Abstract
This paper assesses the impact of three main destabilizing factors on marital stability in Cambodia: the radical reformation of marriage under the Khmers Rouges (KR); the imbalanced gender ratio among marriageable adults resulting from gendered mortality during the KR regime; and, after decades of isolation from the West, a period of rapid social change. Although there is evidence of declining marital stability in the most recent period, marriages contracted under the KR appear as stable as adjacent marriage cohorts. These findings suggest that the conditions under which spouses were initially paired matter less for marital stability than does their contemporaneous environment.
In Cambodia, marriage traditionally follows an extended spouse-selection process, negotiated primarily by the parents of the potential spouses. Like in many other societies, parents’ marital strategy is guided foremost by a concern over the social and economic status of the family their offspring’s marriage associates them with. Parents also pay attention to the individual traits of the potential spouses because they believe that the compatibility of such traits is predictive of the stability of a subsequent union. Shared beliefs about what contributes to marital harmony and to a proper husband-wife relationship, however, are embodied in such compatibility norms, whose efficiency in improving marital stability can hardly be assessed. Even if marriages that deviate from those norms are found to be less stable over time, individuals entering least-favored marriage arrangements are likely to be selected on other potentially disruptive characteristics; moreover, their behaviors might be affected by a self-fulfilling prophecy from people in their immediate surroundings. These concerns are reduced when historical events affect spousal selection for entire marriage cohorts. In the United States, for instance, researchers have linked the greater likelihood of divorce among cohorts who were married during World War II (Preston and McDonald 1979) or the Vietnam War (South 1985) to hastened marriages.
The disruption of Cambodia’s traditional marriage system by the tragic events of the late 1970s was much more drastic. Upon seizing power in April 1975, the Khmers Rouges (KR) swiftly attempted to sever all individual ties, other than those linked to the political hierarchy, by sweeping away the fundamental bases of Cambodian society, such as private land ownership, religious ceremonies, and kinship responsibilities (Carney 1989; Ebihara 1993; Kiernan 1996). To reflect the new collective social organization, the KR organized en masse marriage ceremonies, often without consideration for the families’ preferences and usual matching criteria. Although not all marriages contracted during the KR regime were arbitrarily arranged, the period conditions prevented most marriages from adhering to the typical, careful matching process of previous marriage cohorts.
These marriage practices resumed after the KR regime fell in January 1979, but the short-lived attempt to radically transform Cambodian society left indelible marks, perhaps most conspicuously in its demographic structure. Between 1975 and 1979, executions, epidemics, exhaustion, and starvation killed an estimated 1.5 to 2.0 million people—nearly one-quarter of Cambodia’s 1975 population (Heuveline 1998; Kiernan 1996). The excess of young-adult male mortality over young-adult female mortality also resulted in a low gender ratio among the never-married the population, which may arguably lower women’s probabilities of ever marrying. A low sex ratio may also affect the stability of extant marriages by increasing the remarriage prospects for married men if they were to divorce. Indeed, a cross-national analysis found an association between low sex ratios and high divorce rates (Trent and South 1989).
Although Cambodia’s population sex ratio was again balanced with the gradual addition of new birth cohorts, the next challenge to the stability of the extant marriage system may arise from the end of its political, economic, and cultural isolation. Contrary to the oft-expressed view that modernization destabilizes the institution of marriage and increases divorce and separation rates that were traditionally low, Jones (1997) and Hirschman and Teerawichitchainan (2003) found that marital disruption actually declined in the Islamic countries of Southeast Asia. Even in Thailand, a country whose social traditions more closely resemble those of Cambodia, Hirschman and Teerawichitchainan estimated that the probability of divorce or separation during the first five years of marriage has not increased but instead has remained at the moderately high level of 10% for the first marriage cohorts of 1945 to 1969 (p. 223).
The Cambodian marriage system has therefore experienced three major changes that have been found elsewhere to affect marital stability, but perhaps nowhere experienced as brutally as in Cambodia since 1975: (1) the conditions under which the spouses of the 1975–1978 marriage cohorts were matched, (2) the potentially destabilizing context of a gender imbalance in the marriage market in the post-KR period, and (3) the swift opening of Cambodian society in the most recent period. In this paper, we use retrospective questions from the nationally representative 2000 Cambodia Demographic and Health Survey (CDHS 2000) and from the Mekong Island Population Laboratory (MIPopLab), a demographic surveillance system launched in 2000, to analyze divorce or separation trends over time and across marriage cohorts. In particular, we attempt to isolate the respective impact of these three major changes on the risk of marital disruption.
BACKGROUND
Traditional-Marriage Regime in Cambodia
Cambodia is a Southeast Asian country on the Gulf of Thailand, neighboring (moving clockwise) Thailand, Laos, and Vietnam. Before becoming an independent country in 1953, Cambodia was a French Protectorate and, along with Laos and Vietnam, part of French Indochina. While Cambodia’s modern politics have been closely linked to Vietnam, Cambodia’s marriage regime appears much closer to that prevailing in Thailand than to that in Vietnam, where Chinese influences dominate. The main study of traditional marriage regimes in Southeast Asia suggests that before the introduction of code civil (the legal framework for all civil matter rights, including family law) in the French Protectorate, marriage practices in Cambodia closely fit those embodied in the Siamese Code introduced in Thailand in 1805 (Lingat 1952). One of the main differences with the marriage regimes under Chinese influence noted in that study is that in Cambodia, extended families appeared meaningful only at the highest echelons of society, and marriage was otherwise more commonly destined to form a new and independent elementary unit than to bring another generation to the groom’s extended family. Politically, Frieson (1993:33) maintained that in rural areas, “there are no well defined groups” between the village authority and family households. In practice, both nuclear- and extended-family households are found in rural areas (Ebihara 1993), while nuclear-family households dominate in urban areas (Steinberg 1959).
Either the groom or his parents can initiate contacts with a potential bride’s family, but a matchmaker is often involved in identifying an acceptable match. The bride is typically expected to come from a slightly higher social background than the groom (Népote 1992), and early in the prenuptial approach, parents carefully research the social status of the other party. To achieve a certain sense of harmony, the higher social status of the bride is to be balanced by the individual qualities of the potential groom, such as his education, assiduity at work, or simply abstinence from undesirable behaviors. He is also expected to be older than his bride, but within the same age group, that is, up to roughly seven years older (Népote 1992:135); indeed, the observed age differences between spouses in the 1962 census averaged three years (Migozzi 1973). The norm appears less flexible with respect to marrying an older bride than marrying a much younger bride, which is not favored but is tolerated. Parents also try to assess the compatibility of the potential spouses, in particular by consulting an achar, an elderly religious man who bases his evaluation on the astrological combination of the bride’s and groom’s respective birth timing. Parents are discouraged, however, from marrying a daughter against her will (Lingat 1952).
Once engaged, the potential groom is expected to work for his in-laws (twee bomrae, in Khmer) during a prenuptial period of coresidency of up to a year. This arrangement can be thought of as a form of bridewealth, but it is also intended to test his personal qualities. The wedding is traditionally sealed by the acceptance of gifts from the potential groom to his future parents-in-law. After the elaborate wedding ceremony (Pich 1984), the newlyweds typically continue to reside with the bride’s parents until the parents are satisfied that the new couple can live together as husband and wife. The married couple is then expected to live on their own, with the exception that one of the daughters might remain with her husband in the parental household to help her parents in old age. The families from both sides make commensurate contributions to the establishment of the new household, jointly administered by the husband and the wife. From this point forward, balanced relationships with the relatives on the husband’s side and on the wife’s side appear to set in. Cambodian kinship is thus typically described as bilateral, and the transitional cohabitation with the bride’s parents is seen as a residual practice from the earlier matriarchal civilizations that were once prevalent in the region (Coedès 1971 [1944]), but the characterization of kinship as bilateral remains contested (Népote 1992).
Lingat (1952) described traditional marriage customs as placing a woman under the authority of her husband after they have established independent living. A certain balance is again found in the wife’s protection against excessive authority in that she is allowed to seek a legal divorce unilaterally with a relative ease, whereas a husband can do so only if his wife has been unfaithful. His alternative is simply to leave. If he leaves, he forfeits his rights to the family’s assets that were acquired since marriage, whereas in the case of legal divorce, those assets are divided equally between the spouses. The easy road to divorce for women may explain the moderately high level (about 10%) of divorce and separation in the first five years of marriage for the first marriage cohorts of 1945 to 1969, as observed in Thailand (Hirschman and Teerawichitchainan 2003:223).
Although women appear to enjoy relative autonomy in accepting a potential husband or leaving a current husband, there are potential costs in exercising these options. For a first marriage, the fairly long process of spousal selection and a preference for young wives (under age 25) imply that a daughter cannot veto her parents’ choices too often. A Khmer proverb reminds would-be picky daughters that “you should be married before you are called an old maid.” Attesting to the strength of this injunction, the 1962 census showed that only 2% of women aged 50 years and older were never married and that women’s mean age at marriage was 21.3 years (Migozzi 1973:145). As for women who choose to seek a divorce, they are unlikely to be able to remarry, a stigma that does not seem to apply to divorced men. Stark differences between the proportion of men and women who are divorced and widowed are already visible in the 1962 census data (e.g., in the age group 50–54, 7.4% for men and 27.5% for women; Migozzi 1973:245, table 19) and cannot be adequately explained by the three-year difference in the age at marriage or by small differences in mortality.
Shocks to the Traditional-Marriage Regime
The KR revolutionary marriage
The KR entered Cambodia’s capital city, Phnom Penh on April 17, 1975. They swiftly ordered all urban dwellers to leave for the countryside and all foreigners to leave the country. What they subsequently undertook has been described as the most-radical social transformation ever attempted (Kiernan 1996; Weitz 2003). It certainly stands among the deadliest. While the KR leaders systematically arrested and killed suspected political opponents, some local KR cadres reportedly executed anyone who disobeyed even small orders. Many people also died from exhaustion from long days of indentured labor and the insufficient diet provided by the regime; from the epidemics, particularly among laborers who were sent to clear malaria-endemic forests; and from the absence of modern medicines. Extant estimates of deaths range anywhere from 500,000 to 3 million, but the figures that are the most compatible with subsequent demographic data are in the range of 1.5 to 2.0 million excess deaths (nearly one-quarter of Cambodia’s population in April 1975) for the three years, eight months, and 20 days of the KR regime (Kiernan 1996; Heuveline 1998).
