Environment can be understood as a complex organisation of organic and inorganic elements including air, soil, water and living species, sometimes conceived as the “natural world.” As an object of study, social sciences have largely focused on attitudes towards the environment (Dunlap & Jones, 2002), which equate to environmental concern, a concept typically employed in the empirical literature. Environmental concern can be defined as “the degree to which people are aware of problems regarding the environment and supports efforts to solve them and/or indicate a willingness to contribute personally to their solution” (Dunlap & Jones, 2002, p. 485). Environmental concern has been an important issue for modern societies since the last quarter of the twentieth century, due (at least in part) to the rise of environmental problems (Jones & Dunlap, 1992), including (but not limited to) ozone layer depletion, biodiversity loss, pollution, waste disposal, deforestation, and global warming.

Research on individual differences in environmental concern has largely focused on factors related to pro-environmental values and behaviours, including socio-demographic factors such as age (e.g., Buttel, 1979), level of education (e.g., Weaver, 2002), political orientation (e.g., Dunlap et al., 2001), and income (e.g., Franzen & Meyer, 2010). One of the most widely studied sociodemographic variables in relation to environmental concern is gender (e.g., Zelezny et al., 2000), with research consistently indicating that women are more pro-environmental than men. This phenomenon is known in the literature as a “gender gap” in environmental attitudes and has been widely studied, but unlike age, level of education, political orientation or income, fully conclusive results have not yet been produced (Van Liere & Dunlap, 1980; Weaver, 2002). The dangers posed by the current global ecological crisis can only be fought by a rise in global environmental concern and the enactment of environmental policies, which in turn must be accepted and backed by the population. To design better policies and help increase environmental concern, public institutions need a better understanding of the determinants of environmental concern, and gender is one of the most important factors when it comes to explaining differences in the social arena.

This paper does not focus on differences per se between men and women regarding environmental concern, but on determining under what circumstances that gender gap flourishes. To fully consider the importance of context, this analysis employs hierarchical models as a research technique given that this statistical tool allows us to account for individual and contextual factors simultaneously (de Leeuw & Meijer, 2008). The gender literature on differences between women and men in environmental concern distinguishes between evolutionary approaches where differences between women and men are due to innate disparities (Mies & Shiva, 1993; Schmitt et al., 2008), cultural perspectives that focus on ideology and socialisation as primary forces to explain the gender gap (Blocker & Eckberg, 1997; Hunter et al., 2004), and biosocial explanations taking into account material bases of inequality and the intersection of gender, ethnicity, and class relations in a complex array of sources of domination (Connolly et al., 2020; Wharton, 1991). The current research seeks to shed new light on the study of gendered contexts by analysing the factors explaining differences in attitudes between men and women in environmental concern under these competing frameworks by means of a cross-national study of forty-five countries employing data from the 6th wave of the World Values Survey (WVS; Inglehart et al., 2014) covering the period 2010–2014.

Theories of Gender and Environmental Concern

Socialisation Theory

One of the most widely employed theories to study the gender gap in environmental concern is socialisation theory (Blocker & Eckberg, 1997; Hunter et al., 2004), which is better understood as a body of thought and not a specific, clearly defined theory. Socialisation theory states that men and women are socially constructed through very distinct processes of socialisation that have a profound impact on their pro-environmental attitudes. In this regard, two broad types of gender socialisation are typically recognized: gender traditionalism and gender egalitarianism (Baber & Tucker, 2006; Blakemore & Hill, 2008). Traditional gender socialisation shapes a gendered division of social spaces, spheres, and roles, and enhances sex-typed behaviours and attitudes. Women are oriented toward the domestic sphere and more nurturing, compassionate, and cooperative attributes and roles as caregivers and nurturers, whereas men are oriented toward the public sphere and more competitive and controlling attitudes associated with economic success (Gilligan, 1982; McCright, 2010). Gender egalitarianism, on the other hand, can be defined as “a belief that men and women attain a certain degree of equality within both public and private realms of society, and that women’s status should not depend on their reproductive behaviour” (McDaniel, 2008, p.59), with “less stereotyped interests and more flexible attitudes” about roles in the workplace and in the family (Blakemore & Hill, 2008, p.193).

Socialisation theory argues that traditional gender socialisation encourages women’s pro-environmental attitudes. The theory states that, at a time of global ecological crisis, women who experience traditional gender socialisation often apply those caring and nurturing attitudes to nature, as they feel that nature needs to be protected as well (Strapko et al., 2016). Socialisation theory is linked to an essentialist position, which explains the gender gap in environmental attitudes as the consequence of a special connection of women with nature due to biological traits (Mies & Shiva, 1993) or certain “feminine values” related to the traditional gender-based socialisation (see Stoddart & Tindall, 2011 for a discussion about “ecomaternalism”). In both cases, gender traditionalism, not egalitarianism, acts as a trigger for environmental concern in women.

