Couples’ Relative Resources, Male Power, and Relationship Conflict from a Comparative Perspective

Using two waves of the Generations and Gender Survey for eight European countries, we test under what conditions couples experience high levels of disagreement over time or separate. The results partly support the idea of relative resources, suggesting that a decrease in the status of men in couples (job loss) is significantly associated with high levels of conflict. The transition to high conflict is more frequent when there is a discrepancy between policy and behavior. Social policies designed to meet the needs of working parents in dual-earner couples together with the diffusion of gender egalitarian values can lead to a reduction in unhealthy levels of couple conflict.


Introduction
Conflict is inevitable in any meaningful relationship. It often involves one partner's attempt to solve problems by persuading the other partner to look at things differently. This is considered a "constructive conflict," in which We make a number of contributions to the literature on marital and relationship conflict. First, we study the increase in conflict using longitudinal data, which is the ideal design for examining the causal mechanisms between the loss of relative resources in couples and relationship conflict. Second, we analyze the dynamics of relationship conflict, taking into account a competing risk framework in which unbearable levels of disagreement can lead to divorce or separation. Third, we analyze relationship conflict in countries where policies, gender roles, and identities in family life are substantially different, making our results particularly relevant to the effect of contextual factors.
Our results indicate that men's job loss is related to the transition to high conflict for both men and women. In our view, this demonstrates that challenging hegemonic masculinity-that is, men with lower status than their female partner-is often associated with situations of conflict. Interestingly, we also find that the transition to high conflict is more frequent in countries where the discrepancy between policy design and prevalent family models is largest and where its citizens have more often rather traditional gender egalitarian views. In addition, we find that men and women separate or divorce more often-after controlling for individual-level composition effects-in the three Western European countries of our sample (all "Conservative" welfare regimes) than in Eastern European countries. We discuss the policy implications of our results in the concluding section.

What Causes Conflict in Romantic Relationships? Relative Resources, Gender Roles, and Male Power
Conflicts and disagreements are common aspects of any relationship (Theiss & Leustek, 2016). However, when disagreements arise very frequently, they often reflect situations of hostility, which can be particularly harmful to couples and other family members (McCoy et al., 2013). Studies have shown that there are different issues associated with marital conflicts, such as money management (Papp et al., 2009), household division of labor (Kluwer et al., 1997;Perry-Jenkins & Folk, 1994), the use of flexible work in dual-earner couples with children (Radcliffe & Cassell, 2015),and even women's participation in the labor market (Macmillan & Gartner, 2018). Yet most of these studies have not examined the dynamics of conflict, that is, the causal factors motivating the shift to hostile and unhealthy relationships. Therefore, in this section, we analyze the main theories and empirical evidence that contribute to our understanding of the dynamics of relationship conflict.
Two dominant theories have been used jointly to explain marital conflict: resource theory and feminist theories. The main premise of resource theory is that couples' bargaining power depends on the individual assets of each partner. Resources can either be symbolic like social prestige or material like income or wealth (Blood & Wolfe, 1960). Most theorists using this framework have analyzed the consequences of resource imbalance for couples in different dimensions, such as housework or the gender division of paid and unpaid work (Kluwer et al., 1997;Perry-Jenkins & Folk, 1994). The idea is that the partner with more absolute resources (e.g., he or she has a job or higher status) or with more relative resources in dimensions such as earnings, occupational status, or education can more easily impose his or her preferences in the household because inequalities in the distribution of resources often entail gender asymmetry in power and decision-making (Hobson, 1990;Evertsson, 2007;Killewald, 2016;Nitsche, 2016;Ruppanner, 2010; van Damme & Dykstra, 2018). The theory also assumes that people try to maximize their individual satisfaction at the expense of their partners who have fewer absolute or relative resources. When we apply this theory to the realm of intimate partner violence, the rationale is that violence is instrumental and emerges as a result of a total loss of access to resources or an imbalance in access to resources in the household (Hornung et al., 1981). A lack of power in one aspect of the couple relationship may be compensated for in another aspect.
