A structural VAR approach on labour taxation policies

Abstract

This paper presents a Structural VAR analysis on the employment and output effects of labour tax policies in six European countries for the period 1974–1997. By considering impulse response functions, it turns out that, on average, a shock to the total personal income tax revenues is positively correlated to employment, whereas there is mixed evidence on the output effect. Moreover, the quantitative impact of these effects, especially those related to the output, appears to be quite small. However, by introducing explicitly four labour tax parameters (namely the marginal and average tax rates for the personal income tax and the payroll tax), it turns out that these effects are not negligible after all: for some countries it is possible to conceive labour taxes as policy instruments favouring more employment and a better economic performance. However, the empirical support on the sign of the output and employment effects is mixed, suggesting that the same domestic fiscal policy does not produce the same impact in all the European countries.

Acknowledgements

I wish to thank Gianni Amisano and Massimiliano Serati for their helpful suggestions and comments. I greatly benefited from discussions with participants to presentation held at the 2003 SIEP Conference in Pavia. Financial support from the University of Piemonte Orientale is gratefully acknowledged.

Notes

¹ For labour tax revenue, tax revenue collected by personal income taxes is meant.

² Prices are taken as given and normalized to one.

³ Appendix I reports all the details of the derivation of the system of Equation 4.

⁴ Muysken et al. (1999), by using Dutch data, show that tax incidence depends on the side of the market that is taxed.

⁵ The appendix will describe a possible structure based on some feedback mechanisms.

⁶ The econometric package referred to is MALCOLM (Maximum Likelihood Cointegration analysis of Linear Models) written by Rocco Mosconi (1998). Table A1 refers to the tests on the univariate series of the country specific total personal income tax revenue, real GDP and employment. The test on each countries’ four labour tax parameters are not reported for space convenience but are available upon request from the author.

⁷ As suggested by Dolado et al. (1990) the more general model is started from which contains both a deterministic trend and an intercept in order to be sure that the ‘true’ generation process of the data is nested within the model.

⁸ Note that the Φ₃ (joint insignificance of the trend, the autoregressive term and the constant) is rejected as well. This would imply that West's result should be looked at according to which the statistics should be compared to the values of the Normal table. Yet, since the result of the Φ₂ test is quite robust it is believed that there are enough elements in favour of the conclusion.

⁹ When the role played by each of the four labour tax parameters is evaluated, the vector of the endogenous variables takes the form of:

¹⁰ For Germany the test seems to prefer slightly five lags.

¹¹ For the case of Germany, because of the presence of an intervention dummy which accounts for the 1989 German re-unification the quantilies reported in the table are not appropriate. However, it seems that the findings of a cointegration rank equal to two might be robust to the critical values correction.

¹² For more details on the methodology refer to Amisano and Giannini (1997).

¹³ Labour tax revenues shocks affect positively output in Italy and Sweden. Furthermore, a positive coefficient has been found in the cases of Germany and Sweden.

¹⁴ The Spanish output effect is positive but quite close to zero.

¹⁵ In some cases, e.g. for instance Germany, both the effects turn out to be negative at the end of the observed period.

¹⁶ Table A5 does not report for space constraints the percentage of the variance of the total labour tax revenues explained by the structural innovations in employment and output. This variance is the relevant indicator for evaluating the feedback effects and as expected results to be very small. A table containing these results is available upon request from the author.

¹⁷ For easiness of exposition all the cointegration analyses and the identification of the structural shocks for the four equations model are not reported. These results are available upon request from the author.

¹⁸ The modelling choice presents the cost of considering only a specific portion of tax revenues from labour income. Giorno et al.'s analysis refers to the entire income distribution. However, their income distribution reflects the assumptions made on its shape.

¹⁹ Recall that the empirical findings refer to a model structure that is identified starting from this recursive system and allowing for some feedback mechanism whose restrictions are those accepted by the data.