Introduction
Female political representation is a fundamental democratic value. The proportion of elected women has increased dramatically in the last half-century, but at different speeds in different countries (Hughes and Paxton Reference Hughes, Paxton, Franceschet, Krook and Tan2019). Social scientists have understandably sought to document the factors driving this development. A large literature charts various country-, party-, and individual-level factors that help or hinder women's political entry (reviewed by, for example, Paxton, Hughes, and Barnes Reference Paxton, Hughes and Barnes2020; Waylen et al. Reference Waylen2013). We add to this literature by documenting a new intergenerational channel through which the economic empowerment of one generation of women facilitates the political empowerment of the next.
We use Swedish data that overcome previous barriers to studying intergenerational dynamics. By digitizing archival data, we extend the study period back to the 1970s and 1980s when Sweden experienced a rapid increase in female numerical representation in politics. Administrative records allow us to link these politicians to their parents and observe their parents' employment status, earnings, and the division of earnings between spouses. By pooling these data with data on the full population of Swedish residents, we can compare politicians' parents to other women of the same age living in the same geographic region. This allows us to determine whether male and female politicians' mothers have systematically stronger labour market outcomes than their cohort-region peers.
Fig. 1 motivates the focus on intergenerational dynamics by plotting the female share in the paid labour force (black line) and their shares in local and national politics (grey lines) over time. The rise in female participation in the paid labour force clearly predates the rapid increase in women's political officeholding in the 1970s and 1980s. The resulting pattern of two consecutive waves of empowerment – first economic, then political – indicates that an intergenerational link might exist from one to the other.
Figure 1. Proportion of women in the Swedish workforce and among elected politicians.
Source: Employment is measured using census data between 1960 and 1985 and administrative data between 1991 and 2018. The proportion of elected women is calculated from the complete list of elected officials collected by the authors from public records for 1960–1971 and from the Swedish Electoral Authority for 1973–2018, supplemented with data from Statistics Sweden for the 1960s. In the 1960s, local and national elections were held in different years.
We compare the mothers of male and female politicians and find a dramatic difference in household backgrounds. Female politicians – especially parliamentarians in the early decades when female numerical representation rose the fastest – come from households with economically empowered mothers. This pattern converges over time as female representation in parliament grows. And, while mothers' labour market outcomes differ substantially in the early decades, there are no meaningful differences between male and female politicians' fathers. Women who enter paid and prestigious political offices when the share of women in those offices was previously low tend to have an economically empowered mother.
Our results are informative for several strands of literature, upon which we add nuances to household compositions and female political representation. Recent work shows that breadwinning responsibilities can hinder women's political ambition (Bernhard, Shames, and Teele Reference Bernhard, Shames and Teele2021). Nevertheless, our results highlight that women in breadwinning roles may serve as important role models, encouraging the next generation of women to enter politics. Our results also add to a growing literature on the selection of Swedish politicians. Our focus on gender differences in parental labour market performance is new to this literature, which has focused on parents' social class backgrounds (Dal Bó et al. Reference Dal Bó2017), the impacts of a gender quota on politician quality, and immigrant background (Besley et al. Reference Besley2017; Folke, Freidenvall, and Rickne Reference Folke, Freidenvall and Rickne2015).
Our results provide at least two new insights into the drivers of female political representation. One insight is that policies that strengthen female economic empowerment can help women enter politics not only when the policy is introduced (as shown by, for example, Iversen and Rosenbluth Reference Iversen and Rosenbluth2010; Stockemer and Byrne Reference Stockemer and Byrne2012; Thomsen and King Reference Thomsen and King2020) but also in the next generation. The second insight underscores the importance of same-gender role models in the household for women's political ambitions to form, which holds relevance for the research literature on political ambition and family-based socialization (for example, Bos et al. Reference Bos2022; Jennings Reference Jennings, Dalton and Klingemann2007).
Theoretical Reasons for the Mother-Daughter Link
We draw on previous research to outline why a mother's economic empowerment may help propel their daughters’ entry into politics. Bos et al. (Reference Bos2022) recently theorized that political ambition forms through the joint processes of gender socialization and political socialization. Gender socialization happens as children observe adults in particular roles and make stereotypic inferences about suitability. Political socialization observes that politics is a masculine domain, and these two processes depress girls' political interests and ambitions relative to boys.
Observations in the household are fundamental to social learning (following, for example, Bandura Reference Bandura1977), and the household is an important arena where children observe parents' economic and social roles. A mother can be a role model in many ways. A mother with a greater degree of economic empowerment in the labour market encourages women to work (Fernandez 2013), a mother with a higher education level encourages women to vote (Gidengil, O'Neill, and Young Reference Gidengil, O'Neill and Young2010), and mothers' employment encourages their daughters’ participation in political organizations (Caudillo Reference Caudillo2017). This literature broadly shows that a role model effect from the mother raises the daughter's likelihood of making similar life choices.