Resource theory is very useful in explaining the risk that women experience abuse or conflictual relationships. However, this framework alone cannot explain why imbalanced resources lead to conflict in certain circumstances, as it ignores the importance of gender roles in couples' interactions. Adding feminist theories helps us better understand why changes in relative resources can lead to conflict. At the heart of these theories are two main concepts: (1) gender roles, which prescribe the ideal behavior and attitudes considered acceptable, appropriate, or desirable for women and men in a particular society and time, and (2) "doing gender," which is defined as the daily performance of "a complex of socially guided perceptual, interactional, and micropolitical activities that cast particular pursuits as expressions of masculine and feminine 'nature'" (West & Zimmerman, 2016, p. 126). The main idea is that women and men display their gender in the context of the family. In societies with a traditional gender division of labor, men are expected to perform the male breadwinner role and women the main caregiver role. In the scenario of a new gender balance in which women have higher status and more relative resources, men may use violence to restore their power in the relationship and display hegemonic masculinity (Connell & Messerschmidt, 2005;Macmillan & Gartner, 2018;Wojnicka, 2015). In other words, intimate partner violence and conflict are more likely to occur when men's authority in the household is threatened by women's higher power in the relationship.
Empirical studies using the framework of relative resources have obtained mixed results, and the relationship between individual resources and marital conflict remains unclear. For instance, Villarreal (2007) used resource theory to predict intimate partner violence in Mexico as an example of a country with traditional gender norms (e.g., 40% of Mexican women need their husbands' permission to work outside the home). He explored the association between male violence and female employment. In his study, violence was understood as a compensatory mechanism that allows men to maintain a dominant position in the relationship when they lose power. However, the results did not support the hypothesis of relative resources, but Villarreal found a correlation between women's absolute resources and gender-based violence. That is, an increase in women's absolute resources and independence through paid work reduced the likelihood of male violence. However, this study was based on cross-sectional data and failed to comprehensively explore the dynamics of violence.
In contrast, the study by Hatch and Bulcroft (2004) used data from the United States to study the main determinants of conflict. The authors were particularly interested in exploring the relationship between marriage duration and the levels of disagreement. Apart from that, it turned out that marriage duration has a complex relationship with the frequency of marital disagreements, the authors found that disagreements were more common among young and middle-aged couples and couples with children. However, selective attrition also played a role; couples with frequent disagreements were more likely to divorce or separate. Interestingly, the level of disagreement was higher among dual-earner couples than male breadwinner couples (i.e., employed husbands and housewives). This result was explained by the mediating role of the respondents' gender ideology. Indeed, their agreement with traditional gender roles significantly reduced marital conflict, with women adopting a more submissive role and avoiding or withdrawing from conflict. According to Hatch and Bulcroft (2004), this explained the lower levels of disagreement among older and more conservative couples. The main implication of their study is that disagreements between couples, in general, are expected to increase in more egalitarian societies favoring constructive dialog and differences.
Similarly, Atkinson et al. (2005) argued that gender violence is not so much related to absolute or relative resources, but to the gender ideology of couple members. They found that women's higher income did not pose a threat to their husbands' masculinity as long as the male respondents had an egalitarian gender ideology. Alonso-Borrego and Carrasco (2017) also found that the risk of physical abuse was lower among couples with egalitarian values in Spain. However, both studies only used one case study with crosssectional data and overlooked differences between policies and behavioral outcomes at the societal level. Based on current evidence, it is difficult to determine the relationship between marital conflict in intimate relationships (at the micro level) and socio-economic factors at the societal level. More comparative studies are needed to understand the role of context (Gracia & Merlo, 2016), as we do in this study.
Based on previous research, we test two main hypotheses: a hypothesis at the micro-level, which explores the role of relative resources on relationship conflict, and a macro-level hypothesis between the societal level and the individual level, exploring the role of the national context in the existence of different levels of relationship conflict or separation. First, we expect a decrease in the status of men to be associated with a significant increase in the level of conflict and in the risk of break up: the Gender Status Incongruence Hypothesis. We call it "status incongruence" because men's loss of status within the couple violates traditional gender roles and expectations. Couples' relative resources are measured by the change in the employment and occupational status of each partner over time.
