Size-corrected inference in fiscal policy reaction functions: a three country assessment

Abstract

As a reflection of both strong persistence of debt-to-GDP ratios and correlation of respective innovations with governments’ primary surpluses, standard t-tests in policy reaction functions show actual significance levels that are up to five times larger than their nominal reference. Adopting size-controlled inference by means of Monte Carlo-based and asymptotic Bonferroni critical values, we diagnose fiscal policies in the US and the UK to be sustainable in samples covering more than 100 years. Conditioning on post-WWII subsamples and 5% nominal significance, conventional t-tests signal fiscal sustainability for these countries. In contrast, size-corrected inference hints at a lack of fiscal sustainability and, thus is recommended for the ‘real-time’ monitoring of public debt. The fiscal policy of Portugal is found to lack sustainability irrespective of the considered sample period.