This study challenges the conventional wisdom that elderly, child and overall poverty are divergent. Comparing the official U.S. measure with the Luxembourg Income Study's (LIS) measure, I show that the official measure underestimates elderly poverty by a significant amount and child poverty by a lesser amount. The elderly were considerably more likely to be poor than children in the 1970s, children were more likely to be poor 1984-1997, but the elderly were more likely to be poor in 2000. Both the elderly and children are much more likely to be poor than the overall population. Analyses of 18 rich Western democracies show that overall and child poverty are very strongly positively correlated, while elderly poverty is moderately correlated with those two. Multivariate analyses show some commonalities and some differences in the sources of these three. Two measures of the welfare state significantly reduce overall, elderly and child poverty. While female labor force participation reduces all three, manufacturing employment, economic performance and demographic variables only influence one or two of the dependent variables.