An Estimation of the Effect of Women’s Employment on the Prevalence of Intimate Partner Violence in Mexico

Abstract

Female labor force participation is important for women, children, and societies, but also may have unintended impacts including an increased risk of intimate partner violence (IPV). IPV is a global health, human rights, and development problem with far-reaching economic and societal consequences. Mexico has a very high prevalence of IPV: 43.9% of Mexican women have reported experiencing IPV at the hands of their current partner. The literature on women’s economic participation reveals mixed evidence on whether women’s employment is associated with higher levels of IPV or whether it is protective against IPV. As the effect of women’s work operates differently across contexts, we aim to estimate the effect of women’s employment on their risk of experiencing IPV in rural and urban Mexico. Utilizing the nationally representative 2016 Mexican National Survey on the Dynamics of Household Relationships (ENDIREH), we employ propensity score matching (PSM) to address the potential selection bias between women who are employed and/or receiving a cash transfer with women who are not. We additionally implement inverse probability weighted regression adjustment (IPWRA) to explore this relationship and compare the results with the PSM findings. Three different measures of women’s economic participation are analyzed: whether they had engaged in any productive work outside of the home in the past year, whether they received conditional cash transfers through Mexico’s Prospera program, and whether they received Prospera and worked. Given the high levels of IPV in Mexico and the greater levels of economic participation borne of an increased number of women in the workforce, our results have important potential implications for targeting support to survivors of violence who receive cash transfers and undertake employment in both urban and rural areas.

Keywords

Mexico domestic violence women’s employment propensity score matching inverse-probability weighted regression conditional cash transfers

Background

Violence against women (VAW) is a global public health, human rights, and clinical problem with far-reaching consequences (Devries et al., 2013). It cuts across cultures, religious lines, age, socioeconomic status, and education levels and takes many forms including but not limited to intimate partner violence (IPV), sexual violence (SV), trafficking, forced and early marriages, honor killings, female genital mutilation/cutting, and childhood sexual abuse (World Health Organization [WHO], 2013). In Mexico, the problem of IPV is striking: 44% of Mexican women over 15 years of age report having experienced violence in their current or past relationship, and 25.6% report experiencing some type of IPV in the last 12 months (Instituto Nacional de Estadística y Geografía [INEGI], 2016b). Female labor force participation has been lauded as a way to improve the lives of children, empower women, and to support economic growth (United Nations Women, 2017). However, there is mixed evidence on the relationship between women’s economic participation and IPV, with a range of explanations for why employment and the receipt of conditional cash transfers may increase the risk of violence or protect against it (Vyas & Watts, 2009).

In 2017, 43% of Mexican women were in the workforce compared with 77.6% of men (Organisation for Economic Co-operation and Development [OECD], 2019). Although female participation in the labor force has increased over the past two decades (from 38.9% in 2000 to 44.9% in 2017), Mexico has a relatively low female employment rate relative to other OECD countries (OECD, 2017). In Mexico, 44.9% of the workforce consists of women, which is 15.2 percentage points lower than the average of the OECD countries (60.1%), and substantially lower than the averages in Brazil (55.8%), Chile (50.2%), and Colombia (54.1%; OECD, 2017). Although there are fewer women in the workforce relative to other countries, the possible relationship between female labor force participation and IPV remains important given that many women do work. Moreover, there are other ways in which women can improve their economic status; some women are economically supported through Prospera (previously Oportunidades), the Mexican conditional cash transfer (CCT) program (México Gobierno de la República, 2016). This program targets women in households whose per capita income does not cover the basic food basket or whose members are at risk in terms of nutrition, health, and education (Dávila Lárraga, 2016). The short-term goal of Prospera is to alleviate poverty through the redistribution of income, while the medium-term goal is to improve the well-being of the poor population by ensuring access to nutrition and healthcare (Dávila Lárraga, 2016). Once families are accepted into the program, they have certain health and education-related responsibilities they must comply with to receive their benefits, which depend on their demographic characteristics and the benefits they qualify for (Dávila Lárraga, 2016).

This study employs propensity score matching (PSM) and inverse-probability-weighted regression adjustment (IPWRA) to examine whether women who are formally or informally employed and receiving CCTs in rural and urban Mexico are more or less likely to experience violence at the hands of their intimate partners. This research is important because it highlights the potential unintended consequences of women’s economic engagement and provides insight into some of the risk factors associated with experiencing IPV in Mexico, a topic that has not been sufficiently addressed in the literature.

Female Labor Force Participation in Mexico

The International Labor Organization (ILO) reports that despite significant achievements, the gap in the participation rate of women and men in the labor force in Mexico continues to be extremely high: a 34% difference (ILO, 2019). The gap in the participation rate has decreased only 10 percentage points between 1999 and 2018 (The World Bank, 2018). Although the employment rate of Mexican women has grown 10 percentage points in the last three decades, the rate has slowed recently (ILO, 2019). From 1991 to 1999, it increased 4.6 percentage points, 3.4 points from 2000 to 2009, and from 2010 to 2018, only 1.7 points (ILO, 2019). Such slow progress makes it difficult to close the gap between men’s and women’s labor force participation in the short and medium terms. Currently, the ILO reports that a significant majority of Mexican women workers (79%) are concentrated in the tertiary sector of the economy, much of which is informal and working in services (ILO, 2019).

While female labor force participation has risen moderately in Mexico during the past two decades, many women still face major obstacles to participation. These include traditional gender roles toward work and care and the high amount of unpaid work at home; Mexican women spend 4 more hours per day on unpaid work than men (OECD, 2017). Moreover, there are a lack of policies that facilitate women’s employment, especially by reducing the gender burden in terms of childcare and family-friendly workplace practices (Del Pilar Martínez, 2018). That said, as mentioned, the proportion of women in the workforce is increasing and, by 2010, one third (33.3%) of women were working compared with 29.9% in 2000. According to 2010 Census data, the rate of economic participation of women is very closely linked to community size (INEGI, 2010). In urban areas with populations more than 100,000, over 41% of women work while it is 36% in cities with populations between 15,000 and 100,000, and only 29% in towns with between 2,500 and 15,000 habitants. In rural areas—in communities of less than 2,500 inhabitants—the figure is 17% (ILO, 2019). This is likely related to both opportunities available and community size: rural areas do not offer the same number or types of job opportunities.

In addition, informal employment is highly prevalent in Mexico: around one third (29%) of Mexican women with a job work in the informal sector, 11% in the low-paid domestic service sector, and 51% of the informal self-employed workers are women (OECD, 2012). Protections for workers in Mexico generally cover formal employees, thus, leaving informal workers at a disadvantage (OECD, 2012). For example, parental leave and leave for other family reasons is scarce in Mexico, with only formal workers being covered under the law (OECD, 2012). In 2012, Mexico’s Congress approved a set of reforms aimed at increasing labor market flexibility and improving employment conditions for women and young workers. The Labor reform—which introduced 5 days of paid paternity leave—was an important step toward increasing women’s economic status as it encouraged men to participate more in childcare-related tasks (OECD, 2017).

