1 Introduction

I am very thankful to Diederik Boertien, Sam Jenkinson, and Dimitri Mortelmans, from whom I received very constructive critical and helpful comments.

The first demographic analyses of the historical time trends in divorce rates were carried out more than a century ago (Willcox 1897; Monahan 1940). Up to today, however, no consistent explanation for the long-term trend of rising divorce rates has been offered, and relatively few studies have addressed this issue. While intensive and ongoing research on the determinants of marital instability has resulted in the identification of a large number of risk factors, the question of why there has been a sharp and nearly continuous upward trend in divorce rates in many highly developed countries remains unanswered (see Härkönen 2014 for a review of this research). In many countries, this upward trend continued for more than one hundred years, beginning in the late nineteenth century and ending in the late twentieth century or early twenty-first century (Bennett 2017).

The historical development of divorce rates usually takes the form of an S-curve (Salvini and Vignoli 2011): i.e., divorce rates increase slowly, then more quickly, and then level off. However, the details of this process have varied greatly from country to country. It has, for example, been pointed out that in some countries, the trend of increasing divorce rates came to an end in the 1980s, and may have even turned around. The US seems to be a forerunner in this development, as divorce rates have been decreasing there since the 1980s (Fig. 3.4 in appendix, see Martin and Bumpass 1989; Cherlin 2010; Kennedy and Ruggles 2014; Rotz 2016). In some European countries, divorce rates are still increasing or do not seem to have reached an upper limit (Fig. 3.1). But divorce rates also declined between 2004 and 2015 in a number of European countries, including in the United Kingdom (2004: 2.8, 2015: 1.7), Austria (2004: 2.4, 2015: 1.9), and Belgium (2004: 3.0, 2015: 2.2) (Divorces per 1000 persons, EUROSTAT 2018).

Fig. 3.1
A set of four line graphs illustrates the divorce rate versus time duration from 1960 to 2015 in northern, central, southern, and eastern Europe. It indicates an uneven trend.

(a–d) Crude Divorce Rates (per 1000 persons), 1960–2015 (EUROSTAT 2017). (a) Northern Europe (b) Central Europe (c) Southern Europe (d) Eastern Europe

In the following, we will attempt to review the research on historical divorce trends. We start with a section that focuses on how these trends might be explained (section 3.2). First, we present some methodological considerations on the links between macro- and micro-level factors. Second, we concentrate on the micro level of a marriage, presenting theories that seek to explain how the partners’ attributes and the external social environment are related to marital stability and divorce decisions. Third, we draw on these theories to derive some general hypotheses regarding the macro-level influences on the stability of marriages. Fourth, we examine these broad macro-level influences in more detail, and consider theoretical approaches that seek to explain how long-term socio-structural and cultural changes affect partnerships and the family.

In section 3.3, we discuss the current state of the empirical research on how socio-structural and cultural changes affect divorce rates. We start by exploring the dynamics of this process; i.e., whether period or cohort changes play the larger role. We then look at the four hypotheses developed in the previous section, and we summarize the results of the most influential empirical studies on the question of why divorce rates have been rising and then stabilizing. We cannot, of course, even give an approximately complete picture of all the factors that have contributed to the emergence of this historical trend. The mere fact that this trend started at the end of the nineteenth century, whereas the data and the methodological instruments needed to study such a trend were not developed until the 1980s, clearly restricts our analytical options. Given the societal importance of trends in divorce rates, there are fewer empirical studies on this topic than might be expected. We conclude with some observations and speculations about the forces that have been driving changes in divorce rates (section 3.4).

2 How to Explain Historical Trends in Divorce Rates

Classical sociology was already concerned with the stability of marriage. Durkheim assumed that an increasing division of labor would strengthen marriage as an institution, whereas a liberalization of divorce laws would weaken it. Parsons observed that the increasing structural isolation of the conjugal family from the wider kinship leads to a reduction in support. Burgess and his colleagues argued that the stability of marriage was being undermined by increasing mobility, urbanization, individualization, and pressure on marital partners to adjust to each other (Wagner 1997). While these approaches provided valuable insights into the possible causes of the increase in divorce rates, an analytical multi-level model is needed to explain macro-level variables like the divorce rate. It is, therefore, necessary to distinguish between macro- and micro-level theories, and theories that link these two levels. In sociology, such an analytical schema has been developed by Coleman (1986), and was recently adapted to demography by Billari (2015), who presented a “two-stage view of demography” (Fig. 3.2). Studies conducted at the macro level cannot explain population change, but they can provide novel empirical evidence. The causal relationships that underlie these relationships may be more fruitfully explored at the micro level. Thus, to explain how population structures develop and change, the investigation of actions and interactions at the micro level is needed. The two levels are linked through two types of causal mechanisms: situational mechanisms and transformational mechanisms. In the case of divorce, the transformational mechanisms are an aggregation of individual or dyadic divorce decisions that result in a certain divorce rate at a certain historical time (Fig. 3.2).

Fig. 3.2
A cyclic diagram includes two stages of discovery and explanation. It includes situational and transformational mechanisms at micro and macro level.