The KR’s attempt to radically transform Cambodian society included a frontal attack on the family, which it saw as the core institution of social reproduction. Meanwhile, the powerful familial imagery was “recycled” in describing (1) the intended new society as a one-family society, (2) the higher echelon of the political structure, the Angkar, as providing for the people and requiring their allegiance the way parents cared for and exerted their authority over their children, and (3) the political leader Pol Pot as “brother number one” among the people, that is, the first-born and, as such, the most respected sibling. People were reminded that they were dependent on the Angkar and that their survival was contingent on the benefits they provided to it. Orders were often uttered with a barely concealed death threat: “There is no gain in keeping you; no loss in taking you out.” Thus, local KR cadres controlled and reformed every aspect of daily life, and marriage was no exception.
In spite of the regime’s admonition to end young adults’ deference to their parents, there is ample albeit impressionistic evidence that respect of the elders was not easily uprooted from Khmer culture. Historian Ben Kiernan (1996:215) even suggested that while the KR carried on “systematic assaults on peasant ties to land, family, and religion … it was [their] attack on the family that alienated peasant supporters.” Nevertheless, the separation of family members by age and sex work groups, the long work days, and the abolition of religious ceremonies drastically curtailed family opportunities for matchmaking. Forced marriages have also been reported to have taken place at the time, although their prevalence is unknown (Ponchaud 1998 [1977]:160–161, 294–295; Ngor 1987:292). KR cadres’ marriage offers could not be refused, and it is also possible that toward the end of their reign, the KR sought to respond to depopulation by enforcing marriages. Data we present later provides additional qualitative evidence and, to our knowledge, the first quantitative evidence on marriage under the KR.
Gender ratio in post-KR Cambodia
The KR’s impact on family ties was temporary, and as soon as their regime collapsed in early 1979, many displaced people went on the road searching for family members. When families were again able to resort to their traditional marital strategies, they had to operate in a very different “marriage market,” however, as the strong gender differences in mortality during the KR years resulted in a ratio of only 75 men per 100 women over age 15 in 1980 (Huguet 1992). As is often argued in the case of African Americans (Wilson 1987; for empirical evaluations, see Bennett, Bloom, and Craig 1989 and Lichter et al. 1992), these conditions may have resulted in a shortage of eligible men relative to the number of eligible women, thus reducing the likelihood of marriage for never-married women. In Cambodia, however, the strong stigma against never marrying seems to have induced rather different changes, and an overwhelming proportion of each female cohort still married at least once. Under these conditions, though, parents with unmarried daughters could not be as demanding of their potential sons-in-law, and anecdotal evidence suggests that they also became more aggressive about matching their daughters, at times initiating the process themselves, contrary to the tradition.
A possible way for the low gender ratio to have been accommodated while maintaining almost universal female marriage is, of course, by an increase in the average number of marriages per adult male through increases in divorce rates and in gender differences in remarriage rates. In a cross-national analysis of 66 countries, Trent and South (1989) found an association between low population male-to-female ratios and high divorce rates. This proposition may have to be qualified for Cambodia because divorce was traditionally easier to obtain for women than for men (Lingat 1952). Although a shortage of marriageable men may have the opposite effect on divorce in this context, the stigma attached to remarriage for divorced women suggests instead that women’s demand for divorce may be rather insensitive to the supply of marriageable men. On the contrary, because remarriage was generally accepted for divorced men, the relatively larger pool of marriageable women may have increased their propensity to seek divorce or simply to abandon their current marriage. The most recent census data, from the 1998 General Population Census, confirm the continued gender difference in the likelihood of remarriage for men and women. Over age 15, 1.1% of men and 8.0% of women were widowed, divorced, or separated. Moreover, the proportion of men who are widowed, divorced, or separated increases only slightly, from 1.0% among 20- to 24-year-olds to 3.4% among 50- to 54-year-olds; the corresponding proportion for women increases rapidly with age from 3.8% among 20- to 24-year-olds to 30.4% among 50- to 54-year-olds (National Institute of Statistics 1999a: table A1).
Modernization in Cambodia
Although Vietnamese troops captured Phnom Penh within weeks of entering the country and promptly established a new government made up of Cambodians, the KR continued to fight from the forest-clad hills at the Thai and Laotian borders. The Vietnamese troops remained in Cambodia for 10 years. During these 10 years, peace was gradually restored to most provinces, but land mines continued to kill farmers, among others, returning to claim land. Living conditions nevertheless improved overall, with the setting up of administrative structures and infrastructures, particularly health care and education, along the Vietnamese model. Gender differences in mortality were sharply reduced, and because they had been much more pronounced among adults than among children (Heuveline 2001) under the KR, the gender ratio in the prime age groups for marriage (15 to 30 years) returned close to parity in the early 1990s.
During the decade of the Vietnamese presence, Cambodia’s international economic and cultural ties remained largely limited to the Soviet-bloc countries, as few other countries recognized a government deemed to have been installed by an illegitimate foreign invasion of a sovereign country. The KR retained the Cambodian seat at the United Nations, for instance, although its increasingly clear legacy made its restoration an unpalatable alternative to the Vietnamese occupation. The political impasse ended with the 1991 Paris agreements and the United Nations Transitional Authority in Cambodia’s (UNTAC) organizing of multiparty elections in 1993. The process of political normalization engaged at that time ushered in a period of rapid economic changes that may represent the most recent in a series of shocks to traditional marriage. Because the country was also allowed to rejoin the international economic community, particularly because the United States dissociated its economic policy toward Cambodia from its economic embargo on Vietnam, Cambodia’s reconstruction and a cheap labor force began attracting international aid and foreign investments.
Although better living conditions are expected to relieve some of the marriage stress-ors that may lead to separation, the development of a wage labor market is also expected to improve the alternatives to married life for either spouse, but especially the partner who is the most financially dependent on the other. Empirical evidence supports both arguments, which are referred to as the income effect and the independence effect, respectively (Hannan, Tuma, and Groeneveld 1978). Following Becker’s theory of marital instability (Becker, Landes, and Michael 1977), however, the independence effect has been more thoroughly tested. Huber and Spitze (1980) even expanded Becker’s theory by demonstrating that the likelihood that a wife will consider divorce increases with her work experience. Perhaps the most conspicuous sign of the expansion of work opportunities for young Cambodian women is the spectacular development of the garment industry, which between 1997 and 1999 alone added about 120,000 new jobs (or 2% of the total workforce), mostly near Phnom Penh (Chea and Sok 2001). Comparing annual waves of the national labor force survey (National Institute of Statistics 2003) reveals that the total economically active population aged 10 years and older in 2001 was 17.6% larger than only a year earlier; in addition, in relative terms, the participation in the agricultural sector had declined to 67% of the male and 66% of the female active labor force, from 71% and 79%, respectively, in 1995.
With new foreign investments and aid also came international visitors: the staffs of international organizations that began to channel aid to the country’s reconstruction efforts, members of the numerous nongovernmental organizations that launched programs in the country, and, more timidly at first, Western tourists. How the influx of international nongovernmental organizations affected government circles has already been documented (Boli and Thomas 1997). To our knowledge, however, no scientific study has documented whether Western cultural representations also had an impact on the customs and traditional behaviors of Cambodia’s mostly rural population, that is, beyond the tangible influences of the West on the elites in the capital city. Increased exposure to Western norms with respect to marriage might have induced more romantic unions, which were already noted in the 1950s, but only as an urban phenomenon (Steinberg 1959).
While the extent of the diffusion of love marriages was unknown at the outset of this study, we might expect arranged marriages to prove more stable than romantic marriages. The objective factors of arranged marriages may well endure longer than the personal attraction factors of love marriages. In addition, couples who are married under such arrangements may be more likely to conform to the normative expectation that they will stay married, regardless of the degree of their marital satisfaction. The latter part of the argument is akin to the “commitment to marriage” explanation: in countries where premarital cohabitation has developed as a way of testing the compatibility of potential spouses, divorce rates have concomitantly increased (Lesthaeghe 1983) and those who marry directly are less likely to divorce than those who cohabit first (Axinn and Thornton 1992; Hall and Zhao 1995; Thomson and Colella 1992). While romantic marriages were found to be less stable than arranged unions elsewhere (Whyte 1990; Xiaohe and Whyte 1990), two studies in Southeast Asia documented decreases in the proportion of marriages that ended in divorce during the recent period of economic growth, urbanization, and higher levels of education and female labor-force participation (Hirschman and Teerawichitchainan 2003; Jones 1997). However, with the exception of Thailand—where divorce rates appear to have remained fairly consistent at a moderately high level—these studies considered Muslim countries with marital arrangements that greatly differ from those of Cambodia.
DATA AND METHODS
CDHS 2000
In this paper, we use primarily analyses of data from the CDHS 2000. Between February and June 2000, the CDHS 2000 interviewed a nationally representative sample of 15,351 women aged 15 to 49 (National Institute of Statistics 2001). All women were asked about their background characteristics, reproductive and contraceptive histories, their own and their children’s health, marital status, sexual activities, fertility preferences, and HIV-related knowledge. Our analyses of marital disruption are restricted to the 10,467 ever-married women in CDHS 2000, who also provided the date of their marriage, and background information about their most recent husbands. Additional questions on women’s status, including choice of spouse, were asked of 2,618 ever-married women in one-fourth of the interviewed households.
The CDHS 2000 is the first nationally representative survey to provide such data on marriage in Cambodia (its predecessor, the 1998 National Health Survey, was more limited in geographic and substantive scope). Although it represents a unique research opportunity, the CDHS 2000 is primarily concerned with fertility and health, and it suffers from a few limitations for analyzing marital stability. First, women who were widowed, divorced, or separated were not asked the date their marriage ended. All women were asked, however, when they last had sexual intercourse and, if this occurred within the past 12 months, who their partner was. As we further explain in Appendix A, these data confirmed our anticipation that extramarital sex was very rarely reported in Cambodia. Of the 8,918 ever-married women who reported having had sex in the past 12 months, for all but seven it had been with their spouse. These results suggest that regardless of whether extramarital sex is under-reported in CDHS 2000, the reported date of the last sexual intercourse is seldom after the date of marital disruption. Moreover, 97.4% of married women reported having sex in the past 12 months, suggesting that the reported date of the last sexual intercourse is unlikely to be much earlier than the date of marital disruption reported by currently widowed, divorced, or separated women. This is further validated by the finding that for widowed women, the timing of a marriage’s ending is quite consistent with trends in past male mortality.