However, the relationship between women’s economic participation and violence is unclear as the existing evidence is mixed. A 2010 systematic review by Vyas and Watts identified five studies that showed a protective association between women’s involvement in income generation and past-year violence experience, meaning that women involved in income generation experience less violence (Vyas & Watts, 2009). Six other studies showed a risk association, meaning that women’s involvement in income generation increased the risk of violence (Vyas & Watts, 2009). Research from Ecuador, Egypt, India, Haiti, South Africa, Bangladesh, and the United States shows that different types of economic participation (i.e., microcredit and employment) decrease women’s risk of violence (Bedi, Chhachhi, & Bhattacharyya, 2009; Capaldi, Knoble, Shortt, & Kim, 2012; Gage, 2005; Hadi, 2005; Hidrobo, Peterman, & Heise, 2015; Kim et al., 2007; Kishor & Johnson, 2004; Schuler, Hashemi, Riley, & Akhter, 1996). In contrast, employment, income, and microfinance have been shown to increase women’s chances of experiencing IPV in Peru, the Dominican Republic, Nicaragua, as well as in some of the aforementioned countries such as Bangladesh, India, and the United States (Capaldi et al., 2012; Dalal, 2011; Flake, 2005; Kishor & Johnson, 2004; Koenig, Ahmed, Hossain, & Mozumder, 2003; Naved & Persson, 2005; Rahman, Hoque, & Makinoda, 2011).

Several theories underpin these opposing findings. Explanations such as social exchange theory and relative resource theory posit that violence may be used as a way for a man to control his partner if he lacks economic resources, or that a woman’s increased economic engagement may challenge a man’s status and incite violence against his female partner or spouse (Vyas & Watts, 2009). In contrast, marital dependency theory posits that women being economically dependent on their partners can increase their risk of IPV, and has been supported by empirical evidence in Mexico (Villarreal, 2007; Vyas & Watts, 2009). Gender resource theory provides a more nuanced perspective: namely, that the risk of violence depends on the partner’s gender ideologies, with more egalitarian views protecting women (Vyas & Watts, 2009). There is evidence to support these conflicting theories, and our research aims to present more robust empirical estimates in a new setting.

The literature on the impact of cash transfers on women’s risk of IPV presents similarly mixed findings (Amnesty International, 2012; Bobonis, Castro, & Morales, 2015; Hagen-Zanker et al., 2017; Hidrobo & Fernald, 2013; Perova, 2010). While cash transfers have been proven to have a positive impact on the well-being and opportunities of women and girls, particularly in education and employment, the picture regarding IPV is not so clear. Research from Peru indicates that women who are CCT recipients experience a reduction in IPV (Perova, 2010). Another study from Ecuador shows that for women with a primary education or less, a cash transfer increases emotional violence when the woman has equal or more education than her partner, whereas women who have more than a primary education experience less psychological violence if receiving a CCT (Hidrobo & Fernald, 2013).

The research on women’s economic status and IPV in Mexico is limited (Bobonis et al., 2015; Maldonado, Nájera, & Segovia, 2005; Terrazas-Carrillo & McWhirter, 2015; Villarreal, 2007). Terrazas-Carrillo and McWhirter (2015) found that age, number of children in the household, income, education, self-esteem, family history of abuse, and “controlling behavior of the husband” were all significant predictors of IPV.¹ They conclude that employment status was only a significant predictor when husband’s controlling behavior was excluded from the model, indicating that spousal controlling behavior may be a mediator between employment status and IPV (Terrazas-Carrillo & McWhirter, 2015). Another study found that employment reduces women’s risk of violence (Villarreal, 2007). However, both studies utilize regression methods, which can be subject to potential selection bias (discussed below). Another qualitative study showed that the monetary benefits that women receive from Oportunidades (the precursor to Prospera) did not seem to cause an increase in violence, largely because the money is intended to help the children (Maldonado et al., 2005). Yet another study conducted on the relationship between male perpetrated IPV and CCTs shows that while short-term estimates of physical and emotional violence rise, long-term estimates of IPV victimization are not different between women receiving CCTs and those who are not (Bobonis et al., 2015). In addition, women who participated in Oportunidades appeared less likely to suffer psychological violence at the hands of their intimate partner (García Aísa, 2014). Overall, the existing evidence in Mexico seems to point to women who are economically engaged experiencing less partner violence than women who are not.

Method and Data Analysis

Propensity Score Matching

We utilize two methods to analyze our data: PSM and IPWRA. PSM estimators have been widely used in evaluation research to estimate average treatment effects in observational settings that are at risk of treatment-selection bias (Abadie & Imbens, 2009; Rosenbaum & Rubin, 1983). In observational studies, a group of participants is exposed to a treatment while others are not. However, randomization does not take place, and the data are observed after participants have been exposed to the treatment. PSM is especially useful for dealing with observational data due to its capacity to reduce bias and confounding when random assignment to treatments is not possible. When matching on propensity score values, it is possible to pair individuals in the data from both the treatment and control group ex post, based on their propensity score or a participant’s probability of belonging to the treatment group given its observational characteristics. The average difference between these matched participants is what is then referred to as the average treatment effect on the treated (ATET; Rosenbaum & Rubin, 1983).

The basic idea of PSM is that a vector of several covariates can be reduced to one dimension and is then given a score by calculating the conditional probability of assignment to a treatment given the values of observed characteristics or covariates and all treatment confounders (Rosenbaum & Rubin, 1983). This score is then used as if it were the only confounding covariate. However, while it is no longer necessary for individuals to match exactly on each covariate, there is a need for an individual to be paired with someone with a similar propensity score. Thus, it is also important to test whether participants with the same propensity score have a similar distribution of observable covariates (or characteristics), independent of their exposure to treatment. This check can be performed by examining the balance in the observable characteristics of participants across treatment and comparison groups (after the technique has been implemented) (Austin & Stuart, 2015). Finally, to draw causal inferences from these data, PSM requires treatment and control groups to overlap in an “area of common support” to pair participants with the same propensity scores (Imbens & Rubin, 2015). Thus, individuals that fall beyond this region must be discarded from the analysis.

Vyas and Heise (2014) argue that PSM is a suitable methodological approach to measuring risk of IPV among formally employed women and nonemployed women due to potential selection bias. We, accordingly, use PSM to account for the potential underlying differences (or selection) between women who work and/or receive CCTs and those who do not (Rosenbaum & Rubin, 1983). In the context of this study, PSM estimators present several advantages over other possible methods. First, instead of matching on several covariates with close or exact values, units with dissimilar covariate values can be matched on equal or similar propensity score values (Abadie & Imbens, 2009; Imbens & Rubin, 2015). A second advantage of the PSM method is that it balances the distribution of the observed baseline covariates between treated and control units with the same propensity score (Abadie & Imbens, 2009; Austin & Stuart, 2015).

