Retirement anticipation - gendered patterns in a gender-equal society? a study of senior workers in Norway

ABSTRACT How do different people imagine their own retirement? In this article, we use a representative survey from 2019 to investigate whether female and male senior workers in Norway differ in the ways they anticipate their lives as retirees, and whether covariates of various images differ across gender. We find that women are more likely than men to anticipate spending time on family-related activities, on hobbies and volunteering and on traveling, but less likely to plan for bridge employment and worry about empty time. Health and income are key determinants in men’s images, while partners and children are important for women.


Introduction
Retirement has become a major phase of life for most people in industrialized countries. For many, retirement is welcome, as it opens new possibilities to fully manage their own time and pursue new life projects, free from the time constraints and responsibilities that are associated with paid work. For others, the notion has a sense of dread. For them, retirement implies leaving all the rewards associated with work and coming home to long stretches of time filled with nothing (Grødem & Kitterød, 2021;Karp, 1989;Kojola & Moen, 2016;Vickerstaff, 2006). How different people approach this final phase of life -a phase that will in some cases stretch over 30 years -can be important for their quality of life as retirees, and potentially affect their retirement decisions (Moffatt & Heaven, 2017;Noone et al., 2010). Understanding better how older workers imagine retirement, and how this varies by gender and class, will also help us to better understand the social meaning of retirement in a historical period where this social institution is in flux (Kojola & Moen, 2016;Sargent et al., 2013).
Because women and men tend to balance work and other commitments differently over the life course, it seems likely that anticipation of retirement will vary by gender. One might expect women to be less eager to retire than men and also engage less in preparing for post-work roles and activities, because they have had more punctuated careers and are still making up for lost time when they reach the stage of the life course when men start thinking about retirement (Jefferson, 2009;Loretto & Vickerstaff, 2015). However, since women typically have balanced different roles over the life course, they probably have more relationships and identities to fall back on once they leave employment (Loretto & Vickerstaff, 2012). It is also possible that in countries with a high degree of gender equality, there will be no significant gender differences: as men and women increasingly live similar lives, they will also approach retirement in similar ways.
Generally, how older workers in different life situations imagine their own retirement is largely an understudied topic (Amabile, 2019). There are relatively few studies; moreover, most existing studies draw on qualitative data. Also, most existing research has little to say about gender and gender differences: either because they focus solely or mainly on one gender, or because gender is not a central variable in the analysis. The aim of this article is to supplement this qualitative research with evidence from a survey, carried out in Norway in late 2019. 3,097 individuals aged 30-61, all living in Norway, answered a set of questions about their perceptions of the current pension system, trust in the system, plans for retirement and anticipation of life as retirees. Anticipation was measured by a number of statements about how they anticipated their lives would be, and what they expected they would do, when they retired. These statements, and their covariates, are the focus of the analysis here. We restrict the analysis to respondents aged 50 and over who were in employment at the time of the interview. The aim of this study is to investigate how women and men anticipate retirement, with regard to both feelings (positive/negative) and prospective activities, whether there are systematic gender differences in these respects, and whether the covariates of different images vary across gender.

