4.1. Population-level demographic patterns

The temporal trend in cohort fertility rates (average number of offspring per birth year, Fig. 1A) aligns with previous Estonian estimates based on census (Klesment et al., Reference Klesment, Puur and Valge2010) and survey (Katus, Reference Katus2000) data. The pattern of peaking and subsequent decline in birth rates among cohorts born around 1960 differs from that observed in countries that experienced a post–World War II baby boom (Katus et al., Reference Katus, Puur and Põldma2004). The absence of a post-war baby boom in Estonia has been attributed to population loss, political violence and the lack of post-war economic improvement (Gortfelder et al., Reference Gortfelder, Puur and Klesment2025).

Secular trends in the timing of first births, too, were consistent with previously published data for the Estonian population (Katus & Puur, Reference Katus and Puur2006; Nugin et al., Reference Nugin, Kannike, Raudsepp, Nugin, Kannike and Raudsepp2016). The pattern of advancement – and subsequent delay – of AFB up to around the 1965 birth cohort (Fig. 1B) is similar to trends observed in Greece, the Czech Republic, Hungary and Slovenia. In Nordic countries, postponements of first births continued in cohorts born after 1970 (Hellstrand et al., Reference Hellstrand, Nisén, Miranda, Fallesen, Dommermuth and Myrskylä2021). In contrast, in several other European countries and the United States, the turning points to delay reproduction in women occurred earlier, among cohorts born between the 1930s and 1950s (Beaujouan et al., Reference Beaujouan, Zeman and Nathan2023; Frejka & Sardon, Reference Frejka and Sardon2006).

One of the factors that sustained early family formation and childbearing in Estonia up to the cohorts born after 1965 may have been the housing allocation system under the state socialist regime. New dwellings were typically distributed according to administrative regulations, and to become eligible for housing, a couple generally had to live in conditions that fell below a certain minimum standard, particularly in terms of per-capita floor space. In this context, the birth of a child increased young couples’ chances of advancing in the housing queue. The persistence of these mechanisms until the late 1980s likely explains why the shift towards postponing childbearing occurred relatively late in Estonia (Klesment et al., Reference Klesment, Puur and Valge2010).

Despite the largely parallel trends in AFB and ALB, the reproductive pace increased among both men and women from the 1920s cohorts through to the 1970 birth cohort. In the following birth cohorts, IBIs started to decline. The decline was particularly abrupt among men, which is likely attributable to increased under-registration of fatherhood among those born in 1970s (see Fig. 1A).

The gender gap in cohort average offspring number has been increasing most rapidly among parents with lowest educational level throughout the study (Fig. 2). This pattern indirectly supports the idea that men with the lowest education prevailed among the unregistered fathers. Such interpretation would be also consistent with findings that the prevalence of single motherhood increased with declining stigma against out-of-wedlock births, growing female economic independence and shifting norms about marriage, cohabitation and divorce (e.g. Kutsar et al., Reference Kutsar, Kasearu and Kurrikoff2012; Schnor & Jalovaara, Reference Schnor and Jalovaara2020).

4.2. Sexually antagonistic selection on educational attainment

Over the whole study period, women with primary education had an average of 2.14 children, which is 0.4 more than the average of 1.74 children among women with tertiary education (Fig. 2A, Table 2). This difference is nearly twice as large as the 0.22-child gap observed between the highest and lowest educational attainment groups in the 1958 British National Child Development Study cohort (Nettle & Pollet, Reference Nettle and Pollet2008). Among the birth cohorts with the highest (1959) and lowest (1926) fertility, the difference in the number of children between the most- and least-educated women was even more pronounced – 1 child and 0.68 children, respectively (Table 2).

