The poor mental health of divorced individuals (Afifi et al., 2006; Blekesaune, 2008; Breslau et al., 2011; Hughes and Waite, 2009; Liu and Chen, 2006;
Meadows et al., 2008; Strohschein et al., 2005; Wade and Pevalin, 2004) was seen, for a long time, mainly as a result of the causal effects of divorce, although the role of selection was also acknowledged (Bulloch et al., 2009;
Hemström, 1996; Johnson and Wu, 2002; Kessler et al., 1998; Lund et al., 2004; Murray, 2000). Despite extensive research on the determinants of divorce, surprisingly few studies have directly assessed the effect of poor
mental health on the risk of subsequent divorce (Butterworth and Rodgers, 2008; Idstad et al., 2015; Kessler et al., 1998; Lyngstad and Jalovaara, 2010;
Mojtabai et al., 2017), while more studies have examined changes in mental health before and after divorce (Blekesaune, 2008; Booth and Amato, 1991;
Bulloch et al., 2009; Hope et al., 1999; Johnson and Wu, 2002; Mastekaasa, 1995; Wade and Pevalin, 2004).
Going through the process of separation is stressful, and related marital conflict can raise the level of symptoms of poor mental health (Blekesaune, 2008; Wade & Pevalin, 2004; Waite et al., 2009; Willitts et al., 2004). It is thus difficult to evaluate whether poor mental health immediately before divorce reflects genuine selection out of marriage or the process of divorce (Bulloch et al., 2009; Butterworth and Rodgers, 2008; Kessler et al., 1998).
Accordingly, estimates on the effect of poor mental health on divorce risk are smaller when the time gap between measured symptoms and divorce is longer (Hope et al, 1999; Idstad et al., 2015; Mojtabai et al., 2017).
One way to evaluate the importance of selection compared to marital conflict and the stress of going through divorce is to explore changes in mental health immediately after divorce, and compare these to the changes observed before divorce. In the case of selection out of marriage, we would not expect mental health to deteriorate or improve after the divorce. Examining these changes requires the use of longitudinal data with multiple observations both before and after divorce. A study using US panel data and measuring psychological distress found no support that a high level of symptoms predicts a higher probability of divorce, but the level of symptoms increased immediately before divorce and declined thereafter, reaching pre-divorce levels during the next two years (Booth and Amato, 1991). Analyses using one more wave of the same data and a slightly different measure of poor mental health verified the increase in symptoms before divorce, but a post-divorce decline was only seen among those initially happy with their marriage (Johnson and Wu, 2002). The second study also found support for a small selective effect already evident five years before divorce (Johnson and Wu, 2002). A quite similar measure of poor mental health was used in a panel study exploring three waves of Norwegian data (Mastekaasa, 1995). Controlling for the level of symptoms up to four years before divorce reduced the effect of divorce on subsequent mental health, whereas controlling for symptoms four to eight years before divorce had no impact. This was concluded to support temporary anticipatory effects rather than selection (Mastekaasa, 1995).
Because the time gap between waves was three to four years, it was impossible to evaluate more precisely when the level of symptoms started to increase before divorce.
Two studies using data from multiple annual waves of the British Household Panel survey and measuring mental health with the GHQ found a sharp increase in symptoms of poor mental health immediately before divorce (Blekesaune, 2008; Wade and Pevalin, 2004). At the time of divorce, about 55% of individuals reported poor mental health, the proportion being about
32% two years before divorce and less than 30% two years after divorce, and less than 20% among the continuously married (Wade and Pevalin, 2004).
Among women, the mean level of symptoms started to already increase four years before divorce, followed by a rapid increase during the 18 months immediately before divorce among both men and women (Blekesaune, 2008).
Despite the increase in poor mental health immediately before divorce, both the severity and stability of symptoms predicted divorce up to four years later (Wade and Pevalin, 2004). Nevertheless, the sharp increase in symptoms of poor mental health was followed by a similar decline during the 18 months immediately after divorce, and four years after divorce the mean levels of symptoms were similar to those observed four years before divorce (Blekesaune, 2008), supporting an anticipatory effect instead of selection out of marriage.
Taken together, these previous studies suggest that depending on the measure of poor mental health, studies should record changes in mental health up to five years before divorce to evaluate the role of selection compared to the effects of marital conflict and anticipation of divorce. The time between repeated measurements should be as short as possible.
A critical shortcoming of these longitudinal studies is that they only use data on one of the partners. When studying the effect of mental health on divorce, it is essential to use data on both partners. If we only have information on one partner, we fail to identify half of the couples where only one partner is in poor mental health. With individual-level data it is also impossible to examine the effect of spousal health similarity on the risk of divorce.
Previous studies using data on both partners showed that couples with only one spouse in poor mental health had a lower risk of divorce than couples with both spouses in poor mental health, and a higher risk than couples with two healthy partners (Butterworth and Rodgers, 2008; Idstad et al., 2015;
Merikangas, 1984), but findings on the magnitude of this effect are contradictory. In a small clinical sample, couples with two mentally ill spouses had a significantly higher risk of divorce than couples with only one mentally ill partner (Merikangas, 1984), suggesting that spousal similarity in poor health multiplies the risk of divorce. In an Australian sample, the effect of both partners with mental health problems was additive (Butterworth and Rodgers, 2008), and in a large Norwegian sample lower than expected based on the main effects of both partners’ symptoms of poor mental health (Idstad et al., 2015). The results were also contradictory in showing whether the wife’s or the husband’s poor mental health had a larger effect on the risk of divorce. It also remains unclear whether the risk of divorce reaches a constant level at some time after the onset of poor mental health.