Surveys are often used to obtain data on ethnic differences in health care utilization. This study examines the cross-cultural validity of such data. Survey data were linked to an insurance register, for 3054 residents of Amsterdam, the Netherlands. Concordance was estimated using Cohen's kappa statistics (agreement adjusted for chance agreement) and loglinear analyses regarding hospitalization, ambulatory physiotherapy, and prescription drugs. Self-reports yield higher utilization rates than registration, but irrespective of ethnicity and without statistical significance. Concordance between self-report and registration is fair [kappas (95% confidence interval): 0.77 (0.72-0.82), 0.71 (0.66-0.75), and 0.58 (0.54-0.63) for hospitalization, physiotherapy, and prescription drugs]. Concordance is lower, though mostly without statistical significance, for immigrants, especially from Morocco and Turkey. Health care utilization is somewhat overestimated on the basis of self-report. This has no systematic impact on estimates of ethnic differences in health care utilization but adds measurement error to such comparisons. Copyright (C) 2000 Elsevier Science Inc.