The educational gradient for reproductive success in women retained its direction throughout the 52-year study period, although the magnitude of selection gradients at different educational levels varied considerably. Women’s reproductive rates peaked among cohorts born from 1959 to 1962, during which the average offspring number exceeded 2 in all educational groups except those with tertiary education. These findings are consistent with 2011 population census data, showing that throughout the twentieth century, Estonian women with primary education bore 0.5–0.75 more children on average than women with tertiary education (Tiit, Reference Tiit and Tiit2013). Educational gradient in offspring number is consistent with nearly universal findings that reproductive performance in women associates negatively with educational attainment (Beaujouan et al., Reference Beaujouan, Zeman and Nathan2023; Berrington et al., Reference Berrington, Stone and Beaujouan2015; Hellstrand et al., Reference Hellstrand, Nisén and Myrskylä2024; Kirk et al., Reference Kirk, Blomberg, Duffy, Heath, Owens and Martin2001; Nisén et al., Reference Nisén, Klüsener, Dahlberg, Dommermuth, Jasilioniene, Kreyenfeld, Lappegård, Li, Martikainen, Neels, Riederer, te Riele, Szabó, Trimarchi, Viciana, Wilson and Myrskylä2021; Sobotka, Reference Sobotka2017; Van Bavel et al., Reference Van Bavel, Klesment, Beaujouan, Brzozowska, Puur, Reher, Requena, Sandström, Sobotka and Zeman2018). The convergence of reproductive rates among women of all educational levels in the birth cohorts of the 1970s mirrors trends observed in Denmark, Norway and Sweden. In all these countries, this convergence was caused by the decline of reproductive rates among the least educated, while the reproductive rates of the most educated women either increased or stabilized. In contrast, women with the lowest educational attainment in Finland maintained a reproductive advantage until the birth cohorts of the 1970s (Jalovaara et al., Reference Jalovaara, Neyer, Andersson, Dahlberg, Dommermuth, Fallesen and Lappegård2019). The reproductive advantage of the lowest-educated women over all other educational groups was also retained in the Czech Republic through birth cohorts from 1900 to 1970 (Zeman, Reference Zeman2018).

In contrast to women, men with tertiary education had the highest average offspring number (1.98), wheras those with primary education had the lowest (1.64; Table 2), followed by men with secondary (1.73) and vocational education (1.91; Table 2). The difference in the number of offspring between men with the highest and lowest levels of education (0.34 children) was smaller than that observed among women, but still 3.8 times larger than the 0.09-child-gap observed in the 1958 British birth cohort (Nettle & Pollet, Reference Nettle and Pollet2008). Notably, the educational differences in male reproduction in our study were less clearly distinguishable than in women until the 1950s birth cohorts (Fig. 2). Sexual antagonism in selection on educational attainment thus became distinctly evident in cohorts born from the mid-twentieth century onward. It should be recalled, however, that our analyses likely overestimate the strength of selection against poorly educated men, particularly in the later-born cohorts, which would eventually somewhat weaken the sexual antagonism in selection on educational attainment.

The increasing educational gradient in male reproductive success among cohorts born since the 1950s can likely be attributed to the growing value of male education in the mating market since the 1990s. During the Soviet era, monetary returns on education were minimal: in 1989, university-educated employees earned only 8% more than the economy-wide average wage, and workers with only primary education earned just 3% less than the average (Noorkõiv et al., Reference Noorkõiv, Orazem, Puur and Vodopivec1998). In contrast, the transition to a market economy dramatically increased the returns from education. Men with primary education experienced disproportionately large income losses and higher unemployment rates during the 1990s (Noorkõiv et al., Reference Noorkõiv, Orazem, Puur and Vodopivec1998). This effect of paternal education had tangible consequences on offspring outcomes. For example, among a subset of Estonian schoolchildren born between 1980 and 1987, those with fathers who had primary education were shorter, had smaller heads and reported lower resource availability compared to children of fathers with education beyond the primary level (Lauringson et al., Reference Lauringson, Veldre and Hõrak2020). Notably, these fathers were, on average, 3.3 cm shorter than more highly educated fathers (see also Valge et al., Reference Valge, Meitern and Hõrak2019).

