Women’s Participation in Contract Farming

Abstract Smallholder farmers in lower-income countries often lack access to agricultural inputs, services, and markets. This holds especially for female farmers, with important negative implications for agricultural productivity, child welfare, and rural development. Contract farming is promoted as a means to improving farmers’ access to inputs, services, and markets – and thereby household income and welfare. Could contract farming also reduce prevalent gender disparities? And does it matter who within the household holds the contract? Here, we address these questions and explore patterns, drivers, and implications of women’s participation in contract farming. For this purpose, we use a unique dataset that is nationally representative of smallholder farmers in five African countries, which is the exception in this literature. Moreover, the data allow us to differentiate between different forms of women’s participation in contract farming, which is also an exception in this literature. We differentiate between female-headed and male-headed households and the gender of the contract holder. We find that participation rates among women are lower than those among men – but higher than previous case studies suggest. Our results regarding the importance of the gender of the contract holder for household living standards are inconclusive, for both male-headed and female-headed households, and there is great heterogeneity across countries. We conclude that the topic merits further exploration and discuss directions for future research and implications for policy.


Introduction
Achieving gender equality is a key development goalbut remains a great challenge (UN, 2019). This holds especially for rural areas of lower-income countries where traditional gender norms, gender-based violence, early marriage, and poor education opportunities are often commonplace (Raney, Croppenstedt, Anriquez, & Lowder, 2011;UN, 2019). A large share of the population in these regions depends on agriculture for their livelihoods. Female farmers are often particularly disadvantaged in terms of access to agricultural inputs, services, and markets (Quisumbing et al., 2015;Raney et al., 2011;World Bank, 2011), limiting their agricultural productivity and hence the food available for farm households and markets (Raney et al., 2011). Increasing women's agricultural productivity and income are goals in their own rights, but also key to improving other dimensions of household and child welfare and, thus, economic opportunities for future generations (Doss, 2013;Duflo, 2003;Thomas, 1990;World Bank, 2011) Here, we ask whether contract farming schemes that contract female farmers could help increase women's access to agricultural inputs, services, and marketsthereby increasing women's as well as households' income and welfare. Contract farming refers to a pre-harvest agreement between farmers and buyers (Little & Watts, 1994) and has received renewed attention in the scientific literature and among policymakers, given its potential to help address common market failures facing smallholder farmers (Maertens & Swinnen, 2012;Mishra, Kumar, Joshi, & D'Souza, 2018;Ruml & Qaim, 2021). Contract farmers are often offered farm inputs (typically on credit) and extension services that might otherwise be unavailable but are crucial to increase the quantity and quality of farm output. Similarly, agreements regarding output prices and quantities, which are often part of contracts, can reduce risks and increase planning security for farmers (Barrett et al., 2012;Bellemare & Lim, 2018) Most studies support the notion that contract farming can improve farmers' productivity, income, and overall welfare (Bellemare & Bloem, 2018). This does, however, not hold true in all situations (Mwambi, Oduol, Mshenga, & Saidi, 2016;Olounlade et al., 2020;Ragasa, Lambrecht, & Kufoalor, 2018). For example, contract farming can increase vulnerability to market shocks or contract breach by contractors (Oya, 2012). More generally, contracts differ in terms of their features, which can lead to heterogeneous outcomes (Ruml & Qaim, 2020. Additionally, and importantly in the context of this study, agricultural technologies, policies, and value chain developments can affect women and men differently, given the gendered distribution of tasks, responsibilities, and decision-making-power within households and societies (Doss, 2001;Maertens & Swinnen, 2012;Oduol et al., 2017). Thus, a closer look at gender and contract farming is relevant, also for policy.
