Evaluating the macroeconomic forecasting performance of the Ministry for Finance and Employment

Abstract

This working paper evaluates the macroeconomic forecasting performance of the MFE for forecasts produced between 2004 and 2022 in the Update of the Stability/Convergence Programmes (spring vintage) and in the Draft Budgetary Plans (autumn vintage). The authors’ assessment of this evaluation rests on three pillars: forecasting accuracy, unbiasedness, and forecasting efficiency. Inter-institutional forecasting performance evaluation is also undertaken by benchmarking the results obtained for MFE with those carried out for the COM and CBM forecasts.

The authors find that the MFE underestimates real and nominal GDP growth and expenditure components in the spring and autumn forecasting vintages. This is also the case when evaluating the forecasting accuracy of the COM and the CBM forecasts. When considering forecasting biasedness, the authors find evidence of downward biases in real and nominal GDP forecasts of the MFE, especially when excluding crises years. Indeed, some bias was evident in all expenditure components save for gross fixed capital formation. Evidence of forecast biasedness was also present for the COM and the CBM.

The authors did not find evidence of strong inefficiencies in MFE macroeconomic forecasts, indicating that the information available when the forecasts were produced was adequately factored in. Nonetheless, when assessing the relationship of past data outturns with forecast errors, the authors identified elements of forecasting inefficiencies in private and public consumption expenditure in year t.

An important finding from this study is the effect of statistical revisions on the results presented in this paper. While the authors did not estimate the causal effect of statistical data revisions on forecasting error performance, it is noted that statistical data tends to be revised upwards. Indeed, the authors find an upward systematic bias in most of the components analysed, which leads to a downward bias in forecast errors.