We sent Professor Deniz Sanin a series of questions about her research, focusing on the employment of women within the coffee industry in Rwanda, exploring connections between women's paid work and domestic violence, as well as examining broader aspects of Rwanda's socio-economic landscape.
Thank you so much for taking the time to speak with us today. To start off, could you please introduce yourself and your research background?
Hi! Thank you for having me! I am an Assistant Professor of Economics at the University of South Carolina. I am currently visiting the Economics Department at Harvard University in Spring 2024. I received my Ph.D. in Economics from Georgetown University in May 2022. During the last year of my Ph.D., I was also a research fellow in the Women and Public Policy Program at the Harvard Kennedy School. I did my master’s in economics at Duke University. Although I have been in the US since 2013, I was born and raised in Istanbul, Turkey, and did my undergraduate there (major in Economics and minor in Mathematics)
My research investigates how:
the structure of and changes in the labor markets,
laws, government policies, and institutional environment
affect individual and household decisions, primarily in developing but also in developed countries.
I am particularly interested in studying domestic violence and women's employment. By studying the effects of labor markets and institutions on women's lives, my research aims to understand the determinants of gender inequality and potential policies to reduce it. Most of my work combines administrative and spatial data with household surveys and uses both theory and natural experiments.
Impact of The Coming up of Coffee Mills on Women’s Employment in Rwanda
What made you choose Rwanda as the geographical location of your research?
I was already working on the country. I have another paper, “Do Domestic Violence Laws Protect Women from Domestic Violence? Evidence from Rwanda” (I am currently editing it, so the title is subject to change!), which studies the domestic violence legislation in the country. In 2008, Rwanda became the first country in Sub-Saharan Africa to pass a comprehensive domestic violence law (Hebert, 2015), and I study the effect of the law on couples in the paper. While working on the paper as a Ph.D. student at Georgetown, one day, I saw a bag of coffee from Rwanda in Starbucks. That sparked my interest because I know that Rwandan women work in the coffee industry, and I wanted to work on the effect of women’s employment on domestic violence. After Googling Rwandan coffee, I met with their amazing coffee industry!
How prevalent is domestic violence in the world and in Rwanda?
About 1 in every 3 women worldwide have experienced physical and/or sexual violence from their partners in their lifetime (World Bank, 2015). 41.5% partnered women reported ever experiencing DV as of 2019 in Rwanda (Rwanda National Institute of Statistics, 2019).
Can you walk us through what Rwanda's National Coffee Strategy in 2002 was? What was the economic and political context in which it was launched? How did it change the coffee industry in the country?
In 2000, Paul Kagame came into power and prioritized economic growth to rebuild the country after the Rwandan Genocide (1994). He launched the Vision 2020 program in 2000 (Boudreaux, 2011). The program outlined a list of goals that the government aimed to achieve by 2020. One of the main goals was to transform agriculture into a high-value sector. In light of this goal, the government adopted the National Coffee Strategy in 2002, which aimed to shift to high-quality, wet-processed coffee production to participate in the international specialty coffee market (Boudreaux, 2011). A coffee mill is where coffee cherries, harvest of the coffee tree are processed into high-quality coffee beans so that they can be prepared for export. Between 2000-2006, governmental institutions collaborated with USAID, universities in Rwanda and U.S., and private sector partners under the Partnership to Enhance Agriculture in Rwanda through Linkages (PEARL) project. The project helped farmers establish cooperatives, find loans, and build mills in their communities in a few locations. After the project, farmers continued establishing cooperatives and building mills in their communities nationwide. From 2002 to 2012, the number of mills increased from 5 to 213.
What are some characteristics of the coffee mills that came up in the aftermath of this policy? How did that affect your research? How did this transform the employment and income of households? How did this transformation impact men and women separately? Why do women tend to be more likely to be employed within these coffee mills than men? What were the implications of this?
A mill opening provides two key features for my identification strategy (to find the causal effect of a mill opening on couples): Spatial and Time.
Spatial Variation
A mill only serves coffee farmers (mostly couples) that reside within its catchment area, a buffer zone around the mill. The reason why a mill has a specific catchment area is that coffee cherries will rot if not transported to a mill within a few hours of harvest. To visualize, the circle is the catchment area, and the farmer Couple X is in the catchment area.
Time Variation
A mill also creates a time variation: before and after a mill opening. Both before and after a mill, there is a gendered division of labor in the coffee value chain, and women engage in multiple labor-intensive tasks.
Before a mill opening, the wife harvests the cherries with the husband. Then, she processes the harvest as an unpaid family worker at home as a female-dominant task. The husband sells the home-processed, thus low-quality coffee in the local market for a low price as a male-dominated task.
After a mill opening, there is no change in harvesting. The husband sells the cherries to the mill for a high premium because mill-processed coffee yields a high-quality product that is sold in the international market. The wife does some of the processing tasks at the mill as a wage worker. So, based on the variations, the mill provides wage employment opportunities for the wife and increases the cost of the wife’s incapacitation for the husband.
