Male employment and female intra-household decision-making: a Mexican gold mining case study

Au Yong Lyn, Audrey Article — Published Version Male employment and female intra-household decisionmaking: a Mexican gold mining case study Review of Economics of the Household Provided in Cooperation with: Springer Nature Suggested Citation: Au Yong Lyn, Audrey (2020) : Male employment and female intra-household decision-making: a Mexican gold mining case study, Review of Economics of the Household, ISSN 1573-7152, Springer US, New York, NY, Vol. 19, Iss. 3, pp. 699-737, https://doi.org/10.1007/s11150-020-09520-z This Version is available at: https://hdl.handle.net/10419/288579 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Rev Econ Household (2021) 19:699–737 https://doi.org/10.1007/s11150-020-09520-z Male employment and female intra-household decision-making: a Mexican gold mining case study Audrey Au Yong Lyn 1,2 Received: 1 May 2020 / Accepted: 15 October 2020 / Published online: 18 November 2020 © The Author(s) 2020; This article is published with open access Abstract This study explores the effect of economic booms in male-dominated industries like mining on female intra-household decision-making power. Using the 2007–2008 global financial crisis as an exogenous event which led to a gold mining boom in Mexico, I find thatwomenlivingingoldendowedmunicipalities experienced higher decision-making power contrary to some theoretical predictions. These results appear to be consistent with unitary household bargaining models which assume income pooling, as female decisionmaking power increased despite no changes in female labor force participation and an observed increase in male employment. Findings from a separate survey additionally show that while women residing in gold endowed states had higher decision-making power, they were also more likely to suffer from intimate partner violence (IPV). This suggests that a woman’s intra-household decision-making authority is not necessarily negatively correlated with her risk of IPV as posited in feminist theory. Keywords Gold mining ●Intra-household decision making ●Global income shocks ● Male employment JEL-Codes J12 ●D13 ●F62 ●L72 1 Introduction Mexico has a long-standing history of gold mining, where the practice of mining gold dates back to the pre-Hispanic times and contributed greatly to Latin America’s economic expansion during the colonial era. It was not until the last two decades however, that the Mexican gold mining industry took off. Prior to the 2000s, Mexico *Audrey Au Yong Lyn [email protected] 1 ETH Zurich, KOF Swiss Economic Institute, Leonhardstrasse 21, 8092 Zürich, Switzerland 2 Munich Graduate School of Economics (MGSE), Ludwig-Maximilian University of Munich, Kaulbachstrasse 45, 80539 Munich, Germany 1234567890();,: predominantly focused on silver production, as the country had traditionally been the number one producer of the world’s silver. Due to market speculation of the impending 2007–2008 financial crisis, global gold prices started increasing in 2003 and spiked sharply between 2006 and 2011, with Mexican gold production following the same trend. This event, together with the influx of foreign direct investments (FDI) for mining explorations, inevitably generated a shift in focus from silver to gold mining in Mexico during this time period (Secretaría de Economía, 2013). 1 According to the National Institute for Statistics and Geography (INEGI), the production of gold in Mexico increased three-fold at an average rate of about 5.3% each year between 2000 and 2011. Relative to the world however, Mexico’s global share in gold production during the mining boom only increased from about 0.3% in 2000 to 2.8% in 2011 (The Observatory of Economic Complexity (OEC)). Notwithstanding, Mexico’s rich endowment in precious metals like gold in addition to the gold mining boom due to the 2007–2008 global financial crisis, provides an ideal setting for studies that aim to understand how male employment stimulated by mineral-led activities, impact female intra-household welfare outcomes like decisionmaking and intimate partner violence (IPV). For many centuries, superstition has kept Latin American women away from mining as it was believed that if a woman went near a mine, it would become jealous, hide its wealth and cause catastrophes (Arcos et al. 2018). Lutz-Ley and Beuchler (2020) additionally note that because mining takes place in remote areas, miners are subject to long journeys, uncomfortable settings, and risky conditions. For these reasons, the mining industry has typically been male-dominated. In Mexico, despite efforts to increase female participation in the mining sector, the proportion of women in mining only increased from about 6% in 2004 to 13% in 2009, and subsequently decreased to 11% in 2014 (National Institute for Statistics and Geography (INEGI), 2020). 2 Figure 4of the appendix also plots the absolute number of employees in gold mining by gender, between 1994 and 2014. The graph shows that while male employment in gold mining increased sharply especially between 2009 and 2014, female employment in gold mining only rose marginally during the same time period. Subsequently, in a mining report by the African Union (2009), it was suggested that booms in such male-dominated sectors would result in a decrease in women’sintrahousehold decision-making power. The underlying logic is that as the mining sector flourishes and more men than women gain employment, women become increasingly reliant on their partners’incomes. Alternatively, busts in the mining industry could also decrease women’s decision-making power through poorer employment prospects related to industrial crowding out effects. In a study on the U.K. by Aragon et al. (2018), coal mine closures were found to be associated with an increase (decrease) in male (female) employment due to the crowding out of female-dominated sectors like manufacturing. In the context of oil extraction in the U.S. however, Maurer and Potlogea (2020) did not observe any gender-biased crowding out effects, contrary to 1 S and P Global (2019) estimates that the Latin American region receives approximately 28–30% of the global budget in mining explorations each year, with six mining powerhouses, Peru, Brazil, Chile, Argentina, Ecuador and Mexico accounting for the lion’s share (90%) of the region’s budget. 2 INEGI Data on the share of men and women in mining is only available for the years 1999, 2004, 2009 and 2014 since the economic census is only conducted once every five years. 700 A. Au Yong Lyn Aragon et al. (2018)findings. Ultimately, the impact of changes in extractive industries (EI) on female employment and in turn, intra-household decision-making outcomes remains inconclusive and renders further investigation. To date, only one paper by Tolonen (2018), has explicitly explored the relationship between mining and spousal decision-making dynamics. In a cross-country study on Sub-Saharan Africa, the author found no effect of local gold mine openings on women’s intra-household bargaining outcomes. This in turn raises the question of whether the commonly assumed negative relationship between gold mining and intra-household female decision-making power is consistent with real world data. This paper therefore endeavors to further test the hypothesized negative relationship between booms in male-dominated industries like mining, and female decisionmaking power in a Latin American country like Mexico. From a cultural and socioeconomic perspective, the Sub-Sahara African region and Latin America are highly distinct. The impact on female decision-making outcomes generated by a mining boom could therefore be different across geographic regions. 3 Tolonen’s(2018) findings from her cross-country study on Sub-Saharan Africa may thus not be generalizable to countries like Mexico in the Latin American region. In addition, since more indigenous women in Mexico have been documented to live in mining communities and are simultaneously poorer and subject to traditional gender stereotypes, understanding the impact of economic booms in male-dominated sectors like mining is important for enhancing and facilitating gender equality efforts among these particular demographic groups (Lutz-Ley and Buechler 2020). This paper contributes to existing literature in additional ways. First, no previous studies to my knowledge have used exogenous movements in global commodity prices to evaluate the effect of a mining boom on female intra-household decisionmaking power and intimate partner violence (IPV). In particular, this study exploits the sharp rise in world gold prices between 2003 and 2011, as well as the differences in gold endowment across municipalities and states in Mexico as sources of variation. The paper also utilizes two different data sources, one at the municipality level (MxFLS) and the other at the state level (ENDIREH), to identify changes in not only female bargaining power, but also IPV outcomes during the gold mining boom respectively. To date, only two studies on Sub-Saharan Africa have analysed the impact of mining on IPV. In a cross-country analysis on Sub-Saharan Africa, Kotsadam et al. (2016) found no significant relationship between both factors, whereas in a more location-specific study on Eastern Democratic Republic of Congo (DRC), Rustad et al. (2016) discovered that women who lived closer to artisanal and smallscale mining were more likely to experience sexual violence. In this study on Mexico, due to the detailed information on IPV provided by a state-level (ENDIREH) data set, I am able to segregate IPV into four different forms (physical, sexual and emotional abuse and threats of violence), unlike previous studies by Kotsadam et al. (2016) and Rustad et al. (2016) that have only examined harder types of IPV like physical and sexual violence. 