Imagined retirements and retirement timing
Retirement as a social institution developed only in the 20 th century. Still, even in this short period of time, notions of what this life phase entails have shifted. When first institutionalized, retirement would typically be seen as a well-earned period of rest in a relatively short period between the end of employment and the end of life, or an institution suitable for older workers with diminishing productivity who needed to give way to younger cohorts (Phillipson, 2019). With increasing longevity, and increasing emphasis on leisure, consumption, and individual choice in all phases of life, the social meaning of retirement is changing (Gilleard & Higgs, 2000;Sargent et al., 2013). On the one hand, there is the expectation of longer working lives, underpinned by a certain concept of "active aging." On the other, there is the growing cultural expectation of a more flexible and self-actualizing later life (Moulaert & Biggs, 2013). Sargent et al. (2013, p. 4) have suggested that retirement is being reinvented, and that this reinvention can take two forms. The most radical form would be to challenge or reject the notion of retirement as a distinct period in individuals' lives, in favor of an ideal in which individuals combine paid work, family and leisure in varied ways throughout the life course. A less radical reinvention involves maintaining the notion of retirement as a distinct phase of life, but with changes in timing, the activities pursued, and the meanings associated with this period of life.
The ongoing discussion of the changing meaning of retirement is generally relatively silent on the importance of gender. The standardized life-course conceptualized by Kohli (2007), where individuals moved from education to employment to retirement at given normative ages, was after all a standard for men. Women's normative life course has traditionally been different, more centered on households, children and unpaid work, sometimes with periods of paid employment in between. The emerging standard, where both men and women's lives center on employment and women have their own pension accrual, is likely to influence the perceived meaning of retirement for both women and men. It also potentially changes the perceived relationship between employment and retirement, as both women and men integrate paid employment, family life, unpaid projects and leisure activities over the entire life course.
The changing meaning of retirement-discourse has also been criticized for placing too much emphasis on individual choice (Macnicol, 2015), and for downplaying the many insecurities individuals face in their late careers and transition to retirement. Still, with the uncertainty, individuals imagine, dream, and sometimes plan for retirement. As most studies of retirement behavior center on "health and wealth" (Amabile, 2019), we however have limited knowledge about what these images may be and how they vary by key background factors such as gender, family factors, health and class. Studies of retirement anticipation typically find that individuals' sentiments about their current work situation was key to understanding how they anticipated retirement: those who had high job satisfaction were less likely to anticipate retirement positively (Davies et al., 2017;Fouquereau et al., 2018;Karp, 1989;Vickerstaff, 2006). The same went for those who felt that they had unfinished agendas at work. Those who were "done" were more likely to happily transition into retirement (Karp, 1989;Vickerstaff, 2006). A different predictor in these studies is the financial situation: older workers who felt financially secure, and confident they could afford the life-style they desired after retirement, were more positive toward the prospect than those who were not.
The studies so far all aim to illuminate who looks forward to retirement, and who are ambivalent or even negative to the prospect of retiring. All the qualitative studies have both male and female informants, but rarely discuss gender differences systematically. One exception is Loretto and Vickerstaff (2012), whose findings suggest that married women are more likely to give social reasons for intending to carry on working, whereas men are more focused on the financial dimension. Vickerstaff and Cox (2005, p. 88), too, study gender differences and find that men dominate among the informants who did not want to retire and avoided thinking about it, while women dominate among those who have been overtaken by events, such as changes in domestic circumstances, financial problems or organizational policy. Studies that focus solely on women (e.g., Sherry et al., 2017) find that women, like men, often approach retirement with "trepidation." From these studies, it may seem that professional identity and job satisfaction strongly influence feelings about retirement for both women and men, and that women and men with similar positions in the labor market are likely to have similar feelings about retirement as well. What these qualitative studies then do not reflect, is that women and men tend to systematically dominate in different parts of the labor market and have different working trajectories. Professional men with full-time uninterrupted careers are more representative for the male experience than women in similar positions are for the female. For this reason alone, the qualitative studies should be supplemented by analyses of representative surveys containing questions about retirement anticipation.
The academic literature is even scarcer when we turn from positive vs. negative emotions about retirement, and toward how individuals actually imagine spending their time in this phase of life. Price and Nesteruk (2010) interviewed women who had already retired, and identified five retirement pathways: family-focused, service-focused, recreation-focused, employment-focused, and disenchanted retirements. Those disenchanted with retirement had experienced health problems, financial insecurity, death of a spouse or significant caregiving responsibilities. For discussions of imagined activities in retirement among those still working, we have to turn to the literature on retirement planning. Here, studies of financial planning and preparation dominate (e.g., Boisclair et al., 2017;Foster & Heneghan, 2018), but some studies also aim to map which activities individuals plan for. Such activities can include bridge employment, self-developmental leisure and social leisure (to use the categories proposed by Eismann et al. (2019)). One finding seems to be that the more opportunities older workers had in the respective domains, the more likely they were to have plans in this domain (Eismann et al., 2019, p. 744). Van Solinge et al. (2021) point out that work holds both extrinsic (e.g., monetary) and intrinsic (e.g., daily routine, social contacts, sense of purpose) meanings, and suggest that the form of loss individuals feel will influence how they aim to "compensate" during retirement.
In this study, we first explore how women and men respond to each single item in the survey, which will give an indication on how they may perceive similar activities differently. Second, we use multiple regression analyses to investigate whether and to what extent possible gender differences apply when demographic and socio-economic factors are accounted for, and whether these factors influence differently how women and men imagine their own retirement. In order to reduce complexity somewhat, we construct additive indices from some of the single items in the survey, which we use in the regression analyses. The analysis is mainly explorative, as there is limited research -in particular, limited research using quantitative data -on women's and men's retirement anticipation, at least in a Norwegian context.