We also expect that the risk of relationship conflict and separation depends on the national context. Macro-level factors, such as prevalent models of work-family arrangements (i.e., partners' working hours in paid and unpaid work) and family policies, may explain the levels of relationship conflict. The national context can influence the levels of relationship conflict by favoring family life or, conversely, by creating tension within the family. In particular, we predict that in countries where family policies are aligned with prevalent family-work arrangements, the likelihood of experiencing a transition to high conflict will be lower than in countries where both variables are aligned: the Family Policy Model Gap Hypothesis. Family-work arrangements refer to the main type of employment model in couples, for instance, whether most couples in a country follow the male breadwinner family model (only men work full-time and women are full-time homemakers), the oneand-a-half breadwinner model (men work full-time and women part-time), or the dual-earner family model (both partners work full-time). The family policy model gap refers to the discrepancy between the common family model in a country and state support for that model. Thus, in a society where the dual-earner family model prevails but families cannot count on generous state support to reconcile paid work and family responsibilities, we expect the large discrepancy between daily needs and the constraints of a good work-family balance to put pressure on couples. Therefore, the outbreak of high levels of conflict and divorce will differ based on the level of discrepancy between prevalent work-family arrangements and state support to achieve a good work-family balance. Table 1 includes several indicators for the countries analyzed. It reveals substantial differences in terms of state support for families, prevalent work-family arrangements, and gender values. According to these indicators and as indicated in the last column (Family Policy Model Gap), we identify three main contexts: (1) Low discrepancy country: strong male breadwinner model, characterized by very limited state support for family responsibilities and relatively traditional gender values at the societal level. This group is represented by Georgia, with very low scores for divorce tolerance and access to divorce, low scores for women's actual participation in paid work, and high scores for adherence to male breadwinner norms. In this context, male breadwinner families prevail, combined with non-existent state support for women to combine work and care and non-existent informal support for women to do paid work. In general, this system works well, unless the behavior of couples does not fit this male breadwinner model or men do not assume their (financial) responsibilities. In this case, tensions may arise, but we expect that in general, women will not question their intra-household division of labor, will not express their disagreement with certain situations (Hobson, 1990), and above all will not leave their partner (even if they may be in a situation of financial stress).
(2) Average discrepancy countries: one-and-a-half breadwinner model, characterized by couples with average scores for traditional gender role values, divorce tolerance and access to divorce, women's actual participation in paid work, and adherence to male breadwinner norms-Note that these countries score the highest in our analytical sample because we have no typical egalitarian countries (e.g., the Nordic countries) in our sample. Public spending on families and state support to reconcile work and care is higher than in the previous group. This group is represented by Austria, Germany, and France. The most egalitarian nation in the dataset is France in terms of attitudes and scope of care services and financial support. Intuitively, this one-and-a-half breadwinner model should work quite well in this cluster of countries, as formal family support via policies is relatively in line with couples' expectations of the state; that is, the state supports women financially and in kind to stay at home (at least until the children are of school age). Thus, from this perspective, there should be less tension and conflict in general in these societies. However, when comparing the level of egalitarian norms with the average behavior of couples, we see a difference between these two indicators. Couples seem to generally adhere to more egalitarian norms of the division of labor than what they show in their actual behavior, with most women working part-time. Therefore, we expect some dissatisfaction with the intra-household division of labor and disagreements expressed about it, eventually leading to separation for some couples.
(3) High discrepancy countries: dual-earners family model, characterized by couples with relatively traditional gender role values, high divorce rates, and high female participation in the labor force. The countries represented in this group are Eastern European countries: Bulgaria, Russia, Lithuania, and the Czech Republic. They all have in common virtually no financial and care support from the state for families, although in the Czech Republic (which we group in this cluster because of its high divorce rate and the predominance of the dual-earner work-family arrangement), care services have remained considerably high since the fall of the wall. Overall, this gap between policy and behavior/culture is likely to create tension and conflict in couples, especially when one partner loses his or her job. Financial precariousness may occur in combination with many other family problems, such as alcohol addiction (Malyutina et al., 2004), making it difficult for families to make ends meet. Therefore, we expect that especially in these "regime" types, relationship conflict (and the risk of separation) will be high.
Bulgaria is also included in this cluster despite its "hybrid" model. In this country, adherence to egalitarianism is relatively low and the different work-family arrangements are of equal size, that is, dual-earner families and families in which both partners are unemployed. Note also that this country has a relatively high number of female breadwinner families (refer to Appendix, Table  A1). These hybrid arrangements are combined with some family support, especially long parental leave, although the earnings compensation is rather low. For this relatively low support context with a combination of different workfamily arrangements, we also expect fairly high levels of conflict. In short, we expect more relationship conflict in the Eastern European countries (Bulgaria, Russia, Lithuania, and the Czech Republic), where there is a discrepancy between the prevalent family model and state support, and less couple conflict in Western nations (Germany, France, and Austria) and Georgia, respectively, where practices and policies are more aligned.

