So far, we have established a relationship between male-biased sex ratios in the 19th century and present-day outcomes for which we expect masculinity norms to play an important role: violence; suicide and help avoidance; occupational gender segregation; and opposition to sexual minorities’ rights. We now unpack what underlies this long-term relationship. First, we establish that our results reflect the persistent effect of masculinity norms. We do so by ruling out other explanations and by presenting direct evidence that masculinity norms constitute the mechanism that links historical sex ratios to present-day outcomes. Second, we investigate the strength of different persistence mechanisms that may explain the long-term impact of historical sex ratios.
7.1 Interpretation: masculinity norms or other factors?
7.1.1 Conservatism
The 2017 referendum on same-sex marriage was a politically charged event. Conservative political parties took position against legalization, and religious organizations were also heavily involved in the campaign. Is the relationship between historical sex ratios and present-day attitudes towards same-sex marriage really specific to attitudes towards homosexuality or merely a reflection of a legacy of sex ratios on social conservatism and political preferences more broadly?
Table 6 shows evidence in favor of the former: broad political attitudes, which go beyond the single issue of rights for homosexuals, are unaffected. Column 1 shows that the coefficient associated with the historical sex ratio does not have a significant effect on the share of votes for conservative partiesFootnote 35 in the general election in the year immediately preceding the same-sex marriage referendum. Hence, general conservatism cannot explain our results.
Table 6 Alternative mechanisms Full size table
7.1.2 Crime in general
We argue that the historical sex ratio has forged a locally variegated culture of male violence. Column 2 of Table 6 shows that our earlier results on violent crime and male aggression are not driven by local differences in the prevalence of crime in general: the results show that rates of property crime are unrelated to the convict sex ratio.
Cultural underpinnings of violence will act very differently on premeditated versus non-premeditated crime. Assaults are mostly non-premeditated and often result from quickly escalating confrontations, often over what seems to the initiator of the assault as a grave insult to his masculinity or lack of respect (e.g., Wolfgang (1958); Goffman (1959); Wilson and Daly (1985)). Property crime is much more premeditated, less responsive to impulse, and more reflective of a calculation of costs and benefits (Pinker, 2011).
The differentiated long-term effect of sex ratios on assaults versus property crime is, in fact, similar to the situation in the US South, where the Scots-Irish culture of honor still contributes to high rates of homicide and assault, but not to other types of crime, such as property crime (Grosjean, 2014). It is therefore reassuring that we do not find evidence for more widespread crime in areas that were more male-biased in the past, but only evidence on violent crime, one of the costly manifestations of hegemonic masculinity.
7.1.3 Industrial composition
One potential mechanism of persistence may be industrial specialization. Although the convict sex ratio was not systematically correlated with industrial composition during the convict era (Table 1), heightened masculinity norms may have influenced industrial composition in the intermediate period. This could then have propagated masculinity norms to the present-day. However, using 1933 census data on employment in 21 industries, we do not find any evidence that convict sex ratios influenced industrial composition in the intermediate period (Table 12).
7.1.4 The China shock
In the U.S., the increase in deaths of despair (Case & Deaton, 2020), particularly among men, has been linked to the deterioration in economic circumstances partly caused by rising international manufacturing competition, especially from China (Autor et al., 2013, 2019). One may worry that spatial variation in the sensitivity of local male employment to the rise of China may confound the relationship between historical sex ratios and present-day manifestations of masculinity norms.
To investigate this, we follow Autor et al. (2013) and construct a granular measure of how exposed local male employment was in 1991 to the sudden increase in Chinese imports between 1992 and 2006. The Australian Census allows us to calculate, at the level of Local Government Areas (LGAs), the proportion of men employed in various industries in 1991, at the start of China’s rise to economic prominence.Footnote 36 We then multiply these initial LGA-level shares with subsequent increases in Australia-wide imports from China.
The results in columns 3 and 4 of Table 6 show that male employment in manufacturing in 1991 (column 3) and exposure to import competition from China (column 4) are both unrelated to the convict sex ratio. In line with this orthogonality, Table 11 shows that our main results are robust to controlling for local gender-specific import shocks due to China’s rapid emergence as an economic powerhouse.
7.1.5 Institutional differences and legislation
The different states in Australia were independent colonies until 1901. Only New South Wales, Tasmania, and in later periods Western Australia were convict colonies. The colonies became different states today, which vary in their criminal legislation and, until recently, in legislation that affects sexual minorities, in ways that could be correlated with historical circumstances. For example, South Australia, which never harbored convicts, was the first state to decriminalize homosexuality in 1975, and Tasmania the last, in 1997. All our specifications include state fixed effects that remove the influence of time-invariant state characteristics or differences in legislation across states.