The Norwegian context
Norway is an interesting case in this context for two reasons: the comparatively high levels of gender equality and the flexibility older workers have with regard to retirement. The Scandinavian countries, including Norway, have long been regarded as frontrunners on the path to gender equality (Lister, 2009), but most studies of gender equality/gender differences have focussed on the labor market (e.g., Reisel et al., 2019) or on the divisions of labor in the care for young children (e.g., Bjørnholt & Stefansen, 2018). We do not know to what extent the increasing demand for gender equality has translated into similarities and variations in how women and men in their 50s and 60s imagine retirement, but a front-runner country like Norway provides a good case for exploring this. Also, countries where employees actively chose how and when to retire encourage workers to imagine futures with and without continued employment, which suggests that workers aged 50 and older in a country like Norway may approach retirement more actively than older workers in countries where they are less likely to have a choice.
Norway has for a long time been an affluent country with a tight labor market, and women's employment rates are almost as high as men's in all phases of life. The women who are included in the survey sample used here were born between 1958 and 1969. The oldest of these women have probably had at least some of their children while parental leaves were still short and child-care coverage rates were low, while the youngest women have benefitted from the expansion of both since the late 1980s (Grødem, 2015). Women will on average have had more career breaks and more spells of part-time work than men, but employment rates of mothers are comparatively high and the same is true for men's involvement in housework and childcare (Hook, 2006;Hook & Wolfe, 2012). However, the Norwegian labor market is still highly gender segregated with women dominating in the public sector and in education, health and social work, and men concentrated in the private sector and in manufacturing and finance (Teigen & Skjeie, 2017).
The employment rates among older workers in Norway are among the highest in the Organization for Economic Cooperation and Development (OECD) area (Eurostat, 2020). In 2019, the labor force participation rate of Norwegian men and women aged 55-64 were 78.4% 69.5%, respectively. This was well above the OECD average of 64.4 for men and 55.8% for women (OECD, 2020). Unemployment is much less common among the old than among the young in Norway, and only 1.2% of 55-74-year olds in the labor force were unemployed by the end of 2019 (Statistics Norway, 2020). When older male and female workers in Norway wish to remain in employment, therefore, the odds of them being able to make this decision for themselves are higher than in many other countries.
Adding to this, Norway implemented a comprehensive pension reform in 2011, that removed the old concept of a fixed retirement age and made retirement fully flexible between ages 62 and 75, but on actuarially neutral terms. This means that the earlier one starts to draw pension, the lower the annual amount will be. A formula provides each individual a notional "pension savings account," to which 18.1% of their annual income is credited. When the person retires, their annual pension is determined by the size of this account (pension wealth) and the number of years she is expected to live past retirement (the life expectancy adjustment). The reform will take full effect for cohorts born in 1963 or later, while cohorts born between 1944 and 1962 have their pensions calculated by a combination of old and new rules (for more details on the pension reform, see Pedersen (2017)).
The pension reform contains a number of elements that in effect benefit women, including pension accrual for care work, a guarantee pension for persons with low or no lifetime earnings, and a cap on pension accrual at around 1.5 times the average annual income (Halvorsen & Pedersen, 2019). The occupational pensions that have been reformed in the wake of the reform of public pensions are also regulated in ways that protect women, in the sense that they are mandatory and must be uniform for all employees within a firm (Pedersen et al., 2018). Pensions are comparatively high in the public sector, where women dominate (Hippe & Pedersen, 2019). Women claim lower pensions than men in Norway on average (Halvorsen & Pedersen, 2019), but are still comparatively well-protected. Moreover, the overall poverty rate among old-age pensioners is comparably low in Norway (Ebbinghaus, 2021).
Average actual retirement ages have increased in Norway after the pension reform, particularly among men in the private sector (Bjørnstad, 2019). Still, they remain below the previous formal retirement age at 67 years. For those who are in employment at age 50, the average age for withdrawal from paid employment was 65.7 years in 2018 (Bjørnstad, 2019). Men retire about a year later than women, thus average age for withdrawal is just over 66 for men and just over 65 for women.
For workers in the private sector, there is no coordination between pension and incomes from work, and it is possible to work and draw pension at the same time. This option is open to all workers who have acquired sufficient notional pension savings to give them a pension higher than the guarantee pension. Those who have not, may have to postpone pension uptake until they can claim guarantee pension at age 67. Around 55% of men in the age group 62-66 who are still in employment have chosen to draw pension while working, while the same is true for around 25% of women (Pensjonspolitisk arbeidsgruppe, 2019). One reason women are less likely to do this than men, is because fewer women have the required minimum level of pension accrual. A further reason can be that men are more committed to the "provider" role, and start drawing pensions to secure future incomes for their families in the event of their own death. We do not know what lies behind the large gender disparity, but it is clear that men and women use the flexibility in the new system in different ways. However, there is still little research on whether and in what ways retirement expectations and images of retirement vary by gender among senior workers in Norway.

