Wirtschaft und Gesellschaft 42. Jahrgang (2016), Heft 4 ments.48 Alternatively one could argue that dividend payments are an indication of increasing “shareholder value” orientation, inducing a “downsize and distribute” strategy that will supress wages and employment.49 Household debt has been found to reduce wage share arguably through increasing financial vulnerability that has an adverse effect on workers’ willingness to engage in collective action.50 WELFARE is social government spending at the individual level as explained in the previous section. This variable is measured at the country level and is the same for all sectors. INEQUALITY is country level inequality measured as the Gini coefficient or the income share of the top one percentile, again the same for all sectors. We apply two main estimation techniques. Our baseline estimation is performed using the within estimator (also referred to as Fixed Effects Estimator), while we estimate the variance-covariance-matrix of the remainder error term using the approach developed by Driscoll and Kraay (1998). Therefore, standard errors are fully robust with respect to serial correlation within countries, cross-sectional correlation across sectors as well as general heteroscedasticity. Our main robustness controls are conducted with a first difference estimator. This has the additional advantage that potential non-stationarity concerns are taken care of given that all our variables are unambiguously stationary in first differences.51 Since there is reason for concerns regarding the endogeneity and specifically reverse causality for our measures of technological change and globalisation, and because the effect of other variables will most likely be manifested with a time lag, all explanatory variables enter the equation with a lag. It would be preferable to employ a General Method of Moments estimator to tackle the issue of endogeneity as well as the dynamic nature of the wage share. However, due to the limited number of cross sections in our single country estimations this estimation method is not appropriate. With regards to endogeneity concerns we employ the second best approach by using lagged values of the explanatory variables.52 In addition to the pool of all sectors, separate regression analysis will be performed for sector groups disaggregated as high skilled and low skilled sectors in manufacturing and services separately. In separate regressions we employ four alternative measures of the wage share for robustness check: i) the after tax wage share calculated as explained in the previous section; ii) compensation of employees as a ratio to value added, i. e. the wage share without the adjustment for self-employed workers; iii) wages and salaries as a ratio to value added – this is a measure of primary market distribution since it excludes all redistribution measures including social security contributions; iv) a sample without the outliers in which we drop all observations where the wage share exceed 1. 572