We apply a twin design to examine the relationship between health and education andincome. The estimated associations between health and education and income, controlling forunobserved endowments, at the twin-pair level, are lower than estimates obtained via ordinaryleast-squares (OLS) on the same sample. Thus, OLS-based effects of education and incomeare biased, exaggerating the contribution of education and income to health inequality. Themain part of health inequality is explained by within-twin-pair fixed effects, incorporatingfamily background and genetic inheritance. It appears that education and income policieshave less to offer for reducing health inequality than is usually assumed.