Data
To answer our research questions, we use data from a web-based survey, "Survey on Work and Pension," carried out in autumn 2019 by survey institute Kantar on behalf of Institute for Social Research. The survey was designed by the authors, who are both affiliated with Institute for Social Research. The net sample contains 3,097 respondents recruited from the "Gallup Panel," which is a representative sample of about 40,000 individuals who regularly participate in different types of online surveys conducted by Kantar. Since our interest here is in images of retirement and retirement plans, however, we analyze only responses from individuals who are between 50 and 61 who were in employment at the time of the interview (N = 1,249). 1 The proportion in employment is 82% of the relevant age group; 77 and 86% of female and male respondents, respectively. Kantar has constructed a weight that compensates for a slight under-representation of certain groups (affecting gender, age, education, and geographical location). The analyses in this paper are based on the weighted material, but the number of respondents is presented unweighted.
The survey contained a number of questions about respondents' trust in the current pension system, their opinions about the main redistributive elements in the system, and their thoughts about their own retirement. The section where we asked respondents to think ahead to their own retirement was developed on the basis of existing literature on retirement decision-making and retirement planning, as reviewed above. For practical and financial reasons, there was, however, a limit to how many items we could include, and we settled on 11 statements. Since the literature shows that older workers can approach retirement both with anxiety and with enthusiasm, we were careful to include both "activities"-statements and "sentiments"-statements in our questionnaire. As for activity statements, we prioritized questions on family-related issues, broad leisurely activities like hobbies and volunteering that typically do not require large financial resources, as well as traveling and bridge employment. Informants were asked "If you think ahead to your future life as an old age pensioner, how well do you think the following statements will fit your situation?" Possible answers were "very well," "well," "neutral," "poorly," "very poorly." The 11 statements were: (1) I will feel that I have too much time (2) I will travel and see the world (3) I will spend a lot of time with my family (4) I will miss work (5) I will help my family by taking care of children, redecoration and other tasks (6) I will be tired and low on energy (7) I will work a little for pay (part-time, odd jobs etc.) (8) I will spend more time on hobbies (9) I will spend more time on voluntary work (10) I will enjoy the freedom (11) I will be in poor health Almost all our respondents aged 50 and older responded to each item. The number of nonresponse (N) varied between 2 and 12. Those few non-responses have been recoded into "neutral." All the statements have been standardized so that "very well" is coded 5 and "very poorly" is coded 1.
The two items concerning health and energy levels stand out from the others in that they not only measure sentiments and expectations, they concern a topic the respondent can do little about. In the preliminary analysis, we found that these two items are very strongly correlated with self-reported current health situation. Therefore, we have dropped these two items from the analyses. We have also dropped the statement about missing work because preliminary analysis revealed that people's answers to this question varied little with gender, family situation, health and socioeconomic factorsthe most important covariates in our analyses. The empirical discussion thus relies on eight statements.
Given our interest in possible gender differences, we have first explored whether and in what ways women and men differ in their images of retirement. Next, we have run all analyses separately for women and men in order to examine which factors are important for the ways in which women and men anticipate retirement, and whether the covariates of the different images vary across gender.

Dependent variables
As Table 1 suggests, our informants do not have strong opinions about how they intend to spend their lives in retirement. This makes data "fuzzy," and leaves us with a methodological problem: the items do not cluster neatly into dimensions, as we could have hoped (cf. Eismann et al., 2019). This is an argument for analyzing each item separately, at the risk of drowning any overarching pattern in details. Since our approach is largely explorative, we proceed in an inductive and pragmatic fashion: we combine items that correlate and also theoretically capture the same underlying phenomenon, into indices (3), while analyzing the remaining items that separately (2). The indices are informed by theoretical considerations, based on existing literature on planned and imagined activities past retirement (e.g., Kojola & Moen, 2016;Loretto & Vickerstaff, 2012;Vickerstaff, 2006). The composition of the indices is explained in Table 2. The correlations (PCC) between the items in each of the three indices are fairly high (see Table A1), which supports our assumptions that they can be merged into indies that capture the same underlying phenomena. The relevant correlations are 0.555 (between items 3 and 5), 0.301 (between items 8 and 9), and 0.326 (between items 1 and 10), which agrees with the minimum level of 0.3, as recommended for this type of variables (Hellevik, 1977, p. 268). Chronbach's Alpha satisfies the recommended level only for the family index. 2 In addition to the three indices, we analyze two single items: travel (item 2) and paid work (item 7). These do not fit in easily theoretically in either index.
We thus include five dependent variables in analyses (see Table 2), all varying from 1 to 5. The three indices all have nine values, and the two single items have five. Because they are all continuous variables, we use OLS in the multivariate analyses. 3

Independent variables
The independent variables in the analysis partly come from the questionnaire and partly from background information previously collected for all participants in the Gallup Panel. Our explanatory variables include demographic and socioeconomic factors, while we control for sector of employment and the anticipated time of retirement. Descriptive statistics for the independent variables are presented in Table A2, for the whole analysis sample and separately for women and men. We include the following independent variables: Gender dummy. Women (1), men (0).

Age
continuous and squared (in order to capture non-linear effects).
Children dummy. We differentiate between respondents with at least one child (1), and those with no children (0). Respondents were asked to include both biological and adopted children, and both children in the household and children who had moved away from home. Item 8: I will spend more time on hobbies Item 9: I will spend more time on voluntary work Relish Leisure (index) Item 1: I will feel that I have too much time Item 10: I will enjoy freedom Travel Item 2: I will travel and see the world Work for Pay Item 5: I will work a little for pay Item 1 has been inverted so that higher values indicate that the statements fits poorly. We have standardized all indices by adding the scores and dividing the sum by two.
Partner dummy. We distinguish between those who live with a partner (married or cohabiting) (1) and those who do not (0).

Health
dummy. Those who rated their current health as "good" or "very good" are coded as 1, others as 0.

Education
(highest completed level): We separate between basic/high school/vocational education (reference), university 1-4 years, and university 5 or more years.