Here, we contribute to the literature on contract farming (see above), gender and contract farming in particular (Adams, Gerber, & Amacker, 2019;Mitra & Rao, 2021;Navarra, 2019;von B€ ulow & Sørensen, 1993), and related bodies of literature on female farmers' entrepreneurship (Dohmwirth & Liu, 2020;Lecoutere, 2017) and participation in higher-value agrifood value chains (Coles & Mitchell, 2011;Fischer, Patt, Ochieng, & Mvungi, 2020;Oduol et al., 2017;Oduol & Mith€ ofer, 2014) The evidence on gender and contract farming is limited and overwhelmingly qualitative (Adams et al., 2019;Mitra & Rao, 2021;Navarra, 2019;von B€ ulow & Sørensen, 1993) and focuses on intra-household bargaining and the allocation of household labor. Several studies suggest that women are often heavily involved in the work related to the contract crop, without receiving or controlling the same share of the generated income (Dolan, 2002), which their husbands may use for personal purposes such as financially supporting a second wife (von B€ ulow & Sørensen, 1993) Women's labor input can, however, also increase their bargaining power, control over income, and self-esteem (Adams et al., 2019;Raynolds, 2002), sometimes leading to greater remuneration for their labor (Carney, 1988). Contract farming may also promote the adoption of labor-saving technologies, leading to reduced labor demand and allowing women to engage in other, for example, non-farm activities (Debela, Ruml, & Qaim, 2022;Ruml & Qaim, 2021).
Thus, the literature has broadened our understanding of how contract farming can affect the allocation of household labor and overall household income. Yet little attention has been paid to the gender of the contract holder and whether it matters for household welfare. This is partly attributable to data limitations, as most (especially quantitative) studies are based on data collected among (male) household heads and without information on the gender of the contract holder. Relatedly, the fact that women might be the contract holders (even in male-headed households) might have been largely overlooked because it is more common that men are the contract holders (Dolan, 2001;Maertens & Swinnen, 2012). Here, we address this gap using unique, nationally representative data from five African countries, which include information on who within the household holds the contract. More specifically, we address the following issues: First, we explore gendered patterns in participation rates as well as determinants of participation in contract farming, differentiating by the gender of the household head and gender of the contract holder. Secondly, we explore whether the gender of the contract holder and gender of the household head matter for household living standards (here proxied using poverty levels). We provide hypotheses and more detailed conceptual considerations in the next section. To compare different groups of households (that is, male and female-headed households without contract holders, with only male contract holders, with only female contract holders, and with both male and female contract holders, yielding eight groups in total). For this purpose, we use descriptive statistics, OLS regressions, and Coarsened Exact Matching (CEM). The latter is a matching method commonly used in development economics research (Bertoni, Curzi, Aletti, & Olper, 2020;Green, Subramanian, Vickers, & Dorling, 2015;Nilsson, 2017) that helps to create more similar comparison groups. Although commonly applied in this and related bodies of literature to address selection bias (Meemken & Bellemare, 2020), we remind readers that matching approaches depend on the assumption of selection on observables (Cameron & Trivedi, 2005).

Conceptual framework
As outlined above, we are interested in gendered differences and seek to explore the following questions. First, are households where women hold contracts underrepresented in contract schemes, and what determines their participation in contract farming? Second, does the gender of the contract holder matter for household living standards? In what follows, we provide conceptual considerations and hypotheses regarding both questions.
First, why would gender matter for participation rates? As discussed above, female farmers often face particularly high barriers to accessing agricultural inputs, services, and markets (Quisumbing et al., 2015;Raney et al., 2011;World Bank, 2011). Given these disadvantages and related gender roles and norms, participation in contract farming can be more difficult for women. For example, land ownership is often a prerequisite for participation, and women are often not the owners of households' agricultural land (Raney et al., 2011;Sulle & Dancer, 2020;World Bank, FAO, & IFAD, 2008). Also, participation in contract farming typically requires up-front investments and marketable surplus, both of which are more difficult to make and produce with limited access to resources and inputs. Also, contract farming often involves traditional cash or export crops. These crops and related institutions (for example, farmer organizations and markets) are considered the domain of men in many rural societies (World Bank et al., 2008), further increasing women's barriers to participation. Gendered roles in agriculture can change when the male head of male-headed households deceases or migrates. When women become de facto household heads, they sometimes take over male-dominated tasks and enter maledominated domains (Meemken & Qaim, 2018;Oduol et al., 2017;Oduol & Mith€ ofer, 2014).
