Consensus Economics, the world's leading international economic survey organization, has a longer history than ESPF, and its surveys were analyzed in earlier studies concerning Japan (Chen et al., 2016). The participants of the ESPF panel are twice as many as for the Consensus Economics panel for Japan. Moreover, the ESPF not only has annual multidimensional data but also quarterly multidimensional data.
This paper analyzes the forecasts of gross domestic product (GDP) growth and inflation contained in ESPF for the period 2004－2016. Annual forecasts are for the current and the following year. Quarterly forecasts are for up to four quarters ahead. The forecasts comprise an unbalanced three-dimensional panel with multiple individual forecasters, target years, and forecast horizons (Davies & Lahiri, 1995). We evaluate the accuracy of the forecasts with average root mean squares of forecast errors (RMSE) using preliminary actuals, then we analyse systematic forecast errors using modified Holden-Peel (1990) regressions. We examine the revision process, and try to discriminate between forecast errors that arise from unforecastable macroeconomics shocks and forecasts errors that are idiosyncratic errors, as performed by Capistrán & López-Moctezuma (2014) for Mexican data. We also conducted efficiency tests for fixed-event forecasts using forecast revisions introduced by Nordhaus (1987). We expect residuals in the Nordhaus (1987) efficiency test to show the business cycle components in Japan. We hope to show whether quarterly GDP growth rate forecasts by the ESPF are accurate, and can be relied upon to guide business and policy decisions.
Reference
Capistrán, C. & López-Moctezuma, G. (2014). Forecast revisions of Mexican inflation and GDP growth. International Journal of Forecasting, 30, 177-191.
Davis, A. & Lahiri, K. (1995). A new framework for analyzing survey forecasts using three-dimensional panel data, Journal of Econometrics, 68, 205-227.