Tax Burden, Tax Mix and Economic Growth in OECD Countries

Tax Burden, Tax Mix and Economic Growth in OECD Countries PAOLA PROFETA RICCARDO PUGLISI SIMONA SCABROSETTI June 30, 2015 FIRST DRAFT, PLEASE DO NOT QUOTE WITHOUT THE AUTHORS PERMISSION Abstract Focusing on the 34 OECD countries over the period 1970-2013, we study the relationship between the standard of living (GDP per person of working age) and fiscal variables that measure the overall tax revenue over GDP and the tax mix. We run regressions with country and time fixed effects on 5-years averages of the relevant variables, with very conservative estimates of the standard errors (twoway clustering). We find that tax revenue over GDP is negatively and significantly correlated with GDP per person of working age. Regarding the tax mix, we only find a positive and significant relationship between property taxes and GDP per person of working age. Differently from the most recent literature, we do not find any evidence of a positive correlation between a shift from direct to indirect taxation and standards of living. Keywords: Economic growth, taxation, tax mix, wealth taxation, democracy, robust methods. JEL classification: E62, H20, P50 We thank Paolo Longo for research assistance in data collection. Università Bocconi, Milan, Italy. email: paola.profeta@unibocconi.it (corresponding author) Department of Political and Social Sciences, University of Pavia, Pavia, Italy. email: riccardo.puglisi@unipv.it Department of Law, University of Pavia, Pavia, Italy. email: simona.scabrosetti@unipv.it 1

1 Introduction In a time of tight public finance conditions, because of the high level of sovereign debts, there is a large and expanding debate on the effects of taxation on the level and growth of GDP. Moreover, scholars and policy makers alike for a given level of fiscal pressure are interested in whether the tax mix, i.e. the distribution of overall tax revenue across different taxes, has a separate influence on the growth performance of a country. In this paper we investigate the relationship between tax structure and standard of living in OECD countries in the period from 1970 to 2013. Consistently with the theory according to which growth is a medium, long-term phenomenon we analyse a panel of 5-year averages of economic, fiscal and political data, instead of focusing on short-term dynamics. Moreover, to be robust and conservative in our estimates we run regressions with country and time fixed effects, and two-way-clustered standard errors (at the country and time level, see Cameron et al. 2011). Our results can be summarized as follows: we find that overall tax revenue is negatively and significantly associated with the level of GDP per person of working age. When we look at the tax mix, the only statistically significant findings are a positive relation between property taxes and GDP per worker, and a positive relation between corporate income taxes and GDP per worker, but this last only in the case of lagged explaining variables. On the other hand, we do not find any evidence of growth-enhancing effects of a shift from direct to indirect taxation (see below for further details). Wealth taxes, both on the ownership and the transfer of property, are attracting more and more attention by scholars and policy makers. Prominently, Piketty (2014) claims that the progressivity of wealth taxation should be enhanced in order to deal with the structural problem of inequality; in partial contrast to the Mirrlees review (Institute for Fiscal Studies 2011) Piketty and Saez (2012, 2013) suggest that in an optimal tax design there is room for both inheritance an capital income taxation. In fact, the underlying stylized fact regarding taxation on wealth transfers (tax on bequests and donations) is that there has been a steady decline relative to GDP for the last 3/4 decades, at least in G7 countries (Profeta et al. 2014). There is also a lively debate on the perceived need to lower the so called tax wedge on labour, since this could be beneficial to employment because of a stronger labor demand. Those who propose this tax decrease, first and foremost the European Commission itself (European Commission 2006), typically advise to finance it by increasing taxes on wealth. The closest contribution to ours is a recent study by Arnold et al. (2011). For a sample of 21 OECD countries over the 1971-2004 period they find that a shift from direct to indirect taxation is positively and significantly correlated with per capita GDP; they find a similar result for wealth taxes as well. 1 However we depart from their study in several directions, which we list here. First, instead of considering both the short-term dynamics of adjustment and the long-run steady-state equilibrium, we directly use 5-year averages of our variables and solely focus on the long-run equilibrium. On this time-collapsed dataset we simply run regressions with country and year fixed effects, which we think are more consistent with the theoretical literature on economic growth, i.e. a long-term phenomenon. 1 In their baseline specification consumption and property taxes are merged in the same variable, while in subsequent specifications they are separately included as regressors, and they are both found to be positively and significantly correlated with per capita GDP. 2

