There was no significant association between the self-reported gender equality index and self-rated health, even though women who belonged to the most equal tertile had higher odds of reporting good health compared to those who belonged to the least equal tertile.
For men, there was a statistically significant association between their perception of gender equality in their partner relationship and rating their health as good. For women, the associations were not statistically significant but showed the same directions as men. It is possible that perceiving good health and perceiving equality in your relationship are two sides of the same coin. The question "How do you rate the gender equality in your relationship?" might be understood in a different way than the way that we meant it to be understood. If the question is understood as "Do you have a good relationship?" then it is easy to understand why it is closely linked to self-rated health. Men and women might interpret this question in different ways.
Inconsistency between the two measures
The differences between self-perceived and self-reported gender equality show different patterns for men and women. Men perceived higher gender equality than they reported, whereas for women we observed the reverse - women perceived lower gender equality than they reported. Men seemed to regard themselves as gender equal in spite of tangible, measureable indicators showing the contrary. It might be difficult for men to acknowledge these differences, as they constitute the beneficiary group. This conclusion could be supported by the fact that women perceive less gender equality even when they report a relatively high consistency of measurable indicators in their daily practice. These two findings highlight the risk of failing to capture the core of the concept of gender equality. Power dynamics are known to be central in relationships [6, 37]. Presumably, the distribution of power is central for how people perceive gender equality, and power is difficult to describe and above all difficult to measure.
There might even be a larger difference between the two measurements, as we have been very generous in judging the division of households as equal. If more than 50% of the items were rated as shared equally, we recorded the couple's relationship as equal. Accepting 49% inequality and still ranking the relationship as equal must be considered very indulgent.
This study shows clearly that none of the applied measures is ideal. They tell different stories but are part of the same reality. Men obviously overrate self-perceived gender equality, suggesting that a single question on gender equality does not generate a valid result. On the other hand, the index used overrates equality among women, most likely because important aspects are lacking. A possible solution to this problem might be a combination of measurements - a genuinely gender-equal partnership needs to meet two criteria: (i) it should satisfy reasonable external requirements and (ii) it should be assessed by both partners as equal.
This inconsistency in how gender equality is understood and how it is expressed is observed not only when comparing men and women at an individual level, but also when comparing perceptions of gender equality on structural levels. For instance, although there is a political consensus in Sweden that gender equality is beneficial, there are differences between the parties in its definition . When a societal norm becomes sufficiently strong, it is very difficult to question it publicly, even for political parties. And in the family sphere, societal norms steer decisions even relating to matters that are individual and private.
The fact that 80% of parental leave is taken by mothers is often explained as the most rational choice for the family financially. In a recent study, however, differences in a couple's market productivity did not affect the time spent on household work. This study of market productivity used a theoretical model that assumed that rational economic behaviour explains the division of paid and unpaid work in a family; however, this could not be demonstrated .
The strong societal consensus on the goals for gender equality might lead to a duality - men and women both want to be progressive and gender equal, at the same time as they are acting rationally and solving domestic problems in the most practical way, which often means conforming to traditional family divisions, with household chores remaining the responsibility of the woman [40, 41]. This could possibly also be part of the explanation for the differences found in the current study between the two measures of gender equality. Individuals are often assumed to act in accordance with their self-interests, i.e. what is best for them. Miller suggests that there might be a need for more systematic and direct examination of variance in rational self-interest . The extent to which an individual's actions and/or attitudes reflect a deliberate consideration can vary and this must be taken into account .
Further research on variation in self-interest is necessary. The interests that control people's preferences are not always obvious. There might also be differences within groups that are greater than those between groups: even though evaluation of gender equality requires comparison between men and women, we must be aware of possible differences within each group.
The differences between people's own perceptions of gender equality and the self-reported index, attempting to visualize the national policy, can be interpreted in at least two ways. The first interpretation is that people are sufficiently competent to judge themselves whether or not they are gender equal. This view is supported by the idea of the free will and the individual's right to choose. The free will is, however, a complex matter and philosophers have long discussed the liberal paradox formulated by Sen . However, in a couple there is the problem of gender and its traditions and practice . As gender is a matter of relations between people and as women and men are socialized to lead somewhat different lives, it is difficult to speak of free choice: it must always be a question of interpreting different meanings and of having different perspectives. The second interpretation is that this is a good example of couples' ways of "doing gender" in a context where gender equality has high status . Regarding an unequal division of household work as gender equal could be a way of maintaining the prevailing systems - i.e. it could be a way of allowing two different norm systems to exist side by side - on one side, the discourse on gender equality as a goal of Swedish society, and on the other side the discourse of the traditional family norm system with different obligations for men and women. Many families with small children will recognize the situation where all available energy is needed to make everyday life function as smoothly as possible . By describing this situation as self-elected and preferable, the situation is justified, not least to oneself. Moller-Okin argued that the family is not, and cannot be viewed as, the "non-political" area. Social justice is a political goal in a democratic society, and theories of justice need to apply their standards even to the family. Moller-Okin also showed that when the family is argued as belonging to the private sphere, almost all justice theorists assume that "individual" in a family means the male head .