Given the foregoing, we hypothesize (i) that men are the contract holders in most households that participate in contract farming, (ii) that in somebut fewercontract households, both women and men hold contracts, and (iii) that households where only women hold contracts are underrepresented, and most of these households are headed by women. Our hypothesis that women are underrepresented in contract farming is in line with previous case studies that use non-nationally representative data, mostly from household heads. For example, previous studies suggest that <10 per cent of contracts are issued to female household heads in the vegetable sector in Meru, Kenya (Dolan, 2001) and mango and beans contract holders in Les Niayes and Senegal River Delta areas in Senegal (Maertens & Swinnen, 2012). Regarding determinants, we hypothesize that relatively better-off households (for example, higher levels of education) are more likely to participate, irrespective of the gender of the household head or contract holder.
Secondly, we are interested in the question of whether the gender of the contract holder matters for household living standards. As discussed above, most available studies suggest that contract farming improves farmers' productivity, income, and overall welfare (Bellemare & Bloem, 2018). Theoretically, households with female contract holders might benefit less from contract farming than households where men hold the contracts. This might, for example, be the case when women are offered contracts with less beneficial features than men. Also, given gendered cropping patterns, gender roles, and gender disparities in terms of productivity and access to inputs and services (see above), female contract farmers might sell smaller quantities or products of lower commercial value or lower quality through contracts compared to their male counterparts. On the other hand, it might also be the case that contract farming helps address market and policy failures that affect women more than men, thus, possibly leading to larger income gains among households with female contract holders. Thus, it remains an empirical question whether the gender of the contract holder matters for household income. Yet given that contract households are typically found to benefit, we hypothesize that households, where both women and men hold a contract, benefit more than households where only women or only men hold a contract.
CGAP data are nationally representative of smallholder farm households in the aforementioned countries, where smallholders are defined as farmers with 'up to five hectares OR farmers who have <50 heads of cattle, 100 goats/sheep/pigs, or 1000 chickens' Anderson, Marita, et al., 2016;Anderson et al., 2017;Riquet et al., 2017).
Households were sampled using a multistage sampling strategy. In the first stage, about 200 enumeration areas were selected in each country. Enumeration areas are small geographical units (typically villages), where the categorization was adopted from previous censuses. For CGAP surveys, only enumeration areas with agricultural households were considered. In the second stage, a list containing all agricultural households was prepared for each selected enumeration area. About 15 smallholder households were randomly selected from each list. Table 1 provides an overview of the sample by country.
As part of the CGAP survey, several household members were interviewed using different questionnaires, including a household questionnaire and a multiple respondent questionnaire. The multiple respondent questionnaire was conducted with all household members above the age of 15 years who contribute to household income and contains detailed information on agricultural practices, including whether individuals are contract holders. The household questionnaire was typically conducted with the household head and contains questions on household welfare (for example, household income) as well as a household roster that captures information on individual household members (for example, age, gender, and level of formal education).

Measurement of dependent and independent variables
Contract farming is the key variable of interest in this study. Contract farmers are defined as farmers who have a contract to sell any of their crops or livestock products. Since the survey is not specially designed to analyze the implications of contract farming, we lack information on the details of the contract. For example, we do not know which specific crops contract farmers produce under contract. Similarly, we do not have information on the degree of formality (written or oral contract), services provided (for example, extension, credits, farm inputs), or the type of contractor (for example, traditional buyers, cooperatives, exporters, or supermarkets). This also means that we capture a wide range of different contracts with diverse features.
Our main variable of interestwhether individuals hold a contractwas derived from the question 'Do you have a contract to sell any of your crops or livestock?' This question was part of the multiple-response questionnaire. Therefore, we have multiple responses per household, whenever a household has more than one member above the age of 15 who contributes to household income. Similarly, more than one household member may hold a contract. The gender of each respondent and whether the respondent is the household head was determined based on the household roster.