Inverse-Probability-Weighted Regression Adjustment

To further adjust for confounding and to test our PSM estimates, we employ the IPWRA method. IPWRA estimators have the advantage of remaining asymptotically unbiased, meaning they are more accurate, even in studies that use observational data like ours (Abadie & Spiess, 2016; Austin & Stuart, 2015; Curtis, Hammill, Eisenstein, Kramer, & Anstrom, 2007). Similar to PSM, IPWRA removes bias and confounding from observational studies (Austin & Stuart, 2015). However, unlike PSM, this is achieved by weighting the outcome measures by the inverse of the participant’s probability of belonging to the treatment group given its observational characteristics, or its propensity score (Abadie & Spiess, 2016). These weights are used for covariate adjustment to estimate corrected regression coefficients, giving us more precise ATET without the need for common support (Abadie & Spiess, 2016).

The IPWRA method offers better estimates of the causal effect by combining regression adjustment (RA) and inverse probability weighting (IPW; Abadie & Spiess, 2016). Accordingly, for IPWRA estimators, two models are used: one to predict treatment status and one to predict outcomes. This means that only one model must be correctly specified for the regression coefficients to provide consistent average treatment effects (StataCorp, 2013). Thus, this procedure has been referred to as “doubly robust” in the sense that if one model is misspecified, the other should still hold (Austin & Stuart, 2015). For the IPWRA estimator, we use a logit model to predict treatment status (worked last year, received a CCT, worked last year and received a CCT) as a function of the model covariates. Once again, as in the case of PSM, the overlap assumption of balance across covariates must hold.

IPWRA is a novel technique that produces more robust estimators and that can be used in conjunction with PSM as in this study. When the PSM and IPWRA results vary greatly, it is likely due to selection bias (Abadie & Imbens, 2009). By using IPWRA as a robustness check, we are able to obtain more reliable estimates of the treatment. To our knowledge, we are the first to combine PSM and IPWRA in IPV research in Mexico to handle selection bias issues caused by observed heterogeneities.

Our review of the literature suggests that studies about the relationship between labor status and IPV in Mexico have rarely implemented quantitative approaches, with the exception of two studies that used traditional regression methods (Terrazas-Carrillo & McWhirter, 2015; Villarreal, 2007). We utilize our findings to present potential areas of improvement to the Prospera program and some policy recommendations and, thus, hope that our research will contribute to the evidence gap in this setting both by using new methodologies and by including a programmatic and policy perspective.

Data Source

This study analyzes data from the 2016 Mexican National Survey on the Dynamics of Household Relationships (Encuesta Nacional sobre la Dinámica de las Relaciones en los Hogares or ENDIREH). The ENDIREH is a nationally representative survey administered by the National Institute of Statistics and Geography (El Instituto Nacional de Estadística y Geografía or INEGI; INEGI, 2016b). This survey is particularly relevant to our study because it collects information on violence experienced by women aged 15 years and older that has occurred in the school, work, community, family, and intimate settings as well as some information on labor force participation (INEGI, 2016c). We believe that the nationally representative nature of the survey helps to account for some of the heterogeneity and diversity among Mexican women, specifically education level, socioeconomic status, ethnicity, age, and geography.

Sample

Women are initially screened and given one of three surveys based on the following relationship categorizations: 1 = Married/Cohabiting; 2 = Separated, Divorced, Widowed; 3 = Single (INEGI, 2016b). The sample for this study was taken from the subset of women who are married or cohabitating (N = 72,855). This follows precedent set by comparable analyses of the IPV module data from the Demographic and Health Surveys (Hindin, Kishor, & Ansara, 2008; Vyas & Heise, 2014). In addition, the sample is limited to women aged 15 to 64, to capture the working age population as defined by the OECD (OECD, 2018). To control for regional economic and social conditions that characterize the Mexican population throughout the national territory, we divided our sample geographically into five socioeconomic regions.² The 2016 survey response rate was high (85.7%), and the subsample size of partnered women who are currently at a working age corresponded to 66,943 respondents (embracing 91.9% of married or cohabitating women who were of a working age; INEGI, 2016b).

The binary variables that indicate whether a woman has experienced IPV were constructed by analyzing the Married/Cohabiting survey questions. Partnered women who reported having experienced a physical IPV incident during the last year once or more than once were coded as having experienced physical IPV.³ Similarly, partnered women who responded once or more than once to any of the five questions measuring sexual IPV over the past year were coded as having experienced sexual IPV.⁴ The dependent variable Intimate Partner Violence was then constructed as a dichotomous variable capturing partnered women that experienced physical IPV, sexual IPV, or both. Although psychological and economic IPV were included in the survey, these dimensions were excluded because there is not wide agreement on how to measure these types of violence (WHO, 2013). Given that this study aims at analyzing the effect of women’s employment and/or economic support through CCTs on the risk of experiencing sexual or physical partner violence, this study considers the following key treatment variables. The first binary variable captures whether or not the respondent reported working outside the home during the past year. To account for another factor that could influence women’s economic status, a second binary variable is included that shows whether or not the respondent is a recipient of Prospera. Finally, a third dichotomous variable shows women that have worked outside the home during the past year and receive Prospera. PSM and IPWRA estimates are derived using Stata’s psmatch2 and teffects routines, respectively.

Data Analysis

As noted above, to measure the effect of women’s economic status on their risk of experiencing IPV, we employ two different estimation methods: (a) PSM, and (b) IPWRA. We ran two different models for each method, one for the subsample of rural women and one for the subsample of urban women. As described, PSM requires taking into account the set of key variables that affect both the outcome (experiencing intimate partner violence) and the treatment variables (worked last year, received a CCT, worked last year and received a CCT). Theory and scholarly literature across different regional settings point to the following socioeconomic factors that have been shown to affect women’s employment status: age, level of education, whether or not the respondent is Indigenous, number of children, socioeconomic region, household socioeconomic status, and partner’s employment status (Kwagala, Wandera, Ndugga, & Kabagenyi, 2013). We also account for women-specific survey weights as these are likely related to the probability of responding to the survey, and they might capture other relevant factors that we are not controlling for (DuGoff, Schuler, & Stuart, 2014). Moreover, because women’s employment dynamics in Mexico have been shown to vary significantly across local labor markets in rural and urban settings, results will be presented independently for women residing in urban and rural areas (urban Mexico, N = 49,127; rural Mexico, N = 17,818). PSM and IPWRA are appropriate methods to account for the differences in the rural and urban labor markets in Mexico, and provide a nuanced perspective on this issue.

