Revisiting the Porter Hypothesis: A Nonparametric Analysis on the impact of Pollution Abatement Technologies on firms' performances

Nonparametric regression models are designed to relax the Gauss-Markov assumptions needed to obtain an unbiased and consistent estimator from the traditional parametric regression. The rationale is to let the function be defined by the data locally, without imposing a linear relationship or higher orders polynomials to fit possible non-linearity, at global level. This paper has the aim to investigate the Porter Hypothesis with the use of nonparametric analysis using kernel regression, in particular the local constant estimator developed by Nadaraya and Watson (1955; 1956) and the linear extension proposed by Stone (1977) and Cleveland (1979). The use of kernel to deal with discrete variables is extremely useful to study the effect of the introduction of pollution abatement technologies, used as a proxy assessing for policy stringency, over the value added and hence to test the effect of regulations on firm’s performances: in doing so, starting from the estimator designed by Aitchinson and Aitkens (1976), the extension proposed by Li and Racine (2007) is used. The nonparametric analysis provides a model with a better goodness-of-fit, furthermore the value of the bandwidth referred to the introduction of pollution abatement technology obtained through Kullback-Leibler cross-validation, underlines heterogeneity between groups, and suggesting the positive effect of the introduction of environmental regulation on the performance of firms, leading to the so called Porter hypothesis.