In the current study, the effect of socialisation on environmental concern was examined as a possible cause for the gender-gap in environmental attitudes. I employ an indirect approach using a proxy variable to test socialisation theory given that the circumstances and processes of the two systems of socialisation (traditional and egalitarian) cannot be directly observed. It is inferred that, on average, macrolevel structures related to gender inequality both promote and reflect gender role socialisation.

Macrolevel gender equality might also shape public opinion by exposing more citizens to women in less traditional roles. Moreover, macrolevel gender equality makes women less dependent on men, hence more capable of pushing for egalitarian gender norms and institutions. Societies with higher levels of gender equality are therefore argued to facilitate feminist attitudes more (Yu & Lee, 2013, p. 592).

Thus, it is understood that the likelihood of finding an individual characterised by traditional gender socialisation is higher in countries with higher levels of macro-level gender inequality. This does not mean that an individual belonging to a strong patriarchal country will necessarily hold non-egalitarian views (that would be an ecological fallacy); I argue instead that the number of individuals socialised with less gender-egalitarian attitudes is higher in countries casting higher scores of macro-level gender inequality. Of course, a person might develop egalitarian gender attitudes even after undergoing traditional gender socialisation: Suffragists were born in an age when macro level gender inequality was much more widespread and yet they managed to develop egalitarian gender attitudes. Also important, not all individuals living in low gender-egalitarian countries will have experienced traditional gender socialisation, but in each country that shares non-egalitarian values, the average population is likely to have incorporated those values into their lives through socialisation. Overall, a macro level indicator of non-egalitarian attitudes was considered a good proxy for gender socialisation, though caution is warranted when drawing from conclusions from findings based on these indicators.

Economic Salience

According to “economic salience theory,” the explanatory mechanism for the gender gap is women’s and men’s positions in the marketplace: whereas women are more likely to work outside the formal economic market, as homemakers or caregivers, men are more associated with the labour market. Participation in the formal labour force enhances a “marketplace mentality” because it socialises individuals on the importance of employment and the “necessary” part of nature exploitation in economic dynamics. “Marketplace mentality” is supposed to be more common among men and corresponds to more “un-ecological attitudes” (Merchant, 1980) favouring economic exploitation of natural resources and prioritising economic growth over the well-being of nature. The “motherhood mentality,” by contrast, is understood to be more frequent in women and fosters values related to care and nurturing that ultimately correspond with more pro-environmental attitudes.

Working as a homemaker is more likely to involve engaging in practices of environmental preservation within the home, such as recycling or worrying about the quality of primary goods (Hunter et al., 2004). Furthermore, homemakers play an important role in the care of family pets at home, a task that puts them in direct and daily contact with animals. Some studies point out that caring for pets raises awareness of animal welfare (Kendall et al., 2006). Grier shows how, from the mid-nineteenth century onwards, pet ownership “was transformed into a morally purposive act” (1999, p. 96) which was supposed to raise moral values (especially in children).

In many countries, homemaking is a social role more often performed by women (Baxter, 1997), and although men have increased their share of housework since the 1960s, most of the workload is carried out by women in most countries (England, 2010). Some scholars propose that the underrepresentation of women in the formal labour force (in comparison with men) is a consequence of cooperation between patriarchy and capitalism working as a single institution of control, generating economic positions and productive roles, and determining whether women or men perform them (Wharton, 1991). In any case, the theory states that the experience of different forms of economic participation shapes the degree of environmental concern. In the present study, I examine economic salience theory by testing the effect of homemaking on environmental concern. If this theory is correct, the social role of homemaking (regardless of the sex of the homemaker) should be linked to higher scores on environmental concern. In the current study, a wide range of countries with representative samples and control for contextual effects is used.

Parenthood Status Theory

Parenthood status theory, also known as the “parental roles hypothesis” (Davidson & Freudenburg, 1996), or “eco-mom hypothesis” (Hutner, 2011), argues that women become more pro-environmental due to their experience as mothers. According to the theory, the “parenthood effect” has a different outcome for women and men bearing on their sex-typed roles in families. When they have children, women develop more concern for health and safety that fosters pro-environmental attitudes, whereas men develop or strengthen a “breadwinner role” that runs in the different direction, focusing primarily on securing the material living conditions for the household (McCright, 2010). Nevertheless, it is not always clear whether the parenthood role effect is a consequence of a distinct socialisation process (Blocker & Eckberg, 1997; Strapko et al., 2016), the experience of those roles (McStay & Dunlap, 1983; Blocker & Eckberg, 1997; Shields & Zeng, 2012), or a mix of both (Xiao & McCright, 2012). Thus, for some authors Parenthood Status theory is a particular case of the socialisation hypothesis, whereas for others it is part of the “gendered social roles” argument or structural theories.

