Fiscal Policy and TFP in the OECD: A Non-Stationary Panel Approach

We analyse the in uence of fiscal policy on TFP and per capita output in a panel of OECD countries since 1975. We focus on the effects of government size, government deficits and the composition of taxes and expenditures. Compared to existing studies, our contribu- tion is double. First, we are able to identify both direct and indirect effects of fiscal policy on TFP. The latter stem from the in uence of taxes and expenditures on countries' access to and efficient use of the world stock of technology and knowledge. A second contribution is methodological. The role of the worldwide level of technology introduces a common factor (and therefore cross-sectional dependence) in individual countries' TFP. This common fac- tor is unobserved and most likely non-stationary. The existing empirical literature on fiscal policy and growth largely neglects the econometric complications that may arise from cross- sectionally correlated error terms due to unobserved (and potentially non-stationary) common factors. This leads to inconsistent estimates if the unobserved factors are correlated with the explanatory variables and to a spurious regression problem if they are non-stationary. We appropriately deal with these econometric issues by using the Common Correlated Effects Pooled estimator of Pesaran (2006) and Kapetanios et al. (2006). Our main findings are as follows. Through the direct channel, an overall increase in government size reduces TFP and per capita output. Expenditure shifts in favour of productive purposes have strong and ro- bust positive effects on TFP. Shifts in favour of social transfers reduce TFP. Deficit reduction policies raise TFP if they are financed by expenditure cuts. Through the indirect channel, a rise in the corporate tax rate negatively affects a country's access to the worldwide level of technology whereas education expenditures and human capital formation promote this access.fiscal policy, total factor productivity, long-run output level, unobserved common factors, panel data