42. Jahrgang (2016), Heft 4 Wirtschaft und Gesellschaft tion paid by employers to employees, is a better measure of primary market distribution since it excludes secondary distribution. 5.3 Country-level variables With regard to the control variables, union density has a positive effect on the wage share in specification (3) – indeed it is highly significant and renders the effect of intermediate import penetration insignificant. The effect of union density is however not robust at the 1-digit level in specification (6).57 The result is confirmed for sub-pools of manufacturing industries. However, given that the variable is measured at the country level, the reliability of the estimation results by sub-pools is questionable. In order to obtain at least indicative results with union density measured at the sectoral level we performed robustness tests with union density measured at the sector level regardless of our concerns about its reliability as mentioned in section 3. In general results for sectoral union density confirm the results for country-level union density. The positive but not robust impact of union density is generally driven by all sector and skill groups. Furthermore, we experimented with adjusted bargaining as an alternative measure for workers bargaining power. However, given that bargaining coverage stayed at a constant level since the 1970s in Austria the variable created multicollinearity with our fixed effects and we had to drop it. Social government spending turns out to be insignificant or positive for almost all specifications with the exception of estimations for the high skilled manufacturing sectors only where we find an unexpected negative sign for specifications (7) and (8). Nevertheless, like union density, social government spending becomes insignificant for most estimations in first differences, while it is positive for service sectors. Since there are no measures of financialisation at the sectoral level we can only use country-level variables among which household debt and financial payments appear to have a robust negative effect, albeit mostly for estimations in first differences. This finding is robust to the application of different samples, although the highest statistical significance is achieved for the high-skilled manufacturing sector. Similarly we find a negative effect of household debt for the manufacturing sector for the estimations in levels, in both low and high skilled manufacturing sectors alike. Given that lower income workers might be credit constrained and that the recent surge in household debt was mainly driven by the upper-middle class this result seems plausible. It is not entirely clear, however, why workers in the high-skilled manufacturing sector should be stronger affected by household debt than workers in the high skilled service sector. Our specification (8) reflects the argument that personal income inequality is an indicator of the command over resources and power relations, 579