The two-stage view of demography (Billari 2015; original Coleman 1986)

Fig. 3.3
A model diagram illustrates marital instability. It includes investment, marital quality, or commitment, which is affected by dyadic rewards and costs and causes separation, and divorces, which in turn are influenced by external rewards, costs and constrain the desire to remain married.

Exchange theory on marital instability. (Source: Lewis and Spanier (1979, 1982), modified)

2.1 A Micro Model of Marital Instability

A starting point of every attempt to explain historical trends in divorce rates should be a basic decision model of marital instability. A decision model that is widely used to explain marital instability is derived from exchange theory (Thibaut and Kelley 1961; Levinger 1979; Lewis and Spanier 1979; for more details, see Åberg 2009). A marriage is considered as an exchange relationship (Fig. 3.3). Partners exchange different kinds of material and immaterial resources, and this exchange results in a certain level of mutual attractiveness. In particular, partners’ marital investments result in a high level of dyadic commitment (Rusbult 1980). It is, however, assumed that the partners compare the benefits of an existing partnership with the expected benefits of alternative relationships or living arrangements. “The more attractive alternative is not necessarily another lover; it may be going it alone or living in groups other than a nuclear family” (Levinger 1979: 37f.). A marital relationship is dissolved if the benefits of alternative living arrangements exceed the benefits of the existing relationship. The subjective cost-benefit ratio that is applied to an existing marriage is often denoted as marital quality, which is considered a central explanatory factor of marital instability. Marital satisfaction is often used as an indicator of marital quality. Internal investments and barriers may increase the material, symbolic, and affectional costs associated with a dissolution or divorce, and these costs are taken into account in such decision processes (Levinger 1965). The sources of these barriers are located external to the individual, but are imposed on the individual, and may include feelings of obligation to dependent children or to marital ties, moral religious proscriptions or external pressures from primary group affiliations (e.g., kin), community stigma (e.g., rural-urban), and legal and economic barriers (see Levinger 1965: 24f.). The basic exchange model used to explain marital instability differentiates between internal and external events or factors that determine a marital breakup, and includes three proximate concepts: marital quality, alternative attractions, and barriers. Most of these decision models do not distinguish between the decision to separate and its legitimization through divorce. As the separation is the crucial and decisive step of the dissolution process, we modified the original model by Lewis and Spanier (1979, 1982). However, the timing of and the explanatory factors for the two events might differ. For example, whether a separated couple is likely to divorce might depend on the costs of a divorce and the strength of external barriers, particularly the restrictions imposed by current divorce laws.

Fig. 3.4
A line graph illustrates the divorce rate versus time duration from 1960 to 2015 for E U and the United States of America. It indicates a higher divorce rate in the United States of America.

Crude Divorce Rates in the USA and EU-28 Countries (No. of Divorces per 1000 Population), 1960 to 2015. (Source: Clarke (1993, 1995), EUROSTAT (2017), National Center for Health Statistics (n.d.), Plateris (1978)

There are other micro models of marital instability that are extensions or variations of the social exchange model. A very prominent theory is the household economy model and its application to marital instability (Becker et al. 1977). This model assumes that couples separate when the utility they expect to derive from remaining married is lower than the utility they expect to derive from divorcing. Becker argued that a couple will seek to maximize the utility function of the household, and that task specialization in a marriage results in efficiency gains. When the skills of men and women become similar and women’s earnings increase relative to men’s earnings, the benefits the partners derive from marriage decrease, and, as a result, divorce rates increase. According to Becker, marriages are dissolved when the unions turn out to be mismatches that occurred because of a non-transparent partner market, an inefficient partner search, or an underinvestment in marriage-specific capital.

An extension of the social exchange model is the framing model (Esser 1993, 2002). Frames are defined as mental models of a situation or basic orientations in a certain situation. Esser (2002) differentiated between two frames: the framing of a marriage as a “good” marriage; and the framing of a marriage as failed. If the frame of a “good” marriage is valid, the partners do not evaluate their relationship, and thus do not rationally compare the benefits and the costs of their relationship. The dissolution process starts if one of the partners “defines” his or her marriage as “non-satisfying” or “bad.” The partners then start to evaluate their chances of finding a new partner, or to calculate the costs of a divorce. The triggering event for a re-framing is a marital crisis in which the partners select a certain frame. The framing model of divorce is an example of a theory that combines rational decision models with action models rooted in symbolic interactionism.

Factors that indicate or change marital quality, alternatives, barriers, or marital investment levels are called divorce risks. These risks include socioeconomic factors, such as financial and social resources; and cultural factors, such as values, attitudes, and orientations at the individual and at the dyadic level.

2.2 Hypotheses

The historic trends in rising divorce rates can be explained by the following four hypotheses:

  1. 1.

    The quality hypothesis states that increasing divorce rates are a consequence of declines in the quality or the benefits of marriage. It has also been argued that increasing divorce rates are attributable to decreasing levels of investment in marriage; i.e., that partners who invest less in their marriage are less committed to it. It might therefore be argued that the more resources the partners keep for themselves, the more likely they are to be able to afford the costs of separation. Some scholars have observed that the types of resources that are exchanged in marriages change over time.

  2. 2.