Reports of positive associations between educational attainment and reproductive success in contemporary men are less consistent than those for women (reviewed by Nisén et al., Reference Nisén, Martikainen, Kaprio and Silventoinen2013), possibly due to men being understudied in this regard (Kravdal, Reference Kravdal2007). In Nordic countries, the fertility of low-educated men has consistently been the lowest through birth cohorts from 1940 to 1969 (Jalovaara et al., Reference Jalovaara, Neyer, Andersson, Dahlberg, Dommermuth, Fallesen and Lappegård2019; Nisén et al., Reference Nisén, Martikainen, Silventoinen and Myrskylä2014). Similar to the current study, the gap between low-educated men and other educational groups of men has widened somewhat in Denmark, Finland and Norway since the 1950s (Jalovaara et al., Reference Jalovaara, Neyer, Andersson, Dahlberg, Dommermuth, Fallesen and Lappegård2019). Outside the Nordic countries, the association between education and fertility in men varies from positive through flat to negative (reviewed by Nisén et al., Reference Nisén, Martikainen, Silventoinen and Myrskylä2014).

The detection of sexually antagonistic selection on educational attainment in this study confirms previous findings from a sample of Estonian schoolchildren (15,253 girls and 11,842 boys) born between 1937 and 1962 (Valge et al., Reference Valge, Meitern and Hõrak2022). Sexually antagonistic selection on educational attainment is consistent with several previous studies based on population registries (reviewed by Jalovaara et al., Reference Jalovaara, Neyer, Andersson, Dahlberg, Dommermuth, Fallesen and Lappegård2019; Valge et al., Reference Valge, Meitern and Hõrak2022). For example, opposing education–fertility gradients for men and women were observed in registry data for Norwegian cohorts born between 1940 and 1964 (Kravdal & Rindfuss, Reference Kravdal and Rindfuss2008). A similar pattern was found in a study of Finnish twins born between 1950 and 1957 (Nisén et al., Reference Nisén, Martikainen, Kaprio and Silventoinen2013), as well as in a large register-based analysis of birth cohorts from 1940 to 1974 across four Nordic countries (Jalovaara et al., Reference Jalovaara, Neyer, Andersson, Dahlberg, Dommermuth, Fallesen and Lappegård2019). Sexually antagonistic education–fertility gradients have also been reported in the US General Social Survey (Hopcroft, Reference Hopcroft2006), the Framingham Heart Study (Stearns et al., Reference Stearns, Govindaraju, Ewbank and Byars2012) and Swedish population register data (Fieder & Huber, Reference Fieder and Huber2007).

However, the Wisconsin Longitudinal Study in the US found that although educational attainment was phenotypically negatively associated with the number of children in both sexes, genetic predisposition for education negatively affected reproductive success only in men (Fieder & Huber, Reference Fieder and Huber2022). Negative genetic correlations between educational attainment and LRS have also been reported in the UK Biobank (Hugh-Jones & Abdellaoui, Reference Hugh-Jones and Abdellaoui2022; Warrington et al., Reference Warrington, Hwang, Nivard and Evans2021). Similar to the present study, selection against educational attainment was stronger in women than men. Selection against genetic variants associated with higher educational attainment was also reported in a meta-analysis of 62 cohorts of European ancestry (Barban et al., Reference Barban, Jansen, de Vlaming, Vaez, Mandemakers, Tropf, Shen, Wilson, Chasman, Nolte, Tragante, van der Laan, Perry, Kong, Consortium, Ahluwalia, Albrecht, Yerges-Armstrong, Atzmon and Mills2016). Finally, negative selection against polygenic scores for education – along with evidence of evolutionary response to this selection, which was weaker in males than in females – has been documented in 129,808 Icelanders born between 1910 and 1990 (Kong et al., Reference Kong, Frigge, Thorleifsson, Stefansson, Young, Zink, Jonsdottir, Okbay, Sulem, Masson, Gudbjartsson, Helgason, Bjornsdottir, Thorsteinsdottir and Stefansson2017).

