Socio-economic determinants of divorce in Lithuania: specialization hypothesis reconsidered Ausra Maslauskaite Vytautas Magnus University, Lithuania, email: amaslauskaite@ktl.mii.lt Aiva Jasilioniene Max Planck Institute for Demographic Research Vlada Stankuniene Institute for Demographic Research, Lithuania Domantas Jasilionis Max Planck Institute for Demographic Research 1 Socio-economic determinants of divorce in Lithuania: specialization hypothesis reconsidered Ausra Maslauskaite Vytautas Magnus University, Lithuanian Social Research Centre, Lithuania, email: amaslauskaite@ktl.mii.lt Aiva Jasilioniene Max Planck Institute for Demographic Research Vlada Stankuniene Institute for Demographic Research, Lithuania Domantas Jasilionis Max Planck Institute for Demographic Research Abstract Using census-linked data covering the entire married adult male and female population, this study examines socio-economic determinants of first divorce in Lithuania during 2001-2003. The findings suggest that the observed socio-economic differentials in first divorce risk in Lithuania largely support the gender specialization hypothesis. Although the study found that lower education is related to elevated risks of divorce for both males and females, generally being out of the labour market destabilizes marriage and significantly increases the risk of marital disruption for males only. For males this association is sustained across the whole range of groups of economic inactivity. For females being out of the labour market either decreases the risk of first divorce or has no influence. Unemployed and other economically inactive males are the most vulnerable group showing the highest risk of first divorce. Keywords: education, economic activity status, first divorce, Lithuania 2 Introduction The relationship between socio-economic resources and the risk of divorce, especially for females is thoroughly covered in a wide range of publications focusing on determinants of divorce in Europe and in the USA. The results regarding the effects of education and employment on the risk of marital dissolution for females are inconsistent. The evidence about the educational gradient in divorce in Europe is mixed and vary from country to country, whereas in the USA most of the studies suggest a clear negative relationship between education and marital dissolution (Amato and James 2010). For the relationship between employment and divorce, it has been found that a wife’s participation in the labour market has a de-stabilizing effect on marriage (Cherlin 1992; Chan and Halpin 2002; Jalovaara 2001; 2003, Lyngstad and Jalovaara 2010; Amato and James 2010), although there are some studies reporting an inverse relationship (Svarer and Verner 2006). It has been also suggested that influence of female employment on divorce risk is mediated by a number of other confounding factors and that this may lead to both negative and positive relationships (Amato, Booth, Johnson, and Rogers 2007; Kalmijn, De Graaf, and Portman 2004; Sigle-Rushton 2010, Cooke and Gash 2010). Finally, studies focusing on countries with high female labour force participation (e.g., Nordic countries) have found that being out of employment increases the likelihood of divorce (Hansen 2005; Jalovaara 2001). The evidence about the effect of socioeconomic recourses on divorce risk among males is much more consistent. Most of the studies with a focus on males suggest that for males lower socio-economic recourses generate higher propensity to divorce (Oppenheimer 1997; Amato and James 2010, Lyngstad and Jalovaara 2010). Most of the existing evidence on socioeconomic predictors of divorce in developed countries comes from studies conducted in the USA and in Western and Northern Europe. Only a few studies have examined specifics of this link in countries of Central and Eastern Europe. After the fall of the socialist regime, countries of this region underwent radical political and socio-economic changes, which brought about major changes in their educational systems and in the organization of the labour market. On a country-specific level, this region is unequivocally highly diverse in terms of socio-economic development and cultural background, but similar historical experience has created a social context that makes them homogeneous when they are placed in a broader European or worldwide framework. The “hidden story” of divorce determinants in Central and Eastern Europe was and to some extent remains conditioned by the enduring shortage of adequate data, thus limiting the research on divorce in this region predominantly to studying divorce trends and patterns (e.g., Darsky and Scherbov 1995; Scherbov and van Vianen 1999; Avdeev and Monnier 2000; Sobotka and Toulemon 2008; Philipov and Jasilioniene 2008; Stankūnienė 2006). There are a few very recent studies focussing on socioeconomic determinants, but they are mainly based on data from sample surveys (e.g., FFS and GGS surveys). The evidence from these studies is rather contradictory. A comparative study by Härkönen and Dronkers (2006) examining relationships between female education and divorce in 17 countries (including Estonia, Latvia, Lithuania, Poland, and Hungary) concluded that educational gradient of divorce was positive in Poland and negative in Lithuania. The remaining Central and Eastern European countries (Estonia, Latvia, and Hungary) did