The second limitation concerns women who were married more than once. The data on last sexual intercourse prevent us from estimating the date their first marriage ended. Furthermore, we know the date of their first marriage only, not that of their most recent marriage. Therefore women who were married more than once need to be excluded from the analysis. The selective removal of divorced and widowed women will clearly bias our estimate of marital disruption downward. As we explain later, however, our analyses are comparative (across cohorts and over time). Thus, the main question is whether the extent of the bias varies across cohorts and periods. Unfortunately, the bias may well vary across cohorts and periods because the risk of a woman’s removal (through remarriage) increases the earlier her previous marriage ends (the longer her exposure to the risk of remarriage); marriages among earlier cohorts are more likely to end before a given date (to have a longer exposure to the risk of divorce or widowhood). This censoring issue is thus a potential concern, but the extent and direction of the resulting bias can be quantitatively assessed by turning to another, independent data source.
MIPopLab
MIPopLab is located in a rural district near the Phnom Penh agglomeration. MIPopLab data are not nationally representative—there is no sampling, and all residents of the study area are interviewed—but this rural district was selected, in part, because the strong urban-rural demographic gradient exhibited by the 1998 census (National Institute of Statistics 1999b) suggested that using this rural district would reveal more or less average demographic indicators. The project is primarily prospective, with a biannual follow-up of all demographic events in the study area, which has almost 11,000 residents. However, the benchmark census of the area included a retrospective survey of family reconstruction in the post-KR period. The survey consisted of a few qualitative focus group interviews that contrasted spouse selection in the past and the present and a quantitative questionnaire, including complete marriage histories for all ever-married women under age 75 at the time of the survey.
While focus group data are used to provide qualitative evidence on marriage practices during the KR, the quantitative data allow us to simulate how the restriction to women who were under age 50 at the time of the survey and were married only once affects the estimates of divorce rates across cohorts in the CDHS 2000. Of course, in terms of their retrospective component, both the CDHS 2000 and MIPopLab are also restricted in that only the marriage experiences of the women who survived to the time of the interview are reported. This introduces bias to the extent that, as observed in other settings, it is reasonable to expect a survival advantage of married women compared with widows and, even more, with divorced or separated women. Unless the extent of these survival differences varies strongly across cohorts or over time, however, all divorce and separation rates should be slightly underestimated, and the comparisons across cohorts and periods should be relatively robust.
Methods
We analyze the risk of marital disruption, through divorce or separation, as a duration-dependent process that starts at the time of marriage. We first use multiple-decrement life-table techniques with divorce and separation as two decrements of interest and widowhood or the interview as censoring events. We then construct the associated single-decrement life tables (Preston, Heuveline, and Guillot 2001), treating divorce and separation as a single event (marital disruption) and adjusting for the censoring events. Finally, we use proportional hazard techniques (Cox 1972) to estimate the extent to which differences in the underlying risk of marital disruption across marriage cohorts are accounted for by selected characteristics of the spouses.
Assessing the impact of the spouse selection on marital stability from wives’ answers to the question on who had chosen their husband is problematic. The question was posed only to a one-quarter subsample of ever-married women, which curtails our statistical power. More important, however, is that such retrospective reports are subject to ex post rationalizations. The causality of any association between divorce and statements about spousal selection at the time of survey might well run both ways, and instrument variables are hard to find. Instead, our argument is that all couples married during the KR were essentially married under conditions that did not allow for what is traditionally considered a satisfactory, thorough matching process. To back that claim, we later present qualitative and quantitative evidence on marital choice during the KR. We thus use the cohort of marriage as an indicator of a different, presumably suboptimal (from the standpoint of families) process of spouse selection. We compare the risk of divorce or separation across five sets of marriage cohorts: before 1975, 1975–1978 (KR years), 1979–1985, 1986–1992, and 1993–1999. Conversely, we compare the risk of divorce in four different periods: before 1979, 1979–1985 (the period with the lowest gender ratio), 1986–1992, and 1993–2000 (the period of the fastest socioeconomic change).
Although we are interested here in both cohort and period effects, in these analyses, as in all age-cohort-period models, we cannot estimate models that would include the year of marriage, the duration of marriage, and the year of observation. Because duration is clearly a critical dimension of the risk of marital disruption, we estimate either cohort-duration or period-duration models. To the basic model (Model 1), Model 2 adds wife’s characteristics, beginning with her region of residence (the country’s five major regions). Her age at first marriage is represented by three dummy variables for early (before age 18), intermediate (18–24 years), and late (25 years and older) age at marriage. We also include her education in four categories representing the highest educational level attained: (1) completed secondary school or higher, (2) completed primary school, (3) attended (but did not complete) primary school, and (4) did not attend school. Women’s employment is also included with one inactive category (does not work) and four active categories (1a, 1b, 2a, and 2b): (1) self- or family-employed versus (2) outside employment (employed by a nonfamily member) in either (a) the agricultural or (b) the nonagricultural sectors. Assessing the impact of women’s employment on marital stability is complicated by the fact that employment is reported at the time of the survey, when it can be as much a consequence as a cause of marital dissolution. To disentangle the direction of causality, we also run a similar model with widowhood as the outcome.
Model 3 also includes husbands’ characteristics as reported by wives. Reported in lesser detail than for women in the CDHS 2000, men’s employment is distributed in five different categories: (1) does not work, (2) agricultural worker, (3) skilled and unskilled worker, (4) white-collar worker (clerical, technical, services, sales, professional, and managerial), and (5) occupation unknown. Husband’s education is categorized relative to the wife’s education (more, less, or equally educated). Models 3 also include a dummy variable for premarital sex.
Finally, Model 4 includes variables on the number of children. Again, reverse causation is highly likely, and the effect of children on marital stability is difficult to interpret: marital harmony may encourage spouses to have additional births as much as extant children may discourage them from ending their marriage. Morgan, Lye, and Condran (1988) argued that considering the gender of the children provides a simple way of disentangling the higher parity that results from marital harmony—a gender-neutral mechanism—from children’s effect on marital stability—which could be gender specific. We thus kept separate counts of sons and daughters, and we report the results in the following five time-varying categories: (1) no child, (2) only one son (no daughter), (3) only one daughter (no son), (4) only daughters (two or more), and (5) at least two children, including one son. This allows for significance tests of the effect of the number of children (none, one, or at least two) and of having at least one son.
Weighting and dropping women with missing answers on these key covariates brings our sample size to 8,911 women, and the models we present are based on only their first 10 years of marriage (73,704 marriage-years). If some of the covariates in the proportional hazard models have a different impact on marital stability at different marriage duration, the comparison of cohorts or periods might be influenced by the fact that long-duration marriages are observed only among the earliest marriage cohorts and in the most recent periods. Tests of the validity of the proportional hazard assumption are satisfactory when observations are truncated 10 years after marriage rather than at the time of the survey.
RESULTS
Spouse Matching in the KR and Other Periods
Table 1 compares the characteristics of wives and husbands across marriage cohorts. On these observable characteristics, there appears to be relatively little change across cohorts. Late marriage appears to become slightly more frequent, with the proportion who married after age 25 increasing from 10.9% among the 1986–1992 marriage cohorts to 16.8% among the 1993–1999 marriage cohorts. (Censoring prevents including earlier marriage cohorts in the comparison because the sample includes only women under age 50 in 2000.) The proportion of women who completed primary schooling increased from about 1 in 10 in the early marriage cohorts to about 1 in 4 in the later ones. Husbands’ education changed concomitantly, resulting in wider spreads in the male and female educational distribution and slight increases in the proportions of both husbands who were more educated than their wives and vice versa, but there remains a strong tendency for husbands to be at least as educated as their wives. This pattern is produced largely by gender differences in educational distributions. (A random mixing pattern with the underlying educational distributions of husbands and wives would yield 16.7% of wives more educated than their husbands, compared with the prevailing 10.7%.) While the proportion of men working in the agricultural sector is stable and close to the national average (65.4%), the proportion of women working in the agricultural sector is lower (59.6%) and declines from the earliest to the latest marriage cohorts. A plausible explanation for this gender-specific decline is that childbearing temporarily keeps the most-recently married women out of the work force, as suggested by the doubling of the proportion that does not work from 1 in 10 women who were married before 1975 to 1 in 5 women who were married between 1993 and 1999. Finally, the number of children is obviously lower among women in the more recent marriage cohorts because they are likely still having children, but ancillary data suggest that fertility is indeed declining in Cambodia (National Institute of Statistics 2001).
Table 1.