    The hypothesis of decreasing barriers assumes that an increasing divorce rate is a consequence of a downward trend in the barriers to divorce. When the barriers decrease, the external costs of divorce or separation are reduced. For example, a liberalization of the divorce law or a decline in the stigmatization of divorcees is likely to reduce the symbolic costs of a divorce. If these symbolic costs are low, dissatisfied partners will need fewer resources to separate or divorce.

  3. 3.

    The hypothesis of an increasing legitimization of separation states that the length of time between separation and divorce, or the likelihood that a separation is followed by a divorce, might differ over historical time. The increased liberalization of divorce or the legitimization of marital separations might also alter this dynamic. But another potential explanation for shifts in this pattern is that an increasing proportion of separated partners are aiming to remarry. There is very little discussion of this perspective in the divorce literature (König 1978). It is, however, possible that the separation rates have been stable over historical time, but the divorce rates have not.

  4. 4.

    The opportunity hypothesis argues that divorce rates increase when alternate attractions (Levinger 1965) become more accessible to men and women living in a partnership, and the costs of entering into alternative living arrangements decrease. The availability of alternative partners as an important factor in the breakup of partnerships is also emphasized by the so-called macro-structural opportunity theory of marital dissolution (South et al. 2001). As divorce rates rise, the opportunities to find an alternative partner increase. Such a self-perpetuating process can also be driven by the intergenerational transmission of divorce risks. Moreover, alternative opportunities can emerge if the expected additional lifetime that could be used to find a partner is extended. Thus, the more time a person has to find a partner, the better are the alternative opportunities.

2.3 Two Types of Social Change: Cultural and Socio-Structural Change

These four hypotheses focus on possible changes in divorce risks that are likely to be embedded in two types of societal change: cultural and socio-structural change. Approaches that refer to cultural change may, for example, focus on the values of the partners, the meaning of marriage, the partners’ preferences and knowledge, and the external normative constraints. Approaches that refer to socio-structural change typically focus on individual resources, the distribution of these resources between the partners, and the division of labor. Socio-structural and cultural change may lead to a compositional change in the married population, and may change the social environment of married partnerships. Increasing divorce rates have been attributed to both of these broad societal changes (Perelli-Harris et al. 2017).

A cultural change perspective was advanced in the first half of the twentieth century by Ernest R. Groves in his book, “The Marriage Crisis” (Groves 1928). He argued that marriage adjusts to a changing environment, and that the whole purpose of marriage had changed, especially for young people: “(…) many young couples of today do not know what to make of their situation when they find obstacles to their pleasure-hunt cropping up; having entered marriage only for the sake of adding to their happiness, they may feel they have a right to break off their associations since its returns in pleasure are running low (…)” (Groves 1928: 36). Shortly thereafter, Monahan (1940) suggested that “a change in the family mores” might be responsible for the sharp upward trend in the US divorce rate that started at the end of the nineteenth century.

Another broad societal trend is related to socio-structural changes that can be captured under the heading of modernization. At the end of the nineteenth century, scholars were already arguing that the changing roles of women are responsible for rising divorce rates: “Divorces are most frequent where women are most emancipated” (Willcox 1897: 68). Important aspects of modernization are the expansion of higher education, welfare improvements, higher rates of female employment, and improved mobility. Balestrino et al. (2013) argued that in a post-industrial society, marriage is replaced by the welfare state and the market as the main providers of insurance and other commodities. “The only thing that may still make worthwhile to marry someone is the possibility to enjoy romantic love or at least an affectionate relationship. As a consequence, mismatches are not tolerated (…).” (Balestrino et al. 2013: 4). In classical sociology, it is assumed that modernization goes hand in hand with the differentiation of the society. This differentiation includes a “functional loss” of or a “functional change” in marriage. Thus, as people increasingly expect the exchange in a married partnership to be emotionally satisfying, the economic and instrumental functions of the marriage lose their significance. De Graaf and Kalmijn (2006a) call this a “psychologization process.” To profit from an emotionally satisfying partnership, being in a non-marital partnership or a living-apart-together relationship is sufficient. It therefore appears that the meaning of marriage has changed, and that the heterogeneity of living arrangements with a partner has increased.

However, cultural change and socio-structural change (e.g., modernization) are not independent of each other. It has, for example, been observed that the level of modernization is positively correlated with the moral justifiability of divorce (Dülmer 2014). Inglehart and Welzel (2005) argued that in the post-industrialized era, people place more value on self-expression, and assume that having a strong family is no longer necessary for survival (Inglehart and Welzel 2005: 7).

There are several more specific models of cultural or socio-structural change that have been proposed to explain social changes in marriage and the family. The theory of the second demographic transition is a demographic and family change model that was formulated in 1986 (see Lesthaeghe 2014). It argues that the second demographic transition, which started in the 1960s, was fueled by a contraceptive revolution, a sexual revolution, and a sex revolution. The first revolution led to the adoption of efficient contraception; the second revolution led to a decrease in the age at first intercourse; while the third revolution led to a suspension of the male breadwinner model. It was further assumed that these revolutions were accompanied by increasing individualization, secularization (van de Kaa 1987), and post-materialism (Lesthaeghe 1992). Whereas the first demographic transition was characterized by low divorce rates, the second demographic transition was characterized by a rise in divorce rates and in earlier divorce (Lesthaeghe 2014). A crucial element of this model is the assumption that there is no state of equilibrium. Instead, it is assumed that new living arrangements will emerge, fertility rates will remain low, and the population will decline if it is not supplemented by immigration. Finally, the theory treats cultural factors as exogenous factors that stabilize social and economic trends.