not show any statistically 3 significant relationships (Härkönen and Dronkers 2006). Bukodi and Róbert (2003) shed more light on the role of social resources for union stability in Hungary suggesting that education does not influence union disruption, whereas employment increases the risk for females and decreases it for males. Muszynska (2008) studied the case of Russia focussing on the soviet and transitional periods and did not find any statistically significant relationships between employment and union dissolution. A recent study on Poland by Styrc and Matysiak (2012) revealed that being employed was associated with an increased divorced risk among females (after 1989 only), whereas males showed an opposite relationship. This study aims to contribute to the still scarce scholarly evidence on socio-economic differentials in the risk of divorce in Central and Eastern Europe by examining the case of Lithuania. For more than three decades Lithuania has maintained a very high divorce level and stands among the countries with highest divorce rates in Europe (Statistics Lithuania 2011; Council of Europe 2006; GGP Contextual Database 2012). Despite this fact still very little is known about factors determining high divorce rates in these countries. What distinguishes this study from other studies on divorce in the region is that it uses a unique census-linked dataset representing the entire population of Lithuania in the period 2001-2003. Datasets based on the linkage between censuses and divorce records from population registers have been used for divorce research only in a few countries in Europe so far (mostly in Nordic countries) and in none of Central and Eastern European countries. In this study, we examine how the level of educational attainment and economic activity influence the risk of first divorce for adult males and females in Lithuania. The results are interpreted within the framework of gender specialization theory (Becker1981) arguing that gender specialization ensures the stability of marriage through interdependence of a “working man” and a “caring woman”. It is suggested that along with growing female education and employment, gender specialization as well as the interdependence of genders in marriage diminishes. These changes produce a negative impact on marital stability and consequently elevate the propensity to divorce. In this study, we test whether the notions of this theory applies to the Lithuanian context as well as look for other possible explanations for the identified links between socio-economic recourses and the risk of divorce. Data and methods This study uses a unique in this part of Europe census-linked dataset based on the linkage between all records from the Lithuanian Population and Housing Census of 2001 and first divorce records from the Lithuanian population register. The dataset was created in two steps. In the first step, the linkage of individual census records with divorce, death, and emigration records for the same individuals from the census was performed. The data on dates of deaths and emigrations were needed in order to estimate precise numbers of marriage years of exposure to risk. The linkage procedures were implemented using personal identification numbers as unique identifiers (carried out by employees of Statistics Lithuania, who have permission to work with individual-level data). In the second step, individual-level data were transformed into an aggregated multidimensional frequency format that provides aggregated numbers of first divorces and population exposures for every possible 4 combination of available variables. This data structure is particularly suitable for calculation of period risk measures such as divorce rates or Poisson regression relative risks. The data cover all formally married (first marriages only) individuals between the exact ages 15 and 60 who reported themselves as being married and indicated first marriage dates (95% of all reported marriages) at the census. The final dataset includes 41 thousand first (legal) divorces and 3.18 million person-years of marriage years of exposure. The data were split by the following variables: duration of first marriage, marriage cohort, age at first marriage, sex, number of children (information available for females only), education, ethnicity, economic activity status, and urbanrural residence. Information on the key independent variables, education and economic activity status of individuals, as well as on number of children, ethnicity, and urban-rural residence was obtained from the population census records. The variables on duration of first marriage, marriage cohort, and age at first marriage were constructed accordingly using the exact date of marriage from both the census and divorce records and the exact date of birth provided at the census. The education variable consists of three categories: higher (university and non-university) education (ISCED 5-6), secondary (upper secondary) (ISCED 3-4), and lower than secondary (ISCED 0-2). The variable of economic activity status distinguishes between the economically active and economically inactive population. The economically active sub-population is further divided into employed and unemployed, and the economically inactive sub-population is distributed accordingly into these four categories: disabled, housewives/househusbands, students, and