Percentage Distribution of Selected Characteristics, by Marriage Cohorts
| Marriage Cohorts |
||||||
|---|---|---|---|---|---|---|
| Variable | Pre- 1975 |
1975– 1978 |
1979– 1985 |
1986– 1992 |
1993– 1999 |
All |
| All Women Who Married Only
Once (N = 8,911) |
10.9 | 9.4 | 27.6 | 29.2 | 22.9 | 100.0 |
| Age at First Marriage | ||||||
| < 18 | 63.7 | 30.9 | 32.4 | 27.8 | 29.5 | 33.7 |
| 18–24 | 36.3 | 65.1 | 60.2 | 61.3 | 53.7 | 56.9 |
| 25+ | — | 4.0 | 7.4 | 10.9 | 16.8 | 9.4 |
| Woman’s Education | ||||||
| Did not attend school | 35.8 | 34.6 | 37.0 | 26.8 | 26.0 | 31.1 |
| Attended primary school | 52.6 | 54.8 | 52.1 | 46.5 | 44.8 | 49.1 |
| Completed primary school | 10.8 | 10.2 | 9.9 | 24.8 | 25.7 | 18.0 |
| Completed secondary school or higher | 0.8 | 0.4 | 0.9 | 1.9 | 3.6 | 1.8 |
| Woman’s Occupation | ||||||
| Does not work | 10.7 | 12.3 | 14.1 | 16.1 | 21.6 | 15.8 |
| Self- or family-employed in agriculture | 62.7 | 62.2 | 57.4 | 56.7 | 55.0 | 57.7 |
| Self- or family-employed in other sector | 20.4 | 21.0 | 22.8 | 19.5 | 16.2 | 19.9 |
| Outside employment in agriculture | 2.1 | 1.9 | 1.9 | 2.1 | 1.8 | 1.9 |
| Outside employment in other sector | 4.1 | 2.6 | 3.7 | 5.7 | 5.5 | 4.6 |
| Husband’s Education Relative to Wife’s | ||||||
| Less educated | 6.6 | 8.1 | 10.9 | 12.1 | 11.7 | 10.7 |
| Equally educated | 43.4 | 48.0 | 49.3 | 46.7 | 42.8 | 46.3 |
| More educated | 50.0 | 43.9 | 39.8 | 41.2 | 45.5 | 43.0 |
| Husband’s Occupation | ||||||
| Does not work | 3.9 | 2.8 | 3.4 | 3.3 | 1.7 | 3.0 |
| Agricultural worker | 63.4 | 67.4 | 66.2 | 64.9 | 65.1 | 65.4 |
| Skilled and unskilled worker | 8.9 | 8.8 | 11.8 | 12.5 | 16.3 | 12.5 |
| White-collar worker | 20.4 | 19.3 | 16.5 | 17.7 | 15.6 | 17.3 |
| Unknown | 3.4 | 1.7 | 1.9 | 1.6 | 1.2 | 1.8 |
| Premarital Sex | ||||||
| Yes | 3.2 | 5.1 | 5.0 | 4.2 | 4.6 | 4.5 |
| No | 96.8 | 94.9 | 95.0 | 95.8 | 95.4 | 95.5 |
| Number and Gender of Children | ||||||
| None | 1.1 | 2.0 | 1.6 | 2.4 | 14.0 | 4.6 |
| Only one daughter (no son) | 1.3 | 1.3 | 2.5 | 4.4 | 21.5 | 7.2 |
| Only one son (no daughter) | 1.3 | 2.4 | 1.5 | 4.2 | 22.3 | 7.1 |
| Only daughters (two or more) | 3.3 | 2.8 | 4.7 | 10.2 | 8.7 | 6.9 |
| At least two children, including one son | 93.1 | 91.5 | 89.6 | 78.9 | 33.6 | 74.2 |
| Region of Residence | ||||||
| Phnom Penh | 10.0 | 8.6 | 9.2 | 9.5 | 8.3 | 9.1 |
| Plains | 49.7 | 46.5 | 42.1 | 42.1 | 44.6 | 43.9 |
| Tonle Sap | 24.5 | 29.1 | 27.9 | 28.5 | 28.7 | 28.0 |
| Coastal | 6.3 | 4.6 | 8.9 | 7.9 | 7.1 | 7.5 |
| Mountain | 9.5 | 11.2 | 11.9 | 12.0 | 11.4 | 11.5 |
Overall, Table 1 displays only modest and continuous changes in the characteristics of spouses across marriage cohorts, rather than stark contrasts between the KR and other marriage cohorts. However, MIPopLab qualitative data provide ample confirmation that like all aspects of daily life, marriage was drastically different during the KR period. The change was perhaps best exemplified by the reform of the wedding ceremony, as a woman who married during this period recalled:
We were sent to a wedding ceremony in the third village’s hall, [name]. More than 50 couples were married then. For the wedding, each couple just held each other hands and pledged to be loyal to Angkar. (Focus group discussion, MIPopLab, 12/16/2000)
This wedding ceremony sharply contrasts with the earlier elaborate two-day affair, and so did marriage criteria. Individuals often married for instrumental reasons, such as to reduce the risk of being drafted in the “mobile work teams” assigned to the most arduous tasks in the most remote areas. For instance, the same woman explained:
My older siblings were all married, but I was single and was sent to work in a working camp far away from them … I agreed to form my own family so that I could stay in the village near them.
In some other instances, women who married during this period reported feeling forced to accept the spouse presented to them, as explained by another focus group participant:
One night, I could not sleep until midnight. Suddenly, I heard the group leaders discussing:
“Will she marry Comrade X or not?”
“If she won’t, there will be no gain in keeping her, no loss in taking her out.”
I heard that with my own ears and told my mother what I had heard. She told me:
“Dear, marry him now … . If later the country changes, we will think it over again …”
I agreed with her and I married him. I was frightened to death then.
[Other women echoing: “It was just like that during that time.”] (Focus group discussion, MIPopLab, 12/16/2000)
While the CDHS 2000 data do not allow us to measure many of the changes introduced by the KR quantitatively, the prevalence of forced marriages can be assessed from the question on spousal selection in the one-quarter subsample. Of the respondents who married during the 1975–1978 period, 32.4% reported that a nonrelative chose their husband and/or that they did not give their consent to the marriage, compared with 9.6% of those who married in other years (see Figure 1). In addition, the proportion of married women who reported that they and their husband “chose each other” increased from 8.9% among pre-1975 marriage cohorts to 23.1% among the 1993–1999 marriage cohorts).
Figure 1.
Distribution of Ever-Married Women Aged 15–49, by Spousal Selection and First Marriage Cohort
A second question asked how long a woman knew her husband before marriage. The marriage cohorts of the KR years also exhibit the highest percentage of women who reported that they met their husband on their wedding day (59.7%). This situation, however, is not uncommon in Cambodia; indeed 42.4% of women married in other years also report meeting their husband on their wedding day. This doesn’t necessarily indicate that families no longer thoroughly investigate each other because, as we noted earlier, this investigation is often carried through intermediaries. But at least for the cohorts of the CDHS 2000, it confirms anecdotal commentaries that the requirement for the potential groom to work for his in-laws (twee bomrae) has lost strength. While the CDHS 2000 provides us with, to our knowledge, the first quantitative assessment of the prevalence of forced marriage under the KR, the two categorical variables (i.e., spousal selection and how long a woman knew her husband before marriage) also likely mask important differences in the matching process. Our analyses of the MIPopLab focus group discussions suggest that the most common answers, such as a woman’s meeting her husband on her wedding day (59.7% under the KR, 42.4% in other years) or having her husband chosen by her family (40.5% under the KR, 52.9% in other years), may represent very different marriage processes regarding families’ search for a potential spouse for their daughters. Overall, the qualitative and quantitative data at hand strongly support the claim that marriages were contracted very differently during the KR than in other periods and that the conditions during the KR did not allow parents to evaluate the compatibility of potential spouses as thoroughly as before.
Trends in Marital Stability
Figure 2 presents estimated probabilities of marital disruption by marriage duration for each of the five sets of marriage cohorts. The average probability across all cohorts is quite low: less than 1 marriage in 50 is expected to experience divorce or separation during the first five years of marriage. The proportion rises to about 1 in 29 at 10 years of marriage, 1 in 19 at 15 years, and 1 in 17 at 20 years. Although the increase appears fairly dramatic on Figure 2, we must temper this observation by reiterating that these probabilities are quite low. In comparison, Hirschman and Teerawichitchainan (2003:223) showed proportions of marriages that ended in divorce or separation within five years of marriage that are between 5% and 10% for all Thai cohorts who were married between 1945 and 1969. In Cambodia, only the most recent marriage cohorts (1993–1999) appear to reach 5% at five years of marriage, and at about 7%, disruption among these cohorts would appear average by Thai standards.
Figure 2.
Observed Proportions of Marriages Ending in Divorce or Separation, by Marriage Cohorts and Marriage Duration
The divorce or separation levels estimated from CDHS 2000 are underestimated. This underestimation is only slight, however, as demonstrated by simulations using MIPopLab’s complete marriage histories to assess the censoring bias induced by the restriction of the CDHS 2000 sample to women under age 50 and the necessary removal of all women who had married more than once (see Appendix B). Moreover, the simulations show that particular cohorts are not selectively subject to this bias and thus confirm the intercohort trend shown in Figure 2. In particular, Figure 2 suggests that the 1975–1978 marriage cohorts exhibit less divorce and separation than the average marriages at all durations, rising to not quite 1 in 25 during the first 20 years of marriage. The cohorts married under the KR do not appear exceptional but, rather, intermediate in probabilities of marital disruption that increase from the earlier to the later marriage cohorts. Finally, Figure 2 also suggests a possible period effect: for most of the cohort sets, the estimated probabilities increase more rapidly with duration at the longest duration, that is, during the most recent period.
Determinants of Marital Stability
Table 2 presents our estimated rates of marital disruption per marriage-year of exposure (for the first 10 years of marriage only) by marriage cohorts and characteristics of spouses. The table confirms the impression derived from Figure 2: the risk first increases gradually across cohorts, from 0.4 per 1,000 marriage-years among the pre-1975 marriage cohorts to 3.4 per 1,000 marriage-years among the 1986–1992 marriage cohorts, then more markedly to 10.4 per 1,000 marriage-years among the 1993–1999 marriage cohorts. The marital-disruption rate for the KR marriage cohorts of 1975–1978 is about half the rate across all cohorts (1.5 versus 3.1 per 1,000 marriage-years).
Table 2.
