A new framework for analyzing survey forecasts using three-dimensional panel data

An evaluation of the forecasts of the federal reserve: a pooled approach

“…We adapt these ideas to our set up, where we have a large number of targets but relatively few forecasts of each, and also calculate pooled versions of these tests. We show that if we assume the forecast-error decomposition of Davies and Lahiri (1995), the test based on the lack of correlation on adjacent forecast revisions of Nordhaus is invalid. Instead, we present versions which are valid under the Davies-Lahiri setup.…”

“…To this end, we will adopt the decomposition of forecast errors of Davies and Lahiri (1995), except that their forecasts vary across individuals as well as targets and horizons.…”

“…Note that testing based on (8) is not equivalent to the standard approach of testing the forecast errors for each horizon using separate regressions. Davies and Lahiri (1995) adopt a common bias over horizons but allow individual-specific biases. We begin with horizon-specific biases, and test the assumption of a common bias over horizons.…”

Internal Consistency of Survey Respondents' Forecasts: Evidence Based on the Survey of Professional Forecasters*

“…In the economics sphere point forecasts of the conditional mean or some other measure of the central tendency have been the main focus (see, e.g., Mincer and Zarnowitz (1969), Figlewski andWachtel (1981), Zarnowitz (1985), Keane and Runkle (1990), Davies and Lahiri (1995), Stekler (2002) and Fildes and Stekler (2002)),…”

Testing for unit roots in three-dimensional heterogeneous panels in the presence of cross-sectional dependence

“…Empirical examples of three dimensional panels are still relatively rare, but are starting to appear in the literature, see for example, Goldberg and Verboven (2005), who assembled a large three dimensional panel (using make of car, countries and time) to investigate market integration and convergence in the European car market, and Davies and Lahiri (1995) who analyse forecasts of in ‡ation available for di¤erent individuals, across varying time horizons over the period from 1977 through to 1992. Within the context of a three-dimensional (3D) panel, N might denote countries or industries and M might be regions or …rms within that country or industry.…”