3 For instance, in Latin America, the machismo culture is particularly strong and had origins during colonial times, where women were perceived as weak, naive, and had to be protected by men as they could not fend for themselves (Bridges 1980). Machoism is therefore often associated with an increased risk of intimate partner violence (IPV) perpetration, and implies that men of Latino origin engage in harmful behaviors such as drunkenness, violence and infidelity (Gutmann and Viveros 2005; Mancera et al. 2017). Male employment and female intra-household decision-making: a Mexican gold mining case. . . 701 The results from the analyses show that women living in gold endowed states were more likely to suffer from various forms of IPV, though the types of IPV experienced by women were different for those who were poorer and wealthier. Given the spike in domestic violence during COVID-19, this study is therefore particularly relevant as it helps to shed some light on how changes in male employment as a result of economic booms or busts (in the context of COVID-19) affect women’s risk of IPV, especially in low-middle income countries like Mexico. The findings also reveal that women residing in gold endowed municipalities experienced an increase in their decision-making power at home, which was likely to be driven by a rise in male employment probabilities. Contrary to non-unitary household bargaining models (see McElroy and Horney 1981, Lundberg and Pollak 1994 and Lommerud, 2003) that predict a decline in a woman’s household bargaining power along with a concurrent increase in male employment opportunities generated by a mining boom for instance, the results from this study suggest that an increase in a husband’s outside option through better employment prospects relative to his wife’s, may not necessarily hamper her intra-household decision-making power ability. This finding can be juxtaposed against unitary household bargaining models that predict the pooling of household income (Samuelson 1956). From a policy standpoint, it subsequently elucidates how income resources are distributed within the household, which carries important implications for the effectiveness of cash-transfer programs in reducing IPV for example. The rest of the paper is organized as follows: section II provides a background on gold mining in Mexico and discusses relevant theories relating income-generating opportunities to intra-household female decision-making; section III describes the data and empirical method used in this study; section IV presents the main results of the paper, discusses possible channels and additional outcomes like IPV, and reports a set of robustness checks; section V finally concludes. 2 Background and existing literature 2.1 Gold Mining in Mexico This study specifically focuses on gold mining in Mexico. Due to market speculation of the 2007–2008 global financial crisis, the global demand for gold began increasing in the few years leading up to the crisis in 2003. Figure 4of the appendix supports this phenomenon, and shows an increased demand for labor in the gold mining sector particularly for men, between 2004 and 2014. By 2011, the price of gold was approximately six times higher than pre-shock levels in 2002, with the steepest increase occurring between the end of 2006 and 2011. 4 This event subsequently 4 Given Mexico’s economic ties and dependence on the U.S. as an export market, all economic sectors including mining took the hit during the 2007–2008 global financial crisis. This explains the slight fall in gold production between 2008 and 2009. Mining was however the first industrial sector in Mexico to recover from the crisis as the economy started to recover in 2009, which is depicted by the sharp increase in gold production and prices between 2009 and 2011 (Engineering and Mining Journal, 2011). Moreover, the panel MxFLS data used in this study was conducted in 2002, 2005–2006 and 2009–2010, which does not overlap with the dip in gold production between 2008–2009. 702 A. Au Yong Lyn stimulated a gold mining boom across many mineral-led countries like Mexico, as gold production intensified in response to the spike in global gold prices (Engineering and Mining Journal 2011). Figure 1shows the natural log of the prices and production of gold through these shocks between 1998 and 2011. As can be seen, global gold prices increased steadily from 2003 to 2011, and gold production in Mexico followed the same trend. It is worth noting that although the prices of other precious metals like silver rose as well, the increase was not as substantial as compared to gold. While the prices and production of silver nearly doubled between 2009 and 2011, they were relatively constant before that period (see Figure 5in the appendix). Moreover, silver has a much lower value to weight ratio than gold, where the gold-silver price ratio was 84:1 at the depth of the global financial crisis. 5 Therefore despite Mexico being a large producer and exporter of silver, there is little temporal variation to exploit in silver mining production for this study. Accordingly, I test if the production of gold in Mexico responded to world gold prices, by conducting a first-stage analysis of the relationship between the two variables. The t-statistics of the effect of current prices on future production 1,2 and 3 months after (t+1tot+3) are highly statistically significant (18.09, 20.98, 20.96 respectively), indicating that local production responded strongly to global prices. 6 Figure 1additionally shows that local gold production and world prices followed highly similar trends during the sample period. Most of the gold mining in Mexico is conducted by large-scale companies located in the north and central regions. Figure 2illustrates the geographical variation in gold production across the respective Mexican states between 2003 and 2011 specifically. Fig. 1 Trends in world gold prices and gold production in Mexico. Notes: This figure plots trends in global gold prices and the production of gold in Mexico. Each point is the logged value of average prices over each year and the total production in each year. The blue (red) line denotes production (price). Data on Mexico’s gold production come from the National Institute for Statistics and Geography (INEGI), and data on world gold prices is from the London Metal Exchange’s (LME) monthly historical price series 5 Data for the gold and silver price ratio was from the London Bullion Market Association (LBMA) and ICE Benchmark Administration (IBA). 6 Estimation output for the regression is not listed; but available upon request. Male employment and female intra-household decision-making: a Mexican gold mining case. . . 703 The graph shows that three northern states: Sonora, Chihuahua and Durango contribute to the largest shares of gold produced in the country. While there is some gold mining activity in south-eastern states like Guerrero and Oaxaca, the average volume of gold produced in the north is approximately 4.5 times more than in these states. In total, about 14 out of 32 Mexican states are actively involved in the mining and production of gold, albeit the remaining 18 states also do produce gold but in much smaller quantities, contributing to only 0.4% of the total gold produced in Mexico. Typically, the mining of gold is done through open-pit operations that use capital intensive technology, and many gold mining businesses in Mexico are owned by large international corporations or businesses rather than by locals. 7 Much of the Mexican soil is also favorable for the practice of mining exploration and exploitation, where approximately two-thirds of the national territory contain mineral resources that are suitable for mining purposes (Hurtado and Salazar 1999). On average, it is estimated that exploration activities of a mining area take 10 years, and an additional 20 to 30 years for exploitation (The Mexican Geological Service (SGM) 2020). 8 In the Mexican context, the relationship between ejidos and miners has also played a major role in mining production over the last three decades. Ejidos are essentially areas of land that are used for collective agricultural practices, whereby community members are assigned allotments on which they farm on, and have joint ownership over the land with other community members. In 1993, the mining law reform allowed ejidos and properties of agrarian communities to be privatized, meaning that in order for miners to carry out exploitation work on ejidos, they would have to have an agreement with the owners of the land, and have it registered with the agrarian Fig. 2 A graphical representation of gold production across Mexican states between 2003 and 2011. Notes: The graph shows the total production of gold (in kilograms) in each state between 2003–2011. The darker the shaded area, the greater the quantity of gold produced in the state. Data on gold production come from the National Institute for Statistics and Geography (INEGI): Banco de Información Económica (BIE) series 7 Open-pit mining is often used for extracting “hard rocks”such as metal ores like gold, silver, copper, iron, among others. The process begins by excavating large surface areas of the ground to expose and mine the metal ore beneath. Specifically, holes are drilled into the ground using explosives, and hard rock material is then removed layer-by-layer through controlled demolitions. 