Is There a Gender Gap in Support for Distributive Principles in Old Age Pension Schemes?

A number of studies have identified a gender gap in social attitudes, but few have studied the phenomenon in the context of old age pension schemes. Using data from a 2019 Norwegian survey, we examined patterns of support for three distribution principles in pensions (anti-poverty, merit, and equality). Compared to men, women are more supportive of the antipoverty and equality principles, and less supportive of the merit principle. The gender difference for the anti-poverty principle persists in multivariate analyses, but disappears for the merit and equality principles when we control for socioeconomic factors, particularly income. Analyzing men and women separately revealed that partnership status and health interact with gender. Income however appears to affect men and women equally; high-income women are as likely as highincome men to support the principle of merit and reject the principle of equality. ARTICLE HISTORY Received 2 December 2021 Accepted 9 December 2022


Introduction
One puzzle of modern politics is the pervasive gender gap in social attitudes, which has been identified in a number of studies across diverse countries and policy areas (e.g., Bergh, 2007;Finseraas et al., 2012;Shorrocks & Grasso, 2020) and seemingly always in the same direction: compared to men, women are more enthusiastic about a generous, redistributive welfare state. They are more likely than men to support welfare policies benefitting the unemployed, sick, and elderly (Blekesaune & Quadagno, 2003), more often endorse state responsibility for social benefits and elder care (Mair et al., 2016), and score higher on indices supporting social spending in general (Banducci et al., 2016). Women are also more likely to vote for left-leaning parties (Bergh, 2007;Inglehart & Norris, 2000).
A number of studies have attempted to explain this gap in social attitudes as described below. The debate hinges on the relative importance given to differences in women's and men's strategic interests, on the one hand, and to gendered cultural values on the other. As might be expected given the complexity of the issue, a clear consensus has yet to emerge. We suggest that one way to advance this debate is to supplement studies of gender differences in opinions on general themes (e.g., welfare spending or allocating responsibility for covering certain social risks) with studies of gendered patterns in responses to questions about concrete social programs. Specifically, we highlight a social program with clear material consequences for the individual, namely, old age pensions.
In many cases, the design of old age pensions determines individuals' financial security in old age, and their distribution profiles tend to be particularly important for women (Halvorsen & Pedersen, 2019). Still, we know little about women's and men's preferred distribution profiles. This article discusses those preferences in light of the key distribution principles identified in the literature on distributive justice (Miller, 1999): equity/merit, need, and equality. Earlier studies have employed this framework to discuss pension preferences (Reeskens & van Oorschot, 2013) but not with an emphasis on gender. We consider the extent to which women and men agree or disagree that the system should be regulated so that no pensioner is poor (anti-poverty), that those with higher lifetime earnings should enjoy bigger pensions (merit), and that all should receive the same pension amount (equality). These dimensions encourage respondents to consider their own strategic interest as well as general notions about good and fair distributive principles.
The competing influences of interests and values would be hard to disentangle in settings where women and men lead very different lives, so we studied the question in Norway, which is consistently ranked as one of world's more gender equal countries (World Economic Forum, 2021). If we find non-trivial gender differences in this context that persist after controlling for demographic and socioeconomic factors, it will strengthen the argument that women hold fundamentally different values that are relevant to opinions on social issues.
In what follows, we first review the existing research on gender gaps in social attitudes generally and attitudes to pension systems particularly. This literature review informs our subsequent analyses. We then present the Norwegian case in more detail before describing our data, analytical strategy, results, and conclusions.