An economically empowered mother may also affect her daughter through an information effect (Fernández Reference Fernández2013). Observing a mother who works in the labour market provides information about the pay-off from taking a job. This information may overcome the uncertainties that daughters, more than sons, may have about those pay-offs, such as their job opportunities or their employer's reaction if they have children. Women draw on two sources of information to learn about the cost-benefits of work. Observing a larger share of women in the labour force signals a higher pay-off from working and tilts the decision in this direction. Observing that one's mother works for pay is even more important because it conveys a more trustworthy personal signal to overcome uncertainty for the daughter.
Role models and information effects from mothers to daughters may likely exist for the decision to pursue political office. Even if the mother is not a politician, her economic empowerment in the labour market and household is informative about women in professional and public life and her role vis-à-vis her partner. Like the labour market, politics is an external domain to the household in which women likely face more uncertainty than men about the long-term pay-off from entry. For example, recent studies have reported abundant evidence of bias against women in political parties, as well as the added family-related pressures for women who pursue political careers (for example, Folke and Rickne Reference Folke and Rickne2016; Folke and Rickne Reference Folke and Rickne2020; Silbermann Reference Silbermann2015, reviewed by Thomas and Bittner Reference Thomas and Bittner2017).
The strength of the mother-daughter link will likely depend on multiple contextual variables. Fernández (Reference Fernández2013) expects the strongest intergenerational transmission when female overall economic participation is at the intermediate level. This is when uncertainty is the greatest – not the sure failure indicated by women's total absence or success indicated by a large female presence in the labour market. The level of politics may also be important. The mother's behaviour may matter more for national than local political offices because holding these offices requires a greater investment of time and energy, and their prestige and power likely make them more male stereotyped.
Data, Variables, and Methods
We created a complete list of elected municipal councillors and parliamentarians in fourteen elections. This list combines new archival data (1973–1988) with publicly available microdata (1991–2018). We link politicians via anonymized ID codes to administrative data for all permanent residents of Sweden going back to the 1960s. This gives us information on all parent-child links between ID codes, sex at birth, age, and various tax record data for labour market outcomes.
We focus on mothers in the main analysis; the appendix reports the results for fathers. For both, we use three variables to capture parental labour market outcomes. First, employment is a binary indicator that is self-reported in the census data and includes self-employment. Starting in 1991, Statistics Sweden provides a measurement that defines a person as employed if their total annual wage and business income exceeds a pre-specified monetary threshold (see Statistics Sweden 2019 for methodological details).
The second variable is the earnings percentile in the birth cohort's national earnings distribution. Earnings include both annual wages and business income, and we include people with zero earnings in the calculation. The third variable is the wife's share of household earnings, which captures her economic empowerment within the household.Footnote 1
Economic empowerment is a relative concept. We seek to determine whether the mothers of female politicians have a higher economic position in the labour market and household than other women. Therefore, we means-adjust our three variables to capture how a mother (or a father) deviates from other women (men) who are the same age and live in the same municipality. In any year in which we observe a parental outcome, we find the parents’ gender-municipality peers and calculate the average of the same variable among them. We then subtract that peer average from the parent's value. The resulting variables capture percentage point (ppt) deviations in employment, the share of household earnings, and percentile deviations for the earnings' percentile from the gender-peer average.Footnote 2 Our dataset for the full Swedish population always lets us find numerous peers, so this recalculation does not introduce any missing data. Appendix Table A1 reports the means of the adjusted variables and their unadjusted counterparts.
Because our labour market data comes mostly from tax records, it has very few missing values. Yet, we face a significant data truncation for how far back in time we can observe parental outcomes. This disallows an analysis of parental behaviour when the politician was an adolescent. Instead, we measure parents' outcomes as averages of all available observations where the parent is 30–65 years old.Footnote 3 Even this method results in substantial proportions of missing data. These proportions for parliamentarians are 31 per cent for employment status, 38 per cent for the earnings percentile, and 52 per cent for the mother's earnings share, and were even higher in the 1970s and 1980s (63 per cent, 72 per cent, and 84 per cent, respectively. See Appendix Table A2 for proportions of missing data across variables, political levels, and time periods.)
The data is missing because our labour market variables only became available in the 1960s. We cannot observe the employment status for parents who turned 65 before 1960, the earnings percentile for parents who turned 65 before 1968, or the mother's earnings share if either parent turned 65 before that year. This is more likely for politicians elected earlier in our sample period. It is also more likely for relatively older politicians and, therefore, for male politicians (Appendix Table A3 compares the gender and age compositions in our full sample of politicians and the estimation samples with non-missing parental variables). Data is more likely to be missing for politicians' fathers since they are usually older than politicians' mothers (as reported in Table A2).