8 Information on the timeline of mining exploration and exploitation activities was provided by the Mexican Geological Service (SGM) through a public information request system. 704 A. Au Yong Lyn registry (Penman 2016). Under these circumstances, if a consensus between the owner of an ejido and prospective mining business was not reached, miners would not be able to conduct exploitation activities. 9 2.2 Mining, employment and female decision-making power: conceptual framework 2.2.1 Non-unitary household bargaining models Non-unitary household bargaining models in economics predict that external shift parameters such as an increase in female job opportunities for instance, are key determinants of a woman’s bargaining power. 10 These external environmental factors are typically termed as the ‘outside option’, and are a function of an individual’s decision making authority (McElroy and Horney 1981; Lundberg and Pollak 1994; Lommerud 2003). Subsequently, the theory suggests that a change in a woman’s outside option is sufficient to alter her aggregate bargaining power within the household. Improvements in relative or absolute job opportunities for women due to gold mining for example, should thus increase female bargaining power through the outside option. In regions like Sub-Saharan Africa, Kotsadam and Tolonen (2016) found that the intensification of mining activities generated more jobs for women in complementary sectors like the service industry. However, in a subsequent study on Sub-Saharan Africa by Tolonen (2018), the author found no impact of gold mining on spousal decision-making power, indicating that an improvement in a woman’s outside option through an increase in job opportunities may not necessarily guarantee an increase in her intra-household bargaining power. Whether the Mexican gold mining boom affected female decision-making power through employment, therefore remains an empirical question. The extent to which an improvement in a woman’s outside option can be leveraged to her advantage also depends on how male income or employment changes relative to females’during a gold mining boom. The prediction of non-unitary household bargaining models is that if relative male income or employment increases, the intra-household gender wage gap widens and in turn lowers women’s decision-making power. In several studies on the Middle East and the U.K, booms in industries like crude oil and coal decreased female labor force participation due to higher male wages and increased household wealth, which subsequently crowded out female-dominated sectors like manufacturing (Ross 2008; Aragon et al. 2018). Considering that mining is still a male-dominated activity, economic booms in the sector in theory, could disproportionately increase employment prospects for men and decrease female intra-household decision-making power (African Union 2009). 9 For legal documentation of ejidos and mining concessions, see article 27 of the mining law: http://www. diputados.gob.mx/LeyesBiblio/pdf/151_110814.pdf. 10 This study follows the theoretical framework of a non-unitary rather than unitary household bargaining model. A majority of empirical studies find little evidence in support of the unitary household model framework, which has been criticized for its over-simplicity and lack of applicability to real world data (Lundberg and Pollak, 1994; Browning et al. 2006; Bobonis 2009; Cherchye et al. 2012). Male employment and female intra-household decision-making: a Mexican gold mining case. . . 705 2.3 Unitary Household Bargaining Models Unitary household bargaining models on the other hand, assume the pooling of household income which implies that the household demand for goods and leisure are not dependent on the sources of income (i.e. the individual income contribution of the wife and the husband to household wealth), but rather the aggregate household income. Samuelson (1956) suggested that as long as there is a general consensus among household members on the aggregation of household income and preferences, the income pooling assumption is satisfied. Though the unitary model has been widely criticized for its over-simplicity and lack of applicability to real world data (see Lundberg and Pollak 1994; Browning et al. 2006; Bobonis 2009; Cherchye et al. 2012), there is evidence that some developing countries like India, still follow the unitary framework (see Neff et al. 2012; Klasen and Pieters, 2015). Consequently, under the unitary household bargaining model, a woman’s intra-household decisionmaking authority is dependent on changes in the total household wealth, rather than on changes in her absolute or relative income prospects. In this regard, the outside option characterized by better employment opportunities in gold mining for instance, will not affect her intra-household decision-making power over the consumption of goods and services, unlike what is predicted by non-unitary household bargaining models. A total increase in household finances is therefore sufficient to increase her intra-household decision-making power over consumption goods and services. 3 Data and empirical method 3.1 Data 3.1.1 Dependent variables This study draws from a panel data set to elicit the effect of the gold mining boom on female decision-making power in Mexico. Specifically, it utilizes the Mexican Family Life Survey (MxFLS), which fortuitously overlaps with the timing of the gold mining boom as a result of the 2007–2008 global financial crisis. The MxFLS is a nationally representative panel survey containing information on the decisionmaking power of married or cohabiting couples, and interviewed approximately 8400 households and 35,000 individuals in more than 150 municipalities across Mexico. 11 It also boasts low attrition rates with virtually 90% of households re-interviewed and tracked across all three survey waves. To capture the differences in women’s intra-household decision-making power before and after the spike in gold mining production, I use all three waves of the MxFLS conducted in 2002, 2005–2006 and 2009–2010. In particular, I utilize the first wave of the MxFLS (2002) as the pre-treatment sample, and the last two waves (2005–2006, and 2009–2010) as the post-treatment sample given that gold prices 11 In this study, the MxFLS sample excludes women who got divorced (and who may or may not have changed partners). The same women (and partners) are therefore followed over time across all three survey waves to attenuate concerns relating to possible endogenous marriage formation. 706 A. Au Yong Lyn column (2) show an increase of 0.04, 0.06 and 0.04% respectively, with a two standard deviation increase in gold prices. Though the magnitude of these effects is considerably small, the findings nonetheless indicate that economic booms in maledominated sectors like gold mining, do not necessarily worsen a woman’s decisionmaking power at home and could instead lead to an increase in their intra-household bargaining power over certain goods. In order to check that these results are not spurious, I draw from a separate pooled cross-sectional data set, ENDIREH. As discussed in section III.a3, the ENDIREH, which was conducted in 2003, 2006 and 2011 also contains information on women’s decision-making power, though geographic data is only available on the state-level. Subsequently, I utilize this data set as a robustness check to examine if the gold mining boom indeed generated an increase in intra-household female bargaining power as reflected in the MxFLS. 21 Since individuals in the ENDIREH however, were only interviewed over a period of one month between mid-October and midNovember for all three survey rounds, exploiting gold prices as a source of temporal variation is not feasible. Consequently, I define the post-treatment group as individuals interviewed in the 2006 and 2011 survey during which gold prices and production increased sharply, and those interviewed in 2003 as the pre-treatment group. Given that the ENDIREH surveyed individuals about their experiences with IPV and intra-household bargaining power 12 months (1 year) prior to the date of interview in the ENDIREH, responses in 2003 represent 2002 outcomes before gold prices started to rise in 2003. Table 10 of the appendix reports some descriptive statistics of the mean value of all covariates for both treatment and control groups in the ENDIREH. The summary statistics show that women in the control group are more likely to be rural and indigenous, have more children and have higher levels of education compared to the treated group. Given that the Mexican drug war only began in 2007, the large differences in the homicide rate between the treated and control group is expected as the control group was exposed to much less violence prior to 2007. The phase-in implementation of three domestic violence and divorce laws across states and time also explains the difference in mean values between the treated and control groups. For most of the covariates, the difference in means between both groups are statistically significant, indicating the importance of controlling for these covariates in the baseline regressions. Accordingly, I estimate the following difference-in-differences model: Yi;s;t¼B0þB1GendowmentsPostt ðÞþB2X0 i;tþB3Z0 s;tþ; tþπsþεi;s;tð2Þ Where Yi,s,t is woman i’s average decision-making power index in state sduring survey year t,Gendowmentsis the log of the land area claimed by gold mining concessions between 1973 and 1983 in state s,Posttis a binary variable equal to 1 for the survey years 2006 and 2011 during the peak of the gold price increase, and 0 21 The ENDIREH contains a total of thirteen decision-making questions in the 2003 and 2011 surveys, though the 2006 survey only contains eleven questions. For consistency, I exclude the two questions that were missing from the 2006 survey, and subsequently select questions that are similar to the ones asked in the MxFLS. This results in a remainder of six suitable decision-making questions that are then used to construct the average decision-making power index. For example, questions like ‘who makes decisions about your participation in social or political activities?’are excluded as they are not asked in the MxFLS data set. Male employment and female intra-household decision-making: a Mexican gold mining case. . . 