Robustness

There are a number of criticisisms that could be leveled against our results so far. First, with respect to the time period considered – as discussed above – one could argue that in particular the late 1980’s may have been a rather special period in Austrian migration history. This thus raises the question as to how sensitive our results are to a change in the time period of comparison. Second, as also discussed above, focusing on the Austrian Labour Force Survey induces an element of measurement error into our estimates that could impact on our results. Thus one could ask, what the potential effects of these measurement errors are. Third, as also shown above, most of our results indicate that controls for time varying sending country characteristics mostly remain insignificant. One may thus wonder, how these variables impact results. Similarly one could also wonder, what the impacts of controlling for individual characteristics are on results. Fourth, one may also object to using the total foreign born population, rather than active age groups that may be less strongly affected by still receiving education.

Table 5.9: Logit regression results of the probability for a migrant being high, medium or low skilled before and after accession to the EEA (alternative specifications)

1994-96 X EEA 1997-99 X EEA 2000-02 X EEA Pseudo R2 Log Likelyhood Shorten Period to 1991-2002 Low education –0.083 *** –0.078 ** –0.103 *** 0.220 –46,524,492 0.027 0.039 0.035 Medium education –0.026 0.126 * 0.098 0.087 –82,617,851 0.080 0.074 0.084 High education 0.146 –0.041 0.064 0.161 –67,469,212 0.097 0.056 0.079

Exclude national variables

Low education –0.084 *** –0.079 ** –0.088 *** 0.234 –50,313,377 0.026 0.036 0.024 Medium education –0.032 0.113 * 0.081 0.093 –86,418,076 0.073 0.067 0.056 High education 0.143 * –0.062 0.034 0.161 –70,501,932 0.087 0.043 0.056

Focusing only on sending countries with more than 200 observations

Low education –0.132 *** –0.085 ** –0.114 *** 0.308 –80,930,002 (0.034) (0.039) (0.043) Medium education –0.014 0.170 ** 0.138 ** 0.108 –116,400,000 (0.060) (0.050) (0.056) High education 0.095 –0.044 –0.001 0.214 –78,626,673 (0.064) (0.027) (0.037)

Excluding individual variables

Low education –0.081 *** –0.071 ** –0.092 *** 0.054 –65,477,492 (0.027) (0.034) (0.030) Medium education –0.034 0.099 0.062 0.074 –96,930,198 (0.077) (0.067) (0.074) High education 0.122 –0.028 0.059 0.097 –82,232,946 (0.088) (0.050) (0.064)

Only 20 – 65 year olds

Low education –0.064 ** –0.055 ** –0.079 ** 0.093 –46,529,115 (0.023) (0.027) (0.024) Medium education –0.056 0.131 ** 0.088 0.091 –84,394,947 (0.083) (0.068) (0.074) High education 0.133 –0.067 0.030 0.142 –71,645,741 (0.099) (0.054) (0.073)

Source: Austrian Labour Force Suvey (pooled values 2004-2007), WIFO-calculations. – Reference group are migrants from countries excluding former Yugoslavia, table reports marginal effects of equation 4.1, values in brackets are (cluster corrected) standard errors of the estimate, ***, (**) and (*) signify significance at the 1%, (5%) and (10%) level, respectively. Nobs= number of observations. Country and time fixed effects and other control variables in are not reported.

Table 5.10: Logit regression results of the probability for a migrant being high, medium or low skilled before and after reform in residence law 2003 (alternative specifications)

2003-2005 X

EEA Pseudo R2 Log Likelyhood Exclude national variables

Low education 0.095 *** 0.311 –93,449,110 (0.036) Medium education –0.006 0.123 –131,200,000 (0.036) High education –0.040 0.209 –94,319,581 (0.032)

Focusing only on sending countries with more than 200 observations Low education 0.128 *** 0.307 –76,900,089 (0.046) Medium education 0.001 0.118 –108,300,000 (0.058) High education –0.585 0.193 –76,000,695 (0.323)

Excluding individual variables

Low education 0.023 0.082 –50,825,903 (0.041) Medium education 0.108 0.075 –79,237,040 (0.068) High education –0.106 0.090 –69,137,163 (0.067)

Only 20 65 year olds

Low education 0.149 0.093 –40,083,076 (0.371) Medium education 0.120 0.100 –70,496,821 (0.073) High education –0.532 0.142 –60,769,364 (0.320)

Source: Austrian Labour Force Suvey (pooled values 2004-2007), WIFO-calculations. – Reference group are migrants from countries excluding former Yugoslavia, table reports marginal effects of equation 4.1, values in brackets are (cluster corrected) standard errors of the estimate, ***, (**) and (*) signify significance at the 1%, (5%) and (10%) level, respectively. Nobs = number of observations. Country and time fixed effects as well as other control variables are not reported.

