It is generally believed that intergenerational coresidence by elderly parents and adult children provides old‐age security for parents. Although such coresidence is still the most common living arrangement in many countries, empirical evidence of its benefits to parental health is scarce. Using Indonesian data and a program evaluation technique that accounts for non‐random selection and heterogeneous treatment effect, we find robust evidence of a negative coresidence effect. We also find heterogeneity in the coresidence effect. Socially active elderly parents are less likely to be in coresidence, and when they do live with a child they experience a better coresidence effect. Copyright © 2013 John Wiley & Sons, Ltd.