Rates of Marital Disruption per 1,000 Marriage-Years, by Marriage Cohorts and Selected Characteristics of Spouses (first 10 years of marriage only)
| Marriage Cohorts |
||||||
|---|---|---|---|---|---|---|
| Variable | Pre- 1975 |
1975– 1978 |
1979– 1985 |
1986– 1992 |
1993– 1999 |
All |
| All Women Who Married Only Once | 0.4 | 1.5 | 2.0 | 3.4 | 10.4 | 3.1 |
| Wife’s Age at First Marriage | ||||||
| < 18 | 0.3 | 1.5 | 1.6 | 2.0 | 7.7 | 2.0 |
| 18–24 | 0.5 | 1.5 | 2.0 | 3.3 | 10.2 | 3.1 |
| 25+ | — | 0.0 | 3.6 | 8.3 | 16.0 | 8.1 |
| Wife’s Education | ||||||
| Did not attend school | 0.0 | 1.2 | 2.5 | 3.5 | 11.0 | 3.0 |
| Attended primary school | 0.4 | 1.4 | 1.6 | 3.8 | 11.5 | 3.1 |
| Completed primary school | 1.3 | 2.8 | 2.2 | 2.9 | 8.3 | 3.5 |
| Completed secondary school or higher | 0.0 | 0.0 | 5.0 | 0.0 | 5.7 | 2.5 |
| Wife’s Occupation | ||||||
| Does not work | 0.0 | 1.1 | 0.5 | 2.0 | 6.0 | 1.9 |
| Self- or family-employed in agriculture | 0.4 | 1.4 | 2.0 | 3.0 | 7.0 | 2.5 |
| Self- or family-employed in other sector | 0.7 | 1.3 | 1.4 | 3.8 | 17.0 | 3.4 |
| Outside employment in agriculture | 0.0 | 10.4 | 0.0 | 8.5 | 29.7 | 7.1 |
| Outside employment in other sector | 0.0 | 0.0 | 13.5 | 8.8 | 35.8 | 12.0 |
| Husband’s Education Relative to Wife’s | ||||||
| Less educated | 2.1 | 6.4 | 4.1 | 4.6 | 7.3 | 4.7 |
| Equally educated | 0.5 | 0.1 | 1.5 | 3.6 | 11.5 | 2.9 |
| More educated | 0.0 | 2.0 | 2.0 | 2.9 | 10.0 | 2.9 |
| Husband’s Occupation | ||||||
| Does not work | 3.8 | 0.0 | 5.1 | 10.5 | 25.4 | 7.2 |
| Agricultural worker | 0.4 | 1.5 | 2.2 | 3.0 | 9.8 | 3.0 |
| Skilled and unskilled worker | 0.0 | 0.9 | 0.0 | 4.0 | 8.4 | 2.7 |
| White-collar worker | 0.0 | 1.0 | 2.3 | 3.4 | 12.7 | 3.1 |
| Unknown | 0.0 | 11.1 | 0.0 | 2.9 | 16.8 | 3.0 |
| Premarital Sex | ||||||
| Yes | 0.0 | 5.9 | 2.2 | 6.0 | 4.1 | 3.8 |
| No | 0.4 | 1.2 | 2.0 | 3.3 | 10.6 | 3.1 |
| Number and Gender of Children | ||||||
| None | 0.0 | 2.7 | 3.6 | 4.4 | 20.2 | 6.2 |
| Only one daughter (no son) | 1.2 | 5.2 | 1.5 | 4.9 | 9.6 | 4.5 |
| Only one son (no daughter) | 0.0 | 1.3 | 3.2 | 4.0 | 10.0 | 4.3 |
| Only daughters (two or more) | 0.0 | 0.0 | 0.9 | 4.0 | 7.1 | 2.3 |
| At least two children, including one son | 0.4 | 0.9 | 1.7 | 2.5 | 4.3 | 1.8 |
| Distribution of
Marriage-Years (N = 73,704) |
13.0 | 11.1 | 32.7 | 32.4 | 10.9 | 100.0 |
Also note the unusual effect of a women’s age at first marriage. In Cambodia, marriage at an early age (under 18 years) is associated with a lower risk of marital disruption than marriage at an average age (18 to 24 years), whereas, consistent with traditional norms against late marriage, a first marriage after a woman reaches age 25 is associated with an elevated risk of marital disruption. For younger women, the result is contrary to a relatively well-established finding elsewhere. However, a first marriage before age 18 is not particularly unusual in Cambodia (about a third of married women across cohorts; see Table 1). Furthermore, women typically continue to live with their parents for a few years after marriage, which may soften the potential effect of being a young wife. As for older women, the result may simply signal characteristics that are associated with an elevated risk of marital disruption among women who marry past the preferred age and the husbands they marry when facing the risk of never marrying. Also consistent with traditional matching criteria, a woman faces a higher-than-average risk of divorce or separation when her husband’s level of education is lower than hers. Although male unemployment clearly raises the risk of marital disruption, little consistent variation can be observed across husbands’ occupational categories. Premarital sex does not appear to affect the risk of marital disruption across cohorts consistently. Couples without children and with only one child exhibit a higher risk of marital disruption, regardless of the child’s gender.
Results of the proportional hazard models shown in Table 3 provide a more formal decomposition of the extent to which the differences across cohorts can be explained by differences in the observable characteristics of the spouses. Model 1, unadjusted for any covariates, confirms the gradual increase of the risk of marital disruption with the more recent marriage cohorts. The difference between the marital-disruption risk ratios (RR) of the KR and immediately adjacent marriage cohorts is not significant, but the difference between the KR and more recent cohorts is significant (2.44 for the 1986–1992 cohorts and 7.83 for the 1993–1999 cohorts). Adding individual covariates of the wife, Model 2 yields little change in the size and statistical significance of these differences across cohorts.
Table 3.
Marital-Disruption Risk Ratios, by Marriage Cohorts and Selected Characteristics of Spouses (first 10 years of marriage only)
| Variable | Model 1 | Model 2 | Model 3 | Model 4 | Model 5a |
|---|---|---|---|---|---|
| Marriage Cohorts (1975–1978) | |||||
| Pre-1975 | 0.25 | 0.27 | 0.28 | 0.25 | — b |
| 1979–1985 | 1.37 | 1.32 | 1.28 | 1.22 | 1.58 |
| 1986–1992 | 2.44* | 2.36* | 2.38* | 2.18* | 5.03† |
| 1993–1999 | 7.83*** | 7.80*** | 8.50*** | 7.53*** | 17.40** |
| Wife’s age at First Marriage (18–24) | |||||
| < 18 | 0.72 | 0.73 | 0.71 | 0.83 | |
| 25+ | 1.71** | 1.64* | 1.52* | 1.63 | |
| Wife’s Education (did not attend school) | |||||
| Attended primary school | 0.89 | 0.77 | 0.75 | 0.77 | |
| Completed primary school | 0.56* | 0.44** | 0.43** | 0.40 | |
| Completed secondary school or higher | 0.18* | 0.11** | 0.11** | — c | |
| Wife’s Occupation (self- or
family- employed in agriculture) |
|||||
| Does not work | 0.68 | 0.75 | 0.74 | 0.20† | |
| Self- or family-employed in other sector | 1.57* | 1.88** | 1.90** | 1.66 | |
| Outside employment in agriculture | 2.67** | 2.82** | 2.96** | 2.96 | |
| Outside employment in other sector | 5.09*** | 6.01*** | 5.76*** | 2.66 | |
| Husband’s Education Relative to
His Wife’s (equally educated) |
|||||
| Less educated | 1.72* | 1.75* | 1.67 | ||
| More educated | 0.85 | 0.83 | 0.50† | ||
| Husband’s Occupation (agricultural worker) | |||||
| Does not work | 2.40** | 2.40** | 3.91* | ||
| Skilled and unskilled worker | 0.53* | 0.54* | 1.33 | ||
| White-collar worker | 0.78 | 0.80 | 1.51 | ||
| Unknown | 0.85 | 0.75 | — b | ||
| Premarital Sex (No) | |||||
| Yes | 0.99 | 1.00 | 1.98 | ||
| Daughters and Sons (one daughter, no son) | |||||
| None | 1.61* | 1.31 | |||
| Only one son (no daughter) | 1.01 | 0.65 | |||
| Only daughters (two or more) | 0.44* | 0.41 | |||
| At least two children, including one son | 0.37*** | 0.20** | |||
| Person Choosing Spouse and Consent (self ) | |||||
| Not a family member or without consent | 0.92 | ||||
| Own family members (with consent) | 1.62 | ||||
| Husband’s family members (with consent) | 1.15 | ||||
| Log-Likelihood | −2,022 | −1,971 | −1,956 | −1,933 | −347 |
| Wald Chi-Square | 86.9*** | 184.9*** | 249.0*** | 301.3*** | 133.0*** |
| Global Test for Proportional | |||||
| Hazard Assumption | 3.7 | 10.1 | 14.1 | 15.7 | 45.1* |
Notes: The number of women married only once is 8,911; the number of marriage-years is 73,704. Missing categories are shown in parentheses. Region of residence is also included (coefficients not shown).
One-fourth sample, 1,922 women and 15,719 marriage-years.
Dropped category.
Combined with primary education competed.
p < .10;
p < .05;
p < .01;
p < .001 (two-tailed tests)
Model 2 also reveals that the association between marrying after age 25 and an elevated risk of marital disruption is significant (the reference age is 18–24; RR = 1.71) and is not accounted for by women’s education or occupation; the association between marrying before age 18 and a low risk of marital disruption is not significant. Women’s education is associated with lower risks of marital disruption: the risk of marital disruption for women who completed secondary school was one-sixth that for women with no schooling (RR = 0.18, statistically significant at the 5% confidence level.) As for professional occupation, the estimated risk of marital disruption is notably lower for women who do not work and higher for those who work outside the agricultural sector, but the largest and most significant effects are for women who work for someone other than their family, regardless of the sector (the reference category is self- or family-employed in agriculture; RR = 2.67 in the agricultural sector and RR = 5.09 in other sectors). Although these results are consistent with the expectation that the growing availability of female employment outside the family would make divorce or separation a more “affordable” alternative for married women, the direction of causality is difficult to assess because the occupation is measured at the time of the survey. However, for widows, the same association exists between the risk of marital disruption and a higher propensity to be working, but no stronger association exists between the risk of marital disruption and employment outside the family (data not shown). This finding suggests that no longer being married increases the need to work, but that specific jobs outside the family may have an additional net effect on marital disruption.
Husbands’ characteristics added in Model 3 induce little change in the assessment of the effect of marriage cohorts and wives’ characteristics. The risk of marital disruption is higher when husbands are less educated than their wives (the reference category is the same educational level; RR = 1.72) and when husbands do not work (the reference category is agricultural worker; RR = 2.40). When husbands work, the risk of marital disruption is the highest for agricultural workers and the lowest for skilled and unskilled workers (RR = 0.53).
Model 4 shows that adding the number and gender of children slightly reduces the cohort differences in the risk of marital disruption. Although having more children is strongly associated with a lower risk of marital disruption, the gender of the children is not significantly associated with the risk of marital disruption. The lack of gender effect suggests no strong preference for sons or daughters, which appears consistent with our understanding of the Cambodian kinship system but removes the possibility of testing the direction of the causal link between the presence of children and marital stability. The most recent marriage cohorts have fewer children because of both declining fertility rates and shorter exposure to childbearing (premarital sex is very rare, as shown in Table 1), but the increasing trend in cohorts’ marital disruption is only slightly tempered by the inclusion of children in the model (the reference category is KR marriage cohorts; RR = 7.53 for 1993–1999 marriage cohorts).