7.1.6 Convictism
The extent to which present-day violence, crime, and attitudes towards homosexuality are all stained by Australia’s convict past has been the object of a long-standing and intense debate.Footnote 37 Victorian authorities were so concerned about “blasphemy, rage, mutual hatred, and the unrestrained indulgence of unnatural lust” among convicts that it became one of the main arguments of transportation abolitionists. This in turn has led some to go as far as stating that: “prejudice toward LGBTI people [in Australia] can be summed up in one word: convictism”.Footnote 38
However, we control in all specifications for the number of convicts together with total population, so that our results are immune to any legacy of convictism in and of itself. For assaults and sex offenses, health and suicide, or the share of men employed in male occupations, the coefficient for the number of convicts is not statistically significant. We explore more directly the role played by the share of convicts as a determinant of attitudes towards homosexuality in a short companion paper (Baranov et al., 2020). We show that, contrary to popular opinion, areas with more convicts historically are today more likely to vote in favor of same-sex marriage. This highlights how the convict legacy must be distinguished from that of the radical distortion in sex ratios that convict transportation imposed.
We conclude, having ruled out alternative explanations, that our results reflect how male-biased sex ratios and elevated male-male competition forged a locally variegated culture of male violence, help avoidance, and self-harm, which has persisted until this day. We now turn to additional data that bring more direct evidence that masculinity norms constitute the mechanism that links historical sex ratios to present-day economic, social, and health outcomes.
7.2 Masculinity norms and outcomes: evidence from ten to men
This section provides direct evidence on the relationship between masculinity norms and a range of attitudes and behavioral patterns among Australian men. We use data from the Australian Longitudinal Study on Male Health (Ten to Men), a study of 16,000 boys and men aged 10 to 55 years at baseline.Footnote 39 The study collects comprehensive data on demographic and socioeconomic characteristics; physical and mental well-being; and health behaviors including the use of health services.
Importantly, the second wave of this survey allows us to construct for each respondent a score on the Conformity to Masculinity Norms Inventory (CMNI-22) and thus gauge the extent to which he adheres to a hegemonic masculine identity.Footnote 40 As discussed in Sect. 2.2, the CMNI is a multi-dimensional scale that measures to what extent an individual man’s actions, thoughts, and feelings conform to hegemonic masculinity norms in Western societies, such as emotional control; risk-taking; violence; dominance; self-reliance; and disdain for homosexuals. To create the CMNI score, Ten to Men asks respondents “Thinking about your own actions, feelings and beliefs, how much do you personally agree or disagree with each statement”, followed by statements capturing the dimensions in the CMNI-22. Answers range on a four-point Likert scale from 0 (strongly disagree) to 3 (strongly agree).
“Appendix Table 13” presents correlations between the CMNI-22 score and its primary components of interest. We restrict our sample to adult self-declared heterosexuals (N=13,317). The table shows tight correlations, all with the expected sign, between the various expressions of a hegemonic masculinity identity. We find that the strongest correlates of the overall CMNI-22 consist of norms related to dominance (“I make sure people do as I say” and “I love it when men are in charge of women); disdain for homosexuals (“It is important to me that people think I am heterosexual” and “It would be awful if someone thought I was gay”); violence (“Sometimes violent action is necessary”); and winning (“Winning is the most important thing”).
This survey is useful to relate masculinity norms to the outcomes that we study. Table 8 shows how well the overall CMNI-22 score predicts a number of real-life outcomes measured in Ten to Men. These correspond closely to the outcomes we have considered (and measured using various other data sources). In column 2, each cell is the coefficient associated with the standardized CMNI-22 score in an OLS regression controlling for respondent age (mean=34.9), Aboriginal or Torres Strait Islander indicator (mean=0.03), marital status (6 categories), language spoken at home (9 categories), as well as state fixed effects. Column 3 shows the coefficient on the CMNI-22 score after also adjusting flexibly for household income, respondent education level, and a socio-economic index based on place of residence.
The results confirm that men who adhere to strict masculinity norms systematically self-report types of behavior that align closely with our behavioral outcomes of interest. In particular, in line with our results in Table 2 on violent assault and sexual offenses, we find that men who score higher on the CMNI-22 scale are significantly more likely to admit they have engaged in intimate partner violence. In line with Table 3, we find that these men are also more likely to have thought about, planned, or attempted to commit suicide and are more likely to display signs of depression (as measured with the standard PHQ-9 Depression Score). They also engage in more risky health behavior, including smoking cigarettes, heavy drinking (“Injured while drinking”), and taking hard drugs. In line with medical help avoidance (and our prostate cancer results in Table 3), they are also significantly less likely to have consulted a GP in the past 12 months.
Unfortunately, the Ten to Men survey’s geographic coverage is too limited to enable us to relate norms directly to the historical sex ratio. Yet, as explained in Sect. 4.2, we singled out the CMNI dimension that best predicts the behavioral outcomes that we study (see column 4 of Panel A in Table 8) and commissioned the corresponding question on help avoidance to be included in the Australia-wide HILDA survey. In Panel B, we show that areas that were more male-biased in the past, remain characterized today by a greater prevalence of this masculinity norm. To be precise, a one standard deviation increase in the historical sex ratio is associated with a 2.8% decrease at the mean in a man’s inclination to ask for help.
