Intra-EU Labour Mobility and the Welfare State 87 For the rest of the results in Table 4.8, the explanatory variables are reasonably similar between the two labour migration variables. Trade openness is insignificant for unemployment spending but positively and weakly statistically significant for social assistance and capital openness is insignificant in both models. Left-wing governments are positively and significantly associated with both the unemployment and social assistance replacement rates, as hypothesised. Union density is insignificant in both models, while deindustrialisation is significant and negatively associated with unemployment spending. Finally, unemployment has no statistically significant relationship with either social protection programme. 4.5 Robustness To check the sensitivity of our results to the model specification employed above, we have chosen to run an Error Correction Model as a robustness check. Table 4.9 shows that there are no major changes in the key variables of interest and our assertion that the compensation hypothesis is a potential mechanism for the way that immigration may change the welfare state. However, the overall picture is slightly more mixed. While the lagged variable for CEE labour migration is still significant and positive for the change in unemployment spending, which is very much in line with the idea that increased risk in the labour market increases demand for compensation from the welfare state, we also find a small negative between CEE labour migration and the change in pension spending. For WE labour migration in Table 4.10, we find a negative association with incapacity spending (albeit on the edge of statistical significance), a link not found in our previous model specification in Table 4.7, and a positive one for family spending which is in-line with the previous model (Table 4.7).
RkJQdWJsaXNoZXIy MjY0ODMw