Second, in order not to inflate the precision of our estimates, we conservatively compute two-way clustered standard errors (i.e. at the country and time level simultaneously). 2 Third, we explore the heterogeneity across different areas, by focusing on the entire sample of 34 OECD countries and also on the sub-sample of those developed countries for which we have data for a longer period (from 1970 to 2013). 3 Notice also that, differently from Arnold et al. (2011), the time period that we are investigating also includes the years immediately before and during the subprime and sovereign debt crises of 2007-2012. Fourth, we include in our empirical analysis political and democratic variables, which may have a potential role both in affecting the tax structure and the economic performance of a country: to the best of our knowledge this is a relatively neglected topic in the literature. The paper is organized as follows: Section?? presents the data, while Section?? displays the results, with our robustness checks. Section?? concludes. 2 Data We collect economic, fiscal and political data on the 34 countries that are currently members of the OECD, for the period from 1970 to 2013. 4. Our dependent variable is GDP (expenditure approach)divided by population of working age, while our explanatory variables can be grouped as follows: variables which capture socio-economic and demographic factors which are likely to be correlated with GDP: level of investment over GDP, years of schooling (as a measure of human capital), growth rate of the population, openness to foreign trade (sum of exports and imports over GDP); variables which measure the overall tax revenue and its composition: the ratio of total tax revenue to GDP; the ratio to total tax revenue of several tax sources, i.e. direct taxes, direct taxes on individuals, direct taxes on corporations, property taxes, and social security contributions. We consider the overall amount of revenue, i.e. we sum the revenue at the central, regional and local level; variables which describe the political context: The Polity 2 index of democracy, the Freedom House indexes of civil liberties and political rights, and the Fraser index of economic freedom. Summary statistics of all variables are reported in Table 1. Table 1 near here 2 See Cameron et al. (2011) To put it in a different way, in the methodological trade-off between efficiency and robustness of econometric estimates we strongly prefer the latter (in the Angrist-Pischke 2009 sense). 3 See Profeta et al.2013. The full sample is an unbalanced panel since there is no comprehensive data for new OECD members, namely new EU countries. 4 The 34 OECD countries are: 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, United Kingdom and United States. 3

Regarding revenue as a percentage of GDP, the average is around 32 percent, with a minimum of 9 percent and a maximum of almost 50 percent. On average the largest tax item in our sample is social security contributions (around 33 percent), while individual taxes are on average 27 percent. Property taxes as a share of total tax revenue are the lowest, with an average of about 6 percent. The Polity 2 index of democracy displays a sizeable amount of variation, since it spans from a minimum of 9 to a maximum of 10, but the average is around 8, i.e. countries are mostly on the democratic side of the spectrum. 3 Empirical strategy and results We estimate the following equation: Y it = α i + β t + γ (economic vars) it + δ (fiscal vars) it + ζ (polity2) it + ɛ it (1) where α i and β t are respectively a country and a time fixed effect, (economic vars) it, and (fiscal vars) it are respectively vectors of economic and fiscal variables, while ɛ it is the error term. In order not to inflate the precision of our estimates, following Cameron et al. (2011) we conservatively compute two-way clustered standard errors. The reason for this choice is that even in the presence of country and time fixed effects-, the error terms might still be correlated within each country and across countries during the same five-year period. Our baseline results are shown in Table 2. Table 2 near here We find a negative and statistically significant relation between total tax revenue and standards of living. Moreover, there is a positive and statistically significant relation between property taxes and standard of living. On the other hand, there is no significant relationship between the other tax items and GDP per person of working age, and differently from the most recent literature-, we do not find any evidence of a positive relationship between a shift from direct to indirect taxation and standard of living (the so-called OECD view). Regarding our socio-economic variables, openness shows a strong and positive association with GDP per person of working age. Finally, political variables do not display any statistically significant relation with the standards of living. 5 The same results still hold when we estimate the equation (??) by using lagged explanatory variables (see Table 3). However, in this case we also find a positive and statistically significant relationship between GDP per person of working age and corporate income taxes. Table 3 near here 5 The tables only show the estimated coefficients for the Polity 2 index. However, also the coefficients referred to all the other political variables (the Freedom House indexes of civil liberties and political rights, and the Fraser index of economic freedom) are always not significant. 4

Given that the full sample is unbalanced (data on new EU countries generally start from 1995 onwards), to check the robustness of our baseline results, we replicate our analysis on the sub-sample of OECD countries for which we have data covering the entire 1970-2013 period. 6 Results are shown in Table 4 and 5 which respectively correspond to Table 2 and 3, i.e. with contemporaneous and lagged explanatory variables. Table 4 near here Generally speaking, we confirm the results displayed in Table 2 and 3. If anything, the positive and significant correlation between standard of living and country openness is now statistically significant at the 1 percent confidence level, and the same is true for the negative relation between standard of living and revenue over GDP. 4 Conclusions Our empirical analysis suggests that a lower fiscal pressure is related with higher standards of living (GDP per person of working age). However, we also find that the tax mix is not economically neutral, since a larger share of property taxes over total tax revenue is found to be positively and significantly correlated with standards of living. The main departure from the previous literature is that we find no evidence of positive correlation between a shift from direct to indirect taxation and standards of living (the so-called OECD view). From a policy perspective, if taken at faced value, our findings suggest that a growth enhancing fiscal package would be composed by a lower fiscal burden, with a stronger focus on wealth taxation. 6 The sub-sample includes the following countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Portugal, South Korea, Spain, Sweden, Switzerland, Turkey, United Kingdom and United States. 5

