Facebook: For preference dimensions that are systematically biased toward the same gender across the globe, differences between men and women are larger in more gender-equal countries

The Gender Gap in Preferences: Evidence from 45,397 Facebook Interests. Angel Cuevas et al. Southern Methodist Univ, October 7, 2021. http://faculty.smu.edu/kdesmet/papers/GenderGapFB.pdf

Abstract: This paper uses information on the frequency of 45,397 Facebook interests to study how the difference in preferences between men and women changes with a country’s degree of gender equality. For preference dimensions that are systematically biased toward the same gender across the globe, differences between men and women are larger in more gender-equal countries. In contrast, for preference dimensions with a gender bias that varies across countries, the opposite holds. This finding takes an important step toward reconciling evolutionary psychology and social role theory as they relate to gender.

1 Introduction

Do gender differences in preferences get attenuated or accentuated in more gender-equal societies? On the one hand, evolutionary psychology theory posits that gender equality accentuates differences by facilitating the expression of innate preferences that set men and women apart. On the other hand, social role theory posits that gender equality attenuates differences by eroding gender stereotypes and norms. Using data on the prevalence of a comprehensive set of 45,397 interests by gender across most countries of the world, this paper takes an important step towards reconciling both theories. Our premise is that innately gender-specific interests should mostly conform to evolutionary psychology theory, whereas other interests should mostly conform to social role theory. We find strong evidence consistent with this premise.

Our data on the prevalence of interests by gender and country come from Facebook. The social media company observes each of its almost three billion users’ online activity, not just on its own platform, but also on all websites and apps where it has a presence. In addition, it tracks many of its users’ offline activities by relying on GPS. Through their online and offline activities, users reveal their preferences and interests to Facebook. Using this information to assign interests to users, Facebook has unintentionally created the world’s largest database on preferences. By querying this database through Facebook’s publicly available Marketing API, we collect for most countries of the world the number of male and female users interested in 45,397 different topics. Because the data are at the level of populations (e.g., Canadian men or Ghanaian women), they do not entail any individual privacy issues. Compared to other potential data sources on preferences, Facebook data have two key advantages. First, the interests are broad and comprehensive in their scope, ranging from religious beliefs and sports, to political positions and cuisine. Second, in contrast to surveys, Facebook interests constitute a bottom-up revealed measure of preferences, covering whatever users find interesting, rather than what social scientists deem important. We start by computing for each country the cosine distance between the interest frequency vectors of men and women. This gives us a country-level metric of the overall difference in interests between genders. When regressing this metric on the degree of gender equality, we uncover a weak positive association between a country’s gender equality and the interest gap between men and women. Because different interests may sometimes reflect the same underlying preferences, we use singular value decomposition of the data matrix to identify the main latent preference dimensions. When recomputing our distance metric in this lower-dimensional subspace, we find a slightly stronger positive association between a country’s gender equality and its gender gap in preferences. Next, we differentiate between gender-related and non-gender-related interests. We say that an interest is gender-related if it displays a systematic bias toward the same gender across the globe. More specifically, if in more than 90% of countries an interest is more prevalent among the same gender, then we refer to it as gender-related. For example, “cosmetics” and “motherhood” are universally more common among women, whereas “motorcycles” and “Lionel Messi” are universally more common among men. Conversely, we say that an interest is non-gender-related if its gender bias varies across countries. More specifically, if an interest is more common among men in at least 30% of countries and more common among women in at least another 30% of countries, then we refer to it as nongender-related. For example, “world heritage site” and “physical fitness” do not display a systematic gender bias across the globe. When exploring the relationship between a country’s gender equality and the difference in interests between men and women, we uncover a sharp distinction between gender-related interests and nongender-related interests. More gender equality is associated with greater differences between men and women for gender-related interests, whereas the opposite is true for non-gender-related interests. As an alternative way of classifying interests, we use singular value decomposition to differentiate between gender and non-gender dimensions of preferences. For a preference dimension to be gender-related, we require the relative positions of men and women along that dimension to be similar across countries. With this alternative method, we confirm the paper’s central result: more gender-equal societies tend to be associated with greater differences in gender-related preferences but smaller differences in non-gender-related preferences. To interpret the paper’s main empirical finding, we turn to two seemingly contrasting theories (Falk and Hermle, 2018). Evolutionary psychology argues that men and women differ in areas where they faced different adaptive problems in their evolutionary history (Atari, Lai and Dehghani, 2020). In societies with more equal gender rights, men and women are able to more freely express their innate predispositions, so that preference differences between men and women should widen (Buss, 1989; Schmitt, 2015; Atari, Lai and Deghani, 2020).1 Social role theory, instead, argues that gender differences stem from gender socialization, social norms and sociocultural power structures (Schmitt et al., 2017). Since greater equality of gender rights erodes these norms, preference differences between men and women should narrow. While many papers on gender differences have been framed as a debate on the relative merits of evolutionary psychology and social role theory, these two views are not necessarily competing. Rather, their predictions apply to different preferences – evolutionary psychology to preferences that are innate and social role theory to preferences that are socially constructed. How does the difference between innate and socially constructed preferences relate to our paper’s main result? We argue that for preferences to be innate, they must display a systematic bias toward the same gender across the globe. As such, we can interpret our gender-related interests as potentially innate. In contrast, non-gender-related interests display a gender bias that varies across countries, and must hence be socially constructed. Using this interpretation, our findings are consistent with the predictions of both theories: in more gender-equal countries, differences between men and women are larger for innate (gender-related) preferences and smaller for socially constructed (non-gender-related) interests. Our interpretation depends crucially on the way we classify interests, and hence requires caution. We refer to gender-related interests as potentially innate, because we cannot discard the possibility that some of these interests might be socially constructed. Of course, this would require the process of social construction to occur in the same way in all countries. While in general this seems quite unlikely, in some cases the process of globalization might have led to the homogenization of socially constructed norms across countries. In other cases nature might have given rise to universally held gender norms in the distant past that then persisted through nurture despite no longer having a biological basis.2 For example, historically the relative physical strength of men and women was an important determinant of the division of labor between genders. As a result, universal gender norms emerged that associated some professions with men and others with women. Although technology has eroded these gendered patterns of comparative advantage, the gender norms might still survive.3 While ultimately such norms still have an innate origin, they are no longer subject to biological determinism. This paper is related to several strands of the literature on gender differences in preferences. Closest to our work is the large literature in psychology, sociology and economics that studies whether differences in values, attitudes and personality get accentuated in societies that are more gender-equal. Most empirical studies in this area have focused on gender differences in personality characteristics (Costa et al., 2001; Kaiser, 2019; Mac Giolla and Kajonius, 2019), cognitive abilities (Lippa, Collaer and Peters, 2010), education (Stoet and Geary, 2018), basic human values (Fors Connolly, Goossen and Hjerm, 2020), and specific cultural, behavioral and moral values (Falk and Hermle, 2018; Atari, Lai and Dehghani, 2020). Many of these studies find evidence of divergence between men and women in more gender-equal societies. For example, countries that are more gender-equal are found to exhibit greater sex differences in care and fairness (Atari, Lai and Dehghani, 2020), altruism, trust and risktaking (Falk and Hermle, 2018), and the big five personality traits (Mac Giolla and Kajonius, 2018). Some other studies find the opposite or argue that this relation is not robust. For example, Guiso et al. (2008) show that in societies with greater gender equality the math gender gap narrows, and Kaiser (2019) argues that the gender divergence in personality traits disappears after controlling for ecological stress factors such as hunger and disease. Our paper differs from this previous work in three respects. First, our data cover a broad crosssection of countries. Second, while most studies have focused on particular traits, values or abilities, we focus on 45,397 interests. Because of a lack of comprehensive data on interests and preferences, previous research has been unable to fully compare the predictions of evolutionary psychology and social role theory. Third, while these papers look at the effect of gender equality on differences in preferences, they do not address the possibility of causality running the other way. We deal with this potential endogeneity concern by taking an instrumental variable approach. Our results are suggestive of a causal interpretation of the paper’s main finding.