Another prominent approach focuses on the long-term deinstitutionalization of marriage (Cherlin 2004). According to this model, some of the factors that drive this deinstitutionalization process are identical to the factors that have been identified as elements of the second demographic transition. Deinstitutionalization is defined as a weakening of social norms that shape the partners’ behavior, accompanied by the emergence and spread of new living arrangements as alternatives to marriage. This process can occur in combination with several other trends associated with cultural change, such as shifts towards individualization, self-actualization, and higher levels of personal autonomy. The process of deinstitutionalization is accompanied by decreasing pressure on partners to marry or to stay together.

A combined cultural and socio-structural approach that focuses on the female revolution and increasing gender equality has been proposed by Esping-Andersen and Billari (2015). At the beginning of the process, the traditional family and the male breadwinner model prevail. Driven by the female revolution, and especially by the increasing participation of women in education and employment, a diffusion process of gender egalitarianism takes place that leads to a transition phase with more normative disorder and uncertainty. In this phase, divorce rates increase sharply. A new equilibrium can emerge only if two conditions are fulfilled: first, a critical mass of people have to promote gender equality; and, second, families have to adapt to new expectations. This process leads to a new phase of developed gender egalitarianism, which then becomes a dominant part of the social order. This new normative order is associated with a decline in divorce rates. The result of this diffusion process is a (reversed) U-shaped association between divorce rates and indicators of gender equality.

A very specialized approach refers to so-called self-reinforcing processes. According to this approach, the likelihood of a single divorce depends on the divorce rate or the proportion of divorcees in the couple’s social environment. It is assumed that individual decisions to perform a certain behavior depend on the number of individuals who exhibit this behavior. This is also the central assumption of threshold models (Granovetter 1978). Another potential self-reinforcing process is the intergenerational transmission of divorce; i.e., a process in which divorce rates are transmitted from generation to generation. Heaton (2002) argued that the intergenerational transmission process has created a feedback loop.

Finally, selectivity processes might be relevant, as certain macro conditions that are perceived as jeopardizing marriage might prevent people from marrying. Under such conditions, it is likely that only people who are in stable relationships will choose to marry. This tendency might lead to a compositional change in the married population, which could in turn result in decreasing divorce rates (Martin and Bumpass 1989; Kennedy and Ruggles 2014).

3 Empirical Findings

In this section, we will focus on empirical micro-level studies that examined to what extent period or cohort effects are explained by marital quality, alternative opportunities, or barriers to marriage. Micro studies designed to investigate changing divorce rates are often performed in three steps. The exploratory first step is to analyze whether certain factors are related to the divorce rate. For example, many studies have investigated whether women’s employment status is significantly related to the divorce rate. Recently, numerous studies on the socio-structural determinants of divorce have been published. If such a divorce risk is identified, a second step is needed to check whether the prevalence of this factor has changed over historical time, or whether the direction or the strength of the association between this factor and the divorce rate has changed over historical time (Heaton 2002). In the last step, the question of whether period or cohort differences in divorce rates can be explained by this factor or its association with the divorce rate is answered.

3.1 Explanation of Cohort and Period Effects

Period effects are proxies for current influences, whereas cohort effects are based on past influences. The question is not whether these effects really exist, but which type of effect is stronger. When the period effects in a given year are strong, they may change the historical trend in divorce rates. These effects can influence marriages of any duration at a particular point in time as a consequence of events such as the start of a severe economic recession or a war, or a change in the divorce law. Monahan (1940) called this explanatory perspective the “Situational Aspect”. He further introduced the “Biographical Aspect of Divorce”, stating “that the time at which a marriage is contracted has important bearing upon the success of the marriage” (Monahan 1940). Here, Monahan identified cohort effects. Cohort effects are assumed to exist if marriages are more or less vulnerable from the start, because, for example, people are marrying during periods in which the partner market is unfavorable, resulting in a high proportion of unfavorable partner matches. Cohort effects have thus been described as a “kind of moving average of period effects” (Wils 1990). Two types of cohort changes can be differentiated: compositional changes and behavioral changes. Compositional effects can result from a change in the prevalence of divorce risks among the married. Behavioral effects occur if the size or the direction of individual risk factors varies according to cohort membership.

One of the first studies that investigated whether historical divorce trends can be described as period or cohort effects was conducted by Carlson (1979) for the US. He argued that the baby boom cohorts are responsible for rising divorce rates. Preston and McDonald (1979) also found evidence of strong cohort effects, while Ono (1999) later showed that the rate of marital dissolution varies by period and marriage cohort. These scholars avoided the identification problem by using direct measures for one of the three time dimensions. The findings of Carlson (1979) were not confirmed by Thornton and Rodgers (1987), who argued that period effects dominated cohort effects. Thornton and Rodgers (1987) found that in the US, during the historical periods of the 1930s and 1970s, rapid changes in the divorce rates affected all marriage cohorts, birth cohorts, and age and duration groups in similar ways. Similarly, Lutz et al. (1991) showed for Finland that “period rather than cohort effects have dominated the increase in divorce.” They used aggregate data (Finnish vital statistics) and estimated a duration-period-cohort (DPC) model that isolates the factors that are likely to affect divorce rates. However, it cannot be ruled out that the observed period effects simply reflect fluctuations of a long-term trend that could be explained by either cohort or period effects (Thornton and Rodgers 1987).