Income
Individual gross annual income, based on a question with categorical response alternatives. We use "under NOK 400,000" as reference, because very few have incomes below NOK 300,000. We include "Missing" as a separate category, because about 7% of the analysis sample declined to respond to this question.

Working hours
We distinguish between part-time work (1-36 hours) (reference), full-time work (37-40 hours) and long hours (41 hours or more). Respondents reported their normal weekly working hours, including extra hours and work from home.

Sector of work
dummy, distinguishing between public (1) and private sector (0). We include this as a control because pension benefits (see explanation above) differ across the sectors, as do probably also norms concerning when to retire. Because women more often than men work in public sector, controlling for sector may attenuate possible gender differences in images of retirement.

Anticipated time of retirement
We distinguish between four categories; Early (reference), Medium, Late, and Don't know. 4 We include this control variable because we assume that people who plan to retire early are likely to have more clear visions of their life in retirement compared to those who plan to retire later, as well as those who do not know when they will retire. Conversely, and in accordance with the role of "jump-factors" 5 pointed out in research on timing of retirement (Grødem & Kitterød, 2021;Jensen, 2020), those who have clear plans for retirement and specific tasks that they would like to spend more time on, may be more likely to anticipate early retirement than those with fewer and vaguer plans. Moreover, as women tend to "jump" more often than men do (Jensen, 2020), controlling for anticipated time of retirement may impact possible gender differences in peoples' visions for retirement. Looking at the descriptive statistics in Table A2, we note that women and men are about the same age, but women are slightly more likely than men to have at least one child and less likely to live with a partner. Moreover, more women than men have a university degree, but women generally have lower incomes than men, work shorter hours and more often work in the public sector. Roughly equal proportions (about nine in ten) of both genders report that their health is good. As for anticipated time of retirement, women are somewhat less likely than men to fall in the "Medium category," but more likely to fall in the "Don't know" category. Still, we note that fairly high proportions of both genders fall in the "Don't know" category (29 and 21%, respectively). This agrees with international as well as Norwegian studies suggesting that many senior workers have only vague images of retirement and that late career and retirement plans mature over the years prior to retirement (Furunes et al., 2015;Moffatt & Heaven, 2017).