Potential Confounding Contextual Factors
An important contextual change that happened during the years of interview of the GGS (2004-2013) concerns the economic situation. We portray changes in this situation by showing the unemployment rate per country over the years 2004 till 2013 (refer to Appendix, Table A2). A brief look at the Table immediately shows that most countries do not experience an economic downturn with rising unemployment rates during this period, but in contrast, decreasing unemployment rates. We observe only for Georgia and Lithuania a severe increase in the unemployment rates over the studied GGS period (for both from 2006 to 2009). For men, scholars have found that the unemployment context does not impact upon the individual unemployment effect on separation. For women, loss of a job when the national unemployment rate is low may be less stressful than during periods of high unemployment (Solaz et al., 2020). Therefore, we expect to not find a large influence of the macrolevel context on men's unemployment effect. In contrast, for women, if we find a moderating effect of the unemployment context, it would be weak and only occurring in (reference country) Georgia and Lithuania.

Data and Methods
The study is based on wave 1 (2004)(2005)(2006)(2007) and wave 2 (2007-2013) of the GGS for eight countries (Bulgaria, Russia, Georgia, Germany, France, Austria, Lithuania, and the Czech Republic). We select men (N = 8,490) and women (N = 11,541) aged 18-60 years who are present in both waves and with the same partner to analyze the change over time in their level of conflict. After attrition and excluding missing values on independent variables, we are left with an analytical sample of 8,199 women and 5,665 men. In this analysis, we take into account the fact that during the 3 years between the two waves, conflicts between partners may have been so high that the couples broke up. This implies that we do not observe the transition to high conflict for these couples. To control for the possibility of separation (rather than staying in the relationship), we perform competing risk analysis on the person-wave file, that is, multinomial logistic regression with three outcomes: (1) stay in a calm situation (no conflict or low conflict); (2) transition to high conflict; and (3) separate.
Theoretically, we could distinguish a fourth competing risk: attrition. However, this outcome is difficult to add to this analysis, as we include variables that change between waves 1 and 2. As a sensitivity analysis, we use multinomial logistic regression with attrition as the fourth possible outcome and include only wave 1 variables (refer to Appendix, Table A3 for the correlates of attrition). Attrition is relatively high in some countries and ranges from 15% for women in Georgia to 80% for men in Lithuania. We find that attrition is related to low-educated couples, being older, traditional, less committed to marriage, cohabiting, having separated parents, and having poor health. This suggests that in general, more disadvantaged people drop out between waves 1 and 2, although being older when the union was formed are also factors indicating attrition. If disadvantaged people leave the GGS panel and this is positively related to a higher likelihood of conflict or separation, we are left with a selective panel of more advantaged men and women. If on top of that, we expect the loss of male power to be more common in the more disadvantaged population, it seems likely that we underestimate the relationship between the loss of male power and our outcomes of interest. Most importantly, we believe that key in this estimation is the finding that more traditional respondents are more likely to drop out in the second wave. This leads us to expect that, if our sample suffers from any bias, we will underestimate the impact of status inconsistencies in couples on the transition to conflict or separation because we observe less traditional couples (that are expected to be more sensitive to status inconsistencies).
We analyze men and women separately because the level of disagreement is only captured at the individual level (i.e., not couple data). We know the respondents' perception of disagreement, but we do not know their partners' feelings. However, comparing men and women appears relevant as they report slightly different answers (see Tables 2 and 3). Women report more transitions to a hostile situation and to separation, while men are slightly more mobile in socio-economic terms between the two GGS waves than women.

Disagreement Index and Transition to a High Conflict Situation
We approach couple conflict through the severity of disagreements in daily interactions. To measure the level of disagreement in the relationship, we create a disagreement index of eight items related to interactions: household chores, money, use of leisure time, sex, relationships with friends, parents, and in-laws, having children, and drinking alcohol. A ninth item related to child-rearing issues is analyzed separately only for couples with children. The reliability of the scale is 0.81 in waves 1 and 2.
The transition to a high conflict situation occurs when in wave 1 the respondent reported a conflict level less than 2.5 out of 5, while (s) he reported a level of 2.5 or more in wave 2. Therefore, we construct the following variable: 0 "stable or transition to a low level of conflict" (<2.5 in both waves); 1 "transition to a high level of conflict" (from <2.5 to >2.5); and 2 "separate." Figure A1a and b in the Appendix present the distribution of the disagreement index by country. Some countries show a more right-skewed distribution of the level of disagreement than others. Notice the similarity of the distributions between men and women in the same country (even though this is not couple data).

