Tax policy and economic growth: does it really matter?

Obviously shifting taxes away from labor could negatively affect redistribution. In this paper, we do not take into account redistributive issues. Thus, we leave aside the large literature on optimal taxation which studies the trade-off between efficiency and equity in relation to taxation and the tax mix.

As well argued by Mathé et al. (2015) endogenous growth models are the best theoretical baseline to study the connections between tax policy and economic growth. The predictions of endogenous growth models are empirically investigated—among others—by Kneller et al. (1999) and Gemmell et al. (2014). More recently, Jaimovich and Rebelo (2017) show that the link between taxation and economic growth can be nonlinear, mainly because of different incentives to invest that entrepreneurs face in low and high tax contexts.

Similar results are obtained by Arachi et al. (2015) in a sample of 15 OECD countries over the period 1965–2011.

See also the results in Sanzo et al. (2017) on a sample of 20 OECD countries and the review by Shinohara (2014).

On EU countries see, e.g., European Commission Services (European Commission Services 2006) and Bernardi (2013).

When clustering the standard errors the only mildly significant finding is the negative correlation between a revenue-neutral shift from indirect to personal income taxes and per capita GDP.

Data are available for all years and countries from 1995 to 2014 for the following 34 OECD countries: Australia, Austria, Belgium, Canada, Chile, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Slovakia, Slovenia, South Korea, Spain, Sweden, Switzerland, Turkey, UK and USA. For the period 1971–2014, data are available only for 23 countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany (Western Germany, up to 1989 only), Greece, Ireland, Italy, Japan, South Korea, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, UK and USA.

Corporate income taxes refer to category 1200 (taxes on income, profits and capital gains of corporates).

On this point, see, for instance, Arnold et al. (2011).

Estimates of steady-state coefficients as well as of the parameters of the production function are computed as the ratio between a and \(\phi \).

Results are obtained by means of the very recent Stata routine xtdcee2, proposed by Ditzen (2016), which estimates a heterogeneous coefficient model in a dynamic panel with dependence between cross-sectional units. This routine is the evolution of the xtpmg Stata command (Blackburne and Frank 2007).

We thank an anonymous reviewer for pointing this out.

These countries are: Austria, Belgium, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Slovakia, Slovenia and Spain.

A potential concern with clustered standard errors is the relatively low number of clusters (countries) in our sample. However, as discussed by Cameron et al. (2008), this would imply that the cluster option does compute standard errors that are smaller than the correct ones, thus inducing over-rejection of the null hypothesis. But this bias goes in our direction, since our results with clustered standard errors by and large do not reject the null hypothesis of zero correlation between economic growth and tax variables.

To understand whether the financial crisis played any role in guiding these results, we rerun the same regressions for the pre-crisis time period, i.e., from 1971 to 2007. Results are similar and show no significant relationship between tax shift and economic growth.

Again, we rerun the same regressions for the pre-crisis time period, i.e., from 1995 to 2007. Results are similar to those obtained on the whole time period, as they show no significant relationship between tax shift and economic growth.

When separating out consumption and property taxes, only a shift from income to property taxation is positively and significantly correlated with GDP.

One must notice that—when clustering the standard errors—the coefficient on the accumulation of physical capital in the long run is no longer statistically significant.

The human capital variable enters with a positive sign in the regressions but is not statistically significant at ordinary confidence levels. This result is similar to Mankiw et al. (1992) and Islam (1995).

One must notice that—when clustering the standard errors—the coefficient on the accumulation of physical capital is statistically significant only in two specifications out of five, while the coefficient on the population growth rate is no longer statistically significant.

Results are available upon request.