A serious disadvantage of using period measures has to do with the timing of separation or divorce in the life course. On the one hand, if divorce is postponed to higher ages, divorce rates will be underestimated. Scholars have found evidence of a “gray divorce revolution;” i.e., an all-time high in the rate of later life divorces (Brown and Wright 2017). In light of this trend, it is important to keep in mind that measures may be influenced by timing effects. If divorce increasingly occurs later in marriage, period divorce measures will be misleading (Martin and Bumpass 1989: 39; Kennedy and Ruggles 2014). On the other hand, during periods when divorce rates are rising sharply, divorce is very likely to take place earlier in the life course, which can lead to an overestimation of the “real” divorce intensity (Arránz Becker 2015: 530). It thus appears that the increase in divorce rates can be partly explained by an accumulation of divorces in the younger age groups, and that the leveling off of or the decrease in divorce rates may be due to a delay effect.

Many scholars have summarized these findings by concluding that period effects are stronger than cohort effects (Cherlin 1992:32; Härkönen 2014; Lyngstad and Jalovaara 2010; Villiger 2017). However, this statement is not fully justified. Most studies that considered changes in the divorce rates controlled for the year of birth, the year of marriage, or the historical period. These types of studies generally found that either the cohort or the period variables had strong effects. Typically, clear-cut period effects are observed following changes in the divorce law or recessions. The findings of analyses by Cohen (2014) conducted for the US for the period from 2008 to 2011 suggest that the recession had a negative effect on divorce rates. But the claim that period effects exist might be less plausible when a longer period of time is considered. Previous research based on micro data found either significant period effects (e.g., Goldstein 1999; Heaton 2002; Kalmijn et al. 2004; Ruggles 1997a; South 2001) or significant cohort effects (e.g., Diekmann and Engelhardt 1999; Diekmann and Klein 1991; Teachman 2002; De Graaf and Kalmijn 2006b; Härkönen and Dronkers 2006; Wagner et al. 2015); or considered both cohort and period measures (e.g., Salvini and Vignoli 2011; Villiger 2017; Wolfinger 2011).

3.2 Divorce Risks

3.2.1 Marital Quality and Its Predictors

In this chapter, we will look at studies that examined indicators of the partners’ evaluations of their marriage, like marital satisfaction; as well as interaction variables and socioeconomic and cultural variables that are likely to affect marital quality. It is well established that marital satisfaction is one of the strongest predictors of marital dissolution and divorce, and that the role of marital satisfaction tends to be especially large when the barriers to divorce are low and there are attractive alternative opportunities. In Italy, for example, the barriers to divorce are high, which leads to large numbers of separations and few divorces. It can be argued that increasing divorce rates result from a continuously increasing proportion of unhappy marriages.

The first studies that examined the influence of changing values or attitudes on separation and divorce were not carried out until the 1980s. Changes in the value placed on self-realization or in attitudes about the importance of marriage might affect how much partners invest in their relationship, which could in turn have an impact on marital quality. Thornton (1985) showed that while being affiliated with Catholicism or fundamentalist Protestantism is associated with reduced approval of marital dissolution, attitudes toward dissolution have little influence on subsequent rates of dissolution. Amato et al. (2003) investigated changes in marital quality in the US between 1980 and 2000. They found that although there were no significant changes in marital happiness and divorce proneness over this period, there was a significant decline in marital interactions. They concluded that some trends lead to increased marital quality, while other trends lead to decreased marital quality (Amato et al. 2003). In light of these findings, it is unlikely that changes in marital quality help to explain rising divorce rates. However, since divorce rates have been shown to be lower among couples characterized by cultural homogamy or a shared religious affiliation (e.g., Kalmijn 2007; Wagner and Weiß 2003), it is evident that cultural variables influence the stability of marriages.

The crucial question is whether cultural variables or the prevalence of “bad” marriages can explain cohort or period effects. Kalmijn et al. (2004) have shown for the Netherlands that having emancipatory values positively affects dissolution rates. But strong period effects persist if this variable is included in multiple regression models. Esser (2002) showed that a variable that indicates whether the respondents have ever experienced a marital crisis contributes substantially to the explanation of cohort effects on divorce rates. This finding provides some support for the framing model. So far, the popular thesis that individualization processes are responsible for increasing divorce rates has not been examined.

Age at marriage has been shown to be positively associated with marital stability. It appears to be among the most important predictors of marital stability, independent of the historical period (Thornton and Rodgers 1987; Martin and Bumpass 1989; Lampard 2013). A higher age at marriage is clearly associated with higher marital quality, as factors such as the partners’ maturity, vulnerability to rapid changes during (late) adolescence, competence to take on marital roles, partner search time, and financial resources vary depending on their age at marriage, and should influence the quality of their marital interactions and satisfaction levels (Amato et al. 2003; Abalos 2017).