713 otherwise. X0 i;tand Z0 s;tare vectors of individualand state-level covariates such as a woman’s rural and indigenous status, age, age squared, length of relationship with her partner, number of children, education level, the homicide rate per 100,000 inhabitants, the timing of domestic violence and divorce law reforms across Mexican states and time trends interacted with state characteristics such as proximity to coast, protected areas, presence of ejidos, population density and physio geographic classification to account for heterogeneous market trends associated with prior state conditions. ∅tand πsis the survey year of interview and state fixed effects respectively, and εi,s,t is the error term. The results from estimating Eq. (2) are shown in Table 2. As can be seen, the same qualitative conclusion from the MxFLS data set can be drawn, where the estimated coefficient of a woman’s overall decision-making power index is positive and statistically significant. In particular, women living in gold endowed states experienced increases in their aggregate decision-making power by 0.0005 (0.003/0.640) of a standard deviation relative to the mean. Similar to the findings from the MxFLS, this result appears to be driven particularly by household expenditures where women’s decision-making power over this category of goods increased by about 0.013 (0.011/0.832) of a standard deviation. The findings using the ENDIREH data set in Table 2also indicate that women in gold endowed states experienced lower bargaining power over the decision of which partner should use contraception, with estimated coefficients showing a decline in a woman’s decisionmaking authority over this matter by 0.001 (0.006/0.622) of a standard deviation when compared to the mean. For clarification, this questionnaire is distinct from the one in the MxFLS (Table 1) which asks women about the general use of contraception. The estimate for a woman’s bargaining power over this respective matter (general contraceptive use) using the ENDIREH in Table 2, reveals estimates that are statistically insignificant and zero-bound, different from the significant and positive estimate observed in Table 1using the MxFLS. Overall, the effect of the gold mining boom on women’s bargaining power over general contraceptive use remains ambiguous, albeit there is some evidence pointing to a decline in their decisionmaking authority over which partner in the relationship should use contraception. The effect of the gold mining boom on the remaining decision-making categories which do not show a statistically significant relationship are presented in Table 11 of the appendix. Taken together, albeit the estimated coefficients from the analysis using both the MxFLS and ENDIREH are quantitatively small, they support the basic conclusion that the gold mining boom improved overall intra-household female decision-making power. Moreover, this effect was likely to be driven by the increase in bargaining power over household goods and services such as food expenditures. 4.2 Possible channels It is important to examine the effect of the gold mining boom on labor market outcomes and other determinants of welfare, to better understand the underlying mechanisms behind the increases in female decision-making power. Accordingly, Table 3presents results from an analysis of the effect of the gold mining boom on male and female employment probabilities in the MxFLS. It also shows the impact of the gold mining boom on individuals’perceptions of safety outside of the home 714 A. Au Yong Lyn given the reported association between drug-related violence and gold mines in Mexico (GIATOC, 2016). Across all specifications, the estimates reveal that men living in gold endowed municipalities experienced an increase in employment prospects during the mining boom. Female employment on the other hand, appears to have been unaffected. In particular, male employment probability estimates are statistically significant at the 10% level and positive in the baseline specification in (2), where the estimated coefficient shows an increase in male employment of 1.2 percentage points (or 1.5% when compared to the sample mean). Coefficients for female employment probabilities however, are insignificant and close to zero, indicating that while the gold mining boom had a positive impact on male employment, it did not have any effect on female labor force participation. Altogether, these results suggest that the observed increase in female decisionmaking power was likely to be driven by increases in male employment during the gold mining boom. Unitary household bargaining models lend support to this finding as women’s intra-household decision-making power increased despite no observable changes in their employment probabilities. This indicates income pooling which is a central prediction of unitary models of the household, where the consumption of goods by the husband or wife depends solely on the total household income rather than on the sources of income. The same household income effect has also been documented in other developing countries like India (see Neff et al. 2012; Klasen and Pieters, 2015), where the elasticity of female labor supply increased in response to Table 2 The effect of the mining boom on female decision-making outcomes (ENDIREH data) (1) (2) Mean (std. dev.) DMP index (overall) 0.001 (0.001) 0.003** (0.001) 0.640 Observations (N) 191,768 186,506 DMP index (HH expenditures) 0.004 (0.002) 0.011*** (0.002) 0.832 Observations (N) 190,246 185,107 DMP index (who uses the contraception) −0.004* (0.002) −0.006*** (0.002) 0.622 Observations (N) 115,055 112,242 DMP index (contraceptive use) −0.001 (0.002) −0.001 (0.002) 0.594 Observations (N) 116,173 113,328 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s rural and indigenous status, age, age squared, length of relationship with her partner, number of children, education level, and state-level controls like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Column (1) does not include any controls, and column (2) adds baseline controls and time trends interacted with state characteristics like proximity to coast, protected areas, presence of ejidos, population density and physio geographic classification (heterogeneous market trends). All specifications include state and survey year fixed effects. ***p< 0.01 **p< 0.05 *p< 0.1 Male employment and female intra-household decision-making: a Mexican gold mining case. . . 715 higher household wealth. Subsequently, since male employment in gold endowed municipalities increased, suggesting a rise in aggregate household income, women simply could have had more household wealth to allocate expenditures to household goods like food. In addition, the 2002 MxFLS pre-treatment survey wave shows that women had higher decision-making power over household goods and services compared to men (0.404 for women versus 0.176 for men), lending support to the idea that since women have conventionally had more control over household matters, the increase in women’s bargaining power over this particular set of goods and services could have simply been driven by an improvement in their husbands’ employment opportunities. Table 3also reveals some evidence of a decrease in men’s perception of safety in gold endowed municipalities during the boom, though there were no changes in women’s safety perceptions. The estimated coefficients for men’s fear of assault are significant and positive at the 10% level, and show that the gold mining boom increased men’s fear of getting assaulted outside of their home by 2.9 percentage Table 3 The effect of mining on labor market outcomes and perceptions of safety (MxFLS data) (1) (2) Mean Men: Employed 0.018*** (0.007) 0.012* (0.006) 0.803 Observations (N) 8420 6968 Fear of assault 0.028* (0.016) 0.029* (0.016) 0.296 Observations (N) 8747 7042 Women: Employed 0.003 (0.005) 0.001 (0.004) 0.190 Observations (N) 10,471 9131 Fear of assault 0.017 (0.014) 0.014 (0.012) 0.370 Observations (N) 10,471 9131 Baseline controls No Yes Heterogeneous trends No Yes Individual FE Yes Yes Month FE Yes Yes Notes: Standard errors are clustered at the municipality level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and municipality covariates like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. An analogous set of covariates corresponding to men are included in the male sample. Column (1) does not include any controls, and column (2) adds baseline controls and time trends interacted with municipal characteristics such as rural-urban status, transport network quality, community quality and the presence of ejidos (heterogeneous market trends). All specifications include individual and month fixed effects. The differences in sample size between men and women is due to the fact that fewer men were interviewed in the MxFLS than women. ***p< 0.01 **p< 0.05 *p< 0.1 716 A. Au Yong Lyn points (or 10% when compared to the sample mean). This result is plausible, and could possibly be explained by the increase in drug-related violence in gold endowed municipalities as some Mexican drug cartels reportedly switched into gold production, and since mining is still predominantly a male economic sector (GIATOC, 2016). Men in gold endowed municipalities were therefore likely to be exposed to more dangerous surroundings as a result. Subsequently as expected, the baseline regressions show a strong and significant positive correlation between men’s fear of assault and the municipal homicide rate. A first-stage regression analysis of the relationship between the homicide rate and the interaction term of gold prices and endowment, also reveal statistically significant t-statistics (2.86), indicating a positive relationship between the gold mining boom and violence. 