Model 5 adds reports on who chose the spouse and is therefore limited to the one-fourth subsample that answered the question, thus preventing a direct comparison with Models 1 to 4. None of the results on who chose the spouse are significant, but women who reported having been married without their consent or to a spouse who was not chosen by a member of their or their husband’s family are less likely to have experienced divorce within the first 10 years of their marriage than those whose husband was chosen by their husband’s family, or, more traditionally, by their own family. The significance of the other covariates is diminished by the sample restriction, but the direction and magnitude of the effect is not greatly affected by the inclusion of the spousal-choice variables.
Table 4 displays the same four models by period of marital disruption, rather than by marriage cohorts. These models provide a more direct test of the period effects, but the effects of other covariates are very similar to those shown in Table 3 and need not be repeated. The risk of marital disruption appears to increase over time; the 1979–1985 period (the period with the most imbalanced gender ratio), however, is not statistically different (at least at the 1% level) from immediately adjacent periods (RR = 0.53 for pre-1979 marriages; RR = 1.84 for 1986–1992 marriages). The strongest period effect is in the most recent period, with a statistically significant RR of 6.27.
Table 4.
Marital-Disruption Risk Ratios, by Period and Selected Characteristics of Spouses (first 10 years of marriage only)
| Variable | Model 1 | Model 2 | Model 3 | Model 4 | Model 5a |
|---|---|---|---|---|---|
| Year of Divorce (1979–1985) | |||||
| Pre-1979 | 0.49 | 0.54 | 0.55 | 0.53 | — b |
| 1986–1992 | 1.97* | 1.94* | 1.94* | 1.84* | 1.32 |
| 1993–2000 | 6.37*** | 6.43*** | 6.81*** | 6.27*** | 6.82** |
| Wife’s Age at First Marriage (18–24) | |||||
| < 18 | 0.73 | 0.73 | 0.70 | 0.86 | |
| 25+ | 1.74** | 1.67* | 1.55* | 1.71 | |
| Wife’s Education (did not attend school) | |||||
| Attended primary school | 0.87 | 0.76 | 0.74 | 0.81 | |
| Completed primary school | 0.52** | 0.41** | 0.40** | 0.43 | |
| Completed secondary school or higher | 0.19* | 0.12** | 0.11** | — c | |
| Wife’s Occupation (self- or
family- employed in agriculture) |
|||||
| Does not work | 0.70 | 0.77 | 0.76 | 0.21† | |
| Self- or family-employed in other sector | 1.60* | 1.91** | 1.91** | 1.54 | |
| Outside employment in agriculture | 2.69** | 2.75** | 2.88** | 2.95 | |
| Outside employment in other sector | 5.04*** | 5.90*** | 5.65*** | 2.60 | |
| Husband’s Education Relative to
His Wife’s (equally educated) |
|||||
| Less educated | 1.72* | 1.77* | 1.58 | ||
| More educated | 0.85 | 0.82 | 0.53† | ||
| Husband’s Occupation (agricultural worker) | |||||
| Does not work | 2.16** | 2.14** | 3.88* | ||
| Skilled and unskilled worker | 0.54* | 0.56* | 1.43 | ||
| White-collar worker | 0.78 | 0.80 | 1.49 | ||
| Unknown | 0.89 | 0.80 | — b | ||
| Premarital Sex (no) | |||||
| Yes | 0.99 | 1.00 | 1.78 | ||
| Daughters and Sons (one daughter, no son) | |||||
| None | 1.59* | 1.32 | |||
| Only one son (no daughter) | 1.02 | 0.70 | |||
| Only daughters (two or more) | 0.43* | 0.41 | |||
| At lest two children, including one son | 0.37*** | 0.21** | |||
| Person Choosing Spouse and Consent (Self ) | |||||
| Not a family member or without consent | 0.91 | ||||
| Own family members (with consent) | 1.67 | ||||
| Husband’s family members (with consent) | 1.17 | ||||
| Log-Likelihood | −2,021 | −1,970 | −1,956 | −1,933 | −348 |
| Wald Chi-Square | 81.3*** | 172.1*** | 244.0*** | 289.1*** | 142.6*** |
| Global Test for Proportional | |||||
| Hazard Assumption | 6.9 | 15.3 | 18.9 | 22.1 | 42.8* |
Notes: The number of women married only once is 8,911; the number of marriage-years is 73,704. Missing categories are shown in parentheses. Region of residence is also included (coefficients not shown).
One-fourth sample, 1,922 women and 15,719 marriage-years.
Dropped category
Combined with primary education competed.
p < .10;
p < .05;
p < .01;
p < .001 (two-tailed tests)
DISCUSSION
The analyses presented in this article were designed to estimate the impact on marital stability of a set of potentially disruptive factors that Cambodia experienced in rapid succession. The three primary factors are (1) a radical change in the way marriages were arranged under the KR (1975–1978), (2) a very low gender ratio in the adult population in the decade or so thereafter, and (3) the rapid opening of a society isolated from Western economic and cultural influences for decades up to the UNTAC years (1992–1993).
Our most striking finding is the lack of a notable difference in the subsequent marital stability experienced by the KR marriage cohorts. In the context of an increasing trend in the risk of marital disruption for the most recent cohorts, the risk among the KR marriage cohorts is slightly higher than the risk among earlier marriage cohorts and slightly lower than among the marriage cohorts that immediately followed. These results were unexpected given qualitative and quantitative evidence, including some that we presented in this article, that the KR radically changed the institution of marriage. Much more modest shocks to the institution in the United States have been reported to result in less-stable marriages during World War II (Preston and McDonald 1979) and the Vietnam War (South 1985), and some of the factors believed to have contributed to this elevated risk of marital disruption were also present in Cambodia, likely on a much greater scale (e.g., “instrumental” motives for marriage, separation of spouses just after the wedding).
To account for unexpected results, demographers’ natural inclination is to search for data problems. Simulations from another, albeit not nationally representative, data source that does not suffer from some of the sampling and questionnaire limitations of the CDHS 2000 strongly suggest that such limitations could not produce these findings (see Appendix B). Another possibility is that some of the forced marriages contracted under the KR were not deemed “real” marriages, ceased as soon as the KR regime collapsed, and were never reported in either source of data. No evidence of this exists in our qualitative data, but had this been the case, the only necessary qualification would be that, if the risk of marital duration was ever higher for the cohorts married during the KR than for the other cohorts, it was only so for the few years after marriage.
Our primary finding seems to imply that marriages relatively quickly “forget their past”: marital stability does not appear to depend durably on the initial characteristics of the spouses or the conditions at the time of marriage. This interpretation rests on our expectation that spouses who married during the KR were more poorly matched than were other marriage cohorts. A slightly different interpretation would be that in spite of their time and efforts during the extended process of spouse selection, families might not really be efficient at assessing the future compatibility of spouses and appreciating the factors that may affect marital stability. Our results do show that the most easily measured among the norms commonly expressed about marriage, such as a preferred age of marriage for women and the respective educational levels of each spouse, are associated with lower risk of marital disruption. However, this observation alone does not suffice to prove the efficiency of the criteria for preferred matching of spouses because of the strong possibility that marriages that deviate from the expressed norms are, at the time of marriage, self-selected on other unobserved characteristics that are correlated with marital disruption or later treated differently by their social environment. Such differences in marital stability are not replicated in the comparison across cohorts presented here, for which self-selection is less of a concern because the matching conditions of entire marriage cohorts differed from those dictated by tradition; differential treatment is also less of a concern, because the “deviant” marriages in the KR cohorts were more often imposed on reluctant families by KR cadres than by the spouses themselves.
Another possibility is that the KR marriage cohorts also benefited from consolidating factors that eventually outweighed the potential effect of a likely suboptimal spouse-matching process on their average marital stability. A valid parallel to the stability of the KR marriage cohorts then would be the finding that in the United States or Canada, couples who marry directly exhibit a lower risk of marital disruption than those who cohabit before marriage (Axinn and Thornton 1992; Hall and Zhao 1995; Thomson and Colella 1992). In a sense, the testing of partners’ compatibility that unmarried couples engage in during premarital cohabitation is functionally similar to the extended spousal-selection process in which families “feel each other out” (Khmer expression cited in Steinberg 1959:84). Whereas the explanation for cohabitation is believed to originate in self-selection on the degree of commitment to marriage, self-selection was hardly a factor in marriages under the KR. More plausibly, the commitment to marriage of the cohorts who were married before or during the KR might have been enhanced by these cohorts’ experience of hardships shared with their spouses. Also consistent with this possible mechanism are the levels of marital disruption for pre-KR and KR marriage cohorts that, even after adjustment for possible biases (see Appendix B), remain much lower than those that have been documented in Thailand (Hirschman and Teerawichitchainan 2003), a country whose marriage system at the time shared key features with Cambodia’s.
Our second finding of a lack of support for a hypothesized relationship is that the low male-to-female ratio in the post-KR adult population did not seem to increase the risk of marital disruption. Although the gender ratio in the adult population was at its lowest at the end of the KR regime (1979), we found that the risk of marital disruption increased over time, with the risk during the 1979–1985 period higher than in the previous period but also lower than in subsequent periods, during which the gender ratio in the adult population was returning toward parity. This finding, again, is at odds with previous research (Trent and South 1989) that found an association between a low gender ratio and a high risk of marital disruption, even with more-modest deviations from parity than in post-KR Cambodia. The effect of the low gender ratio was arguably lagged; that is, the low ratio created the demographic conditions for an increase in divorce, but not until material conditions also improved, allowing more men to “afford” divorce and remarriage. The individual covariates of marital disruption that we uncovered here do not provide support for this mechanism, however: we found that the risk of marital disruption increased when the wife’s, rather than the husband’s, status was higher (when wives were more educated than their husbands, wives were employed but not by a family member, and husbands were working in the agricultural sector or not working).
Thus, the discrepancy with previous findings may originate instead in the different marriage systems of Cambodia and countries considered in previous analyses. As we described earlier, the marriage process in Cambodia is hardly amenable to the formalization of two rational actors on a marriage market in which one would expect a lower “supply” of males to improve their market position. In Cambodia, the marriage process more often than not involves two entire families, with the bride’s family typically of a higher social standing that the groom’s family, each attempting to assert that the groom has the individual characteristics of a good provider for the new family. The families seek a certain balance between the slightly higher social status of the bride and the expected higher individual position of the husband in the family. It is possible, then, that the low gender ratio in the adult population helped to improve the position of men with respect to women, reducing the possibility of a husband maintaining an inferior position to his wife because in Cambodia, such a consolidation of male position appears to be associated with greater marital stability.