The importance of the role of motherhood in society has already been studied for many years in classic sociological writings, particularly by Addams (2002). Many pro-environmental initiatives are carried out by groups or even associations of mothers (Cable, 1992; Hutner, 2011). Other studies have pointed out that mothers show more concern for environmental risk than fathers or women without children (Blocker & Eckberg, 1997). In their extensive review of the literature on gender and perceptions of environmental risk (considering here anthropogenic environmental risks such as nuclear energy and nuclear wastes, toxic contamination, acid rain or threats of genetic alteration), Davidson and Freudenburg (1996) found mixed results regarding parenthood theory, with some studies supporting the theory and others rejecting it. The perception of environmental risk is obviously connected to environmental concern, and therefore, the literature on environmental risk is useful to consider here.

For some scholars, this theory “reduces women and feminine identity to motherhood which then minimizes the other roles and identities women occupy” (Price & Bohon, 2019, p. 424). Of course, women’s increased engagement in the labour market should also have an impact on the “breadwinner role” of men and affect to some extent the validity of the theory. To test the parenthood effect, the present study seeks to ascertain the effect of both motherhood and fatherhood on environmental concern. I build on previous studies by testing an interaction effect between sex and parenthood in a large cross-national sample. In this case, I expect that mothers will report higher levels of environmental concern than fathers.

Class-Gender Effect

The class-gender effect focuses on the position women occupy in the social structure to explain the gender gap in environmental concern (Agarwal, 1992). The assumption of the theory is that women hold less powerful positions in most societies, and that fact has implications for environmental concern because “[I]n an unequal society, the impacts of environmental degradation fall disproportionately on the least powerful. Gendered divisions of labour, land and other resources have meant that women have been uniquely and disproportionately affected by ecological destruction” (Norgaard & York, 2005, p. 507). Thus, according to this theory, women develop pro-environmental attitudes to create safer conditions for themselves and others because they are more often exposed to environmental hazards.

Indeed, many studies show that women report higher risk perception than men. Brody (1984) analysed a representative survey in the US about the perceived risk of nuclear power, finding that women believed nuclear facilities were more dangerous and posed more serious problems to health and life than men did. Melber et al. (1977) examined the results of 103 American representative surveys (some of them representative for the whole country, others for states or municipalities) showing that sex differences were consistent regarding perceptions of risk. Other studies have since then refined their analysis and focus on the specificity of white men and perceived risk. Flynn et al. (1994) asked 1,512 Americans to rate 25 risk items (e.g., cigarette smoking, chemical pollution, nuclear waste) and found that white males have consistently lower risk perception scores than white females, and non-white men and women. The class-gender effect is related to this “white male effect,” which states that white men are less prone to perceive all kinds of risk, including environmental risks. Explanations are manifold, although a broad consensus has been developed about the possible explanation of the “white male effect” as a product of economic and power inequality in society.

The world seems safer and hazardous activities seem more beneficial to white males than to other groups. […] compared with white males, many females and nonwhite males tend to be in positions of less power and control, benefit less from many technologies and institutions, are more vulnerable to discrimination, and therefore see the world as more dangerous (Finucane et al., 2000, p. 170).

Providing some support for this notion, Olofsson and Rashid (2011) found no “white male effect” in Sweden, a country with high “gender-equal” scores in all macro-level indicators of gender equality.

Yet most research points out that well-off individuals are also more likely to be pro-environmental (Jorgenson & Givens, 2014). Middle class individuals with a higher education make up the conventional profile for pro-environmental people (Van Liere & Dunlap, 1980), although they are not as likely as members of the lower classes to experience environmental risks. Furthermore, “economic contingency theory” states that the economically deprived are more likely to put environmental issues to one side when faced with economic crisis (Jones & Dunlap, 1992). For her part, Hochschild (2018) describes how working-class individuals can dismiss environmental issues and embrace high-polluting activities as a result of a mixture of fatalistic religion, governmental distrust, dislike for taxation, and in-group loyalty. The postmaterialist thesis of Ronald Inglehart (1990) states that postmaterialist values (such as pro-environmental values) are more likely to be found amongst the middle and upper classes because they have already fulfilled their basic needs. Inglehart claims that lower classes are more likely to hold materialist values, linked to the basic needs they have not yet fulfilled.

To test the class-gender effect, I focus on the position of individuals in the social structure to see whether the class-gender effect is real or not. Well-off individuals should obtain lower scores for environmental concern, according to this theory.

Ecofeminism

Ecofeminism refers to an array of theoretical approaches devoted to the study of the link between feminism and environmentalism (for review, see Agarwal, 1992). In the current study I take a constructivist approach to understanding the gender gap in environmental concern (Mellor, 1996), which has become prevalent in recent decades (Buckingham, 2004). Constructivist ecofeminism argues that both the domination of women by men and environmental degradation are part of the same process that simultaneously controls and exploits women and nature (Merchant, 1980; Mies & Shiva, 1993). For its part, feminism identifies and addresses false dualisms such as reason/nature, reason/emotion, or human/nature and thus supports the existence of a dual subjugation (Peek et al., 1996). If the struggle for gender equality is viewed similarly to the struggle for the preservation of nature, it follows that gender egalitarianism should be positively related with environmental concern (Mellor, 1996), for example, individuals holding more gender egalitarian attitudes should cast higher levels of environmental concern. This does not mean that individuals must necessarily consider themselves “feminists” to be aware of the existence of that dual subjugation.