Results
We start with a descriptive overview of how our informants identified with each of the eight statements regarding what their retirement period might look like. This is shown in Table 1, for the whole sample and separately for women and men.
The main message in this table is that many respondents hold only vague ideas about how they will spend their time in retirement. The proportion who give the neutral answer, coded here as 3, is around one in three on most items. The averages, too, is bigger than 4 only for one item (Item 10: I will enjoy freedom). The high "neutral "proportions for all the items and the low proportions with the lowest/ highest scores suggest that most of our respondents do not have very clear visions of how they will feel about retirement, nor how they will spend their time. We further note that men tend to choose the "neutral" category more often than women, indicating that they have even less clear images of how they will spend their time as retirees or have thought less thoroughly through such issues than women. 6 Descriptive statistics for our five dependent variables (three indices and two individual statements) are provided in Table 3. Of the three indices, the Relish leisure index stands out with the highest score, suggesting that even though people may worry about how to spend their days in retirement, this does not occur as a dominant approach in our sample. The Spend Time with Family index also has a fairly high score, indicating that many respondents anticipate family-related activities in retirement. As for the two single items, almost half of the respondents report that traveling fits well with their anticipations, while fewer, 31% imagine that they will work for pay when retired. Comparing across gender, we note that women are more likely than men to anticipate spending time with the family, spending time on hobbies and volunteering, and travel in retirement and express more strongly that they will enjoy leisure. However, men are more likely than women to imagine that they will do some paid work. Thus, even in a country like Norway, with comparably high levels of gender equality in the labor market as well as in the division of unpaid family work, there seems to be clear gendered patterns in the ways in which senior workers imagine their lives as retirees. We now turn to the question on whether and to what extent these patterns apply when demographic and socio-economic variables are accounted for.
We find that the bivariate gender differences largely hold up in multivariate models. We ran five regression models for each dependent variable, controlling for demographic and socioeconomic factors. We started with a basic model with only gender as an independent variable (model 1), and added various controls in the subsequent models 2-5, as explained in Table 4. Because we are primarily concerned with possible gender differences in senior workers' visions of retirement, we only report the estimates for gender and not for the other covariates (see Table 5). Complete models are available upon request.
We find that according to all five models, women are more likely than men to anticipate to spend time with family in retirement. The relationship is weakened slightly when we control for other factors, but even the most comprehensive model (model 5) provides a strong positive estimate for gender (p < .001). This means that even when women and men have similar family-, health -, and socioeconomic characteristics, and similar plans regarding when to retire, women Each variable was scored on a five point scale from "1 = Fits poorly" to "5 = Fits very well." For all variables a higher score indicates greater agreement with the variable. more often than men foresee family-related activities as retirees. A similar pattern emerges when it comes to the anticipation of new projects in retirement, such as spending time on hobbies and volunteering. Turning to the index of relishing leisure time in retirement, we also find positive gender effects in all models: even when demographic and socioeconomic factors are taken into account, women are more likely than men to look forward to spending their time as they want and thus less likely to fear boredom and loss in retirement. For the two single items (travel and work for pay), we find different patterns: For travel, the coefficient for gender is weak or nonexistent in models 1-3 model, but strengthened when we enter controls for socioeconomic factors. This implies that when we compare women and men with similar socioeconomic resources, women are more likely to respond that they wish to travel when they retire. For work, all models suggest that women are less likely than men to say that they will work for pay as retirees.
Summing up, our analyses so far suggest clear gender differences in senior workers' retirement anticipations in Norway, and this is true for family-related activities and spending time on hobbies and volunteering as well as for traveling plans and bridge employment. Moreover, women seem to worry less about empty days in retirement than men do. The clear gender differences is an argument for proceeding with separate analyses for women and men, to see whether covariates differ across the various dimensions of retirement anticipation for women and men. For instance, is income a more  important predictor for traveling plans than for spending time with the family and for spending time on hobbies and volunteering, which supposedly require less economic resources? And do we observe similar patterns for both genders? We ran separate regression models for women and men, and for each of the five dependent variables we show results from two models. The first one (model 6) includes all the independent variables except anticipated time of retirement, while the latter variable is added in the second model (model 7) (see Table 4 for details). We wanted to check for timing of retirement separately, as we do not know whether this is a covariate (those who are eager to travel plan to retire early) or an aspect of the outcome (those who know they may have to retire early make extensive plans, including for traveling). Results for women and men are reported in Tables 6 and 7, respectively. Only statistically significant results (p < .1) are included. We note, however, that the small sample sizes (N = 539 and 710 for women and men, respectively) make reaching statistical significance at conventional levels difficult.
Looking first at the anticipation of family-related activities in retirement, as measured by the Spend Time with Family index, we find that actually having a family -a partner and/or children -is a strong predictor for both women and men. For men there is also a statistically significant association with health, suggesting that poor health weakens men's aspirations for family life after retirement. This effect is not present for women. For both genders, planned timing of retirement matters: there is an association between planning to retire early, and anticipating time with the family. However, and interestingly, introducing timing of retirement does not weaken the associations between the other covariates and the outcome. Retirement timing is thus associated with anticipating time with the family independent of other covariates.
Turning to the index on plans for hobbies and volunteering in retirement, we find a somewhat different pattern. For both genders, this dimension is positively associated with education: those with education at the college or university level are more likely to anticipate taking on new projects. Also, there are unsystematic associations with income, but in opposite directions for men and women: increasing incomes are associated with more anticipation of new projects for women, but less for men (most coefficients are however not statistically significant at the generally accepted level). Something similar is true for health: being in good health makes men more eager to spend time on hobbies and volunteering, but women less so. This is a surprising outcome and an unexpected gender difference. The extent to which men (but not women) foresee more time on hobbies and volunteering in retirement is related to their anticipated time of retirement in that those who have uncertain retirement plans have lower scores on this index than those who plan to retire early, but including this variable in the model does not affect the effects of the other variables.
As for relishing leisure time in retirement, the analyses reveal a strong positive, although non-linear association with age for both genders (Tables 6 and 7). They also reveal a positive association with partner status for women, but not for men. There is a strong correlation between looking forward to leisure time in retirement and the anticipated time of retirement, in that those who plan to retire early are more than others looking forward to free time.
The multivariate analyses of images of traveling in retirement yields few statistically significant results for women, except from a positive association with partnership status and a negative relationship with anticipated retirement time (Table 6). Expecting higher earning women to report more traveling plans than those with less income, we were surprised to find only modest and not statistically significant effects of income in the analyses. Suspecting that partnered women may expect to benefit from their partners' income in retirement, we added an interaction term between partnership status and income in the analyses (results not shown), in order to explore whether traveling plans are related to income for women without a partner, but not for partnered women, but no such patterns were revealed. Turning to the analyses of men's traveling plans in retirement (Table 7), we find a curvilinear relationship with age -starting out as negative and eventually turning positive, a positive association with health, income and working hours, and, as for women, a negative correlation with anticipated   We included a separate category with the very few observations with missing information on sector of work, but do not show the results.   time of retirement. Surprisingly, and unlike for women, men's traveling plans appear to be unaffected by the presence of a partner. Looking at senior workers' anticipations of performing some paid work in retirement, which is the only dependent variable where women display a lower score than men, we find somewhat diverging patterns across gender in the multivariate analyses. For women, both model 6 and model 7 reveal a negative relationship with partnership status (Table 6), which may reflect that women typically have an older partner with whom they plan to coordinate their work exit (Syse et al., 2014). Model 6 also shows a negative association with income and public sector employment, and a positive association with education. The latter relationship may reflect that highly educated women have better access to flexible working arrangements than the less educated and perhaps also have stronger work commitments. The negative association with income may suggest that higher earning women do not to the same extent as those who earn less need to work in retirement in order to obtain a sufficient income. As for the negative association with public sector employment, we believe that this reflects sector specific rules as well as norms and practices with regard to retirement behavior. When timing of retirement is included in model 7, neither of these latter associations are statistically significant at conventionally accepted levels. Turning to the results for men, we find a positive association with educational attainment in both models. The most expansive model also shows that those who intend to retire early are more inclined to plan for some work compared to those who intend to retire later and those who do not know when they will retire. Apart from this, the multivariate analyses reveal no clear patterns with regard to men's plans for bridge employment.

