Further Discussion of the Effects of No-Fault Divorce on Divorce Rates

Abstract

In response to my comment (Glenn, 1997) on their 1995 study of the effects of no-fault divorce on divorce rates (Nakonezny, Shull, & Rodgers, 1995), Rodgers, Nakonezny, and Shull (1997) introduced a new method for estimating the effects. Application of the method led the researchers to lower their original estimates of the effects but still to conclude that around 57,000 extra divorces per year in the whole U.S.... are directly attributable to the implementation of no-fault divorce law (p. 1028). Although Rogers, Nakonezny, and Shull pointed out that any effects of the implementation of no-fault divorce during the 3 years after implementation may not have persisted, the figure of 57,000 extra divorces per year is being cited by activists who want to rescind no-fault divorce or limit access to it (e.g., Howard Center, 1998). The new method uses linear regression to project a divorce rate for each state for each of the 3 years after implementation of no-fault divorce from the linear trend line established during the 10 years prior to implementation. The estimate of the effect of no-fault divorce is the mean of the observed rates for the 3 postimplementation years minus the mean of the projected rates for those years. The mean of the estimated effects for the 50 states is .23. Unfortunately, this method confounds any effects of implementation of no-fault divorce with the effects of other influences that brought about the divorce boom of the 1960s and 1970s and that led to a levelling off of divorce rates after the late 1970s. The unstated assumption on which the method is based is that no influences other than the adoption of no-fault divorce affected the rate of change in the divorce rate or produced change when there had been none during the 14-year period covered by the data for each state. This assumption is untenable. In the states that adopted no-fault divorce during the early years of the divorce boom, the divorce rate was virtually stable during the early part of the 10-year base period used for the projections but in most cases started to rise well before no-fault provisions were implemented. In other words, Rodgers, Nakonezny, and Shull made linear projections from nonlinear trends--rarely a useful procedure. The Rodgers-Nakonezny-Shull method estimates positive effects of no-fault divorce on divorce rates only during the first 10 years of the divorce boom. (See Figure 1.) For the 11 states that initiated no-fault divorce during 1975-1979, the mean estimated effect is -.73, even though there is no apparent reason why implementation of no-fault divorce should have decreased the divorce rate in any state at any time. This rather bizarre estimate resulted from the fact that divorce rates increased steeply in all states during the 10-year base period used to project the postimplementation rates for the 11 states that inaugurated no-fault divorce during 1975-1979 but then increased less steeply during 1975-1979 and levelled off after that period. The pattern of differing estimated effects by period shown in Figure 1 is largely, if not entirely, an artifact of the method used to arrive at the estimates. How the Rodgers-Nakonezny-Shull method inevitably estimates substantial positive effects for the states that implemented no-fault provisions during the early years of the divorce boom is illustrated by the data in Figures 2 and 3. I selected the seven states that implemented no-fault divorce in 1971 for this illustration because the mean positive effect estimated by Rodgers, Nakonezny, and Shull is greatest for that year, except for 1968, when only one state implemented no-fault divorce (Figure 2). To estimate the effects of influences other than those from no-fault divorce, I compiled parallel data for the 13 states that implemented no-fault divorce after 1974 and for which divorce rates are available for all years from 1961 through 1974 (Figure 3). During 1961-1970, the base period for the projection, the mean yearly increase in the divorce rate in the states that implemented no-fault divorce in 1971 was . …