52 5.3 Estimation results for individual ‘high-wage’ countries27 Our estimation results for France, Germany, the UK, the US, Denmark, Italy and Spain partially confirm our results for Austria, while they differ in interesting aspects as can be seen in Tables 10 to 16. 28 27 The results in this section are based on Guschanski and Onaran, The causes of falling wage share and prospects for growth with equality in a globalized economy, Project Report for INET, (forthcoming). However, estimations for Austria and thereby comparison with Austria are not part of the project. 28 Data availability differs across countries, especially with regards to capital stock data for France and the UK where our cross sections are reduced to eight and eleven sectors as opposed to 19 for Austria for specifications (1) to (3). Furthermore we lose ‘the coke and refined petroleum products sector’ when we apply the first difference estimator for the UK in specification (7) and (8) because it has only 1 observation where all the data is available after cleaning. Exclusion of this sector does however not alter out results. We are able to increase the number of our cross sections to 11 if we estimate specifications (1) to (3) for France using data at the 1-digit level. However, this poses a trade-off since our import data is available at the 2-digit level and therefore requires aggregation and because previous results have indicated that the effect of intermediate import penetration is better observed at a highly disaggregated sectoral composition. However, our results are robust for estimations at 1- or 2-digit levels with respect to intermediate import penetration. Similar considerations apply to the US, where availability of data on the capital stock for the service sectors limits our sample and Spain where there is only very limited data on FDI. In fact, for Spain our sample is reduced to two to three observations per sectors, which in turn creates collinearity between several of our country level variables. For this reason we drop government spending from specification (7) and (8) while we estimate specification (8) without our financialisation variables. The data issues in combination with the limited availability of variables accounting for financialisation is also reason for the reduced number of cross-sections in our first difference estimations.