22 Lastly, these findings also provide evidence of the male-breadwinner stereotype, where despite men being more afraid of getting assaulted or attacked during the gold mining boom, male employment probabilities still increased, with no observable changes in the female labor supply. 4.3 Additional outcome: intimate partner violence (IPV) Since the state-level ENDIREH data set also contains information on women’s experiences of intimate partner violence (IPV), I examine the effect of the gold mining boom on several types of gender-based violence as an additional outcome in the study. Following Bobonis et al.’s (2013) approach, IPV incidences are grouped into the four main categories: physical, sexual and emotional abuse, and the threat of violence, with a total of eight questions classified under physical violence, three questions under sexual violence, thirteen questions under emotional violence, and two questions under the threat of violence. 23 For physical and sexual violence, I create dichotomous variables equal to one if a woman experienced a single incident in the past year. Considering that emotional violence is more likely to be subject to personal interpretation, I construct an emotional violence indicator equal to one if a woman answered “yes”to one incident, but stated that it happened multiple times, or if she answered “yes”to at least two emotional abuse questions. Lastly, the threat of violence indicator is equal to unity if a woman answered “yes”to at least one threat of violence question. Accordingly, I estimate the same equation in (2), but replace the dependent variable with dichotomous indicators of the four different types of IPV. The results from the analysis are displayed in Table 4, and show that while women in gold endowed states had increased decision-making power at home, they were also more likely to suffer from all four forms of IPV. Specifically, the benchmark specification in column (2) shows that women living in gold endowed states experienced approximately 0.9, 0.1, 0.3 and 0.4 percentage points more emotional abuse, threats 22 Output is not listed, but available upon request. 23 For a detailed breakdown of the specific IPV questions and categorization, see Bobonis et al.’s (2013) online appendix: https://www.aeaweb.org/articles?id=10.1257/pol.5.1.179. In 2003, only twenty-nine questions on IPV were asked compared to the 2006 and 2011 survey rounds which had thirty IPV questions. I therefore only include the twenty-nine IPV questions from 2003 across all survey years to ensure consistency in the IPV measures. Male employment and female intra-household decision-making: a Mexican gold mining case. . . 717 of violence, physical abuse and sexual abuse respectively. 24 Compared to the mean values during the pre-treatment period before the gold mining boom, these figures represent an average increase in all four types of IPV by approximately 3.0% (emotional abuse), 3.4% (threats of violence), 3.8% (physical abuse) and 5.5% (sexual abuse). In a cross-country study on male-female job opportunities and IPV, Bhalotra et al. (2020) documented similar effect sizes of female unemployment on the incidence of physical violence. In their paper, a 1% increase in female unemployment was associated with a 2.75 and 2.87% increase in physical abuse respectively. The positive relationship between the gold mining boom and women’s risk of domestic violence could potentially be a result of male psychological stresses related to increased levels of fear. As shown in Table 3, the gold mining boom had a negative impact on men’s perception of safety, which was likely due to increased drug-related crime in mining areas. These findings can therefore be juxtaposed against several studies by Angelucci (2008) and Heise and Kotsadam (2015) for instance, that underscore psychological stress and trauma as drivers of domestic abuse. The rise in female intra-household decision-making power along with a concurrent increase in domestic violence is a striking finding, which could also possibly be explained by theories on male-backlash. The male-backlash effect, first proposed by Macmillan and Gartner (1999) suggests that when a wife experiences an increase in her tangible or intangible independence, men may be stimulated to perpetrate more violence as a means of “reinstating dominance and authority over his wife.”Therefore, in order to emotionally compensate for the increase in women’s decision-making power at home despite better employment prospects, men may retaliate through various forms of IPV. Notwithstanding, given that women’s IPV outcomes are from a more spatially aggregated analysis (state-level) compared to decision-making outcomes conducted at the municipality level, one should remain cautious about attributing the rise in women’s IPV risk solely to these mechanisms related to the gold mining boom. Because more spatially disaggregated data on IPV in Mexico is not available, it is not feasible to check if the mechanisms that happen at a more localized level, like at the municipality level in the MxFLS for example, are comparable to those that occur at a more aggregated level. Lastly, in order to provide a more detailed breakdown of the effect of the gold mining boom on IPV, I create a dummy for each individual question on physical abuse, sexual abuse, emotional abuse and the threat of violence, and conduct an analysis analogous to Eq. (2). Results from the analysis are presented in Tables 13–16 of the appendix. In a subsequent analysis, I examine the impact of the gold mining boom on women’s IPV outcomes across various socio-economic groups, as wider educational gaps between women and their spouses have been documented to increase IPV risk (Hidrobo et al., 2016; Heath, 2014). Additionally, because rural areas may have weaker policing, or have greater proportions of indigenous populations where the 24 According to Cameron et al. (2008), standard errors of the main IPV estimates are likely to be underestimated due to the small number of clusters (30 states), which is below the recommended benchmark of 40. As a robustness check, I therefore present wild-cluster bootstrapped standard errors in Table 12 of the appendix, which show that standard errors increase slightly as expected, and estimates remain statistically significant. 718 A. Au Yong Lyn machismo culture is likely to be stronger, women residing in secluded places may also be subject to more IPV than those living in more urbanized areas. To test for the presence of such heterogeneous effects, I divide women from the ENDIREH into four sub-samples according to their education level (low or high) and their residential status (rural or urban). 25 Accordingly, I examine the effect of the gold mining boom on each of the four sub-samples of women by repeating the estimation in Eq. (2), with the results presented in Table 5. The findings show that the indicator for physical violence is positive and significant for women living in rural areas and those with low education levels, where the probability of physical abuse is approximately 0.7 and 0.4 percentage points (8.4 and 4.8% in comparison to the mean) higher for rural women, and low educated women respectively. In addition, the results reveal that women from lower socio-economic classes were also more likely to experience emotional abuse with estimated coefficients significant and positive across the two different groups of women (1 and 0.8 percentage points for rural and low educated women respectively). Overall, less educated women appear to suffer more from sexual abuse, facing an increase in the probability of this type of abuse by about 0.7 percentage points. Table 4 The effect of the gold mining boom on intimate partner violence (IPV) (ENDIREH data) (1) (2) Mean Emotional Abuse 0.002 (0.002) 0.009*** (0.002) 0.297 Observations (N) 168,787 163,902 Threat of Violence 0.001** (0.001) 0.001** (0.001) 0.029 Observations (N) 191,940 186,666 Physical Abuse 0.002** (0.001) 0.003** (0.001) 0.080 Observations (N) 192,011 186,730 Sexual Abuse 0.003*** (0.001) 0.004*** (0.001) 0.073 Observations (N) 191,665 186,413 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s rural and indigenous status, age, age squared, length of relationship with her partner, number of children, education level, and state-level controls like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Column (1) does not include any controls, and column (2) adds baseline controls and time trends interacted with state characteristics like proximity to coast, protected areas, presence of ejidos, population density and physio geographic classification (heterogeneous market trends). All specifications include state and survey year fixed effects. ***p< 0.01 **p< 0.05 *p< 0.1 25 Women are considered employed if they had worked at least one hour in the past week. According to the INEGI’sdefinition, women living in localities with less than 2,500 inhabitants are classified as rural. Lastly, educated women are defined as those who have completed at least junior high school (‘secundaria’), which is Mexico’s basic compulsory education requirement. Male employment and female intra-household decision-making: a Mexican gold mining case. . . 