The gender gap in social attitudes
The literature offers three main explanations of why women and men hold different opinions on social issues: they may differ in their strategic interests, in core values, or in adaptations to family constellations. Studies that employ gender merely as a control variable as in previous studies on attitudes toward pension schemes (e.g. Reeskens & van Oorschot, 2013) are unable to distinguish between the three.
With regard to strategic interests, the argument is that women as a group have different interests than men. Because they earn less and spend less time on paid work and more on family, they are more economically vulnerable than men (e.g., Shorrocks & Grasso, 2020). They also have a strategic interest in supporting public services, as such services alleviate the burden of women's traditional caring tasks (Shorrocks & Grasso, 2020) and provide a degree of security in the case of divorce (Edlund & Pande, 2002;Iversen & Rosenbluth, 2006). Some authors also stress the role of feminism in raising awareness of women's strategic interests, showing that variations in "feminist consciousness" significantly impact the gender gap in attitudes (e.g., Bergh, 2007).
Another strand in the literature on this topic downplays the importance of strategic interests, instead emphasizing gender differences in value orientation. Stereotypically, men value competition and rewarding the stronger, while women value community and compassion for the weak. Gender differences in values may stem from differences in socialization, the roles women assume in life, innate differences rooted in evolutionary biology, or some combination of these. Variations of this explanation contend that women are fundamentally more empathetic (and less aggressive) than men (Kamas & Preston, 2019); it is also claimed that women on average are more risk averse than men (Borghans et al., 2009).
A third explanation highlights the potentially polarizing impact of gendered life experiences. For instance, women may develop different sympathies and priorities because their work puts them in contact with vulnerable groups, engendering a better understanding of their needs (Diekman & Schneider, 2010). Parenthood also appears to impact mothers and fathers differently; a relatively consistent finding in US data is that mothers are more supportive of government social welfare programs than non-mothers, while fatherhood is associated with more conservative views (Banducci et al., 2016;Elder & Greene, 2012). As for partnership, Finseraas et al. (2012) found that unmarried women were more likely to vote left than married women, while no similar correlation was found for men. Thus, both marital status and parenthood potentially interact with the gender gap.
The abovementioned studies all rely on analyses of generic values or voting patterns. We suggest that concrete questions about a specific social security program with crucial material implications for the individual may clarify the true importance of interests, values, and adaptations. In short, we argue that response patterns may differ when individuals' own financial security is at stake. A few extant studies indicate that men and women support different principles in the distribution of old age pensions. Using European data, Reeskens and van Oorschot (2013) found that women are slightly more likely than men to refute the merit principle. A previous study of Norwegian data (Grødem, 2019) complements this finding and shows that women are more likely to embrace the anti-poverty principle, but neither study systematically investigated possible gender differences or whether gender interacts with other aspects of individual situations, such as income or family status.
Recent studies of the Norwegian pension system have more explicitly highlighted gender differences. The outcome variables in these studies were attitudes to pension accrual for care work (Halrynjo et al., 2019) and to the sharing of accrual between husbands and wives (Kitterød & Pedersen, 2020). Women were more supportive of both, and the gender difference was lessened, but not eliminated, by demographic and socioeconomic controls, suggesting a real gender gap in the expected direction stemming from gendered value orientations. Both studies, however, investigated redistributive mechanisms with obvious gendered implications, so it remains to be seen whether this pattern is replicated when we look at general distribution principles in the pension system.
The old age pension is a recurring topic in Norway's political agenda. Existing studies show that, while the populace finds pensions complex and technical, it has a fairly good knowledge of the main distribution mechanisms (Grødem, 2019;Grødem & Kitterød, 2021), so most adults should be able to reflect adequately on the key principles. Below, we outline the main distributive features of Norway's pension system and offers hypotheses on gendered patterns of opinion.