Main Explanatory Variables: Loss of Male Power
We use two variables to capture the loss of male power in the couple (see Table 2 for descriptive statistics): (1) job loss: 0 "no job loss" (baseline); 1 "job loss of the male partner"; 2 "job loss of the female partner." Note that the job loss of both partners is included in the "job loss of the male partner" category because this category is too small to be analyzed on its own. (2) Change in the relative socio-economic status of the partners: 0 "no change in the relative socio-economic status of the partners" (the International Socio-Economic Index [ISEI] scores of the partners are the same between the two waves); 1 "an increase in the ISEI score of the female partner"; 2 "a decrease in the ISEI score of the male partner." Job loss is based on a change to a non-employment activity status of the respondent or the partner. ISEI scores are a standard measure in the literature to capture socio-economic status and are converted from ISCO codes. They combine an education and income component as a proxy for the respondents' occupational status. More specifically, ISEI scores refer to an occupation's main antecedent (education) and main consequence (earnings) as its formative parts. Hence, ISEI scores capture the cultural and economic resources that are typical of the incumbents of a certain occupation. This leads to ISEI scores between 16 and 90. For more information on the construction of ISEI scores, see Ganzeboom et al. (1992) and Ganzeboom's webpage: http://www. harryganzeboom.nl/ISCO88/index.htm. ISCED codes refer to cross-national comparable education levels, constructed by UNESCO (http://uis.unesco. org/en/topic/international-standard-classification-education-isced).

Contextual Variables
Our multinomial logistic regression models include the context to test the Family Policy Model Gap Hypothesis. As previously described, we distinguish three contexts: (1) Low discrepancy country: strong male breadwinner model (Georgia); (2) Average discrepancy countries: one-and-a-half breadwinner model (Germany, France, and Austria); and (3) High discrepancy countries: dual-earners family model (Bulgaria, Russia, Lithuania, the Czech Republic).

Control Variables
As shown in Table 2, we consider various control variables. At the couple level, we use the following controls: relative education of the partners: 0 "both partners have little education (ISCED 0-3)"; 1 "both partners are highly educated (ISCED 4-6)," but "she is more educated (ISCED 4-6) than him (ISCED 0-3)"; 2 "she is less educated (ISCED 0-3) than him (ISCED 4-6)"; marital status: married (baseline), premarital cohabitation, or cohabiting; age of the respondent when the union was formed; no change in the number of children (baseline), having a child added to the household (either by birth, adoption, or return to the household), or losing a child; and change in the respondent's perception of making ends meet (from very easy to somewhat difficult to difficult or very difficult).
In addition, we consider the following individual-level variables: age; age^2; ISEI score quartile or missing (baseline) to capture the floor or ceiling effects of the change in occupational status; health (1 "very good" to 5 "very poor"); average score for the index of egalitarian values for the following items, from 1 "strongly agree" to 5 "strongly disagree" (higher scores indicate more egalitarian values): "In a couple, it is better for the man to be older than the woman"; "If a woman earns more than her partner, it is not good for the relationship"; "In general, men make better political leaders than women"; "A preschool child is likely to suffer if his or her mother works"; "If their parents' divorce, it is better for the children to stay with the mother than with the father" (alpha = 0.63); average score for the commitment index for the following items, from 1 "strongly agree" to 5 "strongly disagree" (higher scores indicate more commitment to marriage): "Marriage is an outdated institution"; "It is fine for unmarried couples to live together even if they have no interest in marriage"; "Marriage is a relationship for life and should never be ended"; and "It is fine for a couple with an unhappy marriage to get a divorce even if they have children" (alpha = 0.60). Table 3 presents the percentages of men and women belonging to each of our outcome categories: stay in low conflict, transition to high conflict, or separate/divorce. Attrition is not included because it would distort the picture of our dependent variable too much. The proportion of couples increasing their level of conflict varies from 2% to 12% for men and from 3% to 11% for women, while the proportion of couples having broken up between the two waves is higher in some countries than in others (refer to Table 3). Indeed, we observe a significant variation between the countries. On one side, we have a group of countries characterized by very few transitions to conflict and separation, Bulgaria and Georgia. On the other side, some countries have unbalanced levels for both outcomes: Austria has a high level of separation but a low level of conflict, while Russia is the opposite, with a high level of conflict but a low level of separation (when based on men's responses). Finally, France, Germany, and Lithuania present medium and balanced levels of separation and deterioration of relationships.