Robustness checks
As a robustness check, we have run all the analyses with observations with non-response on one of the eight statements/items on retirement anticipation, coded as missing, rather than "Neutral." This applies to 51 observations (26 women and 25 men). The analysis sample thus contained 1,198 observations (513 women and 685 men). This alternative specification has almost no effect on the findings. Results are available on request. Moreover, we have run all the analyses on a subsample including respondents aged 55-61, thus excluding the youngest ones (aged 50-54), who may have the vaguest plans for retirement. Essentially, these analyses provide similar results as those shown in the article, although with slightly fewer statistically significant results, mainly due to the smaller sample size (N = 695).

Discussion
Norway is consistently ranked as one of the most gender-equal countries in the world, still, even here, women and men lead different lives. These differences manifest themselves in the ways women and men anticipate retirement and the factors that influence how they imagine their own lives as old age pensioners. We have found that women appear to have fuller, richer images of what their lives will be like when they retire. Even when we control for all available resources, women are still significantly more likely than men to imagine a retirement where they will spend much time around their families and help them out, and where they will engage more in hobbies and voluntary work. When we compare only those with similar resources, women are also more likely to anticipate traveling. Correspondingly, women are more likely than men to anticipate positive feelings in retirement (enjoying freedom and not feeling that they have too much time), and they are less likely to anticipate that they will take up some form of paid employment. Somewhat caricatured, we could say that women look forward to retiring and immerse themselves in new life projects, while men are more inclined to cling to their jobs and imagine retirement as a loss that they will compensate by working at least a little.
This almost persistent gender difference is perhaps the most important finding in this analysis. In addition, the analysis of covariates for women's and men's anticipation rendered some interesting findings although we acknowledge that a larger sample might have produced even more statistically significant results. The patterns that we do find seem to relate to what women and men regard as their available resources in old age. Again, to caricature findings a little, it seems that men see income and health as their key resources, while women look to their partners. This pattern is particularly visible with regard to traveling in retirement: women imagine that they will travel if they have a partner to travel with, men imagine traveling if they are in good health and have a high income. We found no statistically significant association between imagining traveling and income or working hours for women, and no association with the presence of a partner for men. A similar association with health for men is found for the "hobbies and volunteering"-dimension -men are more likely to imagine taking up a hobby or volunteering if they are in good health. For women, surprisingly, the association with health is negative. This finding is difficult to interpret, but it may have to do with the different forms of hobbies and volunteering that are open to, and seen as attractive by, men and women. If men associate "hobbies" with chopping wood or coaching a football team, while women associate needlework or visiting lonely people in the community, the associations are less surprising. This is something that future research can dig deeper into.
One of our outcome variables measures expected feelings in retirement rather than expected ways to fill time -the "relishing leisure"-index. Both genders have high scores on this variable, but women are even more likely to harbor such anticipations than men are, and having a partner increases women's anticipations of enjoying freedom. The presence of a partner does not influence men's anticipations in this area. Interestingly, women with children seem to fear emptiness in old age more than women without children. This is counterintuitive, but may have to do with the sense of loss that can come from being an "empty nester," as the women analyzed here often already have experienced. Notably, such considerations do not seem to affect men.
For both genders, we find that prospective timing of retirement plays a role for most dimensions. This covariation was expected: our respondents have not yet retired, and the youngest of them were more than ten years away from the lowest retirement age. We asked them, in the same survey, to imagine when to retire (at what age), how to retire (fully or partially) and what they would do in retirement. It is not surprising if these "simultaneous imaginings" impacted on each other. It is also unsurprising that the strongest association between this predictor and the outcome is found for "relishing leisure": those who plan to retire early are significantly more likely than others to look forward to spend their days as they like in retirement and conversely, less likely to fear emptiness and boredom. This is true for both genders. It is worth noting, too, that the largest coefficients are found for "have not thought about [when to retire]": those who see retirement as a time of enjoyment and freedom are less likely to push the thought of retiring down the road. This is in tune with findings from a qualitative study (Grødem & Kitterød, 2021).
What was perhaps less expected was that introducing timing of retirement did little to alter the effects of most of the other covariates, although some effects were slightly attenuated and no longer statistically significant at commonly accepted levels when timing of retirement was included in the model. By and large, individuals with the same gender, and similar education, income and family situation seem to have the same visions for retirement, regardless of when they expect to retire. Push/ stay/stuck-factors in the labor market can influence the prospective timing of retirement (Jensen, 2020), but this does not seem to be strongly related to images of what retired life will imply for individuals with different roles and resources.