Direct selection gradients on educational attainment in our study ranged from −0.32 to 0.21 (Table 3), exceeding those reported in the UK Biobank (−0.061 for women, –0.009 for men) (Sanjak et al., Reference Sanjak, Sidorenko, Robinson, Thornton and Visscher2018) and US Health and Retirement Study (−0.057 for women, −0.022 for men) (Beauchamp, Reference Beauchamp2016). However, our estimates on men were comparable in magnitude to selection gradients on male wealth and income reported in three industrial societies (β = 0.10–0.17; Nettle & Pollet, Reference Nettle and Pollet2008). In contrast, marrying a man from a ‘rich’ social class in pre-industrial Finland was associated with a selection gradient of −0.17, compared to the absence of selection when married to ‘poor’ men (Hayward et al., Reference Hayward, Nenko and Lummaa2015).

4.3. Reproductive timing across sexes and educational levels

Differences in reproductive timing across educational levels were considerably larger than the annual fluctuations over the generations studied (Figs. 1 and 3). Temporal trends in the ages of first and last birth were generally parallel across all educational strata and sexes (Fig. 3). The educational differences in AFB were relatively stable until the birth cohort of 1965, after which they started to increase among both men and women (Fig. 3A). These patterns differ from the Norwegian data where difference in AFB between the women with lowest and highest education showed steady increase from birth cohorts of 1940 (5.6 years) to 1964 (7.5 years) (Kravdal & Rindfuss, Reference Kravdal and Rindfuss2008). In Estonia, the corresponding differences were 3.8 and 3.7 years. The educational gap in age at first birth peaked at 5.7 years in the 1977 birth cohort, largely driven by women with tertiary education, whose AFB rose faster than that of women with primary education.

As the vast majority of the studied parents began childbearing during the Soviet era, the relatively small educational gap in AFB up to the 1965 birth cohort may reflect the specific features of state socialist economies – namely, low labour market returns to education, state-guaranteed full employment with highly structured career paths, and extensive public childcare provision. These conditions likely resulted in relatively low and similar opportunity costs of childbearing for women across different educational levels (Klesment & Puur, Reference Klesment and Puur2010). In line with this reasoning, the impact of educational attainment on AFB was less strong in East Germany than West Germany (Kreyenfeld, Reference Kreyenfeld2000).

Among men, educational differences in AFB were less pronounced but still notable. The consistent pattern of later AFB among tertiary-educated men since the 1929 cohort parallels that of women, suggesting that the long-standing effects of extended education on reproductive timing have operated similarly in both sexes. Comparable results were found in a study of six European countries in birth cohorts of 1960–1987 (Trimarchi & Van Bavel, Reference Trimarchi and Van Bavel2018) and in Finnish birth cohorts of 1940–1950 (Nisén et al., Reference Nisén, Martikainen, Silventoinen and Myrskylä2014). These findings also show that an early start to reproduction was not necessary for achieving a high offspring number among men with tertiary education. In contrast to women, cohort-average ages at first birth largely overlapped among men with primary, secondary and vocational education until the 1970s birth cohorts. This indicates that, for men, educational differences below the tertiary level had little effect on the timing of reproduction.