In all, we conclude that male-biased sex ratios instilled strong masculine identities, which then persisted over time and still manifest themselves in a consistent way across political, economic, and social behaviors, attitudes, and norms. We now investigate the persistence mechanisms that underpin these findings.
7.3 Persistence mechanisms
Earlier work on cultural norms discusses two main persistence channels: (i) cultural vertical transmission within families, and (ii) horizontal peer-to-peer socialization (Bisin & Verdier, 2001; Hauk & Saez-Marti, 2002). We investigate each mechanism in turn. First, and consistent with the literature on the transmission of norms about the appropriate conduct and role of women in society (Alesina et al., 2013; Hansen et al., 2015), we find that vertical transmission within families explains part of the persistence of norms about the appropriate conduct of men. Here we also briefly discuss the role of migration. Second, we also document an important role for peer-to-peer transmission in schools.
7.3.1 Vertical transmission in families
To investigate vertical transmission, we follow the approach of Nunn and Wantchekon (2011) and GK, and contrast the attitudes of individuals of different ancestries. The idea is that only Australian parents transmit values that reflect historical Australian conditions. Individual-level information on ancestry is only available in the HILDA dataset. We regress individual attitudes towards same-sex marriage on the historical convict sex ratio, a dummy variable that indicates whether the respondent was born in Australia, and an interaction between these two variables. The coefficient associated with the interaction captures the strength of vertical transmission: it measures whether the local historical sex ratio influences more strongly the attitudes of individuals who are born in Australia, compared with foreign-born individuals. We also include the set of standard individual controls.
The results in the last column of Table 5 show that vertical transmission in families plays an important role in explaining the long-term persistent effect of convict sex ratios on attitudes towards same-sex marriage. The coefficient of the interaction term between the local convict sex ratio and whether the respondent was born in Australia is negative and statistically significant at the 5% level. This confirms that attitudes towards homosexuality of individuals born in Australia are indeed more sensitive to the historical sex ratio as compared with individuals born overseas.
7.3.2 Migration
The coefficient associated with the main effect of the convict sex ratio in the last column of Table 5 is smaller in magnitude than in our baseline specifications, but still significant at the 1% level. This suggests that, although the local historical sex ratio influences the views of Australian-born more strongly, foreign-born are not insensitive to it.
A recent literature discusses the role of migration in perpetuating cultural equilibria. For example, Bazzi et al. (2020) show that selective migration in and out of frontier areas in the U.S. sustained local norms of individualism. Non-selective migration would, to the contrary, attenuate persistence, as it would dissociate local historical conditions from current ones and bias against finding any relationship between historical conditions and present-day outcomes. However, flows of migrants at any given time are typically marginal as compared with the stock of stayers. This implies that horizontal transmission is more immune to migration, as even non-selected migrants will adjust to local norms.
In the context of international migration, a recent paper by Rapoport et al. (2020) shows, accordingly, that migrants adopt local norms. Our results are compatible with both potential explanations. They can be explained either by selective migration—foreign-born individuals selecting into areas where local opinions are similar to theirs—or by horizontal transmission—migrants adopting local values and attitudes.Footnote 41
7.3.3 Horizontal transmission in schools
To investigate horizontal transmission, we focus on peer-to-peer transmission at a young, impressionable age. We use data on bullying in school from LSAC, a longitudinal survey of youths (see Sect. 4). The results in Table 7 show how boys, but not girls, are more likely to be bullied at school in areas that were more male-biased in the past. A one standard deviation increase in the convict sex ratio is associated with a higher likelihood of parents reporting bullying of their sons by 8.5% points. The increase in rates reported by teachers is lower, at 3.6% points, but still statistically significant at the one percent level.
Table 7 Horizontal transmission: historical convict sex ratios and bullying in school Full size table
Our results on bullying suggest two things. First, they lend credence to the idea that hegemonic masculinity norms are enforced through intimidation, with (perceived) weaker individuals and especially (perceived) homosexuals being likely targets. This can further cement a violent, homophobic and emotionally repressed male social order.Footnote 42 Flood and Hamilton (2008) point out how Australian boys and young men who move outside the boundaries of stereotypically masculine behavior are often verbally and sometimes physically attacked.
Table 8 Historical convict sex ratios, masculinity norms and outcomes Full size table
Second, they suggest that masculinity norms are perpetuated through horizontal peer pressure, starting at a young age in the playground. This is consistent with List et al. (2019) who find evidence for large peer-level externalities in non-cognitive skills correlated with violence, such as inhibitory control, among boys.Footnote 43 Gilmore (1990) argues in this context that becoming a man is not so much a process of biological maturation, but instead a critical threshold that boys must pass through testing. Much of this testing takes place at school and in the playground.