To address these issues, in tables 5.9 and 5.10 we thus present estimation results for a number of alternative specifications of equation (5.1) for both the effects of EEA-accession (table 5.9) and reforms of residence law in 2003 (table 5.10). In particular in these tables report marginal effects for the period-EEA interaction terms of specifications in which we limited our sample to migrants that settled in Austria permanently in the time period 1991 to 2002 in order to gauge the potential bias that may result from including migrants from 1988 to 1990 in our estimates. In addition we also present results of estimating equation 5.1 only for migrants coming from

countries for which we have more than 200 observations in our sample both for the EEA- accession and reforms of residence law in 2003. Furthermore, in a further extension we also present marginal effects when focusing only on the population aged between 20 and 64. Finally, we also present a set of results in which we excluded time varying sending country variables (i.e. GDP per capita, share of low skilled population, infant mortality and the Gini- coefficient) from the estimation as well as a specification in which we excluded individual variables (i.e. age, age squared, gender marital status and the dummy for children), to assess the potential impact of co-linearity of these variables with our period – EEA interaction terms. As can be seen from the results of these perturbations in our specification, the impact of these changes is only minor and provides only few additional insights. As with the results above in all results reported we find that after the EEA-accession the share of migrants with a low educational attainment level that settled in Austria from the EEA countries reduced significantly relative to migrants from other countries. Additionally, these results, however, also indicate that this shift away from low skilled migrants may have been associated with an increased share of medium skilled migrants from the EEA in particular from the time period 1997 onwards. When focusing on countries, which provide more than 200 observations in our data set, our estimates for treatment effects are significant in this case.

Similarly also regressions analyzing the impact of reforms of residence law in 2003 on the skill structure of migrants from third countries reconfirm most of our previous findings. These results, however, slightly strengthen the conclusion of a worsening skill structure of migrants after the reforms, since both when excluding sending country variables as well as when focusing only on the large migrant groups in Austria we find a significant increase in the share of low skilled migrants settling in Austria after 2003, while as above there is no significant impact on the reforms on the share of medium or high skilled migrants.

5.7 Conclusions

The results of this chapter thus suggest that after EEA-accession the skill structure of permanent migrants from the EEA relative to that of migrants from other countries improved. We find highly robust evidence that the share of low skilled permanent migrants from the EEA to Austria reduced relative to the share of low skilled permanent migrants from other countries after Austria’s accession to the EEA. With respect to the reform of residence law in 2003, by contrast, our results are much less robust, with some significant results even pointing to an unexpected negative impact of these reforms on the average skill structure of permanent migrants, with increased shares of low skilled permanent migrants from countries outside the EEA. Our interpretation of these results is that the implicit positive impact of the reforms in the migration regime were countervailed by an increased demand for low skilled migration in other areas of permanent migration from third countries, which are unaffected by quota. This reduces the chances of identifying effects on permanent migration, since the reform only covered around one third to one sixth of the actual migratory moves for residence purposes from third countries to Austria.

Our results thus suggest that liberalized migration can have a positive impact on the education structure of permanent migrants if the previous migration regime was strongly focused on low skilled migrants. The results also present case study evidence to warn that reforms of migration law that provide privileged access of highly skilled permanent migrants may not provide the expected results if other elements of the regulation system governing migration counteract these developments.

To what degree can these results be generalized to other countries or time periods (such as for instance the freedom of movement of labour that will be granted to citizens of the new member states if the EU)? We would argue that in all likelihood it is too early to draw firm conclusions on this issue. As pointed out in the introduction there is only very little formal evaluation literature that analyses the effects of changes in migration policy on migration outcomes. Countries, however, differ substantially in their migration laws as well as labour market institutions and these institutions also change over time. It is also likely that the effects of individual migration policies are shaped by the interactions of a number of these institutions. Thus it is hard to tell from one case study alone, which of our findings are general and which are specific to the particular institutional environment of Austria in the time period analyzed in this paper.