As displayed in Figure 1, we differentiate between male-headed households and femaleheaded households. We consider four types of households for male-headed and four for maleheaded households (thus, eight in total), namely households without contracts, with only male contract holders, with only female contract holders, and with both male and female contract holders. Our outcome of interest is poverty. Poverty levels are calculated based on per capita household income. We translated all values into (2016) US$using official exchange rates at the time of data collection and the consumer price index (CPI). 1 Following (World Bank, 2020), the households are classified as poor if the household's daily income per capita is below US$1.90 PPP (2016).

Econometric approach
Our regression analysis is aimed at analyzing the correlates of contract farming as well as the relationship between contract farming and household welfare, differentiating between the eight groups in Figure 1. All groups displayed in Figure 1 are mutually exclusive. Households without contracts are always our comparison group, meaning we have three types of contract households for male-headed and three for female-headed households. For each type of contract households (N ¼ 6), we estimate separate regressions. Starting with factors that determine whether these different types of households participate in contract farming, we estimate regressions of the following type: where C il is a dummy variable indicating the contract status of the household. For example, for female-headed households, the three outcome variables for the three regressions are: only female household members hold a contract (0/1), only male members hold a contract (0/1), both female and male members hold a contract (0/1).
In the three regression, 0 refers to households where no member holds a contract. Since households with no contracts are always our comparison group (see above), all but one type of contract households are excluded from each regression. For instance, considering the regression for female-headed households with only female household members holding a contract (group vi in Figure 1), our comparison group is female-headed households without contracts (group v in Figure 1), and we exclude female-headed households that have only male members holding contracts (group vii) and both male and female household members holding contracts (group viii).
X il represents household i 0 s characteristics including the age and education of the household head, the household size, most important crops grown by the household, number of crops grown by the household, whether the household uses farm inputs (0/1), and land and livestock ownership of the household. d represents country fixed effects, and u is the error term. Equation (1) is estimated using linear probability (LP) models. We estimate pooled regressions (including all countries) as well as country-specific regressions.
To analyze the relationship between contract farming and household welfare (proxied using poverty levels as discussed above), we also consider the eight groups in Figure 1, where, as before, non-contract households are our comparison groups. Our outcome variable is a dummy variable indicating whether household i in location l is living below the international poverty line.
The key challenge here is that participation in contract farming might be correlated with both the outcome of interest and other factors that determine participation in contract farming. In other words, observed differences in terms of household welfare might be attributable to pre-existing, often unobserved differences rather than the result of participation in contract farming. Thus, results might be biased. Different econometric techniques exist to address this self-selection problem, but the lack of panel and experimental data limits options here. Several studies have used matching techniques to create more suitable comparison groups, thereby reducing possible bias due to self-selection (Blackwell, Iacus, King, & Porro, 2009). Yet it should be noted that matching approaches do not help address possible bias due to unobserved heterogeneity.
Here, we use Coarsened Exact Matching (CEM), which is commonly used in the development economics literature (Bertoni et al., 2020;Green et al., 2015;Nilsson, 2017). Similar to other matching methods, observations are matched based on a set of specified control variables. We include the following variables: age of household head, education of household head, land owned (in hectares), and village (that is, cluster) dummies. We exclude some of the variables used in the regressions above (Equation 1), such as the crop grown, whenever they might be short-term adjustments due to participation in contract farming (that is, farmers often grow specific crops that contractors ask them to grow). We do so because obtained coefficients might overwise be an underestimation. In theory, contract farmers could also farm more land as a result of contract farming. Thus, keeping variables (which would indicate longer-term adjustments) is still a conservative approach.