Gender equality and self-rated health
From the previous discussion, we can conclude that self-perceived gender equality among men is not a trustworthy measure. Men who perceive they are gender equal have high odds of reporting good health, but there are no reasons to believe there is a causal relationship. On the contrary, the theory of convergence, even though rather vague, would suggest that gender-equal men could expect to develop some of the ill health typical for women. When applying the index there was no change between the tertiles, and we interpret this to be a more plausible result.
Health is also a concept with many varied explanations. Women in Sweden and many other countries are less healthy than men, almost regardless of the measurement used. They have more days of sickness absence, use more resources of the health care system and report less good health. But at the same time women live longer. So the fact that women acknowledge their ill health may be beneficial for their survival. This is why we have introduced the theory of convergence, suggesting that greater gender equality will benefit both men and women. Women will live healthier lives and men will live longer.
Although Månsdotter acknowledges the potential biological, social and behavioural grounds for the different health patterns of men and women, she also makes clear that the division of male and female affairs is not static. Finding a new pathway to gender convergence might also lead to health convergence for men and women . This development can already be seen in men who participate in their children's upbringing . The long-standing life expectancy gap between men and women appears to be closing in many societies. Men take better care of their health than previously and lead less risky lives, while women adopt more traditionally masculine behaviours such as smoking and alcohol consumption.
For women in the study, health improved with increased gender equality, using both self-reported and self-assessed gender equality; however, the associations were not significant. This pattern was in line with our expectations as we measured self-reported health rather than mortality, and a larger sample might have resulted in significant differences.
In this study, respondents answered the questions individually and on behalf of their partner rather than as a couple. Putting questions to both partners in a relationship is possibly more valid, but very difficult.
The measurement of gender equality in the relationship based on the questions asked in the questionnaire, here called self-reported gender equality, is used as the normative dimension in this study. The responses to the survey questions comprise the respondent's information on how the couple divides responsibilities and time in their family. From this self-reported data we have produced an estimation of gender equality practice in the household, used as a normative comparison to the self-perceived gender equality question. As in our study to be denoted "Gender equal" it was sufficient to answer "We share equally" in 51% of the 14 questions, the risk of overestimating the numbers of gender-equal relationships is obvious. Nevertheless, the number of couples reporting their relationship as gender equal is low compared to how they perceive gender equality in the same relationship. With higher standards for what constitutes an equal relationship, we would have seen even bigger differences.
Although we have used survey data and compared gender equality in quantitative terms, we have been able to show the ambiguity of respondents in relation to this issue. A qualitative research approach would surely enrich the picture and, we believe, capture gender equality in relationships in more diverse ways. Ambiguous as the phenomenon of gender equality seems to be, elaboration on the basis of in-depth interviews would be of great importance. This would enable a deeper understanding of respondents' ideas and perceptions of gender equality in their lives. For method development, the more costly qualitative approach could prove essential; however, the quantitative methods may prove useful when developed with greater sensitivity for ambiguity. In this way, it would be possible to create a cost-effective and easily repeatable quantitative measurement. Our expressed aim to evaluate the gender equality discourse on the basis of both reported practice and self-perceived gender equality involves judging some consequences to be more important than others, and we emphasize that such judgement can and must be challenged. If the way people divide their time, money and responsibility in a relationship is not a sign of gender equality, what is? Have we missed any important issues in the private sphere that add to the issue or are even more descriptive of gender equality?
The use of principal component analysis in deriving the index of equality does not provide any absolute categories of observed inequality, meaning that it is not possible to ascertain the true level of domestic inequality. The index derived by PCA provides a relative comparison between different groups of respondents in terms of other variables, such as health outcome. For example, it would be possible to use the index derived here to study whether health outcome differentiates between households with the least and the greatest equality.
We compared the background characteristics of the included respondents with those who were excluded, i.e. those with complete and incomplete information respectively, and found that respondents with incomplete information tended to be younger, have a lower level of education, and lower income. This might be a source of selection bias in our study.