The probability of divorce in the U.S. has remained constant for the last two decades at about 'half of all marriages.' While this estimate is well established, and marked differentials in divorced rates are well known, there are no reliable estimates of differences in the cumulative probability of lifetime divorce.
Using data from the 1990 June CPS, we document very large differentials by race, age at marriage, and education in the probability that recent cohorts of marriage will end in separation or divorce. Then, using data from the 1995 NSFG, we find important increases in differentials in marital dissolution, and especially in all unions, during this period of stable aggregate rates. These results indicate that examining only at marital transitions obscures the growth in family instability that has resulted among some groups because of an increasing proportion of unions begun as cohabitation.