Both men and women with tertiary education ended reproduction later than their less educated peers, and secular trends largely mirrored those observed in AFB. The result that highly educated women ceased reproduction later than others aligns with a few previous studies on ALB (Singh et al., Reference Singh, Shekhar and Shri2023 and references therein; Varas Enríquez et al., Reference Varas Enríquez, McKerracher and Montalva Rivera2022). The most straightforward explanation is likely the postponement of childbearing until after graduation (e.g. Mattison et al., Reference Mattison, Moya, Reynolds and Towner2018). For highly educated men, a further reason for delayed cessation of reproduction may be their prolonged accumulation of social capital and other resources, which ultimately attract women with high and sustained reproductive potential (Buunk et al., Reference Buunk, Dijkstra, Fetchenhauer and Kenrick2002; Fales et al., Reference Fales, Frederick, Garcia, Gildersleeve, Haselton and Fisher2016). Additionally, such resourceful men may extend their reproductive careers by repartnering with younger spouses, as higher educational attainment can increase the likelihood of forming multiple partnerships (De Graaf & Kalmijn, Reference De Graaf and Kalmijn2003; Maslauskaitė & Baublytė, Reference Maslauskaitė and Baublytė2015; Nisén et al., Reference Nisén, Martikainen, Silventoinen and Myrskylä2014; Prskawetz et al., Reference Prskawetz, Vikat, Philipov and Engelhardt2003; Schnor et al., Reference Schnor, Vanassche and Van Bavel2017; but see Jalovaara et al., Reference Jalovaara, Andersson and Miettinen2022; Kolk & Barclay, Reference Kolk and Barclay2019; Lappegård et al., Reference Lappegård, Rønsen and Skrede2011).

Unlike AFB and ALB, the pace of reproduction exhibited a sexually antagonistic educational gradient: women with the highest education and men with the lowest education had the shortest interbirth intervals. The association between higher education and shorter birth spacing among women has been previously documented in several northern, western and southern European countries (Klesment et al., Reference Klesment, Puur, Rahnu and Sakkeus2014). This pattern aligns with the commonly observed trend that women of higher socioeconomic status tend to exercise greater control over their reproductive behaviour (Varas Enríquez et al., Reference Varas Enríquez, McKerracher and Montalva Rivera2022). Short birth intervals of highly educated women have been explained by a ‘time squeeze’ effect, arising from postponement of entry to motherhood, thereby reducing birth intervals to meet fertility goals, or to minimize the loss of wages, and job market opportunities, and reduce the time spent away from work in childcare (Bartus et al., Reference Bartus, Murinkó, Szalma and Szalma2013; Compans et al., Reference Compans, Beaujouan and Suero García2023; Klesment et al., Reference Klesment, Puur, Rahnu and Sakkeus2014; Rendall et al., Reference Rendall, Harrison and Caudillo2020).

We are aware of two studies that have analysed the IBIs of men concerning educational levels. In Finnish birth cohorts of 1940–1950, men with tertiary education had the fastest reproductive pace, measured as the interquartile range of age at having children (Nisén et al., Reference Nisén, Martikainen, Silventoinen and Myrskylä2014). Among Finnish men born between 1955 and 1975, interbirth intervals of both men and women mostly decreased with increasing educational levels (but this effect was sex-specifically mediated by AFB; Berg & Rotkirch, Reference Berg and Rotkirch2014). Our finding that, unlike women, more highly educated men had the most extended interbirth intervals contrasts with the results of both Finnish studies cited above. This indicates that, in addition to sex, the mechanisms linking parental education to the pace of reproduction may vary even between neighbouring countries.

4.4. Natural selection on reproductive timing across educational levels

The linear selection gradient on women’s age at first birth among women with primary education (β = −0.30) was similar to that observed over all women of the UK Biobank (Sanjak et al., Reference Sanjak, Sidorenko, Robinson, Thornton and Visscher2018). Compared to those reported in pre-industrial societies (Bolund et al., Reference Bolund, Bouwhuis, Pettay and Lummaa2013; Bürkli & Postma, Reference Bürkli and Postma2014; Käär et al., Reference Käär, Jokela, Helle and Kojola1996; Milot et al., Reference Milot, Mayer, Nussey, Boisvert, Pelletier and Réale2011; Pettay et al., Reference Pettay, Helle, Jokela and Lummaa2007), selection gradients on components of reproductive timing in our study were similar in magnitude and direction, though slightly weaker (β = −0.30–0.58; Fig. 4, Supplementary Table S1).