References [1] Angrist, J. D. and Pischke, J.-S. (2009) Mostly Harmless Econometrics: An Empiricist s Companion. Princeton, Princeton University Press. [2] Arnold, J. M., Brys, B., Heady, C., Johansson, A., Schwellnus, C. and Vartia, L. (2011). Tax policy for economic recovery and growth. The Economic Journal, 121 (February), F59-F80. [3] Cameron, A. C., Gelbachb, J. B. and Miller, D. L. (2011). Robust Inference With Multiway Clustering. Journal of Business Economic Statistics, 29:2, 238-249. [4] European Commission services (DG TAXUD) (2006). Macroeconomic effects of a shift from direct to indirect taxation: a simulation for 15 EU member states. Note presented at the 72nd meeting of the OECD Working Party No. 2 on Tax Policy Analysis and Tax Statistics, Paris, 14-16 November 2006. [5] Institute for Fiscal Studies. (2011). The Mirrlees review. Oxford: Oxford University Press. [6] Piketty, T. (2014). Capital in the Twenty-First Century. Harvard University Press. [7] Piketty, T. and Saez, E. (2012). A theory of optimal capital taxation. NBER Working Paper 17989. [8] Piketty, T. and Saez, E. (2013). A theory of optimal inheritance taxation. Econometrica, 81(5), 1851-1886. [9] Profeta, P., Puglisi, R. and Scabrosetti, S. (2013). Does democracy affect taxation and government spending? Evidence from developing countries. Journal of Comparative Economics, 41, 684-718. [10] Profeta, P., Scabrosetti, S. and Winer, S. L. (2014). Wealth transfer taxation: an empirical investigation. International Tax and Public Finance, 21, 720-767. 6

Table 1 Summary statistics of all variables Variable Observations Mean Std. Dev. Min Max gdp/working age pop 1288 35909.79 15083.85 4347.414 107281.1 years of schooling 1492 9.22492 2.057823 2.45 13.02 investment (% GDP) 1354 23.42935 4.22825 11.21038 37.56727 population growth 1487 0.75427 1.016893-2.5 26.4 openness 1354 71.93537 44.62308 9.1023 371.4397 revenue (% GDP) 1287 32.25975 8.395772 9.279 49.536 property taxes (% of tax revenue) 1286 5.708661 3.44467 0.49 21.417 individual taxes (% of tax revenue) 1209 27.27969 11.01428 2.625 61.606 corporate taxes (% of tax revenue) 1209 8.424878 4.568394 0.6 29.343 ss contributions (% of tax revenue) 1286 33.01225 9.930802 12.339 66.667 polity2 index 1364 8.274194 4.421605-9 10

Table 2 GDP, tax revenue and tax mix in OECD countries (1970-2013) VARIABLES (1) (2) (3) (4) gdp/working age pop gdp/working age pop gdp/working age pop gdp/working age pop years of schooling -781.7-343.5-373.7-364.7 [-0.944] [-0.403] [-0.492] [-0.488] investment (% GDP) -62.06-11.57 8.658-23.54 [-0.242] [-0.0492] [0.0454] [-0.106] population growth 2,248 1,640 989.2 867.2 [1.290] [1.161] [1.093] [1.022] openness 145.7** 144.6** 164.7** 166.1** [2.175] [1.995] [2.407] [2.370] revenue (% GDP) - -318.1* -353.6** -363.0** [-1.934] [-2.134] [-2.192] property taxes (% of tax revenue) - - 852.1* 907.7* [1.693] [1.666] individual taxes (% of tax revenue) - - -2.242 6.880 [-0.0136] [0.0409] corporate taxes (% of tax revenue) - - 306.7 303.5 [0.994] [0.960] ss contributions (% of tax revenue) - - -116.7-139.4 [-0.651] [-0.703] polity2 index - - - -72.38 [-0.341] observations 268 264 249 240 R-squared 0.943 0.947 0.956 0.957 n clusters 9 9 9 9 country fixed effects yes yes yes yes time fixed effects yes yes yes yes Note: 5-years averages of all variables; standard errors are clustered at the country and time levels; fiscal data refer to total government z-statistics in brackets *** Significant at 1% ** Significant at 5% * Significant at 10%