719 Next, I estimate the impact of the mining boom on women from upper socioeconomic classes: those who live in urban areas and are relatively more educated. The results reveal that the positive gold mining shock did not have any impact on the probability of physical abuse for women from upper socio-economic classes. In particular, the coefficients for the physical abuse indicator for women living in urban areas and who are highly educated are statistically insignificant and close to zero. These women however, appear to suffer more from emotional abuse and threats of violence than their poorer counterparts. The findings show that urban and high educated women experienced a 0.7 and 1.2 percentage point (2.3 and 4% respectively when compared to the mean) increase in their likelihood of emotional abuse, and urban and high educated women were subject to 0.2 and 0.3 percentage points (6.9 and 12.5% respectively) more threats of violence from their partners respectively. The results also indicate that urban women are more likely to suffer from Table 5 The effect of mining on the incidence of intimate partner violence (IPV) among women from lowerand uppersocioeconomic classes (ENDIREH data) Rural Urban Loweducated Higheducated Emotional 0.010*** 0.007*** 0.008* 0.012*** Abuse (0.003) (0.002) (0.004) (0.004) Mean 0.265 0.306 0.296 0.298 Observations (N) 33,527 130,375 105,714 58,607 Threat of −0.002 0.002*** 0.002 0.003** Violence (0.001) (0.001) (0.001) (0.001) Mean 0.030 0.029 0.034 0.024 Observations (N) 37,430 149,236 120,203 66,918 Physical 0.007** 0.001 0.004** 0.001 Abuse (0.003) (0.001) (0.002) (0.001) Mean 0.083 0.078 0.084 0.074 Observations (N) 37,435 149,295 120,234 66,951 Sexual 0.002 0.004*** 0.007*** 0.002 Abuse (0.002) (0.001) (0.002) (0.002) Mean 0.081 0.071 0.087 0.057 Observations (N) 37,372 149,041 120,027 66,837 Baseline controls Yes Yes Yes Yes Heterogeneous trends Yes Yes Yes Yes State FE Yes Yes Yes Yes Survey year FE Yes Yes Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s rural and indigenous status, age, age squared, length of relationship with her partner, number of children, education level, and state-level controls like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Heterogeneous trends include time trends interacted with state characteristics like proximity to coast, protected areas, presence of ejidos, population density and physio geographic classification. All regressions include baseline controls, heterogeneous market trends, state fixed effects and survey year fixed effects. Sample sizes differ due to the difference in the number of individuals belonging to each sub-sample. ***p< 0.01 **p< 0.05 *p< 0.1 720 A. Au Yong Lyn sexual abuse than high educated women, with a 0.4 percentage point (5.6%) increase in their probability of experiencing this type of IPV. To explain why poorer women may be more susceptible to harder forms of IPV like physical abuse, I draw from a pioneering study by Gelles (1976) which showed that women who had fewer financial resources were less likely to leave an abusive relationship. Given that poorer women may be more hesitant to report domestic violence or file for divorce due to their lower income earning potential, and hence greater reliance on their husbands’incomes, men with partners from lower socioeconomic classes may be less wary about exhibiting harder types of IPV like physical or sexual abuse. On the other hand, women from upper socio-economic classes may be more likely to report domestic violence or file for divorce in cases of wife abuse, given their greater economic independence and earning potential. Partners of women from higher socio-economic classes may thus commit softer forms of IPV that are less tangible in nature and harder to prosecute. 4.3.1 Robustness: sensitivity analyses and alternative measures of DMP To further probe the robustness of the main results, I use an alternative empirical model to examine the effect of the gold mining boom on female decision-making power. Specifically, I estimate a multinomial logit fixed effects model where the decision-making power indicator is a categorical dependent variable ranging from 1 to 3, with the same covariates and fixed effects as in Eq. (1). For instance, if a woman responds that she was the sole decision-maker of that particular good category, it is assigned a value of 3 (woman has full power in decision-making). If the decision was jointly made with her husband, the category is assigned a value of 2, and if the decision was made solely by her husband, the category is assigned a value of 1 (indicating that the woman had no bargaining power over that good at all). The results from the analysis are presented in Table 17 of the appendix, which show qualitatively similar results as Table 1. Using value 3 (woman has full decisionmaking power) as the base comparison category, the findings indicate that with a unit increase in the decision-making power over the money given to a woman’s spouse’s parents ceteris paribus, the logarithm of the probability that the husband makes the full decision (decision is jointly made) in this category relative to the wife making the full decision in this category decreases by 0.246 (0.248). Equivalently, decisionmaking over food expenditures is only statistically significant for the joint decisionmaking category and shows that the odds of the decision being made jointly in this category versus the decision being made fully by the wife decreases by 0.088. Coefficient estimates are not significant for other categories of goods and services. In essence, the results point to the same baseline conclusion that women living in gold endowed municipalities were more likely to partake in intra-household decisionmaking processes compared to those that did not. As an additional robustness check, I conduct a series of sensitivity analyses on the main female decision-making power estimates presented in Table 1. First, I omit municipalities in three Northern states: Oaxaca, Chihuahua and Durango where gold mining activity is the most prominent (see Fig. 2) to ensure that the main results are not driven by mining intensive states in this region of Mexico’s geographical landscape. Altogether, the results presented in Table 6show that the baseline Male employment and female intra-household decision-making: a Mexican gold mining case. . . 721 estimate in column (2) of Table 1increased by 0.002 to 0.007 compared to the main estimate of 0.005, with coefficients remaining significant at the 5% level. Next, I exclude municipalities with high levels of violence, specifically those in the 75th percentile according to the homicide rate. I do so to check if more violent municipalities downward biased the main female decision-making power estimates given the positive correlation between gold mining and drug-related violence, and subsequently, the documented negative effect of violence on women’s intra-household bargaining power (see Tsaneva et al. 2018). The results from this analysis provide some indication that the main coefficient of a woman’s decision-making power in Table 1is underestimated, as excluding violent municipalities increases the estimated coefficient by 0.001 when compared to the baseline estimate of 0.005. This therefore suggests that in the absence of the confounding drug war, women residing in gold endowed municipalities would have had higher intra-household decision-making power than reported in Table 1. In a following robustness check, I include the interaction term of homicide rates, gold endowment and world gold prices. Albeit the movement of drug cartels into gold mining reportedly only intensified from 2012 onwards according to GIATOC (2016), one cannot rule out that the spike in gold mining opportunities during the sample period prior to 2012, was also accompanied by an increase in drug-related crime. Since excluding violent municipalities increases the main decision-making power estimate, and given the evidence presented in Table 3reflecting decreased perceptions of safety among men in the MxFLS, it is important to account for the Table 6 Sensitivity analysis of effect of the gold mining boom on female decision-making power (overall) (MxFLS data) (1) (2) Mean Exclude northern municipalities 0.001 (0.004) 0.007** (0.003) 0.539 Observations (N) 9225 8020 Exclude violent municipalities 0.003* (0.002) 0.006** (0.002) 0.540 Observations (N) 7580 6763 HR*Gendowment*Price 0.000 (0.001) 0.004* (0.003) 0.541 Observations (N) 10,462 9,123 World export value of gold 0.013* (0.007) 0.016* (0.008) 0.548 Observations (N) 7786 4664 Baseline controls No Yes Heterogeneous trends No Yes Individual FE Yes Yes Month FE Yes Yes Notes: Standard errors are clustered at the municipality level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and municipality covariates like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Column (1) does not include any controls, and column (2) add baseline controls and time trends interacted with municipal characteristics such as ruralurban status, transport network quality, community quality and the presence of ejidos (heterogeneous market trends). All specifications include individual