pensioners and others. For more details on these and other control variables used in the analysis (categories, marriage years of exposure, and number of events), see Table 1. Table 1. First divorces and marriage years under risk (in thousands) by education and economic activity status. Lithuanian females and males, 2001-2003 Females Males Divorces Marriage years Divorces Marriage years Education Higher Secondary Lower than secondary 4.2 (21.0%) 13.3 (67.2%) 2.3 (11.8%) 304.3 (20.8%) 1012.8 (69.1%) 148.0 (10.1%) 2.9 (16.1%) 12.1 (67.6%) 2.9 (16.4%) 235.7 (17.2%) 936.7 (68.5%) 195.3 (14.3%) Economic activity status Active, employed Active, unemployed Inactive, disabled Inactive, housewife/husband Inactive, student Inactive, other Unknown 13.7 (69.0%) 2.8 (14.3%) 0.4 (2.1%) 1.7 (8.7%) 0.5 (2.6%) 0.5 (2.3%) 0.2 (1.0%) 993.8 (67.8%) 194.8 (13.3%) 58.7 (4.0%) 150.0 (10.2%) 15.9 (1.1%) 50.2 (3.4%) 1.7 (0.1%) 12.4 (69.1%) 3.5 (19.6%) 0.5 (2.6%) 0.2 (1.2%) 0.1 (0.6%) 0.9 (4.8%) 0.4 (2.2%) 1021.5 (74.7%) 209.1 (15.3%) 56.9 (4.2%) 20.5 (1.5%) 3.4 (0.3%) 52.0 (3.8%) 4.3 (0.3%) Total 19.8 (100.0%) 1465.0 (100.0%) 17.9 (100.0%) 1367.8 (100.0%) 5 The impact of socio-economic status on the risk of divorce was estimated by applying Poisson regression for count data with first divorce as the dependent variable. The model can be defined using the following equation: Dj E je 0 1 x1, j ... k xk , j e ln(E j ) 0 1 x1, j ... k xk , j , where j is a combination of categories of explanatory variables under consideration, Dj is the expected number of first divorces, Ej is number of marriage-years of exposure to risk, x1,j, …, xk,,j – explanatory variables, and ß0, ß1, …, and ßk are effects of independent explanatory variables. For each combination of j, the Poisson regression model estimates expected divorce rates expressed as the ratios between expected number of first divorces Dj and number of marriage-years of exposure Ej. The effects of independent explanatory variables are presented as relative first divorce risks (expressed as expected divorce rate ratios between divorce rates in the categories under study and corresponding rates in the reference categories) and their 95% confidence intervals. In order to identify potential confounding effects of socio-demographic characteristics on estimates of relative first divorce risks by education and economic activity status, the following modelling strategy was chosen. The first initial model (Model 1) includes the dependent variable and either education or economic activity status and additionally controls for the major demographic characteristics: duration of first marriage, marriage cohort, and age at first marriage. Model 1A was estimated only for females. Apart from the variables included into Model 1, this model also accounts for a number of children. Model 2 accounts for the same variables as Model 1 and also includes either the education or economic activity status variable. Further models, in addition to the basic demographic characteristics, correspondingly include ethnicity (Model 3), place of residence (Model 4), and place of birth (Model 5). Finally, Model 6 adjusts for all the available variables. Likelihood ratio tests were performed by comparing Models 1A-6 to the initial basic Model 1. Results The associations between the risk of first divorce and all the selected control variables are provided in Annex 1. Both for males and females, the association found between age at marriage and divorce risk supports the existing evidence that marriages contracted at later ages tend to be more stable. A similar link has been identified between marital duration and divorce risk: divorce is less likely when marriage is older. The pattern of how marital duration affects the risk of divorce is very much alike for males and females. The risk of marital dissolution increases rapidly during the first five years (the increase is particularly abrupt for males), then levels off and starts declining after about eight years of marital life. Divorce among younger marriage cohorts is more frequent than among older ones. For the effect of children (for females only), it has been found that the risk of divorce decreases with an increasing number of children. The arrival of the second child seems to have the strongest positive effect on marital stability. Lithuanians and Russians are most likely to end up their marriages in divorce, while Poles seem to have most stable marriages. However, males and females of unknown ethnicity show the highest risk of marital dissolution. Furthermore, chances of divorce for males and females residing in rural 6 areas are much lower than for urban residents. Place of birth influences the probability of divorce in the same manner: those born in rural areas have lower propensity to divorce than those originating from urban areas. Tables 2 and 3 show the results obtained from the Poison regressions assessing the impact of education and economic activity status on the risk of first divorce for males and females in Lithuania. The findings suggest a relatively significant association between the level of educational attainment and marital dissolution in Lithuania (Table 2). The results of the regression model controlling for the major demographic divorce determinants, duration of marriage, marriage cohort, and age at first marriage, show that females with higher education are at highest