Also related to our work is the literature that seeks to identify some of the key differences in preferences between men and women. Many experimental papers have documented systematic gender differences in risk attitudes, dislike of competition, and social preferences (see Croson and Gneezy, 2009, Bertrand, 2011, and Niederle and Vesterlund, 2011, for excellent surveys). An important, related, question is to what extent these gender differences are a consequence of nature or nurture. Most direct evidence of the role of nature comes from studies that show that male hormones play a role in certain preferences, such as attitudes towards competition and risk-taking, as well as in career choices and activities (Archer, 2006; Dreber and Hoffman, 2007; Sapienza, Zingales and Maestripieri, 2009; Berenbaum and Beltz, 2021). More generally, the consensus points to both nature and nurture mattering. Even in the case of risk-taking, Gneezy et al. (2008) show that gender differences are society-dependent, ruling out a purely nature-based explanation.

Finally, an extensive literature in economics and political science explores how gender differences in preferences affect individual and societal choices. If women and men have different preferences, then greater female participation in political decision-making has wide-reaching consequences. ClotsFigueras (2012) demonstrates that the election of women politicians in India improves educational attainment; Lippmann (2021) shows that in the French parliament female legislative activity focused more on women’s issues and male legislative activity more on the military; and Funk and Gathmann (2015) show that in direct democracy initiatives in Switzerland women make different choices in health, environmental protection, defense spending and welfare policy. Differences in preferences are also relevant within the household. Quisimbing and Maluccio (2000) show that giving more assets to women translates into an increase in spending on offspring in a variety of developing countries. This is an important insight for government policy that often relies on direct cash transfers to improve children’s welfare. An additional effect of greater preference heterogeneity within the household is increased marital instability (Serra-Garcia, 2021). Gender differences in preferences also have important effects on career choices and other labor market outcomes (Bertrand, 2011). Hence, better understanding the evolution of gender differences in preferences is of great interest to economists.

The rest of the paper is organized as follows. Section 2 describes the data, with a special emphasis on the Facebook data on interests; Section 3 analyzes the relation between gender equality and gender differences in interests and preferences; Section 4 explores how this relation depends on whether interests and preferences are gender-related or not; and Section 5 concludes.