Our strongest evidence in support of a relationship is the decline in marital stability that is associated with recent changes in Cambodian society. Contrasting extant evidence had led us to be ex ante agnostic about the direction of a possible change. Similar to other Southeast Asian countries, particularly Thailand, modernization could plausibly reduce traditionally high rates of marital disruption by allowing spouses to have more input in the selection of their future partner. On the contrary, the overwhelming effect here is an increase in such rates. Whereas the traditional regime that has eroded in several other Southeast Asian countries with a Moslem majority (Indonesia and Malaysia) or substantial minority (Thailand) was marked by easy and relatively prevalent divorce, the prevalence of divorce among older marriage cohorts appears low in Cambodia. The recent increase in rates of marital disruption is also consistent with theories that emphasize the potential for economic independence from husbands that was brought about by increases in female education and employment. This explanation is further supported by the finding that women who are more educated than their husbands and are employed outside the family are more likely to experience divorce, whereas widows are more likely to be employed across occupations but not specifically outside the family.
In the end, the CDHS 2000, and arguably any single cross-sectional survey, does not suffice to assess the exact mechanisms that may have contributed to differences in marital stability or to generalize easily across contexts. Some of our findings differ from earlier findings that were most often generated from data on the United States or other Western societies, and our findings may originate from specific aspects of Cambodian society, such as matrilocal marriage or a relative deference to the parental choice of a potential spouse. Nevertheless, with the unparalleled attempt by the KR to reform its society and its institution of marriage that so affected the people who experienced it, Cambodia is precisely the place to expect long-lasting cohort effects. However, the clearly emerging picture from this country is one in which the contemporaneous conditions have had a much greater impact on marital stability than the conditions at the time of marriage.
APPENDIX A. TIMING OF LAST SEXUAL INTERCOURSE AND SEXUAL ACTIVITY
The CDHS 2000 did not collect information on the date marriages ended. The respondents were, however, asked how long ago (in days, months, and years) they last had sexual intercourse. If it was less than 12 months before the survey, they were also asked about their sexual partner. These data thus allow us to examine the prevalence of extramarital sex and long-term abstinence within marriage. If both are rare, the date of last intercourse might be reasonably close to the date of marital disruption for widowed, divorced, and separated women.
We first examine the sexual activity of women who had sexual intercourse less than 12 months before the survey (see Appendix Table A1). Less than 3% of all married women reported that they had no sexual intercourse in the 12 months before the survey. It is thus safe to assume that the date of last intercourse should only rarely be much earlier than the date a marriage ended. If extramarital sex was prevalent, however, it could be much later than that. But the data on the partners of the women sexually active in the past 12 months are reassuring here too because only 4 of the 8,789 married women and 3 of the 129 separated, divorced, or widowed women last had sex with someone other than their husband. All three of the separated, divorced, or widowed women who reported extramarital sex also reported that it occurred as commercial sex. In any event, the date of last intercourse is only rarely expected to be later than the date of last sexual intercourse.
Appendix Table A1.
Sexual Activity in the 12 Months Before the Survey, Women Aged 15–49, by Marital Status
| Variable | Never Married |
Married | Widowed | Divorced
or Separated |
|---|---|---|---|---|
| Sexually Active in the Past 12 Months | ||||
| No | 4,876 | 233 | 851 | 404 |
| Yes | 6 | 8,789 | 65 | 64 |
| Percentage sexually active | 0.12 | 97.42 | 7.10 | 13.68 |
| Sexual Partner(s) in the Past 12 Months | ||||
| Spouse only | 0 | 8,785 | 65 | 61 |
| Commercial sex partner(s) | 6 | 1 | 0 | 3 |
| Other partners | 0 | 3 | 0 | 0 |
Source: CDHS 2000; see Appendix A for details.
APPENDIX B. MARITAL DISRUPTION FROM CDHS AND MIPOPLAB DATA
We investigate two of the limitations of the CDHS 2000 using the data from MIPopLab. These data provide, on a smaller and not nationally representative population, more-comprehensive marriage histories. Specifically, histories were collected for all women under age 75, which reduces the selectivity on a young age at marriage for earlier marriage cohorts. Second, we collected information on all their marriages, not only the most recent one. To be consistent with the CDHS 2000, we right censor below our observations at the end of the year 2000, even though our data were collected in 2001.
Appendix Table B1 compares the rates of marital disruption for each of the four sets of marriage cohorts with different marriage sample bases. The first one is our full sample. The second is restricted to all the marriages of women who, as in the CDHS 2000, were under age 50 at the time of the interview. The third is restricted to the marriages of women who were married only once. The amount of censoring introduced by the one-marriage limit is stronger in the MIPopLab data (152 out of 1,279 or 11.9%) than in the CDHS 2000 (890 out of 40,237, or 8.7%). This finding is consistent with the higher incidence of divorce observed in MIPopLab than in CDHS 2000.
The intercohort trend from our full marriage sample resembles the trend observed from the CDHS 2000, but the level is higher in each cohort: the marital disruption rate increases first gradually, from 2.9 per 1,000 marriage-years among the pre-1975 marriage cohorts to 8.7 per 1,000 person-years among the 1986–1992 cohorts, then more sharply to 23.6 per 1,000 person-years among the 1993–1999 cohorts. Our estimates from the CDHS 2000 sample increased in a similar manner, but only from 1.2 per 1,000 marriage-years among the pre-1975 cohorts to 15.9 per 1,000 person-years among the most recent cohorts. Limiting the sample to women under age 50 at the time of the interview had relatively little impact on post-1975 marriage cohorts, but the number of pre-1975 marriages dropped from 630 to 111, and the rate of marital disruption increased from 2.9 to 5.7 per 1,000 marriage-years. While it might be seen as an effect of selectivity on a young age at marriage, marrying young is not otherwise associated with higher rates of disruption in Cambodia, and it might simply be an artifact of the sample becoming very small.
Marital-disruption rates decrease when the sample is restricted to women who married only once because divorced or separated women are obviously overrepresented among women who married more than once. MIPopLab data show, however, that nearly two-thirds of women who married more than once (96 out of 152) were actually widowed rather than divorced. One might expect the opposite when divorced women are generally much younger than widowed women, but this is not necessarily the case in Cambodia because many young men were killed during the KR. In addition, the remarriage of widows is not stigmatized as is that of divorcees. While the one-marriage censoring reduces the rates of divorce from 8.9 per 1,000 marriage-years to 5.5 per 1,000 marriage-years, it also renders the estimate closer to what we observe on the full sample without an age limit (5.8 per 1,000 marriage-years).
However, our main concern is how the intercohort trend might be affected, that is, whether the relative decrease in marital-disruption rates varies across cohorts. In this respect, the decrease is strongest for the 1979–1985 cohorts, with the rates falling from 7.7 to 3.3 per 1,000 marriage-years. The decrease is the smallest for the most recent cohorts, from 24.0 to 22.5 per 1,000 marriage-years. If the marital-disruption rate was constant across cohorts, we would expect the censoring to have the largest effect on the earliest marriage cohorts because divorced and separated women are exposed longer to the risk of remarriage. However, the increasing trend in marital separation also implies that the most recent cohorts had the highest proportions of divorced and separated women exposed to the risk of remarriage at all. The joint effect of, first, the rising proportions at risk of remarriage and, second, the decreasing duration of exposure to the risk as we move from the earliest to the most recent cohorts might explain why the effect is the strongest for the middle set of marriage cohorts. We should also note that the number of divorces or separations among each set of marriage cohorts is very small, and the differential censoring might be due simply to random variability in this respect.
In any event, these simulations indicate that the upward trend in marital-disruption rates that are observed for women in the CDHS 2000 who were under age 50 and married only once is unlikely to be an artifact of sample and questionnaire restrictions. As shown in Appendix Figure B1, a comparison of the estimated trend from the CDHS 2000 sample of women who married only once to the rates estimated from the full sample and the similarly reduced sample in MIPopLab suggests what the actual trend might look like if we had been able to analyze all the data. The probable trend among all marriages in the CDHS 2000 is estimated assuming that the effect of sample restriction in CDHS 2000 is similar to that in MIPopLab (in relative terms), that is, by prorating the CDHS 2000 estimates by the ratio of rates between the full and reduced sample in MIPopLab. The small numbers in MIPopLab prevent us from making too strong a case that CDHS 2000 estimates should be so corrected, but a comparison of the broken (observed) and solid (corrected) black lines on Appendix Figure B1 provides support for the assertion that marital-disruption rates increase across marriage cohorts.
While these simulations suggest some uncertainty with respect to the actual trends from the 1975–1978 cohorts to the 1985–1992 cohorts, two major divorce trends from the CDHS 2000 are confirmed: (1) marital-disruption rates gradually increase from the earliest to the latest marriage cohorts, with an acceleration for the most recent marriage cohorts (1993–1999), and (2) the 1975–1978 marriage cohorts do not exhibit exceptionally high rates of marital disruption.
Appendix Table B1.
Rates of Marital Disruption in MIPopLab (per thousand marriage-years), by Marriage Cohorts and Sample Restrictions
| All Women Under Age 75 |
All Women Under Age 50 |
Women Under Age 50 and Married Only Once |
|
|---|---|---|---|
| All Marriages | |||
| Marriages | 1,885 | 1,279 | 1,127 |
| Person-years | 30,886.4 | 14,720.6 | 13,685.5 |
| Divorces and separation | 180 | 131 | 75 |
| Rate of marital disruption | 5.8 | 8.9 | 5.5 |
| Pre-1975 Marriage Cohorts | |||
| Marriages | 630 | 111 | 96 |
| Person-years | 17,151.4 | 2,297.6 | 2,321.5 |
| Divorces and separation | 49 | 13 | 7 |
| Rate of marital disruption | 2.9 | 5.7 | 3.0 |
| 1975–1978 Marriage Cohorts | |||
| Marriages | 143 | 113 | 98 |
| Person-years | 2,695.8 | 2,158.5 | 1,980.9 |
| Divorces and separation | 20 | 17 | 10 |
| Rate of marital disruption | 7.4 | 7.9 | 5.0 |
| 1979–1985 Marriage Cohorts | |||
| Marriages | 386 | 344 | 291 |
| Person-years | 6,279.5 | 5,610.5 | 5,117.5 |
| Divorces and separation | 52 | 43 | 17 |
| Rate of marital disruption | 8.3 | 7.7 | 3.3 |
| 1986–1992 Marriage Cohorts | |||
| Marriages | 377 | 367 | 326 |
| Person-years | 3,575.1 | 3,489.1 | 3,198.3 |
| Divorces and separation | 31 | 30 | 17 |
| Rate of marital disruption | 8.7 | 8.6 | 5.3 |
| 1993–1999 Marriage Cohorts | |||
| Marriages | 349 | 344 | 316 |
| Person-years | 1,184.5 | 1,164.9 | 1,067.4 |
| Divorces and separation | 28 | 28 | 24 |
| Rate of marital disruption | 23.6 | 24.0 | 22.5 |
Source: MIPopLab; see Appendix B for details.