Insights from the ecofeminism literature have very often been neglected in the sociological literature on environmental concern (a singular exception is Norgaard & York, 2005), sometimes due to its lack of an empirical approach (see Stoddart & Tindall, 2011; Smith, 2001). Still, the ecofeminist approach was employed by Somma and Tolleson-Rinehart (1997) and Smith (2001) in their analyses of the gender gap on environmental concern. Somma and Tolleson-Rinehart (1997) found that feminism, not sex, was the real predictor for environmental concern in their American sample. Smith (2001) also showed that sex was not directly related to environmental concern, while gender egalitarian attitudes were related to environmental concern.

The Current Study

The aim of the present study is to examine five relevant competing theories regarding the gender gap in environmental concern and test the following hypotheses, which are tested in different models of hierarchical logistic regressions:

  • Hypothesis 1a (H1a): This hypothesis analyses the main effects of sex and the macro level gender inequality on environmental concern. Following socialisation theory, I expect that women respondents and countries with more traditional gender attitudes would endorse higher levels of environmental concern.

  • Hypothesis 1b (H1b): To test the socialisation theory explanation, I evaluate the possible existence of a combined effect of macro level gender inequality and sex. H1b expects that women will endorse higher environmental concern than men in less egalitarian countries because they probably have experienced a more traditional gender role education in those countries.

  • Hypothesis 2 (H2): In order to test the economic salience theory, I predict that homemakers will endorse higher environmental concern than their counterparts.

  • Hypothesis 2 (H2): In order to test the economic salience theory, I predict that homemakers will endorse higher environmental concern than their counterparts.

  • Hypothesis 3a (H3a): Following parenthood status theory I predict that parents will endorse higher scores of environmental concern than their non-parent counterparts.

  • Hypothesis 3b (H3b): Further, following parenthood status theory, I predict that mothers will endorse higher environmental concern than fathers.

  • Hypothesis 4 (H4): Consistent with the class-gender effect, I predict that individuals who identify as lower social class will endorse higher levels of environmental concern.

  • Hypothesis 5 (H5): Consistent with the ecofeminist perspective, I predict that respondents with stronger gender-egalitarian attitudes will endorse more environmental concern than respondents with weaker gender-egalitarian attitudes.

Method

Dataset and Sample

Hierarachical techniques often employ individual and aggregate data in their models. The individual-level data employed in this study have been drawn from the 6th wave of the World Value Survey (WVS; Inglehart et al., 2014). Although the 7th wave was already available, it did not measure environmental concern and thus we used the 6th wave. The dataset is available in the WVS repository (https://www.worldvaluessurvey.org/wvs.jsp). Our model employs data from 45 countries and 51,763 individual respondents all from representative samples. A total of 25,466 (49.18%) of the respondents were men and 26,297 (50.82%) were women. Regarding level of education, 11,546 (22.31%) respondents had a primary level of education, 26.706 (51.59%) had a secondary level of education, and 13,511 (26.10%) had a higher education.

Measures

Environmental Concern

Environmental concern was measured with one item from Schwartz’s Portrait Values Questionnaire that was included in the 6th WVS. This item is the only one of the Schwartz’s Portrait Values that addresses environmental values, and it is commonly used in studies on environmental concern (see, for instance, Echavarren, 2017; Jorgenson & Givens, 2014). The respondents were asked to indicate the degree to which someone who cares about “looking after the environment is important to this person; to care for nature and saving vital resources,” describes them, using the following response options: “very much like me,” “like me,” “a little like me,” “somewhat like me,” “not like me,” or “not at all like me.” Responses were recoded as a binary outcome with the two highest categories (“very much like me” and “like me”) coded as 1 and the remaining 4 categories coded as 0, accounting for 55% and 45% of the sample respectively.

The decision to dichotomise our dependent variable simplifies the analysis and facilitates the presentation of results. Since the dependent variable in the measurement is binary, a hierarchical logistic regression model is used. Results are basically the same in all models whether a dichotomised dependent variable (using a binary hierarchical logistic regression) was employed, or the full original variable (using then a multinomial multilevel logistic regression), but the gain in the legibility of results is considerable.

