Twins are more different than commonly believed, but made less different by compensating behaviors

•Using a U.S. census, twin survival patterns are shown to differ by birth order.•Older/educated/married mothers show ‘compensating behavior’ to reduce the gap.•Hence cautions are warranted for studies that regards twins homogeneous.•A semiparametric quantile-based fixed-effect estimator is devised for the analysis. Twin studies are popular, because twins are believed to be the same/similar in genes and environmental exposures. It is well documented, however, that the firstborns are healthier at birth. We use the entire U.S. record of twin births during 1995–2000 to show that the survival duration parameters differ between twins depending on the birth order. We also find that wiser (i.e., older or educated) or married (i.e., resource-richer) mothers take more care of the weaker, which is a “compensating” behavior reducing the twin difference, as opposed to “reinforcing (the twin difference)” behavior. The systematic survival pattern difference and the mother's intervention against nature send cautions to twin studies that regard twins homogeneous to interpret their findings accordingly. Since the survival duration in our data is 97% right-censored in one year, we devise a quantile-based ‘fixed-effect’ semiparametric estimator that can handle heavy censoring, which is our methodological contribution.