We address the issue of missing data as follows. When comparing parents of male and female politicians, we hold constant the politician's age to compare women and men whose parental information is equally likely to be missing. Appendix Fig. A1 shows that election-by-election gender gaps in proportions of missing data for the dependent variables go to zero when we hold politicians' age constant. We also show that our results for politicians' mothers look identical when we restrict the data to politicians with non-missing data for both the mother and father (Appendix Fig. A4).Footnote 4
The final sample sizes differ slightly between the three labour market variables. For the employment variable, we have 3,348 person-year observations for parliamentarians (1,365 of whom are women) and 129,530 observations for municipal councillors (51,480 of whom are women). Appendix Table A1 reports all sample sizes and means for the parental labour market variables, politicians' demographics, and party composition by gender. The latter shows a similar average age and a larger share of females in parties on the ideological left.
Analysis and Results
We use two methods to compare the mothers of male and female politicians. The first method graphically plots mothers’ economic outcomes for both female and male politicians. These results appear in the top row of graphs in Fig. 2 for parliamentarians and Fig. 3 for municipal councillors. The second method calculates the size and significance of the gaps by estimating the following equation:
where Eie is the maternal outcome for politician (i) in election (e). We regress these outcomes on a dummy variable for female sex at birth (F i) multiplied by dummies for each election year (De). The vector estimates on these interactions (βe) capture the average female-male difference in mothers' economic empowerment in each election. We report these results in the bottom row of graphs in Figs. 2 and 3 by plotting the estimated coefficients (βe) together with their 95 per cent confidence intervals.Footnote 5 This is done for two regression specifications – one with no additional control variables and the other with fixed effects for the politician's party (p ie) and birth year (y i). These controls restrict the variation used to calculate the difference in mothers' outcomes to male and female politicians who are the same age and represent the same party.
Figure 2. Results for parliamentarians' mothers.
Notes: The upper row of the graphs shows developments over time of economic empowerment variables for mothers of male and female parliamentarians. Outcomes for the mothers are measured in administrative records going back to 1960 for employment and 1968 for earnings. For sample sizes, see Table A7, which shows the regression output. The lower row of graphs shows estimated female-male gender gaps in average maternal outcomes for politicians elected in each election from estimating two variants of Equation (1). Party fixed effects are nine dummies for political parties with representation in parliament during the period under study. Fixed effects for the birth cohort are dummies for each birth year. Vertical lines show 95 per cent confidence intervals.
Figure 3. Results for municipal councillors’ mothers.
Notes: The figure replicates the analysis in Fig. 2 for municipal councillors. For sample sizes, see Table A8, which shows the regression output.
The results for parliamentarians’ mothers in Fig. 2 depict a striking difference in household backgrounds between female and male parliamentarians. In the graphs for the over-time development of mothers' economic outcomes, women's mothers outperform their female cohort-region peers in the labour market (black lines). Meanwhile, male parliamentarians' mothers are less economically empowered than their peers (grey lines). This difference was large in the 1970s and 1980s: female parliamentarians' mothers were 5–15 ppt more likely to be employed than other women, placed 10–20 percentiles higher in the national earnings distribution, and earned a 5–15 ppt larger share of household earnings. Records for male parliamentarians' mothers showed the opposite: the grey lines indicate they were less likely to work, earned less if they did work, and earned a lower share of household earnings relative to other women of their age in the same municipality. These gaps disappear in the later decades.
The lower row of graphs shows the estimated gaps between women's and men's mothers (for regression outputs, see Appendix Tables A7–A8). Most of the observed differences in the early decades are statistically significant at the 5 per cent level. These differences increase when the fixed effects for political party and politician's age are included.
We further assess the magnitudes of the estimated gaps by comparing them to labour market statistics for all women aged 18–65 (the earnings percentile does this by design). The results suggest that in the 1970s and 1980s, among parliamentarians, the average woman's mother ranked about 15 percentiles higher than the average man's mother in the women's national earnings distribution. For the employment result, the difference between female and male politicians' mothers (11 ppt) corresponds to 16 per cent of the women's employment rate, which was 69 per cent at the time. For the women's share of household earnings, the gender gap of approximately 10 ppt corresponds to 32 per cent of the average women's share of the household income in those decades, which was 31 per cent.Footnote 6
Figure 3 displays the results for local politicians. The overall pattern echoes that for parliamentarians, but the gaps between male and female politicians are small. Female councillors’ mothers have somewhat higher levels of employment, earnings percentiles, and shares of household earnings compared to other women in their municipality. By contrast, men's mothers differ negatively or not at all. These gaps are just 2 ppt for the employment variable, 1–2 earnings percentiles, and 1 ppt for the household earnings share. The results clearly illustrate small differences in household backgrounds for male and female councillors, even when the number of women among the councillors increased rapidly in the 1970s and 1980s (recall Fig. 1).