Socialization Theory Indicator: Macro Level Gender Inequality

To assess macro level gender inequality by country, the country-level variable Gender Inequality Index (GII) provided by the United Nations (UNDP, 2020) was employed. This index is made up of three dimensions: Reproductive Health (measured by the maternal mortality rate and adolescent birth rates), Empowerment (measured by the proportion of parliamentary seats occupied by women and the proportion of adult females with at least a secondary level of education), and the Labour Market (measured by the participation rate of women in the labour force; UNDP, 2020). GII is a metric variable accounting for gender gaps in major areas of development in each country. It ranges from 0 (women and men faring equally) to 1 (women fare as badly as possible on all measured indicators). GII is a highly reputed index of gender inequality with objective indicators of gender inequality for most countries in the world and used in similar studies of environmental concern to explore cross-national gender differences (Chan et al., 2019).

Economic Salience Theory Indicator: Homemaker Status

I employed the variable homemaker as the indicator for economic salience theory, which was made dichotomous (1 = homemaker and 0 = not homemaker) for ease of interpretation. The original variable was employment status and had eight different categories: full time, part time, self-employed, retired, students, unemployed, other, and homemaker). A total of 14.7% of respondents identified as homemakers and 85.3% identified with one of the other categories. Results were very similar employing the original variable (Please contact first author for these analyses). The variable homeworker includes both women and men, as the number of men working as homemakers has been increasing over the years, and the theory does not make any distinctions between women and men regarding this point.

Parenthood Status Theory Indicator: Have Children Status

I used the variable children as the indicator for parenthood status theory, which assessed whether respondents have children or not. The original variable allowed for designation of up to seven children, and then included a category for “eight or more children.” The mean number of children for the original variable is 2.54. This variable was also made dichotomous (1 = have children and 0 = have no children). Respondents with at least one child comprised 70% of the sample, and respondents without children comprised 30% of the sample.

Class-Gender Effect Indicator: Subjective Social Class

There was no measure of objective income in the dataset, therefore subjective social class was used as the indicator for the Class-Gender Effect. The original survey item asked respondents to place themselves along a five-category scale (“upper class,” “upper middle class,” “lower middle class,” and “working class”). This variable was collapsed into three categories: “upper class” (encompassing the former “upper class” and “upper middle class” categories), “middle class” (former “lower middle class”), and “working class” (made up of the former “working class” and “lower class” categories). Working class accounted for 39.6% of the sample, middle class accounted for 36%, and upper class accounted for 24.4% of the sample. I created two dummy variables (“middle class” and “upper class”) and left out “lower class” as a reference category. Again, results are similar when the original variable was employed (please contact first author for these analyses).

Ecofeminist Theory Indicator: Gender Egalitarian Attitudes

Gender egalitarian attitudes was measured with the gender equity index provided by the WVS as the indicator for testing ecofeminist theory (see Welzel, 2013). The gender equity index is comprised of three items that are initially measured on a 4-point Likert-type scale ranging from 1 (strongly disagree) to 4 (strongly agree): “On the whole, men make better political leaders than women do”; “When jobs are scarce, men should have more right to a job than women”; and “A university education is more important for a boy than a girl.” Responses are converted to a scale ranging from 0 to 1, and these scores are already calculated and available in the WVS for the gender equity index variable. The development of this index implies a normalisation of those variables and a two-step process which is fully described in Welzel (2013). Cronbach’s α was .65 for this measure. Though the traditional threshold for internal reliability employing Cronbach’s α is 0.70, Welzel (2013) provides a detailed discussion of the reliability and validity of this index. In this metric variable, higher scores equate to more egalitarian attitudes.

Respondent Sex

Sex of the interviewees was not asked in the survey but was collected through observation by the interviewers. A total of 49.2% (n = 26,296) of the interviewees were coded as “female” and 50.8% (n = 25,467) were coded as “male.”

Control Variables

Four variables were assessed to include as controls in the analyses: education level, age, political ideology, and ecosystem vitality. Education level was reported as primary education, secondary education, or higher education in the survey. Two dummy variables were created for secondary education and higher education, leaving primary education as a reference category. Age was reported and measured in years. Political ideology was measured on a 10-point scale, ranging from 1 = far left to 10 = far right.

Ecosystem vitality was used as an indicator of country-level environmental quality and represents one of the two major dimensions of the Environmental Performance Index developed by the Yale Center for Environmental Law and Policy measuring ecosystem health and natural resource management (Emerson et al., 2010). Environmental vitality is an index that measures objective indicators of environmental quality taking into account seven dimensions: air pollution (measured by sulphur dioxide emissions per populated land area, nitrogen oxides emissions per populated land area, non-,methane volatile organic compound emissions per populated land area, and ozone concentrations data), water (which includes an index of water quality likewise composed by parameters of dissolved oxygen, electrical conductivity, total phosphorus, and total nitrogen, as well as indices for water stress –the percentage of a territory affected by oversubscription of water resources and water scarcity, the fraction of water overuse), biodiversity & habitat (composed by biome protection –the weighted average of the percentage of biome area that is under protected areas, marine protected áreas, and critical habitat protection, accounting for areas of endangered species catalogued by the Alliance for Zero Extinction), forestry (growing stock change, a volumetric measure of cubic meters of Wood, and annual change in forest cover), fisheries (marine trophic level, trawling intensity), agriculture (agricultural water intensity, agricultural subsidies, and pesticide regulation), and climate change (greenhouse gas emissions per capita, CO2 emissions per electricity generation, and industrial greenhouse gas emissions intensity).