While many studies have shown that age at marriage does not explain increasing divorce rates, there is some evidence that it accounts for decreasing divorce rates, at least in the US. Although Goldstein (1999) showed that age at marriage does not explain the levelling off of divorce rates, Heaton (2002) stated: “All of the decline on dissolution can be accounted for by the rising age at marriage.” Rotz (2016) argued that the increase in the age at marriage among women is the main proximate cause of declining divorce rates in the US (see also Kennedy and Ruggles 2014).

The educational level of the partners is an important individual resource, as it is a proxy for economic, cultural, and social capital, and is related to holding more liberal opinions. Highly educated men and women have better labor market opportunities, but they also can better afford the costs of a separation than their less educated counterparts. The educational level is a standard variable in studies that focus on multiple divorce risks. No study has found that cohort or period effects are reduced if the educational level is controlled for (e.g., Wagner et al. 2015), and there is no empirical evidence that educational attainment among women helps explain why divorce rates in the US have levelled off (Goldstein 1999). Nevertheless, there is some empirical evidence supporting the so-called “trendsetters’ hypothesis” (Salvini and Vignoli 2011), which states that members of the higher social strata were more likely to divorce in earlier historical periods, and that members of the lower social strata followed or are currently catching up. Using Swiss register data, Villiger (2017) showed that the divorce rate increased more among the lower educational strata than among the highest educational stratum. For the US, Martin and Bumpass (1989:43) found that the “likelihood of separation increased at lower education levels,” and that education “has become a more important independent factor affecting marital disruption.” Heaton (2002) and Raley and Bumpass (2003) found similar results for the US. Härkönen and Dronkers (2006) reported that the educational gradient became increasingly negative in many European countries and the US. Bernardi and Martínez-Pastor (2011) found for Spain that the effect of education was weaker during the historical period that followed the reform of divorce law. An analysis for the Netherlands showed that the effect of education changed from positive to negative (De Graaf and Kalmijn 2006b). In sum, there is a clear evidence of a behavioral effect of education on the divorce rate (Matysiak et al. 2014).

A nearly classical explanation for increasing divorce rates attributes this trend to the rise in female employment and the associated decline in the sexual division of labor. First, it has been observed that female employment improves the financial resources of one or both partners, which might in turn lead to increased marital quality and a decreased likelihood of separating. Ruggles (1997b) discussed the mechanisms that link female employment and marital instability. One of these proposed mechanisms rests on the assumption that increasing female employment makes it less likely that couples will have a traditional division of labor. Less mutual dependence may, however, be associated with decreased marital quality (“interdependence hypothesis”). Another potential mechanism is based on the assumption that if the female partner is employed, the financial risks of separating are reduced (“economic-opportunity hypothesis”), and the costs of and the barriers to divorcing are also lower (Ruggles 1997a; Preston 1997) (3). Yet another argument is that when a woman is not a housewife or a full-time mother, her (time) investments in the relationship may be reduced (van Damme and Kalmijn 2014) (4).

The results of empirical research on the effects of female employment on marital instability have been highly contradictory (Oppenheimer 1997). It is unclear whether female employment causes marital instability, given that either selection or anticipation might be responsible for the association between the two factors (Stevenson and Wolfers 2007; Vignoli et al. 2018). The findings of macro and micro studies on the association between female employment and marital stability have differed (see Ruggles 1997b), as have the findings of analyses of this relationship across countries and historical time. A macro study by South (1985) showed that the female labor force participation rate is positively related to the divorce rate. South (2001) later found that the positive impact of female employment on marital dissolution increased over historical time. Research for Spain indicated that the effect of female employment on marital dissolution decreased after 1981, when there was a liberalization of divorce law (Bernardi and Martínez-Pastor 2011). While Ruggles (1997a) found that male and female employment and separation/divorce are associated at the district level, he acknowledged that because his analysis was based on aggregated data, the causality of this relationship was unclear. He thus observed that “the rise in both female market-labor participation and marital instability could be the result of attitudinal changes. The decline of patriarchal authority within the home and the increase of individualistic values stressing self-fulfillment may have resulted in a decline of social sanctions against both marital breakup and female work” (Ruggles 1997a: 464). Under this scenario, female work and marital instability may no longer be causally linked. Kalmijn (2007) found that women’s employment is associated with high divorce rates in Europe. Killewald (2016) showed for two marriage cohorts in the US (married before and after 1975) that the divorce rate is not affected by female employment, and that the wife’s financial situation does not matter. Instead, the results indicated that even for the cohort born after 1975, the breadwinner model is still relevant, as the probability of a divorce is significantly reduced if the husband is fully employed. These findings are partially in line with those of Vignoli et al. (2018), who found that female employment is not associated with marital stability in Germany and Hungary. Moreover, there is clear empirical evidence that in many countries, female employment is not related to marital stability, and that female employment does not account for the cohort effects on the divorce rate (Wagner et al. 2015).