risk of divorce and that those with secondary and lower than secondary education are likely to have more stable marriages (Model 1). Generally, this association holds true until place of residence is introduced into the model. Urban-rural place of residence seems to be the most significant control variable from the set of variables used in the analysis. With the effect of place of residence having controlled females with lower than secondary education turn out to run the highest risk of divorce as compared to the other two educational groups of females (Model 4). An elevated risk of divorce in the group of females with lowest education is also found when the effect of education is additionally adjusted by place of birth instead of place of residence (Model 5). After controlling for all the selected variables, differences in divorce risk between females with higher and secondary education practically disappear, while for those with lower than secondary education an increased likelihood of marital dissolution becomes even more pronounced (Model 6). For males, the initial model controlling for duration of marriage, age at first marriage, and marriage cohort shows different results than for females (Model 1, Table 2). Opposite to females, the least educated males are most likely to divorce, while highly educated males take an intermediate position. Like in the case of females, most stable marriages are found among males with secondary education. The direction of the effect of education does not change after adding any other and even all the control variables but economic activity, which demonstrates that for males this structural factor is of very high importance. Only in the model adjusted for the effect of economic activity, the risk of divorce for the least educated males becomes lower than for the most educated (Model 2). Importantly, in the same way as for females, urbanrural place of residence and place of birth have a strong impact on the education gradient by markedly increasing the risk of divorce in the group of males with lower than secondary education and to some extent also among those with secondary education (Models 4 and 5). Regarding the effect of economic activity on first divorce, our analysis has revealed that unemployed economically active males run a significantly higher divorce risk than employed males, while for females there is completely no between the two groups (Table 3). In the economically inactive sub-population, economically inactive males generally show an increased divorce risk as compared to employed males as well. The group of “other” economically inactive males stand out as being the most prone to divorce. Househusbands are the only group tending to deviate from this trend. However, differences in divorce risk between househusbands and employed 7 males are statistically non-significant in most of the cases, including the model controlling for the entire set of variables (Model 6). The picture for economically inactive females is quite contrary to males. With the exception of disabled persons and students for whom the results are statistically nonsignificant, divorce among economically inactive females is much less frequent than among employed ones. Housewives constitute a group of females with the lowest probability of marital dissolution. A reduced likelihood of divorce among economically inactive females persists even after controlling for all the compositional effects (Model 6). Table 2. Poisson regression relative first divorce risks by education adjusted for different sets of control variables. Lithuanian females and males, 2001-2003. Education category Model 1 Model 1A Model 2 Model 3 Model 4 Model 5 Model 6 (controlled for duration of marriage, marriage cohort, and age at first marriage) (Model 1 + number of children) Model 1 + economic activity status) Model 1 + ethnicity) Model 1 + and place of residence Model 1 + place of birth) (all varsiables) 1.00 0.93*** 1.00 0.94*** 1.00 1.00 FEMALES Higher 1.00 0.87*** 1.00 0.90*** 1.00 0.89*** 1.00 0.88*** 0.84-0.90 0.87-0.93 0.86-0.92 0.85-0.91 0.90-0.97 0.91-0.98 0.96-1.03 0.92** 1.01 0.97 0.93** 1.12*** 1.07** 1.20*** 0.87-0.97 0.95-1.06 0.92-1.03 0.88-0.98 1.06-1.18 1.01-1.13 1.14-1.28 - 1532 (4) p≤0.001 899 (6) p≤0.001 MALES 622 (4) p≤0.001 1230 (1) p≤0.001 1075 (3) p≤0.001 3615 (18) p≤0.001 Higher 1.00 0.93*** - 1.00 0.88*** 1.00 0.95* 1.00 1.02 1.00 1.02 1.00 1.00 Secondary Lower than secondary LR test, Chi-square (df) 0.89-0.97 0.84-0.91 0.91-0.99 0.98-1.06 0.98-1.06 0.96-1.04 1.19*** Secondary Lower than secondary LR test, Chi-square (df) 1.07** - 1.02-1.13 - - 0.95* 1.11*** 1.30*** 1.24*** 0.90-1.00 1.05-1.17 1.23-1.37 1.18-1.31 1.12-1.25 940 (6) p≤0.001 924 (4) p≤0.001 1152 (1) p≤0.001 1264 (3) p≤0.001 3351 (14) p≤0.001 Table 3. Poisson regression relative first divorce risks by economic activity status, adjusted for different sets of control variables. Lithuanian females and males, 20012003. Model 1 Model 1A Model 2 Model 3 Model 4 Model 5 Model 6 (controlled for duration of marriage, marriage cohort, and age at first marriage) (Model 1 + number of children) Model 1 + education) Model 1 + ethnicity) Model 1 + and place of residence Model 1 + place of birth) (all varsiables) 1.00 1.00 1.00 1.00 FEMALES Active, employed 1.00 1.00 1.00 8 Active, unemployed Inactive, disabled Inactive, housewife Inactive, student 1.00 1.01 