Results
We explore the factors determining the transition to high conflict or union dissolution separately for men and women (refer to Table 4 and Table 5). As previously explained, using multinomial logistic regression with three alternative outcomes-(1) stay in low conflict (baseline), (2) transition to high conflict, (3) separation-allows us to avoid the bias that separation may have (continued)  on our results because of the strong correlation between relationship conflict and union dissolution. The models include a wide range of socio-demographic characteristics at the individual and couple levels that are important to control and the main explanatory variables related to the theory of relative resources: job loss and upgrading or downgrading of occupational status. The models also include an additional country-level variable for the family policy model gap. Table 4 and 5 show that men's job loss increases couple conflict according to both men's and women's responses (0.604 [p < .001] and 0.423 [p < .001], respectively). Men are exp(0.604) = 1.83 times more likely to transition to a conflict situation after losing their job between the two GGS waves, while the odds ratio is 1.53 based on women's responses. In contrast, women's job loss is not important for the transition to high conflict. This result, obtained by controlling for other individual characteristics and for the family policy gap model at the country level, is a clear indication of the importance of the loss of male power, which we theoretically relate to as the challenge of hegemonic masculinity and the risk of conflict. This is reinforced by our results (Table 4 and 5) on the role of job loss in the risk of separation: when the husband loses his job, the risk of separation also increases for men (0.562 [p < .05]), but not for women.
When considering our additional measures for the loss of male power (upgrading of her occupational status or downgrading of his occupational status), we observe that for men, these measures are positively associated with an increase in conflict (especially her upgrade = 0.275 [p < .05]), while for women, they are associated with a decrease in conflict, although the coefficients are not statistically significant (Table 4 and 5). However, they are positively associated with separation (upgrading of her occupational status for men = 0.972 [p < .001] and downgrading of his occupational status for women = 0.517 [p < .001]).
Our results also clearly show that the context is important and that the gap between policy and family models at the country level is a determinant of the quality and stability of relationships. Compared with the context of a low discrepancy country (Georgia), the risk of increasing conflict is highest in countries with a high discrepancy (1.067 [p < 0.001] for women and 1.471 [p < 0.001] for men; Bulgaria, Russia, Lithuania, and the Czech Republic), while the risk of separation is highest in countries with an average discrepancy (3.168 [p < .001] for women and 2.921 [p < .001] for men; Austria, Germany, and France). The effect sizes are more or less similar for men and women. That the impact of family policy-behavior discrepancy on separation is not linear may have to do with the relatively large sample size of Bulgaria and its comparatively low divorce rate that makes the cluster of Eastern European countries less divorce probable.
Finally, our models include additional socio-economic and socio-demographic variables. For the socio-economic situation of couples, we observe that for women, there is a positive effect of low job status (0.428 [p < .01] for the first quartile) on the risk of transition to high conflict. Note that although we control for the change in the respondents' perceived economic situation, we do not take into account the actual economic living conditions of the couples in wave 1. Therefore, women with occupational status in the first quartile are very likely to be poor and thus might have more financial problems to fight about. Why we do not find the same for men is puzzling, although the effect is in the same "positive" direction. In terms of couples' relative education, we only find a significant difference between highly educated homogamous couples and low-educated homogamous couples, and the former are more likely to transition to a high conflict situation but less likely to separate (men's responses). The main socio-demographic characteristics are associated with different risks of separation-in line with general data on divorce trends, the presence of children and cohabiting instead of being married are important determinants of the risk of union dissolution-but not really with the risk of experiencing a transition to high conflict. The arrival of a new child in the home (−0.472 [p < .001]), having egalitarian gender values (−1.95 [p < .05]), and being highly committed to the relationship (−0.167 [p < .05]) are factors that protect men from transitioning to high conflict, while cohabitation (0.365 [p < .001]) is associated with a higher risk of experiencing this transition for men. Finally, for both men and women, the increasingly more difficult economic situation seems to increase relationship conflict.