3.2.2 The Hypothesis of Decreasing Barriers

Cultural changes, and especially changes in values, lead to a liberalization of marriage norms, which is in turn likely to reduce divorce barriers. While secularization processes might result in a liberalization of religious norms, the most fundamental expression of a liberalization of the norms that regulate partnership and marriages is a change in divorce law. There are three major steps in this liberalization process: the legalization of divorce, the introduction of no-fault grounds for divorce (e.g., mutual consent), and the introduction of unilateral divorce (i.e., it is no longer necessary for both spouses to agree to the divorce). Such legal reforms clearly result in a decrease in the barriers to divorce, and thus make divorce less costly. Friedberg (1998) showed that unilateral divorce accounted for 17% of the overall increase in divorce between 1968 and 1988 in the US. Using administrative data from 18 European countries, González and Viitanen (2006), estimated that legal reforms accounted for about 20 percent of the increase in divorce rates in Europe between 1960 and 2002. But the question of whether the increase in divorce rates is mainly attributable to changes in divorce laws has yet to be resolved (González 2014). One of the few studies that investigated the effects of divorce law reform was performed by Kneip and Bauer (2009). The authors showed that in Western European countries, the spread of de facto unilateral divorce practices led to a sustained increase in the divorce rate, whereas the expansion of legal rights to unilateral divorce had only short-term effects. Moreover, they found that divorce rates started rising before the legal changes occurred. Therefore, it is likely that the cultural acceptance of divorce is a third variable that affects both divorce rates and divorce law (Hiller and Recoules 2013).

It has also been argued that irrespective of changes in the divorce law, divorce is becoming more normal. As divorce rates increase, the threshold for deciding to divorce decreases (“threshold hypothesis,” see De Graaf and Kalmijn 2006a). An examination of cohort differences in divorce motives in the Netherlands revealed that citing extreme forms of behavior as reasons for divorce (e.g. infidelity, physical violence, drug and alcohol abuse) has become less common (De Graaf and Kalmijn 2006b).

3.2.3 The Opportunity Hypothesis

South (1985) performed a time-series analysis. He argued that the divorce rate in a given year is determined by the level of divorce in the preceding year; i.e., the higher the level of divorce is, the more liberal the climate surrounding divorce is assumed to be, and the larger the pool of marriageable partners is likely to be. While such findings seem to support the assumption that self-reinforcing processes contribute to increasing divorce rates, they do not explain decreasing divorce rates.

Furthermore, only a handful of studies have focused on district-level sex ratios and their connection with individual dissolution risks (e.g., South 1985; South and Lloyd 1995; South et al. 2001; Lyngstad 2011). Moreover, the main findings of these studies seem to differ according to the units of measurement and methods of analysis used. Based on register data from Norway, Lyngstad found that unbalanced sex ratios have a small negative effect on divorce rates, and, thus, that the greater availability of alternative partners seems to stabilize marriages at the micro level. One explanation for this finding is that the partner in a disadvantageous partner market tends to increase his or her marital investments, which in turn leads to a higher level of commitment. Especially in the German context, there appears to be a gap in research on local sex ratios and their potential impact on relationship instability (Stauder 2015: 429, Wagner et al. 2015: 224). At least, Obersneider et al. (2018) have shown that there is no significant effect of unbalanced sex ratios at the level of German districts on union dissolution. Taken this into account, it is not surprising that there are no studies that investigate the association between changes in divorce rates and sex ratio changes.

3.2.4 The Hypothesis of the Increasing Legitimization of Separation

Separation and divorce: In Italy, a marriage breakdown is much more likely to be indicated by a separation than by a divorce (Vignoli and Ferro 2009). However, there is very little research on the relationship between separation and legal divorce. The event of separation can be more harmful and traumatic for couples than the subsequent legalization of the separation (Cherlin 1992). A study for France found that the length of time between separation and divorce changed little between 1968 and 1985 (Leridon 1990). Martin and Bumpass (1989) estimated for the US that five percent of each marriage cohort separate without divorcing. Recently, Bennett (2017) argued that the “extent to which a marital separation is followed by divorce varies considerably among subgroups.” Two years after a separation, 32% of whites have yet to divorce. The period of time between separation and legal divorce tends to be much longer for African Americans than it is for whites. The lower a couple’s socioeconomic status is, the longer the period of time between separation and divorce is likely to be. For Germany, preliminary results have revealed that the length of time between the subjective end of a union and divorce decreases significantly from marriage cohort to marriage cohort. Subjective union dissolution was measured by the question: “How long have you been together with ‘name of the partner’.” Educational level and the presence of children were found to positively affect the length of time between the partner’s definition of the end of the relationship and divorce (Table 3.1 in appendix). While there appears to be a cohort trend toward a decreasing length of time between union dissolution and divorce, socio-structual factors seem to contribute to a postponement of divorce after the subjective end of the relationship. Although these findings do not justify the conclusion that the factors that drive divorce rates are changing, they do indicate that the socio-structural composition of a married couple affects their probability or timing of divorce following a union dissolution.

Table 3.1 Determinants of the length of time between the subjective end of the relationship and divorce (cox regression)a

4 Discussion

The aim of this contribution was to qualitatively summarize the empirical findings on the factors that can influence historical trends in divorce rates. We started with a general model on the relationship between the macro and the micro level, which helps to organize theoretical strategies. Based on exchange theory, we further delineated a micro model of marital stability that distinguishes between the separation of a couple and the legitimization of the separation through divorce. The framing and the household economy approaches provide extensions or variations of the exchange theory. From this theoretical framework, we derived hypotheses that seek to explain changes in divorce rates.