Results and Analysis

This section first examines whether our data satisfy the covariate balancing property needed for the PSM and IPWRA estimators to hold. To match on the propensity score, each value of the score should have the same distribution for the treatment and comparison groups (also known as balance). Thus, before estimating the causal effect, we first analyze the distribution in the likelihood of receiving treatment to check for adequate overlap. Figure 1 shows the distribution of the estimated propensity scores by the three different treatment indicators included in the analysis for urban and rural Mexico. Histograms showing observations that are on- and off-support for both the treated and untreated groups confirm that the area of common support is very strong for worked in the last year, in both urban and rural settings, with only a few observations falling off-support. However, as shown in Figure 1, while there appears to be common support for CCT recipients in rural Mexico, there is weak overlap for the CCT treatment in urban Mexico. This also appears to be the case for women residing in urban areas that worked during the past year and received Prospera. Furthermore, there are fewer observations when analyzing those who experience both treatments: women who worked during the past year and are CCT recipients. Thus, results with regard to the effect of the CCT treatment and the combined variable regarding those who worked and received CCTs on partnered women’s risk of IPV in urban Mexico should be interpreted with caution.

Figure 1.

Common Support: Distribution of the estimated propensity scores by treatment indicators.

Table 1 provides a covariate “balance analysis” for unweighted and IPWRA-weighted covariates for each treatment indicator versus the corresponding controls in urban and rural Mexico. As with PSM, it is important to test for covariate balancing to estimate unbiased effect sizes. If the IPWRA model is well specified, then covariate balance will be achieved. As expected, we find substantial differences on many unweighted covariates between treated and nontreated partnered women in the raw data. However, once we use inverse propensity weights to balance the treated and comparison groups, we obtain good balance on all covariates—nearly all normalized differences are 0.4 or less with the exception of the number of children born to women in rural areas who are Prospera recipients.⁵ It is important to note that these statistics have no standard errors and, therefore, inference is normal (Austin & Stuart, 2015).

Table 1.

Covariate Balance Analysis With IPWRA Weights for Partnered Women in Urban and Rural Mexico by Treatment Indicator.

Characteristics	Worked During the Last Year		Received CCT From Prospera		Worked and Received CCT From Prospera
Characteristics	Standardized Differences	Variance Ratio	Standardized Differences	Variance Ratio	Standardized Differences	Variance Ratio
Urban
Central-North	0.06 (0.01)	1.08 (1.01)	−0.17 (0.00)	0.79 (1.01)	−0.09 (0.01)	0.89 (1.01)
Central	0.04 (–0.02)	1.15 (0.92)	−0.17 (0.01)	0.49 (1.08)	−0.16 (0.01)	0.53 (1.03)
Central-South	−0.09 (0.01)	0.88 (1.02)	0.22 (–0.03)	1.30 (0.97)	0.19 (–0.01)	1.26 (0.99)
South-Southeast	−0.17 (0.02)	0.78 (1.03)	0.27 (0.02)	1.41 (1.02)	0.14 (–0.02)	1.21 (0.98)
SES Level	0.41 (–0.04)	1.22 (0.89)	–0.93 (0.03)	0.52 (1.00)	–0.72 (0.05)	0.47 (0.72)
Age	−0.08 (0.00)	0.71 (0.98)	−0.01 (–0.07)	0.65 (1.03)	−0.06 (–0.01)	0.54 (1.02)
Age²	−0.13 (–0.01)	0.72 (1.02)	−0.06 (–0.06)	0.67 (1.05)	−0.13 (–0.01)	0.54 (1.05)
School Level	0.60 (–0.02)	2.13 (0.95)	–0.85 (0.07)	0.22 (0.78)	–0.73 (0.03)	0.24 (0.79)
Number of children	−0.28 (–0.03)	0.66 (1.06)	0.70 (–0.41)	1.37 (0.15)	0.59 (–0.15)	1.04 (0.27)
Indigenous	−0.11 (0.02)	0.88 (1.02)	0.35 (0.02)	1.38 (1.01)	0.27 (–0.03)	1.29 (0.99)
Partner works	0.20 (0.00)	0.64 (1.00)	−0.01 (0.00)	1.02 (1.00)	0.03 (0.00)	0.93 (1.00)
ENDIREH weights	−0.01 (–0.03)	0.92 (0.84)	−0.05 (–0.03)	0.86 (1.02)	−0.03 (0.01)	1.05 (1.21)
Rural
Central-North	0.07 (0.01)	1.09 (1.01)	−0.25 (0.01)	0.71 (1.01)	−0.13 (0.01)	0.84 (1.01)
Central	0.04 (0.01)	1.31 (1.05)	−0.03 (0.00)	0.77 (1.00)	−0.05 (0.00)	0.70 (1.02)
Central-South	−0.04 (0.01)	0.96 (1.01)	0.14 (–0.03)	1.15 (0.98)	0.11 (0.00)	1.12 (1.00)
South-Southeast	−0.17 (0.01)	0.84 (1.01)	0.26 (0.03)	1.27 (1.02)	0.09 (–0.01)	1.10 (0.99)
SES Level	0.32 (–0.03)	0.93 (0.85)	–0.63 (0.01)	1.00 (0.97)	−0.31 (0.02)	0.76 (0.73)
Age	−0.11 (0.00)	0.72 (0.97)	0.07 (–0.07)	0.65 (1.05)	−0.02 (–0.01)	0.55 (1.01)
Age²	−0.15 (0.00)	0.71 (1.01)	0.00 (–0.06)	0.69 (1.07)	−0.09 (–0.01)	0.55 (1.04)
School Level	0.53 (0.00)	2.28 (0.99)	–0.68 (0.07)	0.26 (0.77)	–0.49 (0.02)	0.33 (0.83)
Number of children	−0.26 (–0.03)	0.69 (1.11)	0.59 (–0.43)	1.33 (0.16)	0.40 (–0.08)	0.94 (0.39)
Indigenous	−0.10 (0.02)	0.93 (1.01)	0.30 (0.02)	1.20 (1.01)	0.20 (–0.01)	1.13 (1.00)
Partner works	0.17 (0.00)	0.68 (1.00)	−0.01 (–0.01)	1.02 (1.02)	0.05 (0.00)	0.88 (1.00)
ENDIREH weights	0.01 (0.00)	1.00 (0.74)	0.01 (–0.03)	0.78 (0.87)	−0.02 (0.01)	0.66 (0.81)

Source. Authors’ calculations based on the 2016 ENDIREH.

Note. Summary information for the mean subset of the variables used to estimate propensity scores and inverse-probability-weighted regression adjustment scores. Table shows “normalized differences,” as suggested by Imbens and Rubin (2015), between treatment and control groups, with unweighted and IPWRA weighted values, and variance ratios. In each column, unweighted standardized differences and variance ratios are shown first, and weighted differences and ratios appear in parenthesis, respectively. The reference category for the regions is the North region. Differences >|.4| are indicated in boldface. IPWRA = inverse-probability-weighted regression adjustment; CCT = conditional cash transfer; SES = socioeconomic status; ENDIREH = Encuesta Nacional sobre la Dinámica de las Relaciones en los Hogares.

