This estimation procedure relies on a more restrictive set of distributional assumptions than the regression specification; moreover, the number of observations available is somewhat low to estimate the full model. Thus, these tests should be interpreted with these caveats in mind. However, by placing a more elaborate model, it allows to examine the relevance of the results after controlling for potentially relevant interactions. Table 6 reports the results. In contrast to the regression specification, the coefficients here are interpreted as having a positive effect if above zero and a negative effect if below zero