Appendix Figure B1.
Rates of Marital Disruption in MIPopLab and CDHS 2000, by Marriage Cohort
Footnotes
The authors gratefully acknowledge the support of the National Institute for Child Health and Human Development (Research Grant No. 1 R03 HD41537). Since 2002, this grant has enabled the work of the Mekong Island Population Laboratory (MIPopLab), a demographic surveillance system whose data are used in some of the analyses presented here. Developed in collaboration with a team of Cambodian researchers, in particular, Lany Trinh-Bo, a doctoral candidate at the Australian National University, MIPopLab has collected data since 2000 with early support from the Mellon Fund for Demographic Training in Developing Countries at the University of Pennsylvania and the Albert Greenstone Fund at the University of Chicago. More details about the individuals involved in MIPopLab and about its activities are available from the first author’s web page (home.uchicago.edu/~pheuveli). Bunnak Poch also wishes to acknowledge a Population Council Postdoctoral Fellowship. The authors thank Stan Becker, David Bishai, Kelly Musick, Lany Trinh-Bo, Linda Waite, and Yu Xie for useful comments, and Ira Elliott for his editing of the final manuscript.
An earlier version of this article was presented at the 2003 annual meeting of the American Sociological Association in Atlanta, August 16–19.
Contributor Information
PATRICK HEUVELINE, Population Research Center at NORC, and The University of Chicago..
BUNNAK POCH, Center for Population Studies, Royal University of Phnom Penh..
REFERENCES
- Axinn WG, Thornton A. The Relationship Between Cohabitation and Divorce: Selectivity or Causal Influence? Demography. 1992;29:357–74. [PubMed] [Google Scholar]
- Becker GS, Landes EM, Michael RT. An Economic Analysis of Marital Instability. The Journal of Political Economy. 1977;85:1141–88. [Google Scholar]
- Bennett NG, Bloom DE, Craig PH. The Divergence of Black and White Marriage Patterns. American Journal of Sociology. 1989;95:692–722. [Google Scholar]
- Boli J, Thomas GM. World Culture in the World Polity: A Century of International Non-Governmental Organization. American Sociological Review. 1997;62:171–90. [Google Scholar]
- Carney T. The Unexpected Victory. In: Jackson KD, editor. Cambodia, 1975–1978: Rendezvous With Death. Princeton University; Princeton, NJ: 1989. [Google Scholar]
- Chea H, Sok H. The Cambodian Garment Industry. Cambodian Development Review. 2001;5(3):1–8. [Google Scholar]
- Coedès G. The Indianized States of Southeast Asia. The University Press of Hawaii; Honolulu: 1971. [1944] [Google Scholar]
- Cox DR. Regression Models and Life Tables. Journal of the Royal Statistical Society. 1972;B34:187–220. [Google Scholar]
- Ebihara M. A Cambodian Village Under the Khmer Rouge, 1975–1979. In: Kiernan B, editor. Genocide and Democracy in Cambodia: The Khmer Rouge, the United Nations and the International Community. Yale University Southeast Asia Studies; New Haven, CT: 1993. pp. 51–63. [Google Scholar]
- Frieson KG. Revolution and Rural Response in Cambodia, 1970–1975. In: Kiernan B, editor. Genocide and Democracy in Cambodia: The Khmer Rouge, the United Nations and the International Community. Yale University Southeast Asia Studies; New Haven, CT: 1993. pp. 33–50. [Google Scholar]
- Hall DR, Zhao JZ. Cohabitation and Divorce in Canada: Testing the Selectivity Hypothesis. Journal of Marriage and the Family. 1995;57:421–27. [Google Scholar]
- Hannan MT, Tuma NB, Groeneveld LP. Income and Independence Effects on Marital Dissolution: Results From the Seattle and Denver Income-Maintenance Experiments. American Journal of Sociology. 1978;84:611–33. [Google Scholar]
- Heuveline P. ‘Between One and Three Million’: Towards the Demographic Reconstruction of a Decade of Cambodian History (1970–79) Population Studies. 1998;52:49–65. doi: 10.1080/0032472031000150176. [DOI] [PubMed] [Google Scholar]
- – – – . The Demographic Analysis of Mortality Crises: The Case of Cambodia. In: Reed H, Kelley C, editors. Forced Migration and Mortality. National Academy Press; Washington, DC: 2001. pp. 102–29. [Google Scholar]
- Hirschman C, Teerawichitchainan B. Cultural and Socioeconomic Influences on Divorce During Modernization: Southeast Asia, 1940s to 1970s. Population and Development Review. 2003;29:215–53. [Google Scholar]
- Huber J, Spitze G. Considering Divorce: An Expansion of Becker’s Theory of Marital Instability. American Journal of Sociology. 1980;86:75–89. doi: 10.1086/227203. [DOI] [PubMed] [Google Scholar]
- Huguet J. The Demographic Situation in Cambodia. Asia-Pacific Population Journal. 1992;6(4):79–91. [PubMed] [Google Scholar]
- Jones GW. Modernization and Divorce: Contrasting Trends in Islamic Southeast Asia and the West. Population and Development Review. 1997;23:95–114. [Google Scholar]
- Kiernan B. The Pol Pot Regime. Yale University; New Haven, CT: 1996. [Google Scholar]
- Lesthaeghe R. A Century of Demographic and Cultural Change in Western Europe: An Exploration of the Underlying Dimensions. Population and Development Review. 1983;9:411–35. [Google Scholar]
- Lichter DT, McLaughin DK, Kephart G, Landry DJ. Race and the Retreat From Marriage: A Shortage of Marriageable Men? American Sociological Review. 1992;57:781–99. [Google Scholar]
- Lingat R. Les régimes matrimoniaux du sud-est de l’Asie: essai de droit comparé indochinois. Tome premier: les régimes traditionnels. Editions de Bocard and Ecole Française d’Extrême-Orient; Paris: 1952. [Matrimonial regimes of Southeast Asia: Essay in Indochinese comparative law. First volume: Traditional regimes] [Google Scholar]
- Migozzi J. Cambodge: faits et problèmes de population. Editions du Centre National de la Recherche Scientifique; Paris: 1973. [Cambodia: Facts and population problems] [Google Scholar]
- Morgan SP, Lye DN, Condran GA. Sons, Daughters, and the Risk of Marital Disruption. American Journal of Sociology. 1988;94:110–29. [Google Scholar]
- National Institute of Statistics . General Census of Cambodia 1998. Analysis of Census Results. Report 1: Fertility and Mortality. National Institute of Statistics, Ministry of Planning; Phnom Penh, Cambodia: 1999a. [Google Scholar]
- – – – . General Population Census of Cambodia 1998: Final Census Results. United Nations Population Fund for the National Institute of Statistics, Ministry of Planning; Phnom Penh, Cambodia: 1999b. [Google Scholar]
- – – – . Cambodia Demographic and Health Survey 2000. National Institute of Statistics, Directorate General for Health [Cambodia], and ORC Macro; Phnom Penh, Cambodia, and Calverton, MD: 2001. [Google Scholar]
- – – – Labor Force Survey 2001. 2003 Available online at http://www.nis.gov.kh/SURVEYS/ LFS2001/index.htm.
- Népote J. Parenté et organisation sociale dans le Cambodge moderne et contemporain: quelques aspects et quelques applications du modèle les régissant. Editions Olizane; Genève: 1992. [Kinship and social organization in modern and contemporary Cambodia: Some features and applications of the model governing them] [Google Scholar]
- Ngor H. A Cambodian Odyssey. Macmillan; New York: 1987. [Google Scholar]
- Pich S. Le marriage cambodgien. Centre de Documentation et de Recherche sur la Civilisation Khmère; Paris: 1984. [The Cambodian marriage] [Google Scholar]
- Ponchaud F. Cambodge année zéro. Kailash Editions; Paris: 1998. [1977] [Cambodia year zero] [Google Scholar]
- Preston SH, Heuveline P, Guillot M. Demography: Measuring and Modeling Population Processes. Blackwell; Malder, MA: 2001. [Google Scholar]
- Preston SH, McDonald J. The Incidence of Divorce Within Cohorts of American Marriages Contracted Since the Civil War. Demography. 1979;16:1–25. [PubMed] [Google Scholar]
- South SJ. Economic Conditions and the Divorce Rate: A Time Series Analysis of the Postwar United States. Journal of Marriage and the Family. 1985;47:31–41. [Google Scholar]
- Steinberg DJ. Cambodia: Its People, Its Society, and Its Culture. HRAF; New Haven, CT: 1959. [Google Scholar]
- Thomson E, Colella U. Cohabitation and Marital Stability. Journal of Marriage and the Family. 1992;54:259–67. [Google Scholar]
- Trent K, South SJ. Structural Determinants of the Divorce Rates: A Cross-Societal Analysis. Journal of Marriage and the Family. 1989;51:391–404. [Google Scholar]
- Weitz ED. A Century of Genocide: Utopias of Race and Nation. Princeton University; Princeton, NJ: 2003. [Google Scholar]
- Whyte MK. Dating, Mating, and Marriage. Aldine de Gruyter; New York: 1990. [Google Scholar]
- Wilson WJ. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. University of Chicago; Chicago: 1987. [Google Scholar]
- Xiaohe X, Whyte MK. Love Matches and Arranged Marriages: A Chinese Replication. Journal of Marriage and the Family. 1990;52:709–22. [Google Scholar]