The Norwegian context
Norway has made great strides toward gender equality and boasts strong institutional structures to support working mothers and fathers (World Economic Forum, 2021). In 2020, 73% of men and 67% of women were in paid employment, but the labor market is strongly gender segregated, with women dominating in the public and men in the private sector (Wagner et al., 2020). Women are more likely than men to work part time, and female-dominated jobs are typically less well paid than male-dominated ones. While men do more domestic work than before, women still do more than men (Ellingsaeter & Kitterød, 2021) and thus tend to accrue smaller pensions than men.
The core of Norway's current pension system is the National Insurance (NI). The system also includes mandatory occupational pensions and individual savings schemes with (very modest) tax rebates (Grødem & Hippe, 2020), but our survey explicitly asked only about the design of the NI. That scheme underwent comprehensive reform in 2011 with the introduction of a Notional Defined Contribution formula for accruing pension rights (Pedersen, 2017).
A key element of the new system is establishing a strong, transparent link between earnings and accrued pension rights, so the merit principle is baked into the system.
The differences in men's and women's typical working careers, however, inspired widespread fear that women would suffer in a purely merit-based approach, so the system has redistributive components that weaken the link between lifetime earnings and pensions, soften the merit principle, and introduce anti-poverty and equality elements. The primary mechanism for softening the merit principle is a cap on accrual for annual earnings. A core principle is that pension calculations are based on defined contributions, whereby all years of accrual between ages 13 and 75 count equally. Only incomes below 1.3% of the average full-time wage are included in the calculation (Halvorsen & Pedersen, 2019), which implies that the pension distribution will be flatter and more equal than that of wages. Given that men are more likely than women to have incomes in the higher brackets, more men than women are affected by the ceiling. Another important mitigating mechanism is pension accrual for care-giving for children below school age, in which child credits are awarded to the lower-earning parent (usually the mother) and correspond to annual earnings of about 80% of average full-time wages (Halvorsen & Pedersen, 2019).
The system also guarantees a minimum pension to everyone from age 67, regardless of previous labor market participation, which can be seen as an antipoverty measure. From an international perspective, this guaranteed pension is very high in relation to the income level of the working population (Halvorsen & Pedersen, 2022). In 2021, 22% of women and 4% of men drew the minimum pension (NAV, 2022).
Norway's current system thus includes elements of all three key distribution principles. The merit principle is found in the strong link between lifetime earnings and pensions, the equality principle in the accrual for care work and the cap on earnings, and the anti-poverty principle in the comparatively generous guaranteed pension. Based on what we know about men and women's preferences for social spending, it is reasonable to hypothesize that women and men may balance the three principles in different ways and that gender may interact with demographic and socioeconomic factors in this respect.
Against this background and on the basis of the extant literature, we expect bivariate analyses to show that women are more likely than men to support anti-poverty and equity principles and are less favorable toward merit principles. If the initial gender gap disappears when we control for demographic and socioeconomic factors (particularly income), it may be interpreted as evidence for the "strategic interest" hypothesis; if it persists, the "gendered values" hypothesis is supported. The emerging literature on gender polarization provides arguments for analyzing women and men separately. If we find that the control variables (especially having children and living with a partner) make women more equality oriented and/or men less so, we recognize support for the polarization hypothesis.

Data
To answer our research questions, we used data from the web-based Survey on Work and Pension conducted in November/December 2019. The net sample contains 3,097 respondents aged 30-61 (born in 1958-1989) recruited from the Gallup Panel, a sample of about 40,000 individuals who regularly participate in online surveys by the survey firm Kantar. An important aim of the survey was to study the respondents' retirement plans as well as their attitudes toward pension system design, which is why the sample was restricted to individuals of prime working age. Kantar weights the data to compensate for a slight underrepresentation of some groups (affecting gender, age, education, and geographical location). Our analyses are based on the weighted material, but the number of respondents is presented unweighted. We excluded 56 respondents who did not respond to one or more of the outcome variables or, in a few cases, who had not provided information on marital or parental status. The analysis sample comprises 3,020 respondents.

Dependent variables
The survey contained a battery of three questions on distribution principles in the pension scheme. The questions were introduced by the following text (translated from Norwegian): "There can be different views regarding the core principles of the old age pension scheme in the National Insurance. To what extent do you agree or disagree with the following statements: (1) The old age pension in the National Insurance scheme should be regulated so that no old age pensioners are poor. (2) Individuals who have had high incomes over their lifetimes should get bigger pensions from the National Insurance than individuals who have had low incomes over their lifetimes. (3) Everyone should receive the same amount of old age pension from the National Insurance regardless of how much they have worked or how much money they have made over their lifetime." Inspired by Reeskens and van Oorschot (2013), we describe the first as the anti-poverty principle, the second as the merit principle, and the third as the equality principle. The respondents answered on a 5-point scale ranging from 1 (agree completely) to 5 (disagree completely) with "neither agree nor disagree" at the midpoint. In the analyses, we reversed the scale so that higher values denote more support and lower values denote less support.