1.01 1.00 1.03 1.01 1.03 0.96-1.04 0.97-1.06 0.97-1.05 0.96-1.04 0.99-1.07 0.97-1.05 0.96-1.04 0.97 0.95 0.98 0.98 1.03 1.01 0.98 0.88-1.08 0.86-1.05 0.89-1.09 0.89-1.08 0.93-1.14 0.91-1.12 0.88-1.08 0.60*** 0.68*** 0.60*** 0.60*** 0.70*** 0.64*** 0.75*** 0.57-0.63 0.65-0.72 0.57-0.63 0.57-0.63 0.66-0.73 0.61-0.67 0.57-0.63 1.07 1.04 1.08 1.06 1.04 1.03 1.00 0.97-1.17 0.95-1.14 0.98-1.18 0.96-1.16 0.95-1.14 0.94-1.13 0.97-1.17 0.83*** 0.88** 0.83*** 0.83*** 0.85*** 0.83*** 0.88** Inactive, other LR Chi-square (df) 0.75-0.91 0.80-0.97 0.76-0.91 0.76-0.91 0.77-0.93 0.75-0.91 0.75-0.91 - 1258 (4) p≤0.001 47 (2) p≤0.001 MALES 493 (4) p≤0.001 962 (1) p≤0.001 928 (3) p≤0.001 2763 (14) p≤0.001 Active, employed Active, unemployed 1.00 1.36*** - 1.00 1.37*** 1.00 1.37*** 1.00 1.42*** 1.00 1.38*** 1.00 1.41*** 1.32-1.43 1.32-1.42 1.37-1.47 1.33-1.43 1.36-1.46 1.19*** 1.20*** 1.27*** 1.23*** 1.26*** 1.09-1.31 1.09-1.32 1.15-1.39 1.12-1.35 1.14-1.38 Inactive, disabled Inactive, househusband 1.31-1.41 1.18*** 1.08-1.30 0.87* - 0.76-1.00 1.33** Inactive, student - 1.10-1.60 - 0.88 0.88 1.07 0.94 1.06 0.77-1.01 0.77-1.01 0.93-1.22 0.82-1.08 0.92-1.22 1.34** 1.25* 1.27* 1.26* 1.21* 1.11-1.61 1.04-1.51 1.05-1.53 1.04-1.52 1.01-1.47 1.37*** 1.38*** 1.36*** 1.45*** 1.37*** 1.41*** Inactive, other 1.28-1.47 1.29-1.48 1.27-1.46 1.35-1.55 1.28-1.47 1.31-1.51 LR 46(2) 838(4) 1151 (1) 1152 (3) 2457 (10) Chi-square (df) p≤0.001 p≤0.001 p≤0.001 p≤0.001 p≤0.001 Note: statistically significant relative risks are marked in bold. *** - p≤0.001; ** - p≤0.01; * - p≤0.05. Discussion The study finds evidence for the differentiating effect of individual socio-economic recourses, measured by the level of educational attainment and economic activity status, on the risk of first divorce in Lithuania in the beginning of the 21st century. One of the major advantages of the study concerns the usage of the census-linked data covering the entire adult population of Lithuania. Thus, all the statistical analyses in this study are based on a very large sample size which allows producing statistically robust group-specific estimates of first divorce risks. Differently from surveys, these data also cover vulnerable population groups such as people in institutions. The study has some limitations which should be considered before interpreting the results. One of the most important shortcomings is related to the fact that both the socio-economic variables used in the analysis are time-constant and refer to the date of the census. We partially overcome this disadvantage by creating very broad categories (e.g., three categories for education) and by focusing on a relatively short period of observation (2.5 years). To check a potential impact of changes in socioeconomic status on our results, we performed a sensitivity analysis restricting the period of observation to one year. This test returned very similar results and confirmed our findings. Finally, another possible drawback of the data is related to a potential bias in marriage years under risk due to underestimation of true emigration levels (only official emigration records were used). However, using indirect estimates of unregistered (undeclared) emigration by Statistics Lithuania (Statistics Lithuania 2008), we found that this undercount may have only a very minor effect on divorce 9 rates in Lithuania. In addition, our study is restricted only to married people who generally have much lower migration rates if compared to other population groups. The study found a negative educational gradient in divorce risk for both males and females. Interestingly, for females this gradient was positive in the initial model controlling only for the union-specific demographic characteristics. The gradient reversed after having the control for urban-rural place of residence and place of birth added into the model. The study has also revealed that generally being out of the labour market destabilizes marriage and significantly increases the risk of marital disruption for males only. For males this association is sustained across the whole range of groups of economic inactivity. Unemployed and “other” inactive males are the most vulnerable groups showing the highest risk of first divorce. For females being out of the labour market either decreases the risk of first divorce as in the case of housewives and in the group of “other” economically inactive females or has no influence as for unemployed and economically inactive disabled females. One of the most common explanations for the negative relationship between individual socio-economic resources and the risk of marital dissolution rests on the relational stress argument, which states that economic hardships create relational tensions within the couple (Fisher and Liefbroer 2006). Marriage in the modern society is characterized by decreasing importance of production complementarities between husband and wife and is increasingly based on consumption complementarities and risk pooling (Stevenson and Wolfers 2007). It is possible that, in societies with a predominant dual earner family model, being out of employment or having poor prospects in the labour market due to low educational capital lead to lower gains to marriage from consumption complementarities and reduces the risk pooling function of marriage. It has been suggested that decreases in consumption complementarities (being