There is one result that stands in sharp contrast to this general pattern. A father's
death greatly reduces the probability that his son will be married. We do not observe
this pattern for other causes of paternal absence in the linear model, but we did find
this effect in the nonlinear model. One possibility is that death affects marital status
differently from other causes of paternal absence. Alternatively, looking at death
may provide a less biased estimate of the effect of paternal absence than the simple
linear model that does not distinguish among causes of parental absence.