Among women, the strength of natural selection on all components of reproductive timing varied across educational levels. In men, selection on AFB showed little variation by education, whereas selection on ALB and IBIs was strongest among those with the lowest educational attainment. Natural selection on reproductive timing was thus generally strongest among parents with the lowest education and weakest among parents with the highest education. Among individuals with secondary and vocational education, the strength of selection was intermediate and, in most cases, of similar magnitude.

Beyond sexually antagonistic selection, empirical evidence that natural selection acts at varying rates across subgroups within modern populations remains limited. Our findings align best with a study in the UK Biobank (Hugh-Jones & Abdellaoui, Reference Hugh-Jones and Abdellaoui2022) showing that natural selection effects on polygenic scores of anthropometric and behavioural traits were stronger in groups with lower income and education, younger parenthood, not living with a partner and more lifetime sexual partners. Outside these groups, the effects of natural selection were weaker and often statistically insignificant. Among women with AFB above 22 years, the direction of selection was reversed. In the US Health and Retirement Study, natural selection on the polygenic scores of health traits was stronger for low-income groups and unmarried respondents (Hugh-Jones & Edwards, Reference Hugh-Jones and Edwards2024). However, unlike the UK Biobank study, they found little evidence for stronger natural selection among people with lower education.

Hugh-Jones and Abdellaoui (Reference Hugh-Jones and Abdellaoui2022) and Hugh-Jones and Edwards (Reference Hugh-Jones and Edwards2024) interpret their findings in the context of the economic theory of fertility (Becker, Reference Becker1960). The theory assumes that parents make rational decisions about how many children to have and how much to invest in them, balancing the economic costs and benefits. It states that higher potential earnings have two opposite effects on fertility: a fertility-increasing income effect (higher income makes children more affordable) and a fertility-lowering substitution effect (time spent on childrearing has a higher cost in foregone earnings). Hugh-Jones and Abdellaoui (Reference Hugh-Jones and Abdellaoui2022) suggest that the negative association between income and education (and traits related to these) versus fertility indicates that substitution effects dominate income effects, explaining why rising income or education levels typically shift preferences towards fewer, ‘higher-quality’ children.

Historical evidence for variation in the intensity of natural selection across population subgroups comes from pre-industrial Finland. Pettay et al. (Reference Pettay, Helle, Jokela and Lummaa2007) found the strongest selection on earlier age at first reproduction in women of the poorest wealth class, whereas selection favoured older age at reproductive cessation among the wealthy and middle-class mothers. The authors interpret their findings in the context of the theory of life-history evolution, which predicts stronger selection for early reproduction under conditions of high mortality and a lower chance of reproducing later in life, as was the case for the poorer women. However, even though the poor women would have benefited the most from early reproduction, they had the latest mean AFB. The authors suggest that poor women might have delayed their reproduction through behavioural means to adjust their child number to match their unfavourable economic circumstances.

The situation in the present study differs in that women with the lowest levels of education – used here as a proxy for the lowest social class – exhibited both the strongest selection for earlier reproduction and, typically to developed societies, the earliest average AFB throughout the study period. Notably, such women (along with the least-educated men) also displayed the strongest selection for the age of last birth. However, both women and, in particular, men with only primary education were also the first to cease reproduction compared to better-educated parents. Stabilizing selection for intermediate IBIs also showed the strongest quadratic selection gradients among minimally educated parents of both sexes (Fig. 4, Supplementary Table S1). At that, men with primary education exhibited the fastest, while women with primary education showed the slowest reproductive pace for most of the study period (Fig. 3).

Altogether, these results indicate that variation in reproductive timing across societal strata does not consistently reflect the selective pressures acting on recent generations. This pattern echoes findings from a study of pre-industrial Finns spanning up to 12 generations, where age at first birth remained relatively high over 350 years, despite strong selection for earlier reproduction and strong genetic correlations with fitness (Bolund et al., Reference Bolund, Bouwhuis, Pettay and Lummaa2013).

