I discuss instrumental variable estimates of the effect of providing unpaid adult care on the caregivers' probability of being employed, using eight waves of the European Community Household Panel. I focus on men aged 40-64 and women aged 40-59 from thirteen Member States, aggregated in two groups of Northern-Central and Southern countries. Previous papers with European data found that IV estimates are more negative than estimates assuming exogeneity of caregiving. I show that this difference is not robust once account is taken of time-invariant unobserved heterogeneity. Indeed, instruments turn out not to be needed, and the estimated effect is negative, but small in both groups of countries. (C) 2011 Elsevier BM. All rights reserved.