I also would like to clarify that there are 2 time variations I use in the paper:
Yearly time variation: Before and after a mill opening (yearly domestic violence data, self-reported)
Monthly time variation: Mills operate during the harvest season, March-July. Before and after the operation of mills (monthly domestic violence hospitalizations)
The Relationship between Domestic Violence and Women’s Employment
Existing literature mentions that women’s employment could both increase and decrease their exposure to domestic violence. Could you walk us through the reasons for both? Do you think there are gaps in literature that currently tries to analyze this link between women’s employment and domestic violence, and why is that so?
This is a very important question because I think if we focus on the mechanisms, then we can understand the relationship between women’s employment and domestic violence better and design better policies against it.
Based on theories on women’s income and DV, providing jobs to women can decrease violence either via an increase in women’s outside options and/or a decrease in the financial stress in the household. This can be thought of as an income effect. On the other hand, it can increase violence due to husbands’ increased incentives to extract women’s resources or through male backlash. There is evidence for both channels in the literature. Yet the studies in the literature either focus on the effects of unearned income such as cash transfers and dowry payments (Angelucci 2008, Bobonis et al. 2013, Hidrobo et al. 2016, Haushofer et al. 2019, Bloch and Rao 2002, Calvi and Keskar 2021 ), or they study employment mainly as an income shock due to their context, source of variation or data (Aizer 2010, Anderberg et al. 2016, Bhalotra et al. 2021., Erten and Keskin 2021a, b, Kotsadam and Villanger 2022.)
Yet, a job is more than income. Recent research suggests that it provides psychosocial benefits and confidence (Hussam, Kelley, Lane and Zahra 2022, Hussam, Kelley, Lane 2023, McKelway 2023.) It decreases the exposure between couples (less time for violence) (Dugan et al., 1999). Moreover, a job is earning income using capacity to work, or health as human capital. (Becker 1962, 2007; Goldin 2016.) This highlights the fact that domestic violence can create economic costs (Adams-Prassl, Huttunen, Nix, and Zhang 2023). This raises the question of whether domestic violence decreases when the husband has economic self-interest in his wife’s work capacity (it is costly for the husband to incapacitate his wife). I find evidence supporting this mechanism in my paper, on top of the increase in the wife’s outside options (and thus bargaining power) and contribution to household resources. I ruled out exposure reduction as a mechanism in my context based on my data.
In the context of Rwanda, how did women’s employment in coffee mills impact women’s exposure to domestic violence? For instance, how did the reasons that meant husbands resorted to domestic violence prior to the establishment of the mills change in light of women’s employment? You mention that most of the community works in mills primarily during May, June and July. Does this seasonal nature of employment mean that the impacts of women’s employment on domestic violence are also seasonal in nature? Or are they more long-term, in which case, why?
According to my results:
Partnered women in the catchment areas (CAs) are 15% more likely to work for cash, 29% less likely to self-report DV in the past 12 months. I also find that mill exposure increases the earnings of each spouse.
As a unique feature of the paper, I use the universe of monthly hospital records on domestic violence in the country. I also show that Hospitals in the CAs are 14% less likely to have a DV patient in a harvest month compared to one month before the harvest season. Moreover, right after the harvest season, when the cost of incapacitation decreases, DV hospitalizations revert to their pre-harvest level.
It is important to note that although there is no change in domestic violence hospitalizations among the hospitals in the catchment area of a mill after the harvest season relative to the pre-harvest (there is a reversion), the post-harvest levels of hospitalizations are still lower compared to the cases from the hospitals outside of the catchment areas.
What are the mechanisms behind the results of your study?
Below are the potential mechanisms behind a decrease in DV. I test each one in the paper and either find supporting evidence for it or rule it out based on my data.
1. Increase in women’s outside options
2. Exposure Reduction
3. Increase in household earnings: Due to women’s and/or husband’s earnings
4. Increase in the cost of women’s incapacitation
First, I find that upon a mill opening, partnered women in the catchment areas are more likely to make decisions on large household purchases by themselves or jointly with their husbands relative to their husbands making the decision for them. This provides suggestive evidence that the mill opening increases women’s bargaining power. I find similar results for contraception usage.
A common question I receive is that the decline in DV can be due to the reduction in the time couples spend together, given that women now go to the mill to work. In order to test this exposure reduction mechanism, I use couples with plausibly no change in exposure to each other before and after a mill. Those are women working in agriculture with husbands in non-agricultural manual jobs, like plumbers, construction workers, etc. Here, the mill is a shock to women’s earnings, not to the couple's shared time. Those couples are already not seeing each other during work hours before a mill, and they continue to do so after a mill opening. Using time use data, I also check that there is no change in husbands’ time spent at work, and women do not report a higher number of hours spent at work. According to results, results on working for cash and a decrease in DV remain where the decline in DV is comparable to the overall decline in DV. This suggests that exposure reduction is not the dominant mechanism behind the results.