Independent variables
We included the following demographic and socioeconomic factors in the multivariate analysis, which came partly from the questionnaire and partly from information stored in the Gallup Panel: gender: dummy, women (1), men (0); age: continuous and squared; marital status: married (reference), cohabiting, single; children: dummy, at least one child (1), no child (0) (both children in the household and those who had moved away were included); education (highest completed level): elementary/high school/vocational (reference), university 1-4 years, university 5 or more years; income (annual individual gross): based on a question with categorical response alternatives; we included "missing" as a separate category, as about 7% of the analysis sample declined to answer this question; working hours (normal weekly hours): not in employment (reference), part-time work (1-36 hours), full-time work (≥37 hours); health: poor (reference), good, very good; the respondents reported whether their health was very good, good, poor, or very poor, but we collapsed the two latter alternatives, as very few considered their health to be very poor.
Descriptive statistics for the whole analysis sample and separately for women and men are presented in the Online Appendix, Table 1. Women and men in the analysis sample were about the same mean age, but women were slightly less likely than men to be formally married and were more likely to have at least one child. More women than men had a university degree, but women generally had lower incomes than men, were somewhat less likely to be employed and to work full time, and were more likely to consider their health to be poor.

Analytical strategy
We first present the distribution for the three statements measuring support for the distribution principles (anti-poverty, merit, and equality), comparing support for the distinct principles in the sample as a whole and among women and men separately (Table 1). Next, adopting a multivariate framework, we explore whether possible gender differences are linked to demographic and socioeconomic factors. For each dependent variable, we provide results from three models: a basic model with gender as the only independent variable (Model 1); a model with gender and other demographic variables (Model 2); and a model with both demographic and socioeconomic variables (Model 3) ( Table 2). Because we are primarily concerned in this part of the analysis with possible gender differences in support for the three (re)distribution principles, we report only the estimates for gender. (Complete models are available upon request.) On the question of whether patterns of support differ for women and men, we conducted separate analyses for each gender and for each of the three dependent variables, including demographic and socioeconomic variables (Table 3).
In the multivariate analyses, we used ordered logistic regression (OLOGIT), which is appropriate for ordered, non-interval outcome variables. All the estimates are reported as odds ratios. Coefficients above 1 indicate a positive effect, and those below 1 indicate a negative effect (compared to the reference category). Table 1 shows that a clear majority of our respondents agree that the pension scheme should be regulated so that no old age pensioners are poor. This is true for both genders, but, confirming expectations, women expressed even stronger support than men for the anti-poverty principle. The average scores are 4.50 and 4.28 for women and men, respectively; 88% of women and 82% of men agreed with this statement.

Support for the three distribution principles: Descriptive results
The merit principle receives far less support than the anti-poverty principle, and support is even lower among women than among men. Forty percent of women and 48% of men agreed with the statement; the average scores are 2.98 Pearson chi 2 (4) = 55.14, p ≤ .001; 2 Pearson chi 2 (4) = 38.21, p ≤ .001; 3 Pearson chi 2 (4) = 36.03, p ≤ .001. and 3.21, respectively. The equality principle won even less support than the merit principle, but around 30% completely or partly agreed that everyone should get the same old age pension amount from the NI. Women were more supportive than men, with an average score of 2.87 (compared to 2.57 among men). The analyses in Table 1 thus reveal that both women and men support the anti-poverty principle the most, while the merit principle receives the second-most support and the equality principle the least. However, and in line with our expectations, the bivariate analyses confirmed that women are more oriented toward anti-poverty and equality than men and less enthusiastic about merit, although the gender differences are modest in all cases. Next, we show whether these gender differences persisted in the multivariate framework.