Overall, these results allow us to confirm that as postulated in our Gender Status Incongruence Hypothesis, the loss of male power in the couple (expressed through men's job loss) leads to a higher probability of transitioning to a high-conflict situation. Our results also confirm that as postulated in our Family Policy Model Gap Hypothesis, when the discrepancy between prevalent family models and state support is large, most couples are forced to adopt dual-earner arrangements in the face of not-fully-egalitarian gender norms and family policies. Therefore, we confirm our initial idea that considering the context when analyzing the change in relationship conflict is necessary when we want to understand the effect of the loss of male power, which can have a negative effect on the quality of relationships based on the existence of a normative context of hegemonic masculinity. 1

Conclusions and Discussion
This research had two objectives. First, we wanted to understand why some heterosexual couples are more likely to experience high levels of couple conflict. Second, we wanted to understand to what extent high levels of couple conflict are related to the discrepancy between prevailing family arrangements (the typical work-family balance of each partner) and practical state support for families at the country level. Based on feminist theories and the theory of relative resources, we hypothesized that the experience of men losing their status in the couple can lead to high levels of conflict, as it challenges expected gender roles in traditional societies. Our results (partly) confirm our Gender Status Incongruence Hypothesis for certain dimensions of relative resources (most importantly, job loss). In particular, the results reveal that when a man loses his job, it increases the likelihood of a couple shifting to a situation of severe relationship disagreements.
We also hypothesized that different levels of discrepancy between prevalent work-family arrangements, policy design, and gender values at the societal level would lead to different levels of relationship conflict. Our sample includes the following categories: "low discrepancy: strong male breadwinner family model in a traditional society with a non-supportive state" (Georgia); "average discrepancy: one-and-a-half breadwinner family model in a rather egalitarian society with an average supportive state" (Germany, France, and Austria); and "high discrepancy: dual-earners family model in a society that more and more develops into a traditional direction of gender role values and with a non-supportive state" (Bulgaria, Russia, Lithuania, and the Czech Republic). We find that in line with our hypothesis, the likelihood of experiencing high couple conflict varies with the level of discrepancy. The results indicate that the transition to high conflict is more frequent in countries characterized by average and high discrepancies between prevalent work-family arrangements and limited state support for working parents, respectively. We also find that a larger family policy model gap increases the likelihood of separation or divorce (although this effect is not "linear" and separation occurs more frequently in the average discrepancy context than in the high discrepancy context).
In addition, we observe gender differences in the perception of conflict; in particular, we find gender differences in the association between transition to conflict and men's job loss. This result is consistent with that of previous studies. For instance, Hatch and Bulcroft (2004) found that men reported more frequent marital disagreements than women, which they interpreted as the result of societies moving toward more gender egalitarian values and men feeling increasingly misplaced. However, other studies (Heaton & Blake, 1999) using data from matched partners have found similar reports of marital disagreements between men and women. Further analysis with larger datasets and longer observation periods may shed more light on gender differences in the perception of conflict.
We want to mention one important limitation of our study: we could not investigate a moderation by gender norms (or values on the individual level) of the relationship between men's job loss and the transition to high conflict between couples. This is because of lack of power due to the small sample size of the available countries in the panel version of GGS. We are looking forward to future research examining this, especially since our sample is relatively traditional with the absence of Nordic European countries. A comparison with the Nordic countries would give us a better idea of the validity of the relative resource theory in more gender egalitarian societies.

Implications
In summary, only certain dimensions of couples' relative resources seem to be significantly related to high levels of conflict (i.e., men's job loss). Even though not all measures of changes in power have a significant impact, the effect of men's job loss is in line with relative resource theory and more specifically with the gender status incongruence hypothesis. This implies that for couples where men are losing power (i.e., men are losing their job), the likelihood is higher that conflict levels rise to severe conflict. In our view, this shows that challenging hegemonic masculinity-that is, men have lower status than their female partner-is often associated with situations of conflict. In addition, we found that couples in certain contexts experience more conflict and have a higher risk to break-up. However, more countries are needed to identify better the specific policies at the country level that help reduce destructive levels of couple conflict due to status incongruence in couples. With more countries, a multi-level analysis could be done and one could look at cross-level interaction effects (i.e., to what extent does men's job loss have a higher impact in certain contexts than in other and why). In any case, although other factors (such as the economic situation of countries or path dependency) may play a role, we can carefully conclude that the main policy lesson of our study is that the transition to high relationship conflict is lower in countries where work-family arrangements are aligned with prevalent gender values and family policies. As a result, broad and generous policies combining work and care for families in a fair and equal way for both partners are essential to avoid conflict and hostile situations between couples, which can eventually lead to separation.