To test the increase in household earnings mechanism, I continue to focus on couples with different occupations. Among these couples, an increase in earnings for women is coupled with no change in husbands’ earnings. Recall that there is a decline in DV among those couples as well. Thus, there is a decline in DV even among couples whose household earnings increased only due to women’s paid employment. This suggests that the increase in the wife's contribution to household earnings is also a plausible mechanism.
To distinguish the incapacitation cost mechanism, I revisit the hospitalization results and use monthly consumption data (results summarized in graphs below). Three key points on this. First, based on the DV hospitalization results, when the incapacitation cost mechanism decreases in August, DV reverts to its pre-harvest level. Second, one month after the harvest season, August, mill-exposed households continue to enjoy a higher consumption relative to the pre-harvest, plausibly due to saving harvest income derived by selling cherries to the mill and the wife’s wages from the mill. Yet, in August, there is no change in the domestic violence hospitalizations compared to the pre-harvest, when mills do not operate. This rules out the income effect for DV hospitalizations, severe DV cases. Third, plausibly, the wife’s outside option (which I take as the utility of being divorced here) in the catchment area is the same in July-August. Women can have savings from their wages; moreover, the husband knows that his wife will have a job opportunity the following year. These 3 points suggest that the seasonality of the cost of women’s incapacitation plausibly drives the seasonality of DV hospitalizations.
Moreover, using data on crops, I show that when the value of the wife’s work capacity is fixed within the year, there is no change in DV hospitalizations within the year, which is in line with the seasonality argument. For this, I use Irish potato regions where there is no major change in the value of women’s work capacity for their unpaid tasks within the year and women mostly do not contribute to household income via paid work. No change in hospitalizations for Irish-potato suitable regions within the year further confirms the incapacitation mechanism.
To wrap up, I provide evidence for three channels.
1. Increase in women’s outside options
2. Increase in women’s contribution to household resources
3. Increase in the cost of women’s incapacitation
Since these channels are active via women’s employment, the paper first concludes that the decline in DV is plausibly driven by women’s employment. As a second conclusion, an increase in the cost of women’s incapacitation mechanism operates beyond other channels.
Is divorce a viable alternative for women undergoing domestic violence in Rwanda, both legally and due to social norms? How does women earning their own income affect their ability to ask for divorce? What is the impact of this on domestic violence?
Yes, divorce is a viable alternative in the Rwandan context. First, based on the laws, according to my previous work, women use their right to divorce their husbands unilaterally if their husbands engage in DV in Rwanda (Sanin, 2021). Divorce rates are higher in Rwanda compared to India as well, which may suggest that getting a divorce may be less of a taboo in the country compared to the Indian context.
[One thing to note: I did not find that a mill opening increases the probability of women being divorced.]
Let me highlight that divorce being a viable alternative is very important in this topic. Theoretically, based on my model, providing jobs to women decrease DV when:
The woman’s threat of divorce is credible based on laws, social norms.
The husband benefits from women’s work capacity.
If divorce is not a credible threat in the context, the job will not increase the woman’s outside option (bargaining power) in the first place. Then, the husband’s incentive to extract the wife’s earnings can either dominate or offset the gains from an increase in the woman’s outside option.
Do you think you would recommend methods of increasing female employment to be considered as a strategy across developing countries for combating the prevalence of domestic violence for policymakers and non-profit organizations working towards this goal?
Based on my evidence on the mechanisms, women’s employment may affect DV via the value of work capacity, not just via income. This complements Hussam et al. (2022). My findings have policy implications which suggests that providing job opportunities to women has more potential to decrease domestic violence compared to a cash transfer in some contexts.
BUT, do the results indicate that providing jobs to women decreases DV in every context? – Unfortunately, NO. As a very recent example, Kotsadam and Villanger (2022) found no effects of providing factory jobs to women on physical DV in Ethiopia in an RCT.
Yet, the mechanisms in this paper shed light on why different contexts yield different results and what the potential conditions for the employment policy to work may be. As aforementioned, according to my theoretical framework, providing jobs to women decreases DV when:
The woman’s threat of divorce is credible based on laws, social norms.
The husband benefits from women’s work capacity.
Both conditions hold for Rwanda. To support the first point, women use their right to divorce their husbands unilaterally if their husbands engage in DV (Sanin, 2021), and divorce rates are higher compared to Ethiopia. To support the second point, couples work together in my context (there is an employer (husband) - employee (wife) relationship within the couples). Also, the wife shares her earnings with the husband: Based on data, a higher share of women decide jointly with their husbands on how to use their earnings compared to Ethiopia. These points suggest that the institutions of a country (formal and informal) and the structure of the job also matter when we are designing policies to combat domestic violence.
Note from Deniz Sanin: All of the references in this script can be found at https://denizsanin.github.io/MyWebsite/Coffee_DenizSanin.pdf. I formed the script (answers) based on the paper, which was written by me.