The share of women in top incomes
While the changes in top incomes, and in particular the rising income share of top earners in many countries, has received enormous attention in recent years, one important aspect of these developments has received relatively little notice: the gender dimension. What share of the top 10% or top 1% is made up of women? How has this changed over time and across countries? Are top-income women and men similar in terms of income composition and observable characteristics? How can we understand the differences?
Part of the neglect is due to a good reason: The top-income literature depends heavily on tax data and taxation is often handled jointly for married couples, making it difficult to separate men and women.1 There is, of course, a large literature on top-earning women (or rather the relative lack of them) observing, as summarised by Marianne Bertrand (2018) in a recent overview, that “[d]espite decades of progress, women remain underrepresented in the upper part of the earnings distribution, a phenomenon often referred to as the ‘glass ceiling’”.2
However, an important insight from the top-income literature is that, especially when it comes to the very top of the income distribution, capital incomes can (and often do) play an important role. This means that more fully understanding the gender dimension of top incomes may require data not just on labour market outcomes but also on other incomes and in particular capital incomes, as well as other individual characteristics.
Evidence from Sweden, 1971–2017
In a recent paper (Boschini et al. 2020), we exploit detailed register data on the full Swedish population starting in 1971, the year when Sweden shifted from joint income taxation of married couples to fully individual income taxation.3 The data allow us to distinguish labour income and capital income (and to some extent also underlying wealth holdings, as well as separating realised capital gains from other capital incomes; see the paper for details).
Figure 1 shows the basic development over time for three top groups – the top 10% (P90-100), the top 1% (P99-100), and the top 0.1% group (P99.9-100) – in the total income distribution and the labour income distribution, respectively.
Figure 1 Share of women in top income groups, 1971–2017
The figure illustrates two basic facts about the presence of women at the top of the income distribution.
- First, the share of women has steadily increased over time but they still remain a clear minority compared to men.
- Second, the higher up in the income distribution one moves, the smaller the share of women.
But comparing the two panels also reveals a subtler point: especially in the earlier decades and in the very top group, there were substantially more women in the top of the total income distribution than in the top of the labour earnings distribution. In the 1970s and 1980s, the share of women in the top 0.1% of the total income distribution is about two to three times as large as in the labour earnings distribution. Figure 2 illustrates this in a different way, showing the average income composition for top-income women and men respectively in terms of multiples of the (tax) population average.4
Figure 2 Average income composition for women and men respectively in terms of multiples of the (tax) population average, 1971–2017
Figure 2 shows that at the beginning of the period, when there were very few women at the top, those women were very different in their average income composition in that capital income was much more important for them. Over time, however, this has changed so that top-income women today, while still being fewer, are much more similar to top-income men on average.5
This development happens over a period when capital incomes become more important for Swedish top earners in general. One might have expected that this would give the initially more capital-intensive top-income women a ‘mechanical advantage’, leading to an increased share of top women. However, a more careful decomposition analysis shows that the prevailing trend over this period was women gaining shares in the top of the labour income distribution, which is what mainly caused the increased share of top women documented in Figure 1 (see section 4 of Boschini et al. 2020).
The gender convergence in income compositions applies to a number of other individual dimensions as well. The population average and the composition in the top-income groups in terms of age, education, and marital status, and so on have changed substantially, but in all of these dimensions, the characteristics of top-income men and women have converged. The one area where differences remain striking is in family and, in particular, partner income characteristics.
Figure 3 shows a summary: top-income women are much more likely to have partners who are also in the top of the income distribution. Out of the top-1% women who are married, 70% have a partner who is at least in the top 10% (and about 30% are also in the top 1%).
For married top-1% men, only 30% have a partner who is in the top 10%, and only a couple of percentage points are in the top 1% (see Boschini et al. 2020 for more details).6
Figure 3 Top-1% women’s and top-1% men’s partners who are in P90-100, 1971-2017
Taken together, this suggests that family circumstances are at least as important as individual characteristics in understanding the relative absence of women in top-income groups. Having a partner who is also focused on a top (income) career is likely to be more demanding (for both parties), and such couples are much more common among top-income women than men.