For those who worked during the last year (columns 1 and 2), we observe important (standardized) differences in the raw data for the level of schooling and the socioeconomic status between treatment and comparison groups in both urban and rural settings. Nonetheless, once we apply inverse probability weights, all normalized differences between the IPWRA covariates are practically zero, while the weighted variance ratios are significantly closer to one. Once again, this tells us that the reweighted output is balanced.

With regard to the Received CCT from Progresa treatment (columns 3 and 4), the table’s output indicates significant imbalances in the standardized raw data in terms of socioeconomic status, school level, and number of children between CCT recipients and nonrecipients. We see improved balance for most of these covariates, nonetheless, the low variance ratio for the number of children indicates that this covariate may not be balanced in our model; the weighted variance ratio still appears to be considerably less than 1.0 (0.4 for urban and rural Mexico). Finally, in the case of those women who both worked during the past year and are CCT recipients, we again observe important differences in the raw data for the level of schooling, the socioeconomic status, and the number of children per women. However, these differences practically disappear once we employ reweighting.

The Effect of Employment and CCTs on Mexican Women’s Risk of IPV

Estimates of the prevalence of physical and/or sexual IPV by the three different treatment indicators included in our analysis are shown in Table 2. In both urban and rural settings, our findings indicate that IPV is more prevalent among partnered women who worked during the past year compared with those who were not employed. Moreover, this effect was significantly higher for partnered women residing in the rural setting; results show a statistically significant 6% to 7% increase in the prevalence of IPV among rural working women versus a 3% increase for those residing in Mexico’s cities.

Table 2.

Risk of IPV Among Currently Partnered Mexican Women.

Risk of IPV %	Urban		Rural
Risk of IPV %	PSM	IPWRA	PSM	IPWRA
Sexual and/or Physical
Worked during the last year	0.036*** (0.005)	0.033*** (0.004)	0.064*** (0.013)	0.066*** (0.007)
Received CCT	0.029** (0.011)	0.033*** (0.007)	0.008 (0.012)	0.021* (0.008)
Worked and received CCT	0.089*** (0.017)	0.071*** (0.011)	0.106*** (0.022)	0.083*** (0.013)

Note. Results shown are estimates of the ATET. Bootstrapped standard errors are shown in parenthesis. IPV = intimate partner violence; PSM = propensity score matching; IPWRA = inverse-probability-weighted regression adjustment; CCT = conditional cash transfer; ATET = average treatment on the treated.

p < .05. **p < .01. ***p < .001.

As compared with employed women, partnered women who receive cash transfers also appear to experience a statistically significant increase in physical and/or sexual IPV in urban settings. This increase is similar in size to the effect experienced by partnered women living in urban Mexico who worked during the past year (3%). In rural settings, however, the effect of CCTs on the increased risk of violence upon partnered women was smaller in size, and PSM estimates did not achieve statistical significance. Interestingly, in both urban and rural settings, there appears to be a large and statistically significant increase in the prevalence of IPV for women that were both employed during the past year and are CCT recipients (7% and 8%, respectively). While it is unclear why we find this, it might be because working outside the home increases violence for all women, and this effect may be stronger for women who receive Prospera given that they may be at risk in ways that are not captured in our data.

We then turn to examine the difference in the risk of past-year IPV between employed and unemployed women depending on the type of work they perform. Table 3 shows the PSM estimated effect-size for each type of employment status. Our results indicate that women in both urban and rural settings with stable employment experience a statistically significant increase in the risk of IPV (3% and 7%, respectively). In contrast, women in urban settings that were precariously employed⁶ experience a statistically significant increase in the risk of IPV of nearly 6%. Self-employment appears to have a similar effect to CCTs in urban areas (2.5% increase); however, self-employed women residing in rural Mexico or those with precarious jobs do not appear to experience an increased risk of IPV. Finally, in both settings, there does not seem to be an association between unpaid household employment and an increased risk of IPV. In ancillary analyses, we found that physical violence is more prevalent among women who work and receive a CCT than sexual violence.⁷

Table 3.

Propensity Score Matching Estimates of the Risk of IPV Among Currently Partnered Mexican Women by Employment Type.

Risk of IPV %	Urban	Rural
Sexual and/or Physical
Stable employment	0.029*** (0.006)	0.067*** (0.012)
Precarious employment	0.057** (0.019)	0.043 (0.026)
Self-employment	0.025* (0.012)	0.036 (0.022)
Unpaid household employment	0.000 (0.048)	0.064 (0.095)

Note. Results shown are estimates of the ATET. Bootstrapped standard errors are shown in parenthesis. IPV = intimate partner violence; ATET = average treatment on the treated.

p < .05. **p < .01. ***p < .001.

Discussion and Implications for Policy

This study builds on Vyas and Heise’s (2014) findings on women’s empowerment and the risk of partner violence and contributes to the literature in a new country: Mexico. We combine PSM and IPWRA methods to better analyze the relationship between women’s economic status—measured in terms of employment and regular cash transfers—and their risk of experiencing violence at the hands of their intimate partners, in both urban and rural Mexico. Our research findings differ from the results of previous studies on this issue in Mexico, as our methods account for potential selection bias and show an increased risk for IPV among women who are employed and receiving CCTs across settings. The risk is accentuated more for employment than for CCTs and is higher in rural areas, and this contrasts with previous research in Mexico that shows that different forms of economic participation decrease the risk of IPV (García Aísa, 2014; Maldonado et al., 2005; Terrazas-Carrillo & McWhirter, 2015; Villarreal, 2007).

Our data suggest that in rural areas, receipt of a cash transfer appears to incur a somewhat lower risk of violence toward female partners. However, the risk of experiencing IPV was slightly higher among urban women receiving the CCT from Prospera. Nearly 55% of recipients of government cash transfers (predominantly Prospera benefits) reside in rural localities while the remaining 45% live in urban areas (INEGI, 2016a). According to relative resource theory, one explanation may be that the women are not receiving a large sum of money, so their male partner may not feel threatened or economically disempowered. Moreover, these transfers may also be seen by male partners to be a less threatening arena of economic participation than employment because they are focused on the family. One other reason why men may not feel threatened by CCTs and why we find a lower risk of violence for women who receive them is that men may be forcing the women to give them the money and, therefore, maintaining control over their partners.