Ecosystem vitality is a continuous country-level variable where higher values equate to good environmental quality and has been employed in other studies on environmental concern (Gallego-Álvarez et al., 2018). If the environmental concern of some respondents is affected by the state of their natural environment, the effect can be better isolated by controlling for this variable. GDP per capita (at purchasing power parity) was taken from the World Bank database.

Analytic Strategy

To test the six hypotheses, I conducted eight logistic hierarchical regressions where the dependent variable is environmental concern in all models. Random intercept models were used in all cases except for models 2, 5 and 8, where random slope models were employed to handle cross-level interaction effects in the equation. Regression coefficients (intercepts or covariate effects) are called “random” when they are allowed to vary randomly across higher-level units (countries in this case). In random intercepts models, only the means of environmental concern (our dependent variable) are allowed to vary across countries. Random slope models relax the assumption that the effect of the variable made random is equal across countries. In our cross-level interaction models, the individual-level variable should be random because we aim to account for differences in countries on that variable and employing only intercept random models in a cross-level interaction affects statistical inference (Heisig & Schaeffer, 2019).

Results

Table 1 provides descriptive information for the main study variables for the total sample and separately by gender. The regression models were based on 51,763 respondents, which includes the total amount of respondents for whom there is complete data for the main study variables and belong to the countries for which there is data on the contextual variables.

Table 1 Descriptive Statistics of Variables
Full size table

I first examined whether there was a difference between men and women on environmental concern. An independent samples t-test confirmed a small but significant difference, whereby women (M = 0.56, SD = 0.49) endorsed more environmental concern than men (M = 0.54, SD = 0.49), t(51,761) = -4.85, p < .001. Table 2 presents the zero-order correlations among all study variables. GII and GDP were significantly and strongly correlated (r = .79), and therefore, GDP and GII were included simultaneously in the full model to control for their shared variance.

Table 2 Correlation Matrix
Full size table

The first stage in hierarchical regression analysis is to run a null model without independent variables. The null model showed statistically significant differences between countries regarding the dependent variable (thus justifying the use of hierarchical analysis) and an intraclass correlation coefficient of 10%, meaning that 10% of all differences in environmental concern in our sample can be attributed to between-country differences.

Table 3 presents the results of all models including odds ratios instead of coefficients. Odds ratios are a measure of the association between the categories of a given independent variable in relation to a dependent variable. Figures higher than one indicate a positive correlation between the independent variable and environmental concern, whereas numbers smaller than one indicate a negative correlation. In all models, the variables have been entered simultaneously in the regression.

Table 3 Summary of Odds Ratio in Mixed Effects Logistic Regression in 45 Countries (N = 51,763) for Determinants of Environmental Concern
Full size table

Tests of Socialisation Theory

Model 1 tested Hypothesis 1a which was based on socialisation theory. According to socialisation theory, women should be more pro-environmental in traditionalist countries where the assumed pro-environmental effects of traditional socialisation for women are more widespread and more effectively transmitted. This first model examined the main effects of the two key indicators to evaluate socialisation theory: respondent sex and gender inequality index. Both variables behaved as predicted by the hypothesis: Women endorsed higher environmental concern than men and countries with greater gender inequality also reflected higher environmental concern, supporting H1a.

I tested the critical prediction for the interaction between macro level gender inequality and sex. H1b predicted a positive coefficient (an odds ratio greater than 1) for the interaction between the gender inequality index and respondent sex. Sociodemographic control variables behaved as expected (although political ideology was non-significant). The interaction effect was significant, but the odds ratio was smaller than 1, therefore H1b was not supported. Overall, these findings indicate that women are more likely to be pro-environmental when they live in countries with lower rates of inequality. Model 1 refutes the socialisation theory and cast the results in the opposite direction.

Model 3 tested Hypothesis 2 which was based on economic salience theory. The theory claims that individuals who work in the formal economic market are more likely to develop a “breadwinner role” that fosters weaker environmental concerns. H2 predicted that being a homemaker (and thus outside the formal economic market) would be positively associated with environmental concern. I included the variable homemaker to compare individuals who are responsible for managing the home to individuals who do not carry that responsibility. Being a homemaker did yield statistically significant differences in Model 3, but in the opposite direction than expected and only for a confidence level of 95%, whereas it is not significant in the full model. Model 2 refutes Hypothesis 2 and suggests that economic salience theory does not adequately account for the gender gap in environmental concern.