one of the benefits of marriages) may lead to stress in the marital relations between two spouses (Stevenson and Wolfers 2007). A dual earner family model is a long-standing tradition in the Lithuanian society (Aidis 2006; Motiejūnaitė 2010). Husband’s and wife’s income pooling was and remains the predominant strategy for ensuring household survival and consumption opportunities. Female employment rates were high during the soviet decades (around 80-85 per cent) and stay at a high level (about 60 per cent) until now (Kanopienė 1998; Statistics Lithuania 2012). The last two decades were marked by rapid transition from the socialist to the market economy and fundamental restructuring of the labour market, which subsequently led to remarkable shrinking of the heavy industry sector, traditionally filled with lower educated male workers. It is possible that due to lower educational resources, these population groups were left with limited adaptive capacities in the new economic reality and faced a higher risk of unemployment. The economically disadvantageous position was responsible for their reduced contribution to consumption opportunities of the family and thus to risk pooling gains of marriage, which in turn was causing a high level of relational stress in marriage. Another explanation for elevated divorce risk among people with lower socioeconomic resources emphasizes inadequate relational capital possessed by lower 10 educated population groups. The research evidence suggests that the higher educated have better communication and conflict solving skills (Amato 1996) that play a crucial role in situations of economic hardship. A higher propensity to divorce observed in lower social classes is also enhanced by low costs of divorce, which is usually the case in societies with high divorce rates. It develops simultaneously: when divorce behaviour becomes non-selective, social and economic costs of divorce decrease, and the educational gradient in divorce from positive turns into negative (De Graaf and Kalmijn 2006; Hoem 1997; Härkönen and Dronkers 2006). The pattern of divorce being more common among socio-economically disadvantageous population groups is also reinforced by traditional normative gender expectations. The fact that normative gender expectations prevailing on the societal and couple level exert great influence on the effect that male and female participation in the labour force has on the risk of divorce has been proved in numerous studies (Cooke 2004; Cooke and Gash 2010; Amato, Booth, Johnson, and Rogers 2007; Sigle – Rushton 2010). Despite predominance of dual earner family model and high female employment rates, the Lithuanian society can be still characterized by traditional gender culture that is resistant to structural gender role modernization (Juozeliūnienė, Kanopienė 1995; Maslauskaitė 2008; Stankuniene and Maslauskaite 2008). As shown by previous studies, normative expectations towards male roles in Lithuania are more homogeneous across generations and sexes, while expectations towards female roles seems to be more heterogeneous, with more egalitarian attitudes expressed in birth cohorts of younger and better educated females (Stankūnienė et al. 2003). In this traditional normative context, male unemployment or being out of the labour market would mean a non-compliance with the dominant masculinity norms (i.e., a breadwinner role) and thus would contribute to increasing the risk of marital dissolution. At the same time, for females becoming unemployed or dropping out of the labour market does not elevate divorce risk. It is possible that in the traditional normative context, a large share of these females accept or/and adapt to their status of economic inactivity by fulfilling expectations based on traditional gender roles. Substantially lower first divorce risk of housewives observed in this study support this hypothesis. However, it is mutually nonexclusive that this pattern comes out of limited opportunities possessed by housewives and other economically inactive or unemployed females to exit unhappy marriage (Becker, 1981). Almost all the above explanations mainly address the direct causal effect of socioeconomic status on first divorce. However, the existence of a reverse association between these two variables is also very likely (i.e., divorce could be leading to lower socio-economic status). Although in the Lithuanian dataset socio-economic status is fixed at the census baseline (before the occurrence of divorce), there could be other (indirect) forms of reverse causation. For example, Jalovaara (2001) suggests that lower socio-economic resources may point to the weakness of marriage related to the lack of interest to accumulate joint resources. Our study shed more light on the importance of socio-economic factors of first divorce in Lithuania and contributed in this way to filling the gap in knowledge of divorce predictors in the country and in the region of Central and Eastern Europe. Unfortunately, a restrictive nature of the census-linked data providing relatively few 11 individual characteristics allowed us neither to test a wider range of alternative explanations nor to more thoroughly examine confounding effects on education and economic activity status. It is possible that some unmeasured factors such as psychological characteristics or family background are important contributors to the propensity to dissolve the first marriage in Lithuania and that additional control for these effects would lead to some changes in the relative risks of the two socioeconomic variables. More comprehensive explanatory studies based on more detailed data and advanced methods employing control for unobserved heterogeneity are required in order to confirm and explain the observed relationships. 