Please keep in mind that a descriptive study like that only shows correlations of variables. That work wasn't a longitudinal study, therefore one cannot convey a causal relationship of the variables <number of partners> and <divorce rate> from it - even if the errors pointed out by jarvist would not be present. It's also quite possible that <number of partners> is only a confounding variable for something different that has not been measured, e.g. certain mood problems, bad parental examples on how to properly/ fairly argue in a relationship etc. That would explain both the number of failed relationships before and after marriage.