Hypothesis 3a expected that individuals who had children would cast higher values of environmental concern than their counterparts. Table 3 shows that children is not significant, indicating no differences between parents and respondents without children regarding environmental concern. Thus, H3a was rejected. Hypothesis 3b expected the interaction effect between children and respondent sex to be positive. The interaction effect between children and respondent sex was not significant in the model, indicating that parenthood did not produce differences between women and men on environmental concern. Thus, neither of the hypotheses were supported, indicating that parenthood status fails to explain the gender gap in environmental concern.

Model 6 tested Hypothesis 4 which was based on the class-gender effect. Class-gender effect argues that individuals at the bottom of the social ladder shoulder the effects of environmental problems to a greater extent and, therefore, they should develop stronger pro-environmental concern. H4 anticipates that individuals belonging to upper and middle class would show less environmental concern than working class respondents (i.e., H4 expects an odds ratio lower than 1 for Subjective social class). Table 3 revealed statistically significant differences regarding the social class variables, but the odds ratios of upper and middle class were higher than 1, meaning that higher social status respondents reported higher environmental concern than individuals at the bottom of the social ladder. As the findings were in the opposite direction from expected, H4 was also rejected.

Hypothesis 5 was based on ecofeminist theory, which states that gender-egalitarian attitudes raise environmental concern as both women and nature are dominated by a sort of dual subjugation from patriarchy and capitalism. Therefore, H5 predicts a positive relationship (odds ratio above 1) between the gender equity index and environmental concern. Model 5 found that gender equity was highly significant and positively correlated to environmental concern, supporting H5.

The last model presented in Table 3 is the full model, which includes all the variables included in all analyses. The pertinence of a full model is to evaluate the stability of previous models and confirm whether the pattern of results changes significantly once all variables are included. In this case, most results retain not only their significance and orientation, but even the size of the odds ratio. I can conclude that the pattern of findings reported in Table 3 remains stable no matter the combination of variables used in the study. Among the control variables, educational level remains significant in this model, as well as age. Regarding the variables related to the hypotheses of the study, the homemaker variable ceases to be significant in the full model, and subjective social class remains significant but in the opposite direction to what the initial hypothesis indicated. The same occurs with the interaction factor between respondent sex and GII, which remains significant in the full model. Despite all the variables included, gender equality remains highly significant in the full model: Favourable attitudes towards gender equality serve to predict environmental awareness.

Discussion

The current study examined five of the most important theoretical approaches to explain the gender gap in environmental concern using data from 45 countries. Each theory proposes a factor that might explain differences between women and men regarding environmental concern: socialisation (socialisation theory), economic roles (economic salience theory), parental roles (parenthood status theory), social position (class-gender effect) and gender egalitarian attitudes (ecofeminist theory). Overall, the findings did not provide support for most of the theories, including socialisation, economic salience, parental status, and class-gender effect, whereas support was found for ecofeminism. I consider the pattern of findings below.

The results from Model 1 (testing socialisation theory), though contrary to what socialisation theory would predict, are consistent with findings presented by Strapko et al. (2016) and Chan et al. (2019), who also noted that traditional gender socialisation was negatively correlated with environmental concern. Of course, the rejection of socialisation theory does not mean that socialisation processes play no role in pro-environmental concern.

The hypothesis regarding economic salience expected homemakers to have higher levels of environmental concern than their counterparts. However, the results are the opposite: individuals who do not work as homemakers, on average, have higher scores of environmental awareness. The fact that homemakers are less environmentally concerned than other professions probably means that discourses and ecological worldviews are more closely related to the realms of the public sphere rather than the private sphere (Liobikienė & Poskus, 2019). In any case, even if the homemaker variable had yielded significant results in line with the hypothesis, the theory would still be a poor approximation to the gender gap in environmental concern, given that the number of women homemakers is already a minority, as seen in Table 1. Homemaker is no longer a category that unequivocally encompasses women, and this affects to the ability of the economic salience theory to explains the gender-gap in environmental concern (Price & Bohom, 2019).

The present study examined two distinct hypotheses within the general theory of parenthood status. On one hand, whether parents have higher environmental concern than non-parents, and on the other hand, whether mothers develop higher environmental awareness than fathers when having children. In both cases, the results show that there are no differences regarding environmental concern between respondents with or without children, or between fathers and mothers. The theory must be rejected as a useful tool for explaining the gender gap in environmental concern. Although it does play an important role in the field of perception of environmental risks (Davidson & Freudenburg, 1996).

The theory of the class-gender effect explains the differences between men and women regarding environmental concern due to the lower average social positions occupied by women on the social scale and the greater environmental vulnerability of these positions. Nevertheless, Table 1 showed that there were not as many differences in the subjective position on the social scale between men and women in the representative samples of our study. On the other hand, Table 3 showed that lower classes had less, not more, environmental concern than higher classes, contrary to what the theory explains. Nonetheless, in many countries, environmental concern can be related to a concern for quality of life (Inglehart, 1990) or even be a lifestyle or habitus linked to higher social classes (Kennedy & Givens, 2019).