12 References Aidis, Ruta. 2006. From Business Ownership to Informal Market Traders: the Characteristics of Female Entrepreneurship in Lithuania. In Friederike Welter, David Smallbone, Nina Isakova (eds.) Enterprising Women in Transitional Economies. Aldershot: Ashgate, pp. 119143. Amato, Paul R. 1996. Explaining the intergenerational transmission of divorce, Journal of Marriage and the Family 58(3): 628-640. Amato, Paul R., and Spencer James. 2010. Divorce in Europe and the United States: Commonalities and differences across nations, Family Science 1(1): 2-13. Amato, Paul. R., Alan Booth, David R. Johnson, and Stacy J. Rogers. 2007. Alone together: How marriage in America is changing. Cambridge, MA: Harvard University Press. Avdeev, Alexandre, and Alain Monnier. 2000. Marriage in Russia: a complex phenomenon poorly understood, Population: An English Selection 12: 7-50. Becker, Gary S. 1981. A treatise on the family. Cambridge: Harvard University Press. Bukodi, Erzsébet, and Péter Róbert. 2003. Union disruption in Hungary, International Journal of Sociology 33(1): 64-94. Burgess, Simon, Carol Propper, and Arnstein Aassve. 2003. The role of income in marriage and divorce transitions among young Americans, Journal of Population Economics 16(3): 455-475. Chan, Tak Wing, and Brendan Halpin. 2002. Union dissolution in the United Kingdom, International Journal of Sociology 32(4): 76-93. Cherlin, Andrew J. 1992. Marriage, divorce, remarriage. Cambridge, MA: Harvard University Press. Cooke, Lynn Prince. 2004. The gendered division of labor and family outcomes in Germany, Journal of Marriage and Family 66(5): 1246-1259. Cooke, Lynn Prince, and Vanessa Gash. 2010. Wives’ part-time employment and marital stability in Great Britain, West Germany and the United States, Sociology 44(6): 1091-1108. Council of Europe. 2006. Recent demographic developments in Europe. Strasbourg: Council of Europe Publishing. Darsky, Leonid, and Sergei Scherbov. 1995. Marital status behavior of women in the former Soviet Republics, European Journal of Population 11: 31-62. De Graaf, Paul M., and Matthijs Kalmijn. 2006. Change and stability in the social determinants of divorce: a comparison of marriage cohorts in the Netherlands, European Sociological Review 22(5): 561-572. Fischer, Tamar and Aart C. Liefbroer. 2006. For Richer, For Poorer: The impact of macroeconomic conditions on union dissolution rates in the Netherlands 1972-1996, European Sociological Review 22(5): 519-532. GGP Contextual Database. 2012. The Generations and Gender Programme, http://www.demogr.mpg.de/cgi-bin/databases/GGP/index.plx?dest=nidi. Hansen, Hans-Tore. 2005. Unemployment and marital dissolution: a panel data study of Norway, European Sociological Review 21(2), 135–148. Härkönen, Juho, and Jaap Dronkers. 2006. Stability and change in the educational gradient of divorce. A comparison of seventeen countries, European Sociological Review 22(5): 501-517. Hoem, Jan M. 1997. Educational gradients in divorce risks in Sweden in recent decades, Population Studies 51(1): 19-27. Jalovaara, Marika. 2001. Socio-economic status and divorce in first marriages in Finland 1991-93, Population Studies 55(2): 119-33. Jalovaara, Marika. 2003. The joint effects of marriage partners’ socioeconomic positions on the risk of divorce, Demography 40(1): 67-81. Juozeliūnienė, Irena, Kanopienė, Vida. 1995. Women and Family in Lithuania. In Lobodzinska, Barabara. (ed.) Family, Women, and Emloyment in CentrasEastern Europe. Westport: Greenwood Press, pp. 155–165. Kalmijn, Matthijs, Paul M. De Graaf, and Anne-Rigt Portman. 2004. Interactions between cultural and economic determinants of divorce in the Netherlands, Journal of Marriage and the Family 66(1): 75-89. 13 Kanopienė, Vida. 1998. Women in Economy. In Suzanne LaFont (ed.) Women in Transition. Voices from Lithuania. NY, pp. 68-80. Lyngstad, Torkild Hovde, and Marika Jalovaara. 2010. A review of the antecedents of union dissolution, Demographic Research 23(10): 257-292. Maslauskaitė, Aušra. 2008. Moterų užimtumas ir lyčių kultūra: lyginamoji Lietuvos ir Europos šalių analizė [Female Employment and Gender Culture: Lithuania in a comparative perspective]. In Moterys, darbas, šeima [Women, Work, and Family]. Vilnius: Vilnius university Press, pp.18-62. Motiejūnaitė, Akvilė. 2010. Female Employment in Lithuania: Testing Three Popular Explanations. Journal of Baltic Studies 41(2): 237-258. Muszynska, Magdalena. 2008. Women’s employment and union dissolution in a changing socio-economic context in Russia, Demographic Research 8(6): 181-204. Oppenheimer, Valerie. 1997. Women's employment and the gain to marriage: the specialization and trading model, Annual Review of Sociology 23(1): 431-453. Philipov, Dimiter, and Aiva Jasilioniene. 2008. Union formation and fertility in Bulgaria and Russia: a life table description of recent trends, Demographic Research 19(62): 2057-2114. Rogers, Stacy J. 2004. Dollars, dependency, and divorce: four perspectives on the role of wives’ income, Journal of Marriage and Family 66(1): 59-74. Ruggles, Steven. 