Correction to Skill specificity and attitudes toward immigration

Pardos-Prado, S. and C. Xena. 2019. Skill specificity and attitudes toward immigration. American Journal of Political Science, 63(2): 286–304. https://doi.org/10.1111/ajps.12406

The number of countries reported in Table 1 in the original publication of Pardos-Prado and Xena (2019) has been found to be incorrect. We are very grateful to Professor Michelle Dion for bringing this issue to our attention.

The error was due to logging GDP and unemployment spending after centering all variables. This inadvertently dropped from the analysis a significant number of observations coded as 0. The loss of country sample size after introducing logged variables was difficult to spot since the software we use to run cross-classified hierarchical models does not report the number of countries.

Reassuringly, the substantive results remain unchanged if GDP and unemployment spending are not logged, and therefore if the full sample is retrieved (13 countries across five waves). A corrected version of Table 1 can be found below. The coefficients of interest (skill specificity and occupational unemployment) remain highly significant across the four model specifications: p = 0.009 in the second model, and p = 0.000 in all other models. In fact, the coefficient of skill specificity is now more precisely estimated (narrower CIs). The sign remains consistently positive, meaning that higher values of skill specificity or occupational unemployment increase anti-immigrant attitudes. Our theory does not involve any country-specific feature, so it probably makes sense that the results are not overly sensitive to changes in the country sample.