Worse than you think: Public debt forecast errors in advanced and developing economies

Abstract

We compile a unique dataset of medium-term public debt forecasts for an unbalanced panel of 174 countries, based on International Monetary Fund (IMF) (for the period 1995-2020) and Economist Intelligence Unit (2007-2020) projections. We find that, on average, (i) there is a positive forecast error (FE) in the debt-to-gross domestic product (GDP) projections-that is, realized debt ratios are larger than forecasts; (ii) the FE increases with the projection horizon and is statistically significant and large-about 10% of GDP at the 5-year horizon; (iii) the magnitude is similar between advanced economies (AEs) and emerging markets and developing economies (EMDEs) and in EMDEs is present irrespective of recessions while for AEs is associated with surprise recessions in the forecast horizon; (iv) FEs are not statistically different between IMF program and non-program cases; and (v) positive FEs are only partly attributable to optimism about growth or the fiscal balance. Looking at the correlates of FEs, we find that FEs are larger during periods of recession, elections, fiscal stress, and high uncertainty and in countries with more economic volatility and public debt.