Regarding preferences for activation, we hypothesized that the effect would become stronger
with increasing skill levels because high-skilled outsiders should have particularly strong
incentives to favor investment in human capital and jobs: what they want first and foremost is
an opportunity to work. Hypothesis 2a is fully confirmed by the data: Model 4 indicates a
positive and significant interaction effect for activation preferences. Figure 2 shows that the
higher the educational level, the stronger labor market vulnerability affects preferences for activation.
The predicted probabilities substantiate this finding: the likelihood of an individual
with the highest degree of outsiderness and tertiary education to strongly support activation (a
score of 7 or more on a scale from 0 to 10) is 52%, while the same individual exposed to the
lowest degree of labor market vulnerability has only a 36.3% chance to strongly support activation.
The difference between a low-skilled individual with highest and lowest degree of
labor market vulnerability is around 15.9 percentage points.
To reiterate, we expect preferences for social insurance to be particularly strong among highskilled
insiders (H2c). Although social insurance is relatively unattractive for all low-skilled in
account of their inherently low contributions, the higher the human capital of a respondent, the
stronger we expect the marginal effect of vulnerability to be. We therefore expect the insider/
outsider divide to widen with increasing levels of education. Figure 2 confirms this interaction
effect graphically. The results corroborate our theoretical reasoning: high-skilled insiders
support the equivalence principle more strongly than high-skilled outsiders. A high-skilled
insider (lowest degree of outsiderness) has a likelihood of 81.9% of agreeing that individuals
with higher contribution records should receive larger old-age pension benefits. For the same
individual being an outsider (highest degree of labor market vulnerability), this probability is
15 percentage points lower.
To conclude our analysis,we examine the relative impact of education and labor market vulnerability
on social policy preferences. Table 5 reports the predicted probabilities of an average
individual at lowest (primary education or less) and highest (tertiary education) level of education
to support each of the three distributive principles, compared with the same individual at
minimum and maximum values of labor market vulnerability.We find that differences between
lowand highly educated respondents are much smaller than differences between individuals with the lowest and highest value of outsiderness with regard to all three social policy preferences. The
importance of labor market vulnerability for social policy preferences is striking: differences in
predicted probabilities between individuals with highest and lowest values of labor market vulnerability
are around 16 percentage points, and differences between individuals with highest and
lowest levels of education range between 8.4 and 13.1 percentage points. The same analyses calculated
for different model specifications of the average individual (different countries, men
instead of women and different ages) lead to the same conclusion that the degree of labor
market vulnerability is more important for social policy preferences than the level of education