Multivariate results
The multivariate analyses (OLOGIT) confirmed that women are significantly more likely than men to support the anti-poverty principle, and controlling for demographics barely affected the impact of gender (Table 2). Controlling for socioeconomic factors in Model 3 slightly reduces the gender difference, but it remains statistically significant (p ≤ .001). In this most expansive model, the estimate (odds ratio) for gender is 1.351, which is somewhat lower than the Table 2. The effect of gender (men = 0/women = 1) on attitudes toward redistribution principles in the old age pension system. Ordered logit, odds ratios, standard error between brackets. Models 1 to 3 (women and men combined). N = 3,020. estimate in the base model (1.751) but still suggests a degree of gender difference in support for the anti-poverty principle. When analyzing women and men separately (Table 3), we found that being single is associated with more support for the anti-poverty principle among both women and men, and for women we find a significant effect of cohabitation (as distinct from marriage). Having at least one child seems to make both women and men more supportive of the anti-poverty principle, but this effect is not statistically significant for women and only marginally so for men. As for the socioeconomic factors, both income and working hours affect attitudes in both genders: support for the anti-poverty principle decreases considerably for higher earners (over NOK 500,000 for both genders) and with longer working hours (compared to not being employed). For men, we also found a marginally significant negative effect of having extensive university education compared to basic or vocational training, but neither women nor men exhibited any association between health and support for the anti-poverty principle.
The merit principle and equality principle (flat-rate pensions) appear in these analyses partly as each other's opposites, so we discuss the results for these two principles together. The multivariate analyses (Table 2) revealed significant gender effects for both outcome variables in the base model that are only marginally affected by controlling for demographic factors (Model 2). Interestingly, however, the gender effects disappear when we control for socioeconomic factors, particularly income, in Model 3.
Analyzing women and men separately reveals few effects of the demographic variables on either item, although having at least one child seems to make women more supportive of the merit principle, while being older makes them less enthusiastic about the equality principle (Table 3). As for the socioeconomic factors, income is the one variable with significant effects in both genders. Respondents with incomes above NOK 500,000 were significantly less enamored of the principle of equality and more likely to agree that high earners should earn bigger pensions. For both items, women and men with incomes above NOK 700,000 stand out, with high estimates (odds ratios) for the merit principle and low estimates for the equality principle. Men's opinions in these areas also seem to be shaped by their health. Men who reported being in very good health were significantly more likely to support bigger pensions for higher earners (merit principle) and significantly less likely to support flat pensions (equality principle) than men who assessed their health as poor.

Robustness check
We repeated the analyses using ordinary least squares regression, treating the three outcome variables as continuous. These analyses yielded essentially the same results as those of the OLOGIT analyses and thus did not change the general conclusions. We also recoded each of the outcome variables to dummies, assigning those who agreed (completely or partly) to 1 and everyone else to 0. Logistic regressions with dichotomous outcome variables provided similar results to those of the OLOGIT analyses. The alternative analyses are available upon request.