and month fixed effects. ***p< 0.01 **p< 0.05 *p< 0.1 722 A. Au Yong Lyn Fig. 7 Coefficient plot for a woman’s relative decision-making power (DMP). Notes: This figure plots the coefficients of the regression of a woman’s decision-making power index on the lag of gold prices (t=1, 2,..,6). The orange ‘x’denotes the highest coefficient value at t-5, and shows a quadratic relationship between lagged price proxy values and a woman’s decision-making power estimate Table 7 Summary statistics of dependent variables Mean SD Min Max N Dependent Variable–Woman’s average decision-making power (DMP) over all goods: Mean DMP index (MxFLS data) 0.5 0.13 0 1 10,470 Mean DMP index (ENDIREH data) 0.7 0.21 0 1 191,768 Dependent Variables–Woman’s decision-making power over (individual decisions) (MxFLS data): Contraceptive use 0.5 0.3 0 1 7992 Spouse’s work choice 0.2 0.3 0 1 10,397 Own work choice 0.6 0.4 0 1 10,369 Money given to spouse’s family 0.3 0.3 0 1 7833 Money given to own family 0.5 0.4 0 1 7927 Large expenditures 0.4 0.3 0 1 10,295 Child’s health 0.5 0.3 0 1 8668 Child’s education 0.5 0.3 0 1 8469 Child’s clothes 0.6 0.4 0 1 7192 Spouse’s clothes 0.3 0.4 0 1 10,409 Own clothes 0.8 0.3 0 1 10,430 Food eaten in the house 0.8 0.3 0 1 10,388 Other Dependent Variables (MxFLS data): Male employment 0.8 0.4 0 1 8964 Female employment 0.2 0.4 0 1 10,478 Male fear of assault 0.3 0.5 0 1 9069 Female fear of assault 0.4 0.5 0 1 10,478 IPV outcomes (ENDIREH data): Emotional Abuse 0.3 0.5 0 1 168,787 Physical Abuse 0.1 0.3 0 1 192,011 Sexual Abuse 0.1 0.2 0 1 191,665 Threat of Violence 0.0 0.2 0 1 191,940 Notes: Summary statistics of dependent variables. The number of observations (N) for decision-making power questions vary, since not all women in the sample have children, and some decisions are neither made by the husband or the wife Male employment and female intra-household decision-making: a Mexican gold mining case. . . 729 Table 8 Summary statistics of independent variables Mean SD Min Max N Independent Variables (MxFLS data): Indigenous 0.1 0.3 0 1 10,478 Presence of spouse during interview 0.1 0.3 0 1 10,455 Age 43.6 12.9 15 101 7990 Age squared 2074 1253 225 10,201 7990 Years of education 9.6 4.3 0 22 10,466 Number of children 1.6 1.5 0 9 10,478 Homicide rate 0.5 0.6 0 3.4 10,478 Assist Law 0.7 0.4 0 1 10,478 Divorce law 0.7 0.5 0 1 10,478 Penal Code 0.9 0.3 0 1 10,478 Independent Variables (ENDIREH data): Rural status 0.2 0.4 0 1 192,055 Indigenous 0.1 0.2 0 1 191,711 Age 40.9 13.8 15 104 191,849 Age squared 1860.1 1253.3 225 108,816 191,849 Length of relationship 21.3 14.0 1 83 189,142 Number of children 3.3 2.3 0 25 190,410 Homicide rate 15.9 19.4 2.0 130.6 192,055 Assist Law 0.8 0.4 0 1 192,061 Divorce law 0.8 0.4 0 1 192,061 Penal Code 0.9 0.3 0 1 192,061 Notes: Summary statistics of independent variables 730 A. Au Yong Lyn Table 9 The effect of the gold mining boom on female decision-making power (other goods and services) (MxFLS data) (1) (2) Mean (std. dev.) Normalized DMP (individual goods): Own clothes −0.002 (0.005) −0.002 (0.007) 0.829 Observations (N) 10,419 9087 Spouse’s clothes −0.007 (0.008) −0.007 (0.010) 0.538 Observations (N) 7125 6144 Child’s clothes 0.000 (0.008) 0.003 (0.012) 0.461 Observations (N) 5880 5092 Child’s education 0.009 (0.007) 0.008 (0.006) 0.516 Observations (N) 7779 6774 Child’s health 0.003 (0.010) 0.007 (0.009) 0.534 Observations (N) 8049 7014 Spouse’s work choice −0.006 (0.009) −0.001 (0.010) 0.208 Observations (N) 10,389 9053 Own work choice 0.004 (0.006) 0.008 (0.006) 0.594 Observations (N) 10,358 9026 Money given to spouse’s parents −0.003 (0.010) 0.008 (0.008) 0.292 Observations (N) 7000 6029 Household expenditures 0.004 (0.006) 0.011 (0.007) 0.386 Observations (N) 10,266 8953 Baseline controls No Yes Heterogeneous trends No Yes Individual FE Yes Yes Month FE Yes Yes Notes: Standard errors are clustered at the municipality level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and municipality covariates like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Column (1) does not include any controls, and column (2) adds baseline controls and time trends interacted with municipal characteristics such as rural-urban status, transport network quality, community quality and the presence of ejidos (heterogeneous market trends). All specifications include individual and month fixed effects. ***p< 0.01 **p< 0.05 *p< 0.1 Male employment and female intra-household decision-making: a Mexican gold mining case. . . 731 Table 10 Descriptive statistics: ENDIREH survey Postt=1 (T) Postt=0 (C) Individual Covariates: Rural 0.196 0.229 Indigenous 0.059 0.092 Length of relationship 21.439 20.708 Age 41.072 39.875 Age squared 1877.746 1769.333 Number of kids 3.181 3.624 Education level 3.579 4.145 State Covariates: Homicide rate 17.366 8.728 Assist law 0.867 0.589 Divorce law 0.861 0.518 Penal Code 0.894 0.822 Observations 160,750 31,305 Notes: ‘T’represents the treated group belonging to the 2006 and 2011 ENDIREH survey and ‘C’represents the control group surveyed in 2003. The table reports mean values for each covariate and group Table 11 The effect of the gold mining boom on female decision-making power (other goods and services) (ENDIREH data) (1) (2) Mean (std. dev.) Normalized DMP (individual goods): Own work choice 0.000 (0.002) 0.004 (0.002) 0.829 Observations (N) 171,794 167,236 Expenditures on own goods 0.003 (0.002) 0.000 (0.002) 0.538 Observations (N) 187,334 182,193 Children’s goods 0.001 (0.001) 0.001 (0.001) 0.461 Observations (N) 162,143 158,935 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the municipality level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p< 0.05 * p<0 732 A. Au Yong Lyn Table 12 The effect of the gold mining boom on intimate partner violence (IPV) with wild-cluster bootstrapped standard errors (ENDIREH data) (1) (2) Mean Emotional Abuse 0.002 [0.004] 0.007** [0.008] 0.297 Observations (N) 168,787 163,902 Threat of Violence 0.001** [0.001] 0.001* [0.001] 0.029 Observations (N) 191,940 186,666 Physical Abuse 0.002** [0.001] 0.003** [0.003] 0.080 Observations (N) 192,011 186,730 Sexual Abuse 0.003*** [0.002] 0.004*** [0.003] 0.073 Observations (N) 191,665 186,413 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state-level controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p< 0.05 * p< 0.1 Table 13 The effect of mining on the incidence of physical intimate partner violence (IPV) (ENDIREH data) (1) (2) Mean Pushed you/Pulled your hair 0.001** (0.001) 0.002 (0.001) 0.061 Observations (N) 191,999 186,720 Tied you up −0.000* (0.001) −0.000 (0.000) 0.002 Observations (N) 191,961 186,684 Kicked you 0.001** (0.001) 0.001 (0.001) 0.019 Observations (N) 191,976 186,698 Hit you with his hands 0.003*** (0.001) 0.002** (0.001) 0.047 Observations (N) 191,979 186,702 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state-level controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p< 0.05 * p< 0.1 Male employment and female intra-household decision-making: a Mexican gold mining case. . . 733 Table 14 The effect of mining on the incidence of sexual intimate partner violence (IPV) (ENDIREH data) (1) (2) Mean Forced sexual relations 0.001** (0.001) 0.001* (0.000) 0.022 Observations (N) 191,607 186,359 Demand sex 0.003*** (0.001) 0.004*** (0.000) 0.069 Observations (N) 191,648 186,397 Used force to have sex 0.002*** (0.001) 0.002*** (0.000) 0.023 Observations (N) 191,530 186,287 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state-level controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p< 0.05 * p< 0.1 Table 15 The effect of mining on the incidence of the threat of violence (IPV) (ENDIREH data) (1) (2) Mean Threatened with weapon 0.001** (0.000) 0.001* (0.000) 0.014 Observations (N) 191,958 186,684 Threatened to kill you 0.001** (0.001) 0.001 (0.001) 0.025 Observations (N) 191,943 186,668 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state-level controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p< 0.05 * p< 0.1 734 A. Au Yong Lyn Table 16 The effect of mining on the incidence of the emotional intimate partner violence (IPV) (ENDIREH data) (1) (2) Mean Destroyed things in house 0.002** (0.001) 0.002** (0.001) 0.046 Observations (N) 191,965 186,692 Threatened to leave 0.002** (0.001) 0.002** (0.001) 0.087 Observations (N) 191,973 186,700 Not allowed to leave home 0.000 (0.000) 0.001* (0.000) 0.036 Observations (N) 191,967 186,964 Accused you of cheating 0.001 (0.001) 0.003** (0.001) 0.069 Observations (N) 191,940 186,667 Made you feel fearful 0.003*** (0.001) 0.002** (0.001) 0.082 Observations (N) 191,954 186,681 Ignored you 0.003** (0.001) 0.003 (0.002) 0.096 Observations (N) 191,963 186,691 Baseline controls No Yes Heterogeneous trends No Yes State FE Yes Yes Survey year FE Yes Yes Notes: Standard errors are clustered at the state level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and state-level controls like the homicide rate per 100,000 inhabitants. Column (1) does not include any controls, and column (2) adds baseline controls and heterogeneous market trends. All specifications include state and survey year fixed effects. *** p< 0.01 ** p<0.05*p<0.1 Table 17 The effect of mining on female decision-making power (multinomial logit), with full decisionmaking power as the base category (MxFLS data) Husband makes full decision Joint decision-making (1) (2) (3) (4) Decision-making power over: Food Expenditures −0.089 (0.076) −0.084 (0.099) −0.063* (0.034) −0.088** (0.041) Observations (N) 5504 4786 5504 4786 Money to spouse’s parents −0.232** (0.102) −0.246* (0.130) −0.206** (0.102) −0.248* (0.129) Observations (N) 3784 3276 3784 3276 Baseline controls No Yes No Yes Heterogeneous trends No Yes No Yes Individual FE Yes Yes Yes Yes Month FE Yes Yes Yes Yes Notes: Standard errors are clustered at the municipality level and reported in parentheses (.). Baseline controls include a woman’s age, age squared, indigenous status, education level, number of children, the presence of her spouse during the interview, and municipality covariates like the homicide rate per 100,000 inhabitants and the introduction of domestic violence and divorce laws. Columns (1) and (3) do not include any controls, and columns (2) and (4) add baseline controls and time trends interacted with municipal characteristics such as rural-urban status, transport network quality, community quality and the presence of ejidos (heterogeneous market trends). All specifications include individual and month fixed effects. ***p< 0.01 **p< 0.05 *p< 0.1 Male employment and female intra-household decision-making: a Mexican gold mining case. . . 