When considering the risk for IPV by type of work, the highest risk is among women who have stable employment in rural areas, followed by precarious employment in urban settings. When comparing all types of employment, the risk of IPV is highest among women who have stable employment in rural Mexico. Given that many economically active rural Mexican women are still employed in the agricultural sector, this higher incidence of violence and the taxing nature of agricultural labor may cause a double burden for these women (Comisión Económica para América Latina y el Caribe [CEPAL], 2014). The only type of employment that did not show an increase in IPV was unpaid household employment, which may also relate to relative resource theory. If a woman is not being paid, then she may not be perceived as being “economically empowered” so there may be less violence at the hands of her partner. A qualitative follow-up study could be helpful to explore some of the relationship dynamics between women and their partners. Moreover, it may be useful to do more quantitative research and disaggregate by different industries/sectors as the risk for IPV may vary depending on this.

In Mexico, the impact of IPV has been documented in the literature. Results from a recent randomized control trial illuminate some of the impacts of violence: Women who experienced high levels of IPV had greater risk of work disruptions (i.e., missing work, losing job, etc.) resulting from the violent incident (Gupta et al., 2018). Moreover, our data show that approximately 16% of the women that reported experiencing either physical or sexual IPV also reported experiencing physical pain or a serious impairment of health as a result. Such injuries are, in turn, likely to lead to work absenteeism. According to the ENDIREH, nearly 4% of respondents who had experienced physical or sexual violence at the hand of their partners reported leaving their studies or work. Another 2.4% said they had to miss work (INEGI, 2016b).

Policy Implications

According to the WHO (2012), there is a need for “comprehensive, multi-sectoral, long-term collaboration between governments and civil society” to respond to IPV, because women who have experienced IPV often need a complex range of support services from healthcare, social services, legal entities, and law enforcement. Mexico has a program titled the Comprehensive Programme to Prevent, Care, Punish and End Violence Against Women 2014-2018; however, reports indicate that policy implementation has been ineffective thus far (Office of the United Nations High Commissioner for Human Rights, 2018; United Nations Development Programme, 2017; United Nations Women, 2017, 2018). Evidence shows that policies are not effectively reaching people to whom they are targeted; 78.6% of the women who have experienced some type of IPV with their current or former partner did not seek help or file a complaint (INEGI, 2016b). Frías (2013) found that of Mexican women who experienced physical and sexual violence at the hands of their partners, only 11.03% went to a government agency that supports IPV survivors, and 22.84% went to the police or public authorities; the main reasons Mexican women did not seek support were fear of a partner, feeling that violence is insignificant, and because of children (Frías, 2013).

Given the importance of formal support in mitigating some of the potential negative outcomes associated with violence, it is crucial to understand what factors might influence a woman’s decision. Coworkers and managers may be able to serve as confidants as part of women’s support network on the job, but they will need to be aware of the existing resources available to women. Research indicates that when women have higher levels of perceived social support, they are more likely to disclose to nonfamily members, and this is crucial as formal support has been shown to improve women’s mental health and safety (Frías & Angel, 2007; Liang, Goodman, Tummala-Narra, & Weintraub, 2005).

Although the likelihood of experiencing violence is lower for Prospera recipients than for women who work, it may also be possible to target women victims who are working or who are receiving CCTs from Prospera to try to improve post-violence service utilization through awareness raising measures. Irrespective of whether the violence was caused due to women’s economic status, employers and CCT programs could help disseminate information to women on available services given the overall low level of service utilization among violence survivors. Another possibility would be to add a gender training component for Prospera recipients and families. There is evidence from South Africa showing that when gender equality training is incorporated into microfinance programs, it can reduce IPV (Pronyk et al., 2006). Adding a gender dialogue component to savings groups has been shown to reduce economic abuse and lower acceptance of wife beating in Côte d’Ivoire (Gupta et al., 2013). One comprehensive review of economic interventions to prevent IPV shows that globally economic strengthening and gender transformative interventions (i.e., microfinance and group-based discussions) show positive effects on lowering self-reported IPV (Gibbs, Jacobson, & Kerr Wilson, 2017).

Limitations

There are several limitations to this study. First, we only consider women who have experienced IPV in heterosexual relationships, excluding potential risks in same-sex relationships, and based on data limitations, it is unclear whether there is nonbinary gender diversity in the sample. Furthermore, the PSM findings for women who receive CCTs in urban areas should be interpreted with caution given that the available data violate the aforementioned assumption of common support for PSM. PSM only allows for controlling on observable characteristics, such that other confounding factors may be affecting the likelihood of working and experiencing IPV. For example, 23.4% of women in our sample who experienced IPV reported that abuse began and worsened for economic reasons, thus, economic factors could confound the relationship. Due to the cross-sectional and temporal nature of the data, it is also impossible to determine the directionality of the effect, such that causality cannot be determined. It is possible that women who are experiencing violence in the home may search for employment as a safety planning strategy to get away from their partner or to gain economic independence. Moreover, our research does not shed light on why levels of partner violence are consistently higher for women who are economically engaged in different ways. A qualitative follow-up study would help to explain some of the nuances in the interactions between partners, and to explore some of the temporal dimensions that this study lacked.

Conclusion

Female economic participation is increasing in Mexico and is crucial for the advancement of women and families and for the development of the country. Our study utilizes appropriate methods to deal with potential selection bias when examining the relationship between economic participation and IPV. Our findings indicate that women who participate economically experience higher incidence of violence. Therefore, there is a need for further research about the risk factors for IPV, including employment and receipt of CCTs. Only upon further understanding these dynamics will policy makers begin to get a handle on some of the potential unintended consequences associated with policies and programs that encourage female economic participation and design the appropriate mechanisms to prevent IPV and link women to formal support.

Footnotes

Acknowledgements

We thank Dr. Peter Ward for his input on this manuscript and Dr. Chandler Stolp for methodological guidance. We also thank Dr. Erin Lentz for her direction on where to publish.

Authors’ Note

This article was first presented at the Association for Public Policy Analysis and Management (APPAM) International Conference in 2018 in Mexico City, and we would like to acknowledge the Lyndon B. Johnson (LBJ) School of Public Affairs for providing us with financial support to attend the conference and present, as the feedback we received was invaluable to the development of this manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the grant P2CHD042849, Population Research Center, awarded to the Population Research Center at The University of Texas at Austin by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. This research has also received support from the grant, T32HD007081, Training Program in Population Studies, awarded to the Population Research Center at The University of Texas at Austin by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

ORCID iD

Sophie M. Morse

Notes

Author Biographies

Ana P. Canedo, MPA, is a doctoral student at the Lyndon B. Johnson (LBJ) School of Public Affairs and a Population Research Center graduate student trainee at the University of Texas at Austin. She holds an MPA from Cornell University, a BA in Economics, and a BA in Political Science from the Mexico Autonomous Institute of Technology (ITAM), Mexico.

Sophie M. Morse, MPP, is a doctoral student at the Lyndon B. Johnson (LBJ) School of Public Affairs and a Population Research Center graduate student trainee at the University of Texas at Austin. She holds an MPP from the Johns Hopkins Bloomberg School of Public Health and a BA in International Studies from Middlebury College.