Although our analyses yield mostly modest and unsystematic effects of the socioeconomic variables, the patterns that we do find echo findings by Eismann et al. (2019), who highlight how retirement plans are affected by available resources. When we ask our respondents to imagine their life as retirees, they seem to consider the resources available to them: do I have a good and stable relationship with my partner and children? Do I have financial maneuvering space? Am I in good health? The extent to which they expect to spend time with their family, take up hobbies, or travel, seems connected to the availability of such resources. We however also see gender differences in what resources respondents rely on: education is important for both men and women, but men seem to place more importance on health and wealth than women do, while partners and children matter more for women. However, the rather weak and unsystematic effects of socioeconomic resources for both genders may be related to the fact that the pension system in Norway is comparatively generous and the poverty rate among retirees comparatively low (Ebbinghaus, 2021), and also that many hobbies and leisure activities require few financial resources in Norway. Asking for more specific hobbies and distinguishing between various types of travel might have provided a slightly different picture. Also, adding activities that are masculine coded, such as maintenance work or hunting and fishing, might reduce the observed gender differences in plans for hobbies or now projects in retirement. Future research may want to ask more detailed questions to better grasp the connections between gender, resources, and images of retirement. We also acknowledge that our results should be interpreted as preliminary, since running a large number of models on the same data set, as we do in this article, increases the risk of Type I error: i.e. concluding that an observed effect is statistically significant when it is not. Our results should thus be subject to verification by futures studies.
Our findings confirm what many qualitative studies, both in Norway and internationally, have found: that most people's plans, aspirations and images of life in retirement are vague (e.g., Furunes et al., 2015;Grødem & Kitterød, 2021;Loretto & Vickerstaff, 2012;Moffatt & Heaven, 2017) Our respondents cluster in the middle on most of our outcome variables, and there are few strong patterns. Above, we noted that this presents a methodological challenge: how do we identify patterns in this hazy landscape? There is also a more substantial challenge, as one may ask to what extent it makes sense to even try to study such elusive topics. It is worth remembering that there are few hard facts when studying social phenomena, thus elusiveness is always a matter of degree. Moreover, as pension reforms are implemented across the world, it may be fruitful to monitor how images of retirement change. If we repeated this study in Norway in 15 years, would we find that respondents are less likely to imagine traveling and enjoying the freedom, and more likely to imagine continued (bridge) employment? If this survey was run in a country with lower pensions or more scattered coverage, would the results be very different? Even investigations of hazy issues, which fail to yield clear patterns at one point in time, may have their merit in times of change. Future research may want to supplement macroeconomic projections on fiscal stability and poverty in old age with studies of widespread images and aspirations, to improve understandings of the question that motivated this article: what sort of social phenomenon is retirement imagined to be in modern societies, and how does this differ between men and women? Notes 1. We excluded 6 observations with missing information of whether or not they had children and one observation with missing information on whether or not he/she were living with a partner or not. 2. Chronbach's Alpha is 0.703 for the first index, 0. 456 for the second one, and 0.478 for the third one. However, we note that Chronbach's Alpha typically underestimates true reliability on two-item scales (Eisinga et al., 2013). 3. As for the two single items we have also conducted ordered logistic regressions, which essentially yields the same results. 4. The categorization is based on responses to two survey questions, the first one being "At what age do you see yourself starting to draw old age pension"? As noted, this is not necessarily the same as when the individual will withdraw fully or partly from employment: it is possible to draw pension and work at the same time. We therefore also check the participants' responses to the question "The pension age in the public pension system is flexible, and one can, on certain conditions, chose to retire at any point between 62 and 75 years of age. One pays for early take-up by accepting lower annual pension amounts. One can also combine work and pension. How do you imagine you will end your working life?" Potential answers are (1) "Work as much as I do now until I stop working and start drawing pension"; (2) "Draw pension and work as much as I do now for a period"; (3) "Draw pension and work less than I do now for a period"; (4)"Have not thought about it". Those who have indicated that they plan to draw pension at 62 and responded with alternative 1 or 3 on the second question, were coded as "Early". The "Medium" category includes those who indicated that they plan to draw pension at 62 and responded with alternatives 2 or 4 on the second question, as well as those who indicated that they planned to draw pension at age 63-66 years. The "Late" category includes those who indicated that they planned to draw pension at ages 67-75 years, while the "Don't know" category includes those who did not respond to question on when they planned to start drawing old-age pension.
5. In discussing factors that impact on the time of retirement, researchers typically distinguish between push, pull, stay, stuck and jump factors (Jensen, 2020). "Jump" denotes voluntary retirement motivated by a search for life conditions that are more fulfilling than paid work. 6. The average numbers of «Neutral» responses are 2.52 among men and 2.06 among women (on a variable ranging from 0-8).