Table 3 GDP, tax revenue and tax mix in OECD countries (1970-2013) (1) (2) (3) (4) VARIABLES gdp/working age pop gdp/working age pop gdp/working age pop gdp/working age pop years of schooling, lagged -730.3-183.9-311.9-262.3 [-0.691] [-0.182] [-0.361] [-0.295] investment (% GDP), lagged -168.9-87.69-49.86-120.1 [-0.682] [-0.350] [-0.255] [-0.560] population growth, lagged 1,578 888.8 110.2 11.01 [1.034] [0.775] [0.200] [0.0223] openness, lagged 166.4*** 165.5** 183.9*** 185.5*** [2.587] [2.348] [3.084] [3.027] revenue (% GDP), lagged - -322.9** -384.8** -437.9*** [-2.192] [-2.216] [-2.617] property taxes (% of tax revenue), lagged - - 970.6* 1,117* [1.678] [1.927] individual taxes (% of tax revenue), lagged - - 26.45 29.66 [0.140] [0.150] corporate taxes (% of tax revenue), lagged - - 367.7* 384.8* [1.649] [1.727] ss contributions (% of tax revenue), lagged - - -18.02-10.85 [-0.0703] [-0.0414] polity2 index, lagged - - - -29.08 [-0.0932] Observations 238 232 219 211 R-squared 0.943 0.948 0.955 0.957 n clusters 9 9 9 9 country fixed effects yes yes yes yes time fixed effects yes yes yes yes Note: 5-years averages of all variables; standard errors are two-way clustered at the country and time levels; fiscal data refer to total government z-statistics in brackets *** Significant at 1% ** Significant at 5% * Significant at 10%

Table 4 GDP, tax revenue and tax mix in a selection of OECD countries (1970-2013) VARIABLES (1) (2) (3) (4) gdp/working age pop gdp/working age pop gdp/working age pop gdp/working age pop years of schooling -590.0 80.11-5.766-23.57 [-0.746] [0.106] [-0.00903] [-0.0366] investment (% GDP) -133.9-72.19-34.28-94.06 [-0.463] [-0.283] [-0.172] [-0.413] population growth 2,323 1,631 918.1 817.8 [1.282] [1.127] [1.135] [1.085] openness 175.7*** 176.4*** 196.2*** 198.9*** [3.199] [2.981] [3.739] [3.717] revenue (% GDP) - -475.2*** -512.1*** -543.2*** [-3.533] [-3.919] [-4.109] property taxes (% of tax revenue) - - 944.9** 1,016** [1.984] [1.986] individual taxes (% of tax revenue) - - 46.56 52.55 [0.276] [0.313] corporate taxes (% of tax revenue) - - 291.6 297.0 [0.857] [0.851] ss contributions (% of tax revenue) - - -67.26-75.93 [-0.349] [-0.370] polity2 index - - - 27.58 [0.125] Observations 230 228 218 209 R-squared 0.940 0.948 0.960 0.961 n clusters 9 9 9 9 country fixed effects yes yes yes yes time fixed effects yes yes yes yes Note: 5-years averages of all variables; standard errors are clustered at the country and time levels; OECD countries with no data before 1995 are excluded; fiscal data refer to total government z-statistics in brackets *** Significant at 1% ** Significant at 5% * Significant at 10%

Table 5 GDP, tax revenue and tax mix in a selection of OECD countries (1970-2013) (1) (2) (3) (4) VARIABLES gdp/working age pop gdp/working age pop gdp/working age pop gdp/working age pop years of schooling, lagged -482.6 249.4 109.1 151.6 [-0.450] [0.248] [0.125] [0.170] investment (% GDP), lagged -166.5-96.98-54.03-138.1 [-0.549] [-0.341] [-0.232] [-0.566] population growth, lagged 1,685 921.0 68.73-39.38 [0.988] [0.736] [0.135] [-0.0855] openness, lagged 188.4*** 190.7*** 211.6*** 214.3*** [3.401] [3.427] [5.638] [5.528] revenue (% GDP), lagged - -441.3*** -509.1*** -573.1*** [-3.306] [-3.260] [-3.812] property taxes (% of tax revenue), lagged - - 1,014* 1,166** [1.759] [2.049] individual taxes (% of tax revenue), lagged - - 97.64 98.26 [0.512] [0.495] corporate taxes (% of tax revenue), lagged - - 392.2 416.2* [1.611] [1.706] ss contributions (% of tax revenue), lagged - - 35.06 49.64 [0.124] [0.175] polity2 index, lagged - - - 6.215 [0.0189] Observations 206 204 195 187 R-squared 0.937 0.946 0.956 0.958 n clusters 9 9 9 9 country fixed effects yes yes yes yes time fixed effects yes yes yes yes Note: 5-years averages of all variables; standard errors are clustered at the country and time levels; OECD countries with no data before 1995 are excluded; fiscal data refer to total government z-statistics in brackets *** Significant at 1% ** Significant at 5% * Significant at 10%