Growth effects of fiscal policies: an application of robust modified M-estimator

Abstract

In the field of economics only nonhigh-quality data is usually available, which can cause the widely used least square estimators (LSE) to be biased and inefficient. Therefore, the present study introduces the robust modified M-estimator (MME) proposed by Yohai et al. (1991). In the case of growth regressions with fiscal variables it can be shown that LSE is biased and inefficient, whereas MME is not. The robust regressions ascertain a stable positive growth effect of public infrastructure and education. Moreover, this study shows that government size has not been detrimental to growth for OECD countries in the past. No growth effects of taxation have been found so that endogenous growth theory is not corroborated in this regard. Consequently, fiscal policies aiming at promoting growth should focus on infrastructure and education.

Notes

¹Note that the term ‘MM-Estimator’ in Yohai et al. (1991) is abbreviated to ‘MME’ in this article.

²Conditional convergence is not accounted for due to various reasons. Most crucially, I have performed a pre-test finding no empirical evidence for conditional convergence between the OECD countries of the sample. The Phillips-Perron tests, with and without a deterministic trend, suggest that the dispersion of the growth rates of labour productivities, measured by the robust median absolute deviation, is stationary across the countries of the sample. In both tests, with and without a trend variable, the null hypothesis of a unit root is rejected at a 1% level (Z-statistics: −25 and −21). These results clearly do not support the hypothesis of conditional convergence between countries of the sample in the observed time period. Moreover, this study does not explicitly test the endogenous growth model. Finally, some authors claim that the standard indicator used for conditional convergence, i.e. the per capita initial GDP, cannot be interpreted unambiguously, which would make this indicator rather useless (Pack, 1994, p. 65; Thirlwall, 2003, p. 45).

³For this the R-packages MASS, sandwich, lmtest and zoo have been applied.

⁴In the following transport and communication infrastructure are shortened to transport infrastructure.

⁵The comparison refers to the following studies: Kneller et al. (1999), Bassanini et al. (2001), Bleaney et al. (2001), Fölster and Henrekson (2001), De Ávila and Strauch (2003), Agell et al. (2006).