Our analysis showed that the existing research on this topic did not really distinguish between separation and the institutionalization or societal legitimation of separation through divorce. This is likely because in most contexts, the only available data on divorce or separation are from official statistics. Most of the existing micro studies failed to measure different states and events in the separation process, including divorce proneness, the subjective end of a partnership, the splitting up of households, and legal divorce. Thus, the number of unhappy but stable marriages is unknown, and we cannot assume that marital separations are increasing because divorce rates are rising. It might therefore be more appropriate to investigate the conditions under which couples separate, rather than focusing on divorce.

The dissolution process at the micro level of the dyad is likely to be influenced by a range of factors. The need to integrate this process into a multilevel framework that specifies how the macro and the micro level are linked to each other, while also accounting for changing socio-structural and cultural conditions, adds to the explanatory complexity. We have to acknowledge that the explanatory power of certain factors differs across historical time, or is applicable to increasing (or decreasing) divorce rates only. Moreover, it is obvious that the state of empirical research is fragmentary. Much of the existing empirical research is not driven by theory, and uses indicators that have an ambiguous meaning, or that are linked to different theoretical concepts. While official national statistics allow us to trace historical divorce rates back to the nineteenth century, such data cannot be used to perform a more detailed analysis of the social forces that have led to changes in divorce rates. Large-scale projects with a representative database that allow researchers to perform detailed analyses of marital instability were not started in the US or in Europe until the 1970s or the 1980s. Although theory-driven empirical research on this topic is lacking, many plausible hypotheses regarding the factors that have driven the upward trend in divorce rates have been proposed. Unfortunately, very few studies have tried to test these hypotheses.

There is very little empirical evidence that the quality hypothesis holds. There is no reason to believe that rising divorce rates are due to an upward trend in marital conflict, changes in the interactions of the partners, or increases in the partners’ resources. Although many studies have found that these factors are associated with marital stability at the micro level, their results do not prove that these factors are associated with changes in the divorce rates at the macro level. Moreover, many studies have shown that the central socioeconomic variables do not mediate the association between cohort or period variables on the one hand, and separation or divorce on the other.

Nevertheless, we cannot rule out the possibility that the historical increase in women’s educational and employment status has had indirect effects on divorce rates at the macro level. For example, these socioeconomic changes might have induced cultural changes that led to rising divorce rates, such as the individualization of women, the liberalization of divorce law, and a decline in the stigmatization of divorcees. This argumentation points to the relevance of certain diffusion processes, and thus an even more complicated explanation of historical divorce trends. The direction of the causal relationships is, of course, disputable, but the model presented by Esping-Andersen and Billari (2015), which we outlined above, could provide an appropriate theoretical framework for such a diffusion process.

The role of culture in explaining divorce trends has been neglected by empirical research. Cultural shifts such as the deinstitutionalization of marriage, processes of individualization, and changes in values, are often cited in explanations of divorce trends, but the acknowledgement of the importance of these developments seldom results in an appropriate research design. Cultural change could be a strong explanatory factor, as it affects marital stability via marital quality and via external normative barriers. In contrast, socio-structural factors, which are mainly individual resources, generally affect marital stability via marital quality alone. There is considerable empirical evidence that the costs of divorce have decreased. A liberalization of the divorce law that affects the barriers to divorce is accompanied by a diffusion process of liberal ideas and attitudes. The problem with the deinstitutionalization thesis is that using it to explain decreasing divorce rates is not easy. It can, however, be applied in this context if we assume that an increasing age at marriage is a consequence of this deinstitutionalization process, and that a higher age at marriage leads to more stable marriages. Moreover, we can assume that the deinstitutionalization of marriage is accompanied by increasing opportunities to enter into alternative living arrangements outside of marriage.

In the US and in many European countries, divorce rates are no longer increasing. This trend is best explained by a diffusion process modeled by Esping-Andersen and Billari (2015), which focuses on the female revolution and increasing gender egalitarianism. A counterargument to Esping-Andersen and Billari is that changes in the employment or educational levels of women cannot explain cohort or period effects on the divorce rate. The shift in values and norms that accompanies the female revolution may, however, have a stronger effect on marital stability than women’s educational levels or employment status. An advantage of this model is that it can be easily related to the theory of the second demographic transition and the deinstitutionalization thesis.

Modeling feedback loops and self-reinforcing processes could lead to a breakthrough in research on this topic. For a number of reasons, it seems obvious that the more divorced people there are, the more marriages there will be that end in divorce. However, such processes have not yet been examined in detail. A similar argument can be made with respect to the transmission hypothesis. How many couples separate because their parents divorced? Is the intergenerational transmission of divorce risks a process that will eventually reach an equilibrium?

Based on our current knowledge, it is not possible to develop any reliable predictions about future trends in divorce rates. First, micro and macro studies have to be integrated more systematically. Second, cultural and socio-structural explanations need to be combined. It is likely that socioeconomic subgroups differ not only in terms of their individual resources, but in terms of their attitudes, their values, and the meaning they assign to marriage (Seltzer et al. 2005).