This clearly relates to, for example, the findings in Fisman et al. (2006) and to recent work on the impact of gender identity on the relative share of men’s and women’s respective contribution to total household income (e.g. Bertrand et al. 2015, Eriksson and Stenberg 2019), suggesting that there is a social norm prescribing that men should earn more than women (as also found in Bursztyn et al. 2017). It also relates to recent work by Folke and Rickne (2020), who find that women who are elected to high political office face a higher probability of divorce, while this is not the case for men. It also relates to work on increasing gender wage gaps among executive managers after having their first child (Keloharju et al. 2019) and in couples more generally (Angelov et al. 2016, Kleven et al. 2019).
Putting our basic results on the share of women in top groups in Sweden next to those in Atkinson et al. (2018), who study the share of women in top groups for eight countries over time periods when taxation has been individual, we see a remarkably similar development over time (Figure 4). The share of women in the top 10% and top 1% has approximately tripled, from around 10% to around 30%.
Figure 4 Share of women in top income groups in various countries, 1971–2017
In a related paper (Bobilev et al. 2019), we explore whether Luxembourg Income Study data can shed light on the presence of women in the top of the income distribution. We find developments of the share of women in top groups (top 10% and top 1%) of the labour income distribution for 28 countries are very similar in trend and levels from around 1980 until today.
The Luxembourg data also allow us to look at partners and family circumstances. Even though there are a limited number of countries where samples are large enough to look at questions like these for the very top groups, we find a consistent pattern of asymmetries among top-income men and women when it comes to family. Having a partner and having children are positively associated with being in top-income groups for men but negatively associated for women, although time interactions suggest that these differences have decreased over time. Also, top-income men are likely to have partners who are not in the top of the income distribution, while this is not the case for top-income women.
Understanding patterns like these is likely to be important for understanding the remaining differences in top-income shares between men and women.
Blau, F D, L M Kahn (2017), “The gender wage gap: Extent, trends, and explanations”, Journal of Economic Literature 55(3): 789–865.
Bertrand, M (2018), “Coase Lecture – The Glass Ceiling”, Economica 85: 205–31.
Bobilev, R, A Boschini and J Roine (2019), “Women in the top of the income distribution – What can we learn from LIS-data?”, Italian Economic Journal, forthcoming.
Boschini, A, K Gunnarsson and J Roine (2020), “Women in top incomes – Evidence from Sweden 1971–2017”, Journal of Public Economics 181.
Fisman, R, S S Iyengar, E Kamenica and I Simonson (2006), “Gender differences in mate selection: Evidence from a speed dating experiment”, Quarterly Journal of Economics 121(2): 673–97.
Folke, O, and J Rickne (2020), “All the single ladies: Job promotions and the durability of marriage”, American Economic Journal: Applied Economics 12(1): 260–87.
Garbinti, B, J Goupille-Lebret and T Piketty (2018), “Income inequality dynamics in France 1900–2014: Evidence from distributional national accounts (DINA)”, Journal of Public Economics 162: 63–77.
Guvenen, F, G Kaplan, J Song (2014), “The glass ceiling and the paper floor: Gender differences among top earners 1981-2012”, NBER Working Paper 20560.
Piketty, T, E Saez and G Zucman (2018), “Distributional national accounts: Methods and estimates for the US”, Quarterly Journal of Economics 133(2): 553–609.
1 A notable exception is Atkinson et al. (2018), which studies the share of women in top groups for eight countries with independent taxation. Also, some recent papers, such as Garbinti et al. (2018) on France and Piketty et al. (2018) on the US, deal with gender inequality in countries where married couples file taxes jointly, but these papers focus on individual labour earnings.
2 See Blau and Kahn (2017) for an overview.
3 The shift was not complete until 1986, as some special types of incomes remained joint after the 1971 reform, but these were so small that they have very little impact on any calculations. See Appendix A of the published paper for details.
4 In the case of Sweden, the tax population for the adult population is very close to the actual population. It was slightly lower in the 1970s but none of the trends are due to changes in this as we calculate income shares in relation to a homogenously defined reference total for the population (as is standard in the top-income literature).
5 One can argue with this comparison since there is a ‘thinning out’ of women as one moves up in the distribution, implying that average men and women in a top group should not be expected to be equal if the income composition also changes gradually. This is true, but since the overall composition is such that capital becomes increasingly important as one moves up the distribution, the average of women who are relatively more concentrated toward the lower part of the group should look less capital intensive, not more, if men and women in the same positions were in fact similar.
6 Part of this is a reflection of there being fewer women in top groups, but that is far from explaining all of the difference.