Taxation and political business cycles in EU economies

Abstract

This paper examines whether incumbent national governments of 11 member states of the European Union manipulated the tax policy instruments at their disposal in order to create national political business cycles, opportunistic or partisan. The empirical evidence, based on data concerning the 1965 to 1997 period, does not support this hypothesis. Rather, it appears that governments have pursued stabilization policies.

Notes

¹ From the fifteen member states of the EU, our analysis excludes Luxembourg due to data unavailability. Austria, Finland and Sweden are not examined either due to the length of our data set (1965–1997) and their so-to-speak late accession (1995) in the EU. A longer time period was not available due to the lack of statistical data concerning various categories of the variables employed in the countries of the sample. For the same reasons, the ten new members of the EU as of spring 2003 are also not examined.

² Theoretically, λ can take any value from zero to infinity. When λ is equal to infinity (zero) the solution to the constrained minimization problem is a linear trend (the original series). Recently, Ravn and Uhlig (1997) strongly recommended a new HP filter adjustment rule, according to which any value of λ between 6.25 and 8.25 is a reasonable choice. Their finding is in agreement with the Baxter and King (1995) proposition that the HP filter approximates the ideal band-pass filter when λ = 10.

³ To avoid overburdening the analysis with symbols, a country-specific subscript is omitted from all variables in (2).

⁴ Variables TYP and TYC include taxes levied on income of individuals and profits (gross income minus allowable tax relieves) on corporations, taxes levied on capital gains and gains from gambling. Variable TP covers recurrent and non-recurrent taxes on the use, ownership or transfer of property. TGS stands for all taxes, other than import and export duties, levied on the production, leasing, transfer, delivery or sales of a wide range of goods and/or the rendering of a wide range of services, irrespective of whether they are domestically produced or imported and irrespective of the stage of production or distribution at which they are levied, i.e. value-added taxes, sales taxes and multi-stage cumulative taxes. TSS includes excise taxes, taxes on profits generated and transferred from fiscal monopolies, customs and import duties as well as taxes on exports, foreign exchange transactions, investment goods and betting stakes and special taxes on services, which do not form part of a general tax. TOS is a residual tax item on receipts from goods and services. Finally, TD = TYP + TYC + TP and TI = TGS + TSS + TOS. All variables in levels are expressed at national currencies.

⁵ For the construction of the dummy variables see also Alogoskoufis et al . (1992).

⁶ Variable d_D assumes the value one from 1967 (April) to 1974 (July) in the case of Greece, and for the years before 1975 (April) and 1977 (July) for Portugal and Spain, respectively.

⁷ In fact, the various versions of Equation 3 were estimated with OLS. To all estimated regressions that performed well in terms of the autocorrelation, heteroscedasticity and CUSUM (cumulative sum of squares) tests, we applied the Newey–West procedure. In all remaining cases we experimented with the ARMA approach. This approach involves an iterative three-stage procedure of identification, estimation and diagnostic checking. Model identification tools such as SACF (sample autocorrelation function), SPCF (sample partial autocorrelation function) were used for identifying adequate models. For model selection we used Schwartz's SBC as the model selection criterion. The optimal order of the model is chosen by the value of m, which is a function of p and q, so that SBC(m) is minimum.

⁸ See, for instance, Saunders and Klau (1985), Graham and Seldom (1990) and Aldcroft (2001).

⁹ The three countries with a proportional political system, notably, Belgium, Italy and the Netherlands, are not examined here.

¹⁰ The detailed estimates of the above cycles of tax- and target variables as functions of the political dummies and other non-political regressors are available upon request.

¹¹ Following different methodologies, Heckelman (2001) and Bratsiotis (2000) cannot find conclusive evidence in support of the rational partisan model in the cases of the UK (for GDP growth and unemployment) and Greece (for inflation), respectively.

¹² The source of the information in this table is estimates of regression (3) in association with the search of electoral cycles. Degrees of freedom problems prevented comparable estimates for the case of partisan cycles.