References

Abadie

Imbens

G. W.

(2009). Matching on the estimated propensity score. Retrieved from https://economics.mit.edu/files/11853

Abadie

Spiess

(2016). Robust post-matching inference (Working Paper). Retrieved from https://scholar.harvard.edu/files/spiess/files/robust.pdf

Amnesty International. (2012). Colombia: Hidden from justice: Impunity for conflict-related sexual violence, a follow-up report. London, England: Author.

Austin

P. C.

Stuart

E. A.

(2015). Moving towards best practice when using Inverse Probability of Treatment Weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Statistics in Medicine, 34, 3661-3679. doi:10.1002/sim.6607

Bedi

A. S.

Chhachhi

Bhattacharyya

(2009). Marital violence and women’s employment and property status: Evidence from North Indian villages. World Development, 39, 1676-1689.

Bobonis

G. J.

Castro

Morales

J. S.

(2015). Conditional cash transfers for women and spousal violence: Evidence of the long-term relationship from the Oportunidades program in rural Mexico. Retrieved from https://www.semanticscholar.org/paper/Conditional-Cash-Transfers-for-Women-and-Spousal-of-Bobonis-Castro/87bc20df9b5da2087364370fefbada6d572314e4

Capaldi

D. M.

Knoble

N. B.

Shortt

J. W.

Kim

H. K.

(2012). A systematic review of risk factors for intimate partner violence. Partner Abuse, 3, 231-280.

Comisión Económica para América Latina y el Caribe. (2014). Population Note 98. Retrieved from https://repositorio.cepal.org/bitstream/handle/11362/37680/1/S1400185_es.pdf

Curtis

L. H.

Hammill

B. G.

Eisenstein

E. L.

Kramer

J. M.

Anstrom

K. J.

(2007). Using inverse probability-weighted estimators in comparative effectiveness analyses with observational databases. Medical Care, 45, S103-S107. doi:10.1097/MLR.0b013e31806518ac

10.

Dalal

(2011). Does economic empowerment protect women from intimate partner violence? Journal of Injury and Violence Research, 3, 35-44. doi:10.5249/jivr.v3i1.76

11.

Dávila Lárraga

L. G

. (2016). How does Prospera work?: Best practices in the implementation of conditional cash transfer programs in Latin America and the Caribbean. Retrieved from https://publications.iadb.org/en/publication/17104/how-does-prospera-work-best-practices-implementation-conditional-cash-transfer

12.

Del Pilar Martínez

. (2018, March). Mexico, lagging in female labor force participation. El Economista. Retrieved from https://www.eleconomista.com.mx/empresas/Mexico-rezagado-en-participacion-laboral-femenina-20180308-0186.html

13.

Devries

K. M.

Mak

J. Y.

García-Moreno

Petzold

Child

J. C.

Falder

. . . Rosenfeld

(2013). The global prevalence of intimate partner violence against women. Science, 340, 1527-1528.

14.

DuGoff

E. H.

Schuler

Stuart

E. A.

(2014). Generalizing observational study results: Applying propensity score methods to complex surveys. Health Services Research, 49, 284-303.

15.

Flake

D. F.

(2005). Individual, family, and community risk markers for domestic violence in Peru. Violence Against Women, 11, 353-373.

16.

Frías

S. M.

(2013). Strategies and help-seeking behavior among Mexican women experiencing partner violence. Violence Against Women, 19, 24-49.

17.

Frías

S. M.

Angel

R. J.

(2007). Stability and change in the experience of partner violence among low-income women. Social Science Quarterly, 88, 1281-1306.

18.

Gage

A. J.

(2005). Women’s experience of intimate partner violence in Haiti. Social Science & Medicine, 61, 343-364.

19.

García Aísa

. (2014). Conditional cash transfers and intimate partner violence among Mexican couples: The impact of Oportunidades on psychological abuse prevalence (Master in Economic Development and Growth). Lund, Sweden: Lund University.

20.

Gibbs

Jacobson

Kerr Wilson

(2017). A global comprehensive review of economic interventions to prevent intimate partner violence and HIV risk behaviours. Global Health Action, 10(Suppl 2), Article 1290427.

21.

Gupta

Falb

K. L.

Lehmann

Kpebo

Xuan

Hossain

. . . Annan

(2013). Gender norms and economic empowerment intervention to reduce intimate partner violence against women in rural Côte d’Ivoire: A randomized controlled pilot study. BMC International Health and Human Rights, 13(1), Article 46.

22.

Gupta

Willie

T. C.

Harris

Campos

P. A.

Falb

K. L.

Garcia Moreno

. . . Okechukwu

C. A.

(2018). Intimate partner violence against low-income women in Mexico City and associations with work-related disruptions: A latent class analysis using cross-sectional data. Journal of Epidemiology and Community Health, 72, 605-610. doi:10.1136/jech-2017-209681

23.

Hadi

(2005). Women’s productive role and marital violence in Bangladesh. Journal of Family Violence, 20, 181-189.

24.

Hagen-Zanker

Pellerano

Bastagli

Harman

Barca

Sturge

. . . Laing

(2017). The impact of cash transfers on women and girls. Education, 42(15), Article 2.

25.

Hidrobo

Fernald

(2013). Cash transfers and domestic violence. Journal of Health Economics, 32, 304-319.

26.

Hidrobo

Peterman

Heise

(2015). The effect of cash, vouchers, and food transfers on intimate partner violence: Evidence from a randomized experiment in Northern Ecuador. American Economic Journal: Applied Economics, 8, 284-303. doi:10.1257/app.20150048

27.

Hindin

M. J.

Kishor

Ansara

D. L.

(2008). Intimate partner violence among couples in 10 DHS countries: Predictors and health outcomes. Retrieved from https://dhsprogram.com/pubs/pdf/AS18/AS18.pdf

28.

Imbens

G. W.

Rubin

D. B.

(2015). Causal inference in statistics, social, and biomedical sciences. Cambridge, UK: Cambridge University Press.

29.

Instituto Nacional de Estadística y Geografía. (2010). Mexico Population and Housing Census 2010. Retrieved from https://www.inegi.org.mx/programas/ccpv/2010/

30.

Instituto Nacional de Estadística y Geografía. (2016a). Encuesta Nacional de Ingresos y Gastos de los Hogares 2016 Nueva serie. Retrieved from http://www.beta.inegi.org.mx/programas/enigh/nc/2016/

31.

Instituto Nacional de Estadística y Geografía. (2016b). Encuesta Nacional sobre la Dinámica de las Relaciones en los Hogares 2016. Retrieved from http://www.beta.inegi.org.mx/programas/endireh/2016/default.html

32.