It is important to note, however, that the traits considered here differ in how selection was measured. Educational attainment is defined for all individuals, including those who never reproduced, enabling standard approach to measuring selection (covariance between trait values and reproductive success). On the other hand, parameters of timing of reproduction (AFB, ALB, IBI) are only defined for parous individuals. Thus, our estimates of selection on reproductive timing describe variation among those who reproduced, but do not capture selection acting through childlessness. If the probability of remaining childless covaries with reproductive timing, our estimates may underestimate the total strength of selection at the population level. Such a scenario is likely, given that, genome-wide association studies have shown that genetic variants associated with later AFB predicted higher chances for remaining childless, especially among women (Verweij et al., Reference Verweij, Mills, Stulp, Nolte, Barban, Tropf, Carrell, Aston, Zondervan, Rahmioglu, Dalgaard, Skaarup, Hayes, Dunaif, Guo and Snieder2019).

Apparent absence of response to selection for heritable life-history traits demonstrates that cultural, economic and social influences can override natural selection (Courtiol et al., Reference Courtiol, Tropf and Mills2016; Tropf, Stulp, et al., Reference Tropf, Stulp, Barban, Visscher, Yang, Snieder and Mills2015). For example, although natural selection strongly favours early reproduction in women, Europe and the United States saw a marked delay in the age of first birth during the latter half of the twentieth century. Likewise, the widespread expansion of female education in developed countries continues, despite natural selection favouring women with lower levels of schooling (Courtiol et al., Reference Courtiol, Tropf and Mills2016; Tropf, Stulp, et al., Reference Tropf, Stulp, Barban, Visscher, Yang, Snieder and Mills2015). An example of how phenotypic selection can obscure or override genotypic selection is provided by the study of Kong et al. (Reference Kong, Frigge, Thorleifsson, Stefansson, Young, Zink, Jonsdottir, Okbay, Sulem, Masson, Gudbjartsson, Helgason, Bjornsdottir, Thorsteinsdottir and Stefansson2017). Among Icelandic men, genetic variants associated with higher educational attainment were selected against, even though higher educational attainment itself was not associated with reduced fertility.

This study provides further evidence that the course of human evolution in modern societies is shaped by conflicting and inconsistent selective pressures. Our findings and those from pre-industrial Finland (Bolund et al., Reference Bolund, Bouwhuis, Pettay and Lummaa2013) indicate that at the phenotypic level, moderately heritable reproductive traits can remain resistant to the forces of natural selection over generations. These findings indicate that, although natural selection can be clearly detected, predictions regarding its demographic consequences and phenotypic responses in human populations remain speculative.

On the other hand, evolutionary changes can sometimes occur rapidly, as demonstrated by a study from Iceland, showing that selection against genetic variants associated with educational attainment can lead to measurable change in the genetic composition of a population in a few generations (Kong et al., Reference Kong, Frigge, Thorleifsson, Stefansson, Young, Zink, Jonsdottir, Okbay, Sulem, Masson, Gudbjartsson, Helgason, Bjornsdottir, Thorsteinsdottir and Stefansson2017). The same study also revealed that these genetic variants directly affected reproductive success, independent of the actual level of education attained. A response to selection can just remain unnoticed at the phenotypic level even if the genomic response has occurred.

Given that the strength of selection varies across sexes and educational groups, it has also been proposed that such dynamics may contribute to increasing inequality in polygenic scores (Hugh-Jones & Abdellaoui, Reference Hugh-Jones and Abdellaoui2022). As those authors emphasize, patterns of natural selection are not the outcome of a single, society-wide phenomenon. Instead, they result from opposing forces, operating in different parts of society and pulling in different directions. This highlights the need for future research to compare selection processes across various societal subgroups – ideally incorporating both phenotypic and genetic perspectives simultaneously, and accounting for the potential role of sexual antagonism.