. At the same time, some of the analysis has also normative implications, though these can be no more than tentative and suggestive. Clearly, it is desirable for central banks to disseminate information and knowledge as equally as possible across agents not least because a high degree of heterogeneity is likely to result in unwanted financial market uncertainty and volatility. In particular the fact that such heterogeneity is linked to regional factors, which significantly influence forecasters’ expectations, raises many issues for policy-makers, such as the choice of communication tools and strategies to enhance a more homogenous understanding of monetary policy. This may be particularly relevant in periods like the present, where some economic regions – and in particular financial centers – are facing a more uncertain and volatile environment than others.