To Aggregate, Pool, or Neither

In November 2007, the Federal Open Market Committee (FOMC) announced a change in the way it communicates its view of the economic outlook: It increased the frequency of its forecasts from two to four times per year, and it increased the length of the forecasting horizon from two to three years. The FOMC does not release the individual members' forecasts or standard measures of consensus such as the mean or median. Rather, it continues to release the forecast information as a range of forecasts, both the full range between the high and the low and a central tendency that omits the extreme values. This paper uses individual forecaster data from the Survey of Professional Forecasters (SPF) to mimic the FOMC's method for creating their central tendency. The authors show that the midpoint of the central tendency of the SPF is a reliable measure of the consensus, suggesting that the FOMC reporting method is also a reliable measure of consensus. For the dates when both are available, the authors also compare the relative forecast accuracy of the FOMC and SPF consensus forecasts for output growth and inflation. Overall, the differences in forecast accuracy are too small to be statistically significant.

We estimate a simple reduced-form model of expectation formation with three distinct deviations from full-information rational expectations that combines underreaction with overreaction to new information. In particular, forecasts are sticky, extrapolate the most recent news about the current period, and depend on the lagged consensus forecast about the period being forecast. We find that all three biases are present in the Survey of Professional Forecasters as well as in the Livingston Survey, and that their magnitudes depend on the forecasting horizon. We also stress the point that using the past consensus forecast to form expectations is a reasonable thing to do if a forecaster is not able to come up with full-information rational expectations all by herself. Finally, we show that forecasters that have more sticky expectations generally tend to rely less on the lagged consensus forecast but extrapolate news about the current period more.

Forecast combination through dimension reduction techniques

ABSTRACTIn this paper, we propose a framework to evaluate the subjective density forecasts of macroeconomists using micro data from the euro area Survey of Professional Forecasters (SPF). A key aspect of our analysis is the use of evaluation measures which take account of the entire predictive densities, and not just the probability assigned to the outcome that occurs. Overall, we find considerable heterogeneity in the performance of the surveyed densities at the individual level. However, it is hard to exploit this heterogeneity and improve aggregate performance by trimming poorly performing forecasters in real time. Relative to a set of simple benchmarks, density performance is somewhat better for GDP growth than for inflation, although in the former case it diminishes substantially with the forecast horizon. In addition, we report evidence of an improvement in the relative performance of expert densities during the recent period of macroeconomic volatility. However, our analysis also reveals clear evidence of overconfidence or neglected risks in expert probability assessments, as reflected in frequent occurrences of events which are assigned a zero probability. Copyright © 2014 John Wiley & Sons, Ltd.

This paper examines asymmetry of the loss function of professional forecasters for output growth, inflation, and exchange rate forecasts, based on the survey of professional forecasts (SPF) data of three South Asian countries (India, Indonesia, and Singapore). Our results provide India’s unbiased output growth forecasts and under-predicted inflation and exchange rate forecasts; Indonesia’s broadly unbiased forecasts; and Singapore’s under-predicted output growth forecasts and unbiased inflation and exchange rate forecasts. Testing the rationality of all three countries’ SPF forecasts, we find that all are rational under an asymmetric loss function but not under a symmetric loss function.

Whereas strong demand is a key factor driving high inflation in the US, inflation in the Euro Area is mainly due to adverse external supply shocks (in Europe, energy prices are much higher due to the war in Ukraine). Standard monetary policy response to such shocks is to accommodate first-round effects, to fight spiralling inflationary expectations in order to prevent second-round effects. Long run inflation expectations — as measured by the survey of professional forecasters — still seem to be well anchored. The ECB’s announced tightening intends to dampen rising household’s inflation expectations. Given the current high uncertainty about the economic outlook, a soft landing calls for modest, data-dependent steps, allowing for a reversal in case the outlook worsens.

This paper investigates the accuracy and heterogeneity of output growth and inflation forecasts during the current and the four preceding NBER-dated US recessions. We generate forecasts from six different models of the US economy and compare them to professional forecasts from the Federal Reserve’s Greenbook and the Survey of Professional Forecasters (SPF). The model parameters and model forecasts are derived from historical data vintages so as to ensure comparability to historical forecasts by professionals. The mean model forecast comes surprisingly close to the mean SPF and Greenbook forecasts in terms of accuracy even though the models only make use of a small number of data series. Model forecasts compare particularly well to professional forecasts at a horizon of three to four quarters and during recoveries. The extent of forecast heterogeneity is similar for model and professional forecasts but varies substantially over time. Thus, forecast heterogeneity constitutes a potentially important source of economic fluctuations. While the particular reasons for diversity in professional forecasts are not observable, the diversity in model forecasts can be traced to different modeling assumptions, information sets and parameter estimates.