735 References African Union. (2009). Africa mining vision. Addis Ababa: AU. Ajzenman, N., Galiani, S., & Seira, E. (2015). On the distributive costs of drug-related homicides. The Journal of Law and Economics,58(4), 779–803. Angelucci, M. (2008). Love on the rocks: Domestic violence and alcohol abuse in rural Mexico. The BE Journal of Economic Analysis & Policy, 8(1). Aragón, F. M., Rud, J. P., & Toews, G. (2018). Resource shocks, employment, and gender: evidence from the collapse of the UK coal industry. Labour Economics,52,54–67. INEGI. (2020). Censos Economicos. Sistema Authomatizado de Informacion Censal (SAIC). Mexico City: INEGI. Arcos, A, R Guzmán, Estefanía. (2018). Brechas de género en la minería. La minería artesanal y de pequeña escala colombiana desde una perspectiva de género. Envigado (Antioquia): Alianza por la Minería Responsable. https://www.responsiblemines.org/wp-content/uploads/2019/02/DIGITALBRECHAS-DE-G%C3%89NERO_M21_C13-compressed.pdf. (assessed 22 April 2020). Beleche, T. (2017). Domestic violence laws and suicide in Mexico. Review of Economics of the Household,17(1), 229–248. Bhalotra, S., Kambhampati, U., Rawlings, S., & Siddique, Z. (2020). Intimate Partner Violence: The influence of job opportunities for men and women. Policy Research Working Paper; No. 9118. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/33232. License: CC BY 3.0 IGO. Bobonis, G. J. (2009). Is the allocation of resources within the household efficient? New evidence from a randomized experiment. Journal of Political Economy,117(3), 453–503. Bobonis, G. J., González-Brenes, M., & Castro, R. (2013). Public transfers and domestic violence: The roles of private information and spousal control. American Economic Journal: Economic Policy,5(1), 179–205. Bridges, J. O. (1980). The Mexican family. In M. S. Das & C. J. Jesser (Eds.), The Family in Latin America (pp. 295–334). New Delhi: Vikas Publishing House. Brown, R., & Velásquez, A. (2017). The effect of violent crime on the human capital accumulation of young adults. Journal of Development Economics,127(2017), 1–12. Brown, R., Montalva, V., Thomas, D., & Velásquez, A (2017). Impact of violent crime on risk aversion: Evidence from the Mexican drug war. No. w23181. National Bureau of Economic Research. Browning, M., Chiappori, P. A., & Lechene, V. (2006). Collective and unitary models: a clarification. Review of Economics of the Household,4(1), 5–14. Cameron, A. C., Gelbach, J. B., & Miller, D. L. (2008). Bootstrap-based improvements for inference with clustered errors. The Review of Economics and Statistics,90(3), 414–427. Cherchye, L., De Rock, B., & Vermeulen, F. (2012). Married with children: A collective labor supply model with detailed time use and intrahousehold expenditure information. American Economic Review,102(7), 3377–3405. Chiappori, P. A., Fortin, B., & Lacroix, G. (2002). Marriage market, divorce legislation, and household labor supply. Journal of Political Economy,110(1), 37–72. El Congreso del Estado de Michoacán (2019). Gobierno Federal vulnera a mujeres al quitar subsidios a estancias infantiles y refugios: Lucila Martínez. http://congresomich.gob.mx/gobierno-federalvulnera-a-mujeres-al-quitar-subsidios-a-estancias-infantiles-y-refugios-lucila-martinez/. (assessed 23 April 2020). Ellsberg, M., Heise, L., Pena, R., Agurto, S., & Winkvist, A. (2001). Researching domestic violence against women: Methodological and ethical considerations. Studies in Family Planning,32(1), 1–16. Engineering and Mining Journal: Global Business Reports (2011). “Mexican Mining”.www.gbreports. com/wpcontent/uploads/2014/12/Mexico_Mining2011.pdf. Gelles, R. (1976). Abused wives: Why do they stay? Journal of Marriage and the Family.38(4), 659–668. Global Initiative against Transnational Organized Crime (GIATOC) (2016). Organized crime and illegally mined gold in Latin America. https://arcominero.infoamazonia.org/GIATOC-OC_Illegally-MinedGold-in-Latin-America-3c3f978eef80083bdd8780d7c5a21f1e.pdf. Gutmann, M. C., & Viveros, M. (2005). Handbook of studies on men and masculinities: Masculinities in Latin America (pp. 114–128). Thousand Oaks, CA: Sage Publications. Heath, R. (2014). Women’s access to labor market opportunities, control of household resources, and domestic violence: Evidence from Bangladesh. World Development,57,32–46. 736 A. Au Yong Lyn Heise, L. L., & Kotsadam, A. (2015). Cross-national and multilevel correlates of partner violence: an analysis of data from population-based surveys. The Lancet Global Health,3(6), e332–e340. Hidrobo, M., Peterman, A., & Heise, L. (2016). 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(3), 284–303. Hurtado, A. C., & Salazar, M. T. S. (1999). México: globalización y espacios mineros. Baética: Estudios de arte, geografía e historia,21,35–56. Klasen, S., & Pieters, J. (2015). What explains the stagnation of female labor force participation in urban India? The World Bank Economic Review,29(3), 449–478. Kotsadam, A., Østby, G., & Rustad, S. A. (2016). Structural change and wife abuse: A disaggregated study of mineral mining and domestic violence in sub-Saharan Africa, 1999–2013. Political Geography, 56,53–65. Kotsadam, A. & Tolonen, A. (2016). African mining, gender, and local employment. World Development, 83, 325–229. Landsberger, H. A. (1957). Hawthorne Revisited: A Plea for an Open City. Cornell University. Lommerud, K. E. (2003). “Battles of the sexes: non-cooperative games in the theory of the family.”In Economics of the family and family policies,46–60. Routledge. Lundberg, S., & Pollak, R. A. (1994). Noncooperative bargaining models of marriage. The American Economic Review,84(2), 132–137. Lutz-Ley, A. N., & Buechler, S. J. (2020). Mining and women in northwest Mexico: a feminist political ecology approach to impacts on rural livelihoods. Human Geography,13(1), 74–84. Macmillan, R., & Gartner, R. (1999). When She Brings Home the Bacon: Labor Force Participation and the Risk of Spousal Violence Against Women. Journal of Marriage and the Family,61(4), 947–958. Mancera, B. M., Dorgo, S., & Provencio-Vasquez, E. (2017). Risk factors for Hispanic male intimate partner violence perpetration. American Journal of Men’s Health,11(4), 969–983. Maurer, S. E., & Potlogea, A. V. (2020). Male‐biased demand shocks and women’s labour force participation: evidence from large oil field discoveries. Economica.https://doi.org/10.1111/ecca.12341. McElroy, M. B., & Horney, M. J. (1981). Nash-bargained household decisions: toward a generalization of the theory of demand. International Economic Review, 333–349. Neff, D. F., Sen, K., & Kling, V. (2012). The puzzling decline in rural women’s labor force participation in India: a reexamination. SSRN Scholarly Paper ID 2066585. Rochester, NY: Social Science Research Network. Observatory of Economic Complexity (OEC). oec.world/en/ (assessed 21 November 2019). Penman, M. (2016). Ambivalent company attitudes and how they shape conflict: Mining conflicts in Mexico’s ejidos. The Extractive Industries and Society,3(3), 754–761. Rangel, M. A. (2006). Alimony rights and intrahousehold allocation of resources: evidence from Brazil. The Economic Journal,116(513), 627–658. Ross, M. (2008). Oil, islam, and women. American Political Science Review,102(1), 107–123. Rustad, S. A., Østby, G., & Nordås, R. (2016). Artisanal mining, conflict, and sexual violence in Eastern DRC. The Extractive Industries and Society,3(2), 475–484. S&P Global Market Intelligence (2019). PDAC Special Edition: World Exploration Trends 2018. https:// www.spglobal.com/marketintelligence/en/documents/world-exploration-trends-march-2019.pdf. (assessed 16 April 2020). Samuelson, P. A. (1956). Social indifference curves. The Quarterly Journal of Economics,70(1), 1–22. Secretaría de Economía. “Minería.”(2013). https://www.gob.mx/se/acciones-y-programas/mineria. (assessed 25 September 2019). Sitio De Servicio Geológico Mexicano (SGM) (2020). https://www.gob.mx/sgm. (assessed 22 April 2020). Stevenson, B., & Wolfers, J. (2006). Bargaining in the shadow of the law: Divorce laws and family distress. The Quarterly Journal of Economics,121(1), 267–288. Tolonen, A. (2018). Endogenous gender norms: evidence from Africa’s gold mining industry’.Research Paper Series 209. Tsaneva, M., Rockmore, M., & Albohmood, Z. (2018). The effect of violent crime on female decision making within the household: evidence from the Mexican war on drugs. Review of Economics of the Household,17(2), 615–646. Velásquez, A (2010). The economic burden of crime: Evidence from Mexico. Economist. Male employment and female intra-household decision-making: a Mexican gold mining case. . . 737