Discussion
This article examined whether a gender gap exists in attitudes toward distribution principles in old age pensions and whether and how demographic and socioeconomic factors differently affect women's and men's attitudes. Based on the literature, we proposed three potential drivers of gender differences in attitudes to social issues: differences in strategic interests, in values, and in life experiences. Our analyses suggest that opinion formation may be more complex than any of these individual approaches imply and may vary both by the nature of the issue and by differences in individuals' assessments of their own situation.
An important finding is that women and men rank the three principles in the same order: first the anti-poverty principle, then the merit principle, with the equality principle as the least popular. Nevertheless, women are consistently more equality oriented, being more likely to agree that no pensioner should be poor (anti-poverty principle), less likely to agree that pensions should be bigger for higher lifetime earners (merit principle), and more likely to agree that pensions should be flat (equality principle). In the multivariate analyses, however, the findings diverge; women's stronger support persists for the anti-poverty principle when we control for socioeconomic factors but disappears for the merit principle and the equality principle. In both genders, the influence of income on these dimensions is very strong, and income is a primary factor in eliminating the gender difference in support for the merit and equality principles.
Our final question concerned whether demographic and socioeconomic factors influence attitudes in gendered ways. We found few indications of such mechanisms, but some gender differences emerged. Married women are less likely to endorse the anti-poverty principle than single or cohabiting women. Cohabiting women, in fact, resemble single women more than they resemble married women in their assessment of anti-poverty measures, while this pattern does not apply to men. This suggests that it is not the presence of a partner that is crucial to women but rather the legal and institutional protection associated with marriage or the more extensive pooling of resources in married than in cohabiting couples (Lyngstad et al., 2011). Formal marriage may be less important to men, as both cohabiting and married men typically have somewhat higher earnings than their female partners in Norway. By contrast, our findings do not support the hypothesis that parenthood makes women more equality oriented but men less so. Both men and women appear to be more supportive of the anti-poverty principle when they have children, although the effect is weak and not statistically significant for women and only marginally significant for men. This somewhat surprising pattern should be understood in the context of Norway's institutional setup (as argued by Finseraas et al., 2012). Because of long parental leaves (including paternity leave) and pension accrual for childcare, Norwegian mothers' interests and values may differ less from non-mothers' than among women in countries with less supportive policies. Correspondingly, Norway's emphasis on paternal care, supported by a long paternity leave, may deepen differences between fathers and men without children.
Our findings also suggest that believing their health to be poor rather than very good influences men's attitudes to the merit and equality principles, while this is not found for women. This may be related to the design of the Norwegian pension system, which rewards long working careers; perceiving their health as poor may lead men to anticipate shorter careers, a prospect that evidently makes them less supportive of the merit principle and more sympathetic to the equality principle. Taken together, our findings suggest that women feel vulnerable -and support equalizing measures in old age pensions -when they are not married, while men feel vulnerable when they are in poor health. This demonstrates the importance of separately analyzing the dynamics of women's and men's attitudes.
Finally, and importantly, support for all three principles is strongly related to income among both women and men; a negative association exists between income and support for the anti-poverty and equality principles, while a positive association remains between income and the merit principle. When controlling for income, women remain more committed to the antipoverty principle, while the gender gap disappears for the two other principles. These patterns suggest different dynamics for what may be seen as high-stakes vs. low-stakes concerns. High-income earners risk nothing by agreeing to a general principle that old age pensions should be above the poverty line. Much more is at stake when they are asked to relinquish the merit principle in pensions or to embrace the equality principle, both of which imply significantly lower pensions to themselves. Our finding of a gender gap only for lowstakes concerns indicates that, when their personal interests are at stake, women are as self-interested as men.

Limitations
It is a weakness of this study that we used data from only one country, so we do not know the extent to which our findings can be generalized beyond Norway. Like most survey data, ours are cross-sectional, which limits our range of analyses. Moreover, the survey was restricted to individuals of working age. We note that the variables in the regression model explain only a modest proportion of the variance in the dependent variables, which may suggests unobserved heterogeneity. An alternative interpretation of the low R 2 is that the respondents did not have strong, consistent opinions on pension issues, inevitably causing "random" effects in the response patterns.

Conclusion
This article contributes to closing the knowledge gap about preferences for pension system design. It lends nuance to the debate on gender gaps in social attitudes by suggesting that different dynamics may be at play for distinct redistribution principles and by showing that women and men may emphasize different concerns in assessing the three principles.
Old age pension systems are more sustainable when based on principles that the population generally agrees are legitimate. Our findings suggest that women and men may approach the notion of legitimacy differently and that partnership status, self-assessed health, and income all interact with gender in opinion formation. These findings are relevant to ongoing discussions of pension reform in many countries and present a challenge for further research. We suggest that future studies on gender gaps in social attitudes ask a broader range of questions, including both high-stakes and low-stakes concerns.

Key point
• Most people support pension systems designed to ensure that no pensioner is poor. • Flat-rate state pensions remain a divisive issue. • Women are, in general, more equality oriented than men. • Self-interest overrides ideology on "high-stakes" issues for both genders. • Gender interacts with family status and health, but the effects are slight.