Downgrading Estimates Controlling for a Key Confounder: Evidence From PIAAC

Abstract

Why do immigrants earn less than equally educated natives? Labour market downgrading—the systematic penalty immigrants face relative to comparable natives—is the standard answer, but one key confounder has been largely absent: measured cognitive skills. Using PIAAC microdata covering more than 230,000 adults across 36 countries, we estimate immigrant penalties while controlling for internationally comparable literacy and numeracy proficiency. Adding the cognitive skill controls reduces the estimated monthly income penalty by roughly 28%, from 6.8% to 4.9%. The residual penalty is heterogeneous: largest in Southern Europe and concentrated in destinations with strict employment protection, incomplete credential mutual recognition and the absence of skill-screened immigration policy. Three complementary identification exercises feasible within PIAAC—Bleakley–Chin age-at-arrival, Adsera–Pytlikova linguistic distance and a Goldsmith-Pinkham–Sorkin–Swift shift-share/Bartik instrument at the cohort × destination level—yield estimates consistent with the OLS pattern; the instrumental variables specification is also consistent with positive selection of immigrants on unobservable traits, which would bias OLS towards zero. The findings point to three policy-relevant levers: language training, credential recognition and skill-screened immigration policy.