Meeting Individual Analyst Expectations

This paper studies the joint dynamics of U.S. inflation and a term structure of average inflation predictions taken from the Survey of Professional Forecasters (SPF). We estimate these joint dynamics by combining an unobserved components (UC) model of inflation and a sticky‐information forecast mechanism. The UC model decomposes inflation into trend and gap components, and innovations to trend and gap inflation are affected by stochastic volatility. A novelty of our model is to allow for time‐variation in inflation‐gap persistence as well as in the frequency of forecast updating under sticky information. The model is estimated with sequential Monte Carlo methods that include a particle learning filter and a Rao–Blackwellized particle smoother. Based on data from 1968 Q4 to 2018 Q3, estimates show that (i) longer horizon average SPF inflation predictions inform estimates of trend inflation; (ii) inflation gap persistence is countercyclical before the Volcker disinflation and acyclical afterwards; (iii) by 1990 sticky‐information inflation forecast updating is less frequent than it was earlier in the sample; and (iv) the drop in the frequency of the sticky‐information forecast updating occurs at the same time persistent shocks become less important for explaining movements in inflation. Our findings support the view that stickiness in survey forecasts is not invariant to the inflation process.

This paper investigates whether a specific type of a recurrent neural network, in particular Jordan neural network (JNN), captures the expected inflation better than commonly used feedforward neural networks and traditional parametric time-series models. It also considers competing survey-based and model-based expected inflation towards ex-post actual inflation to find whose predictions are more accurate; predictions from survey respondents or forecasting modelers. Further, it proposes neural network modelling strategy when dealing with nonstationary time-series which exhibit long-memory property and nonlinear dependence with respect to lagged inputs and exogenous inputs as well. Following this strategy, overfitting problem was reduced until no improvement in forecasting accuracy of expected inflation is achieved. The main finding is that JNN predicts inflation in euro zone quite accurately within forecasting horizon of 2 years. Regarding rational expectation principle we have found a set of demand-pull and cost-push inflation characteristics as exogenous inputs which helps in reducing overfitting problem of recurrent neural network even more. The sample includes euro zone aggregated monthly observations from January 2000 to December 2019. The results also confirm that inflation expectations obtained from JNN are consistent with Survey of professional forecasters (SPF), and thus, monetary policy makers can use JNN as a complementary tool in shortcomings of other inflation expectations measures.

Introduction In recent years the availability of relatively long panel data on expectations obtained from surveying professional forecasters has made it possible to test the validity of different expectations formations mechanisms in a scientific manner. The use of panel data enables the econometrician to decompose forecast errors into macroeconomic aggregate shocks for which forecasters should not be held responsible, and forecast specific idiosyncratic errors and biases for which they should be held accountable. Muth (1961) suggested that theories of expectation formation should be consistent with the economic model being considered, and defined expectations to be rational if they are equal to mathematical expectations conditional on the set of all information relevant for forecasting. Since the rational expectations hypothesis (REH) has wide-ranging implications for economic theory and policy, it is not surprising that several economists have used survey data on expectations in order to test REH directly. Using the American Statistical Association – National Bureau of Economic Research (ASA-NBER) survey of professional forecasters, Zarnowitz (1985) has performed tests of rationality of forecasts of inflation and other macroeconomic variables. He found that the inflation forecasts did not satisfy rationality; even though many of the other forecasts were rational. Keane and Runkle (1990) have argued that the inability of most researchers to accept REH is due to the use of incorrect methodology.

Inflation and Professional Forecast Dynamics: An Evaluation of Stickiness, Persistence, and Volatility

ABSTRACTAn accurate assessment of inflation expectations is crucial for the management of monetary policies. However, expectations are not directly observed and are hence normally inferred either from the interest rate structure or from surveys of professional forecasters. Alternatively, a direct measure may be obtained from consumer surveys. The aim of this paper is to study the formation of inflation expectations in Brazil, using a novel dataset based on the FGV/IBRE consumer survey. Basing our model on the rational inattention hypothesis, we find that individual heterogeneity plays a very significant role in shaping individual expectations; also, Brazilians adjust expectations to current inflation and to a fixed reference value, while professional forecasts do not play a very relevant role.

ABSTRACTExpectations are at the centre of modern macroeconomic theory and policymakers. In this article, we examine the predictive ability and the consistency properties of macroeconomic expectations using data of the European Central Bank (ECB) Survey of Professional Forecasters (SPF). In particular, we provide evidence on the properties of forecasts for three key macroeconomic variables: the inflation rate, the growth rate of real gross domestic product and the unemployment rate.

Economic forecasts, such as the Survey of Professional Forecasters (SPF), are revised multiple times before realization of the target. This paper studies the sources of forecast revisions. By decomposing the fixed‐event forecast error into a rational component due to unanticipated future shocks and an irrational component due to measurement error in acquired information or forecasters' reactions, we derive the conditions under which fixed‐event forecasts that contain an irrational forecast error can still possess the second moment properties of rational forecasts. We show that internal consistency of fixed‐event forecasts depends on the magnitude of the rational and irrational components in the revisions. As such, fixed‐event forecasts subject to irrational error may still be internally consistent, although they are not rational, as evidence in many empirical studies. We illustrate our methodology with the SPF inflation forecasts data. Our results show evidence of a sizeable and heterogeneous irrational forecast error component across forecast horizons and a high irrational‐to‐news ratio for most forecasters. This finding also provides insight into why forecast revision effort is not always fully compensated by revision reward.