Economic growth and tax components: an analysis of tax changes in OECD

Abstract

The paper examines empirically the changes in the tax mix of the OECD countries in response to economic growth from 1980 to 1999. It is found that economic growth, measured by GDP per capita, has had a significant effect on the tax mix of the OECD countries. Analysis reveals that different taxes respond differently to the growth of GDP per capita. It is shown that while the shares of personal and property taxes have responded positively to economic growth, shares of the payroll and goods and services taxes have shown a relative decline.

Notes

¹See, for example, Martin and Lewis (1956), Shin (1969), Bahl (1971), and Chelliah (1975).

²See, for example, Engen and Skinner (1999).

³See Abizadeh and Gray (1985) for more on Wagner's Law.

⁴In light of these findings, we will not venture into the collective choice aspect of the tax components.

⁵See, for example Sato (1967).

⁶See, for example, King and Rebelo (1990) and Stokey and Rebelo (1995).

⁷We are conscious of the distinction made by Engen and Skinner (1999, p. 308) regarding the changes in the level of GDP as opposed to the changes in the rate of growth of GDP. For this reason, we are very clear and specifically use the rate of growth of real per capita income as a proxy for economic growth in our empirical models.

⁸This amounts to the detection of causality between any fiscal policy measure (tax or expenditures) and economic growth. Accordingly we will test for causality prior to testing our main hypotheses, to ensure robustness in our empirical analysis. See, the section on empirical specification below.

⁹Note that defining tax ratio in terms of GDP, although possible, may result in some econometric problems since this will mean that GDP will appear on both sides of the equation.

¹⁰This category also includes estate, inheritance and gift taxes and taxes on financial and capital transactions.

¹¹This category includes value added taxes, general sales and excise taxes and various license fees and use taxes that are imposed at the federal, state and local levels.

¹²These include customs and import duties and export taxes.

¹³An alternative is to use a seemingly unrelated regression (SUR) model instead of running eight different regressions. This would be relevant when separate regressions are related because the errors associated with the dependent variables may be correlated. However, when same set of explanatory variables is used, SUR gives the same results in terms of coefficients and standard errors as separate regressions.

¹⁴Natural logarithms of tax shares are not taken due to zero shares for some taxes in some countries.

¹⁵

¹⁶See Tosun (2005) for an analysis of the changes in tax structures in response to trade liberalization.

¹⁷This dummy variable accounts for the new entrants to the EU between 1980 and 1999. These countries are Austria, Finland, Greece, Portugal, Spain and Sweden.

¹⁸All the years before 1987 take the value 1 so that the natural logarithms of the values are equal to 0.

¹⁹See Hausman (1978) for the original description of this test.

²⁰See Greene (2003, pp. 287–303) for more on this issue.

²¹See, for example, Joulfaian and Mookerjee (1990 and 1991) and Jones and Joulfaian (1991).

²²Note that there is no universal agreement as to whether lag lengths should be chosen to minimize FPE. Jones (1989), for example, demonstrates that, under certain conditions, lag lengths based on some arbitrary and ad hoc method may be preferred.

²³See Armah (1997) for detail calculation of FPE and determining the direction of causality using this method.

²⁴We expect this trend to continue in the foreseeable future as baby boomers, who hold a very high proportion of national wealth, begin to reach their retirement age.