Why bother with testing? The validity of immigrants’ self-assessed language proficiency

• Are self-assessed language skills valid measures of language proficiency?
• Self-assessed language skills correlate moderately with language test scores.
• Self-assessed language skills yield different regression results than test scores.
• Task-specific assessments are not more accurate than general ability estimates.
• Self-assessed language skills are systematically biased in certain groups.

Due to its central role in social integration, immigrants’ language proficiency is a matter of considerable societal concern and scientific interest. This study examines whether commonly applied self-assessments of linguistic skills yield results that are similar to those of competence tests and thus whether these self-assessments are valid measures of language proficiency. Analyses of data for immigrant youth reveal moderate correlations between language test scores and two types of self-assessments (general ability estimates and concrete performance estimates) for the participants’ first and second languages. More importantly, multiple regression models using self-assessments and models using test scores yield different results. This finding holds true for a variety of analyses and for both types of self-assessments. Our findings further suggest that self-assessed language skills are systematically biased in certain groups. Subjective measures thus seem to be inadequate estimates of language skills, and future research should use them with caution when research questions pertain to actual language skills rather than self-perceptions.