Instituto Nacional de Estadística y Geografía. (2016c). Encuesta Nacional sobre la Dinámica de las Relaciones en los Hogares (ENDIREH) 2016 Principales Resultados. Retrieved from http://internet.contenidos.inegi.org.mx/contenidos/productos/prod_serv/contenidos/espanol/bvinegi/productos/nueva_estruc/promo/endireh2016_presentacion_ejecutiva.pdf

33.

International Labor Organization. (2019). Mexico. Retrieved from https://bit.ly/2tbN7YM

34.

Kim

J. C.

Watts

C. H.

Hargreaves

J. R.

Ndhlovu

L. X.

Phetla

Morison

L. A.

. . . Pronyk

(2007). Understanding the impact of a microfinance-based intervention on women’s empowerment and the reduction of intimate partner violence in South Africa. American Journal of Public Health, 97, 1794-1802.

35.

Kishor

Johnson

(2004). Profiling domestic violence: A multi-country study. Calverton, MD: The DHS Program. Retrieved from https://www.popline.org/node/235024

36.

Koenig

M. A.

Ahmed

Hossain

M. B.

Mozumder

A. B. M. K. A

. (2003). Women’s status and domestic violence in rural Bangladesh: Individual- and community-level effects. Demography, 40, 269-288.

37.

Kwagala

Wandera

S. O.

Ndugga

Kabagenyi

(2013). Empowerment, partner’s behaviours and intimate partner physical violence among married women in Uganda. BMC Public Health, 13(1), Article 1112. doi:10.1186/1471-2458-13-1112

38.

Liang

Goodman

Tummala-Narra

Weintraub

(2005). A theoretical framework for understanding help-seeking processes among survivors of intimate partner violence. American Journal of Community Psychology, 36, 71-84.

39.

Maldonado

Nájera

Segovia

(2005). Effects of the Opportunities Program on relationships and families. Director General de Coordinación y Vinculación. Retrieved from http://www.pot.sedesol.gob.mx/work/models/SEDESOL/Resource/1668/1/images/efecto_relaciones.pdf

40.

México Gobierno de la República. (2016). Information about Prospera social inclusion program. Retrieved from https://www.gob.mx/prospera/documentos/que-es-prospera

41.

Naved

R. T.

Persson

L. Å

. (2005). Factors associated with spousal physical violence against women in Bangladesh. Studies in Family Planning, 36, 289-300.

42.

Office of the United Nations High Commissioner for Human Rights. (2018). Committee on the elimination of discrimination against women reviews report of Mexico. Retrieved from https://www.ohchr.org/EN/NewsEvents/Pages/DisplayNews.aspx?NewsID=23344

43.

Organisation for Economic Co-operation and Development. (2012). How does Mexico compare? Retrieved from https://www.oecd.org/mexico/Mexico_final_EN.pdf

44.

Organisation for Economic Co-operation and Development. (2017). Closing the Gender Gaps: It is Time for Mexico to Act. Retrieved from http://www.oecd.org/gender/Closing%20the%20Gender%20Gap%20-%20Mexico%20FINAL.pdf

45.

Organisation for Economic Co-operation and Development. (2018). Working age population (OECD Data). Retrieved from https://data.oecd.org/pop/working-age-population.htm

46.

Organisation for Economic Co-operation and Development. (2019). Employment: Labour force participation rate, by sex and age group (OECD.Stat). Retrieved from https://stats.oecd.org/index.aspx?queryid=54741

47.

Perova

(2010). Three essays on intended and not intended impacts of conditional cash transfers. Retrieved from http://digitalassets.lib.berkeley.edu/etd/ucb/text/Perova_berkeley_0028E_10750.pdf

48.

Pronyk

P. M.

Hargreaves

J. R.

Kim

J. C.

Morison

L. A.

Phetla

Watts

. . . Porter

J. D.

(2006). Effect of a structural intervention for the prevention of intimate-partner violence and HIV in rural South Africa: A cluster randomised trial. The Lancet, 368, 1973-1983.

49.

Rahman

Hoque

M. A.

Makinoda

(2011). Intimate partner violence against women: Is women empowerment a reducing factor? A study from a National Bangladeshi sample. Journal of Family Violence, 26, 411-420. doi:10.1007/s10896-011-9375-3

50.

Rosenbaum

P. R.

Rubin

D. B.

(1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41-55. doi:10.1093/biomet/70.1.41

51.

Schuler

S. R.

Hashemi

S. M.

Riley

A. P.

Akhter

(1996). Credit programs, patriarchy and men’s violence against women in rural Bangladesh. Social Science & Medicine, 43, 1729-1742.

52.

StataCorp. (2013). teffects ipwra—Inverse-probability-weighted regression adjustment. Stata 13 Base Reference Manual. Retrieved from https://www.stata.com/manuals13/teteffectsipwra.pdf

53.

Terrazas-Carrillo

E. C.

McWhirter

P. T.

(2015). Employment status and intimate partner violence among Mexican women. Journal of Interpersonal Violence, 30, 1128-1152. doi:10.1177/0886260514539848

54.

United Nations Development Programme. (2017). From commitment to action: Policies to end violence against women in Latin America and the Caribbean. Retrieved from https://reliefweb.int/sites/reliefweb.int/files/resources/UNDP-RBLAC-ReportVCMEnglish.pdf

55.

United Nations Women. (2017). Facts and figures: Economic empowerment | UN Women – Headquarters. Retrieved from https://www.unwomen.org/en/what-we-do/economic-empowerment/facts-and-figures

56.

United Nations Women. (2017). The long road to justice, prosecuting femicide in Mexico. Retrieved from http://www.unwomen.org/en/news/stories/2017/11/feature-prosecuting-femicide-in-mexico

57.

United Nations Women. (2018). Mexico. Retrieved from http://lac.unwomen.org/en/donde-estamos/mexico

58.

Villarreal

(2007). Women’s employment status, coercive control, and intimate partner violence in Mexico. Journal of Marriage and Family, 69, 418-434.

59.

Vyas

Heise

(2014). Using propensity score matching to estimate an “unbiased effect-size” between women’s employment and partner violence in Tanzania. Journal of Interpersonal Violence, 29, 2971-2990. doi:10.1177/0886260514527823

60.

Vyas

Watts

(2009). How does economic empowerment affect women’s risk of intimate partner violence in low and middle income countries? A systematic review of published evidence. Journal of International Development, 21, 577-602. doi:10.1002/jid.1500

61.

The World Bank. (2018). Labor force participation rate, female (% of female population ages 15+) (modeled ILO estimate). Retrieved from https://data.worldbank.org/indicator/SL.TLF.CACT.FE.ZS?locations=MX

62.

World Health Organization. (2012). Intimate partner violence. Retrieved from http://apps.who.int/iris/bitstream/10665/77432/1/WHO_RHR_12.36_eng.pdf

63.

World Health Organization. (2013). Global and regional estimates of violence against women: Prevalence and health effects of intimate partner violence and non-partner sexual violence. Geneva, Switzerland: Author.