Ecofeminism predicts that individuals with egalitarian gender attitudes hold higher levels of environmental concern., and this was supported in the current study. Attitudes that reflect endorsement of gender equity were significantly linked to environmental concern, and this pattern held when all other independent and control variables were included in the model. This is the only theory included in the analysis that was supported by the data. The variable of gender equality has been rarely included in quantitative analyses of environmental values, and based on these results, it is highly useful in explaining individuals' environmental concern as modern societies are moving towards scenarios where gender egalitarian attitudes are increasingly important both quantitatively and qualitatively (Inglehart & Norris, 2003). It is interesting to note that respondent sex remains significant in the full model as well as in the model testing Hypothesis 5 on ecofeminism. The introduction of the variable of favourable attitudes towards gender equality does not exhaust the differences between men and women regarding environmental concern. If respondent sex had ceased to be significant in this model, it is possible that the introduction of gender egalitarian attitudes (which are correlated with respondent sex as women are more likely to endorse such attitudes) was driving the apparent correlation between respondent sex and environmental concern. However, that is not the case. When considering the contextual variables that seek to explain environmental concern in our models, the strength and sign remain stable across all models, including the full model. The interaction effect between sex and GII shows that the positive effect on environmental concern of being female is stronger in more egalitarian societies. The most evident implication is that the gender gap in environmental concern remains still to be fully understood; however, attitudes around gender egalitarianism may play a critical role here.

Limitations and Directions for Future Research

All data come from representative samples of each country, although it should be noted that by no means does the WVS provide a representative picture of the world. The proportion of Western countries in the sample or the number of developing countries might deliver a skewed portrait of the issue. It is also true however that our results are in line with previous studies working with different datasets. In any case, further research must be conducted (whether national or cross-national) to support (or refute) our findings.

Employing a large cross-national database offers an advantage in terms of generalisation but also requires working with a pre-set list of variables. For example, a much better indicator for subjective social class would have been income or profession. Thus, results regarding the class-gender effect must be interpreted carefully (although they are in line with most studies). Another major limitation is the use of a proxy for studying socialisation theory. Socialisation theory is very elusive and difficult to verify directly. To do so, an investigation should measure not just pro-environmental values but also the very process of socialisation in men and women regarding the environment in a longitudinal panel to measure its impact on pro-environmental values and attitudes in adulthood. In future research, it would be useful to develop questions in cross-national surveys about gender socialisation and gender roles. Questions about gender socialisation should investigate knowledge, beliefs, and expectations about social roles. Questions about socialising practices could be related to games and toys, clothing, room decorations, or expressions of anger and mild aggression in everyday life. In both cases, insights from psychological scales of gender attitudes would be very useful.

Despite the potential of feminism or gender egalitarian attitudes as explanatory variables for environmental concern, only a few studies include them in their analysis, and this should change in future works. Future research should also focus on the effect of variables related to gender egalitarianism on pro-environmental behaviour as well, where the effect is probably larger than on environmental concern.

Practice Implications

There are several policy implications in this study. Our results point to the need and convenience of developing joint policies on the environment and gender equality, which could be mutually reinforcing. Gender equality attitudes should be incorporated into the idea of ecological citizenship (Dobson, 2003) to help develop and encourage private pro-environmental behaviours. Contrary to socialisation theory, where environmental concern pivots on traditionalism, our findings link environmental concern among women to gender egalitarian attitudes. Courses on environmental education should therefore pay special attention to feminist issues in their syllabus.

Economic salience theory and the class-gender effect seem to be no longer useful to study differences between men and women regarding environmental concern. Parenthood status and socialisation theory may well be relevant still, especially if better information about socialisation practices at the family level in cross-national surveys can be obtained. This research shows that gender-egalitarian attitudes and environmental attitudes are correlated to some extent, and this fact should influence environmental research. Ecofeminist tradition should be incorporated into quantitative studies of environmental concern, and scales of gender-egalitarian attitudes (such as the gender equity index) should be included in questionnaires of international databases studying perceptions of the environment.

Conclusion

This study constitutes an important step to understanding the gender gap in environmental concern, as well as bringing ecofeminist thought more squarely into the empirical study of environmental issues with the theoretical insight of sociology and social psychology. By underlining the importance of gender egalitarianism in the study of environmental concern, I follow the earlier recommendations of the founders of environmental sociology (McStay & Dunlap, 1983), and hope this will usher in a new research agenda. Power plays a role here. Ecofeminism points to the potentialities and natural connections between feminism and environmentalism, as both constitute paradigms of thought that challenge current power hierarchies, systems of exploitation, and more importantly, system justification (Goldsmith et al., 2013). Our findings suggest the importance of a more gender egalitarian world for a cleaner future.