We have argued that mothers' economic outcomes are more important than fathers' outcomes for women's political officeholding. Alternative theories might be that women's political empowerment thrives in families where the father or both parents outperform their cohort-region peers. Previous studies suggest that fathers may play an important role in the political socialization of daughters (Lawless and Fox Reference Lawless and Fox2010), and high-resource households might help women overcome disadvantages relative to men when entering male-dominated political organizations. However, repeating our analysis of politicians' fathers does not support this alternative conjecture. The fathers of male and female parliamentarians and municipal councillors outperform their cohort-region peers in the labour market, but there are no systematic or large differences by politician gender (Appendix Figs. A4 and A5).
A further reflection concerns the gradual decline in the gender gaps in the 1980s before the Left Party and the Social Democrats introduced zipper gender quotas in 1994. The share of women grew gradually in this early period, which may have made role modelling and information channels from mothers less important for women's entry into high political office. Two alternative interpretations receive little support in an extended analysis. Neither the politicians' labour market performance nor their parent's education levels appear to account for the main result. Adding control variables for the politician's own labour market outcomes to a two-period version of Equation (1) does not change the results (Panels A and B in Appendix Table A4). The same is true when the mother's education level is held constant, either as a level variable or adjusted in the same way as our main outcomes (Appendix Table A5).
A final robustness test restricts the sample to politicians elected for the first time. This does not change the overall results, indicating that parental differences correlated with the politicians' seniority do not drive our main results (results in Panel C, Table A4). A final empirical extension finds no clear patterns in the ‘quality’ of parliamentarians with more or less empowered mothers, using the quality measurement based on residual earnings developed by Besley et al. (Reference Besley2017) (Appendix Table A6).
Our full results show that female parliamentarians had more economically empowered mothers than male politicians in the decades when the share of women in parliament was at an intermediate level (between 10–35 per cent). While these maternal outcomes differed, paternal outcomes did not. A crude mechanism analysis based on statistical correlations failed to find support for parental education, female labour market outcomes, or partner choice driving our results. A possible residual interpretation might be that economically empowered mothers act as role models and provide moral support for their daughters' political careers (following Fernández Reference Fernández2013). This interpretation might also align with the lack of difference in maternal outcomes for the less prestigious and unpaid local political offices, for which a role model might be less important. Finding the difference in maternal outcomes during the period when the private signal from the mother was more important, that is, at intermediate levels of female participation, is also in line with what we expect from theory. As the level of participation grew over time, the private signal from the mother likely lost importance relative to the stronger public one.
Conclusions
Women's presence in Swedish national and local politics more than tripled from 10 per cent in the 1970s to 35 per cent in the 1980s – before a subset of parties adopted gender quotas in the early 1990s. We study this trend to understand better the driving factors behind women's entry into political officeholding. We uncover a new and dramatic pattern in the family backgrounds of women who entered national office in this period – compared to their male colleagues and women who entered office in later elections or at the local level, female parliamentarians' mothers had a substantially higher level of economic empowerment than other Swedish women.
Our results highlight the importance of women's economic empowerment for women's political empowerment beyond contemporaneous relationships in a single generation. There are also intergenerational dynamics that run from mothers to daughters. This intergenerational transmission implies that policies that increase female economic empowerment will likely have a long-run impact on gender equality in politics. Another conclusion is that the intergenerational link is not uniformly important across time and types of political office. Our results suggest a stronger importance for paid and prestigious national offices and at intermediate levels of women's numerical representation.
Future work could extend the analysis to specific policies that might trigger this intergenerational transmission by increasing female labour market participation and earnings. In the Swedish case, such policies might include individual taxation, generous parental leave, and universal and affordable childcare. While the Nordic countries are an obvious testing ground for this type of research, these results would be highly relevant for the many countries currently in the process of developing and expanding these types of public policies.
Supplementary Material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123423000595
Data availability statement
Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/UGMBIJ
Acknowledgements
We thank Josefine Magnusson for her expert research assistance. We also thank Lena Wängnerud, Maria Oskarson, and Anders Sundell, participants at the Twenty-Eighth International Conference of Europeanists, the European Conference of Politics and Gender, and workshop participants at the University of Gothenburg (GEPOP, History Club, and Politics & Gender) for helpful comments. We are also grateful to the anonymous reviewers and the editor for their comments and suggestions.
Financial support
Financial support from the Swedish Research Council (grant 2019-03153) and The Knut and Alice Wallenberg Foundation (Wallenberg Academy Fellow Grant for Johanna Ricke) is gratefully acknowledged.
Competing interests
None.
Ethical standards
This research was conducted in Sweden and was approved by the Swedish Ethical Review Board.
Open access