1997. The rise of divorce and separation in the United States, 1880-1990, Demography 34(4): 455-466. Scherbov, Sergei, and Harrie van Vianen. 1999. Marital and fertility careers of Russian women born between 1910 and 1934, Population and Development Review 25(1): 129-143. Sigle-Rushton, Wendy. 2010. Men's unpaid work and divorce: reassessing specialization and trade in British families, Feminist Economics 16(2): 1-26. Sobotka, Tomáš, and Laurent Toulemon. 2008. Changing family and partnership behaviour: common trends and persistent diversity, Demographic Research 19(6): 85-138. Stankuniene, Vlada. 2006. Santuokos, istuokos, santuokinis statusas [Marriage, divorce, marital status], in V. Stankuniene (ed.), Lietuvos gyventojai: struktura ir demografine raida [Population of Lithuania: composition and demographic development]. Vilnius: Department of Statistics and Institute for Social Research, pp. 101-115. Stankuniene, Vlada, and Ausra Maslauskaite. 2008. Family transformations in the postcommunist countries: Attitudes towards changes; in Charlotte Höhn, Dragana Avramov, and Irena E. Kotowska (eds.), People, Population Change and Policy Acceptance Study (Vol. 1). Berlin: Springer, pp. 127-157. Stankūnienė, Vlada, Jonkarytė, Aiva, Mikulionienė, Sarmitė, Mitrikas, Algimantas, Maslauskaitė, Aušra. 2003. Šeimos revoliucija? Iššūkiai šeimos politikai. [Family revolution? Challenges for Family Policy]. Vilnius: Institute for Social Resarch. Statistics Lithuania. 2008. International migration of Lithuanian population. Vilnius: Statistics Lithuania. Statistics Lithuania. 2011. Demographic Yearbook 2010. Vilnius: Statistics Lithuania. Statistics Lithuania. 2012. Women and men in Lithuania 2011. Vilnius: Statistics Lithuania. Stevenson, Betsey, and Justin Wolfers. 2007. Marriage and divorce: changes and their driving forces, Journal of Economic Perspectives 21(2): 27-52. Styrc, Marta, and Anna Matysiak. 2012. Job and stable marriage? Effects of women's employment on marital stability in Poland. Paper presented at the European Divorce Research Conference, Helsinki. Svarer, Michael, and Mette Verner. 2006. Do children stabilize Danish marriages?, Journal of Population Economics 21(2): 395-417. 14 Annex 1. Marriage years at risk (in thousands and per cent) by control variables and the relative divorce rate ratios from a model including all variables, Lithuania, adult females and males, 2001-2003 Females Marriage years at Divorce risk risk ratio Duration of marriage (years) <1 4.4 (0.3%) 1 - 1.99 33.1 (2.3%) 2 - 2.99 36.2 (2.5%) 3 - 3.99 36.8 (2.5%) 4 - 4.99 36.6 (2.5%) 5 - 5.99 38.5 (2.6%) 6 - 7.99 83.3 (5.7%) 8 - 9.99 90.7 (6.2%) 10 - 12.99 164.2 (11.2%) > 13 (ref.) 941.3 (64.3%) Marriage cohort Before 1970 132.5 (9.0%) 1970 - 1979 363.0 (24.8%) 1980-1989 498.9 (34.1%) 1990 - 1994 243.5 (16.6%) 1995 - 2001 227.2 (15.5%) Age at marriage (years) < 20 269.9 (18.4%) 20-24 (ref.) 880.3 (60.1%) 25-29 236.0 (16.1%) > 30 78.8 (5.4%) Number of children No children 0.9 (0.1%) 1 child 2 children 412.3 (28.1%) 722.6 (49.3%) 3 or more children 236.2 (16.1%) Unknown 93.0 (6.4%) 0.84 0.66-1.08 1.03 0.89-1.18 1.46*** 1.27-1.67 1.58*** 1.38-1.80 1.62*** 1.42-1.85 1.46*** 1.28-1.67 1.59*** 1.44-1.77 1.49*** 1.36-1.64 1.40*** 1.31-1.50 1.00 0.19*** 0.16-0.22 0.46*** 0.40-0.52 0.94 0.84-1.05 0.97 0.89-1.05 1.00 1.47*** 1.42-1.52 1.00 Males Marriage years at risk 4.1 (0.3%) 32.1 (2.3%) 35.1 (2.6%) 35.8 (2.6%) 35.6 (2.6%) 37.4 (2.7%) 81.8 (6.0%) 88.7 (6.5%) 160.9 (11.8%) 856.3 (62.6%) 76.5 (5.6%) 342.4 (25.0%) 489.5 (35.8%) 238.4 (17.4%) 221.0 (16.2%) 71.1 (5.2%) 786.4 (57.5%) Divorce risk ratio 1.41** 1.09-1.82 1.59*** 1.38-1.84 1.99*** 1.73-2.28 2.24*** 1.96-2.57 2.16*** 1.89-2.48 1.90*** 1.65-2.18 2.04*** 1.84-2.27 1.84*** 1.67-2.02 1.56*** 1.45-1.68 1.00 0.18*** 0.15-0.22 0.43*** 0.38-0.49 0.88* 0.78-0.99 0.89** 0.82-0.97 1.00 1.43*** 1.36-1.51 1.00 0.77*** 0.74-0.81 0.67*** 0.62-0.71 392.3 (28.7%) 1.01 0.64-1.60 1.00 0.64*** 0.62-0.66 0.63*** 0.59-0.67 1.30*** 1.24-1.36 - - - - - - - - 118.0 (8.6%) 0.75*** 0.72-0.77 0.71*** 0.67-0.75 Note: statistically significant relative risks are marked in bold. *** - p≤0.001; ** - p≤0.01; * - p≤0.05. 15 Annex 1 (continued). Marriage years at risk (in thousands and per cent) by control variables and the relative divorce rate ratios from a model including all variables, Lithuania 2001-2003. Females Marriage years at Divorce risk risk ratio Ethnicity Lithuanian (ref.) Russian 1220.0 (83.3%) 89.2 (6.1%) Polish 110.7 (7.6%) Other 44.5 (3.0%) Unknown 0.8 (0.1%) Males Marriage years at risk 1.00 1.03 0.97-1.09 0.70*** 0.66-0.74 0.86** 0.79-0.95 4.99*** 4.17-5.99 1136.4 (83.1%) 86.0 (6.3%) 96.7 (7.1%) 47.6 (3.5%) 1.1 (0.1%) Divorce risk ratio 1.00 1.06 1.00-1.13 0.71*** 0.67-0.76 0.87** 0.79-0.95 7.89*** 6.88-9.05 Place of residence Urban (ref.) Rural 1019.1 (69.6%) 446.0 (30.4%) 1.00 0.70*** 0.67-0.72 940.5 (68.8%) 427.2 (31.2%) 1.00 0.62*** 0.59-0.64 Place of birth Urban, Lithuania (ref.) Rural, Lithuania 610.8 (41.7%) 751.2 (51.3%) 1.00 0.76*** 0.74-0.79 1.01 0.94-1.08 2.09*** 1.79-2.43 563.9 (41.2%) 703.4 (51.4%) 1.00 0.76*** 0.73-0.78 1.00 0.94-1.07 3.32*** 2.91-3.79 Other country 99.0 (6.8%) Unknown 4.0 (0.3%) 97.1 (7.1%) 3.3 (0.2%) Note: statistically significant relative risks are marked in bold. *** - p≤0.001; ** - p≤0.01; * - p≤0.05. 16