CEM estimation involves several steps. The first step is to recode each of the control variables so that somewhat similar values are grouped and assigned the same value. That is, the variables are converted into discrete representations known as bins. This is referred to as 'coarsening' of the variables (Blackwell et al., 2009) Second, a set of strata is then created with each having similar coarsened values of the control variables. Observations in strata that contain at least one treated and one control observation (matched strata) are kept while observations in strata that contain only control or only treated units are dropped from the sample (Iacus, King, & Porro, 2012). Finally, for each of the matched stratum, a weight is computed based on the relative proportion of observations which are then used to estimate the effect of the treatment variable (Sidney, Coberley, Pope, & Wells, 2015), for example using OLS regressions.

Prevalence of contract farming and characteristics of contract holders
Tables 2 and 3 present an overview of the prevalence of contract farming among maleheaded and female-headed households, respectively. About a quarter of the sample households participate in contract farming, with relatively little variation across male-headed and female-headed households (25 and 23%, respectively) and with large variation across  (Meemken & Bellemare, 2020;Oya, 2012). Excluding Tanzania, about 12 per cent of the sample of male-headed households and 6 per cent of female-headed households participate in contract farming. Also, in male-headed households (Table 2), men are more likely to hold the contract than women (columns 3-4), which holds for all countries. In female-headed households (Table 3), however, women are more likely to hold the contract, which also holds for all countries. All of these figures support our hypotheses, namely that women face greater barriers to participation than men, which might be due to gendered roles and disparities in terms of access to inputs and services, as discussed in section 2. Also, we hypothesized that participation rates among women in female-headed households might be higher, given that women might take over male-dominated tasks and roles (Meemken & Qaim, 2018;Oduol et al., 2017;Oduol & Mith€ ofer, 2014). We find that about 84 per cent of all female-headed contract households have female (or female and male) contract holders and that 34 per cent of all male-headed contract households have female (or female and male) contract holders. These figures support the aforementioned hypothesis but also suggest that previous case studies have underestimated women's participation rates (about 10% as summarized in section 2), for example, because they are based on data from household heads and do not capture the contract status of other household members.
It is more common, in most countries, that only men (and not men and women) in maleheaded households (Table 2); and only women (not women and men) in female-headed households (Table 3) hold a contract.
Next, we explore the differences between households with and without contracts. Tables 4 and 5 display descriptive statistics for male-headed households and female-headed households, respectively. As before, we consider the groups displayed in Figure 1.
Male-headed households that participate in contract farming own more land [when male household members hold the contract, as shown in columns (2) and (4)]. Their heads are more likely to have ever attended school, and they are more likely to own livestock [when female members hold the contract, as shown in columns (3) and (4)]. This partly supports our hypothesis that contract households are relatively better off than their counterparts without contracts, although differences are smaller than expected. More pronounced are differences that might well be the outcome of (rather than a precondition for) participation in contract farming. For example, households that participate are more likely to use farm inputs such as fertilizers, seeds, and pesticides, irrespective of the gender of the contract holder. Also, irrespective of the gender of the contract holder, male-headed contract households are less diversified (no. of crops grown) and more likely to name traditional export crops (such as coffee, cocoa, and cotton) or fruits and vegetables as their most important income-generating crops. For female-headed households (Table 5), we have few observations for households where only men hold contracts (column 2) and where both men and women hold contracts (column 4). Thus, we should interpret results with caution, which holds for all parts of the analysis. Thus, we mainly focus on female-headed households with only female contract holders (column 3) in describing results. Similar to male-headed households (see above), female-headed households (column 3 of Table 5) are headed by heads who are more likely to have ever attended school, butunlike male-headed contract householdsthey cannot generally be considered better off than their counterparts without contracts. Yet similar to contract households headed by men, contract households headed by women are more likely to use inputs, are more specialized, and are more likely to grow fruits and vegetables as their main crops, compared to their counterparts without contracts. This, again, might however be the outcome of (rather than a prerequisite for) participation in contract farming. Tables 6 and 7 display correlates of contract farming, again for male-headed and femaleheaded households separately. As described (see Equation 1), the dependent variable is a contract dummy, and we run separate regressions for all 3 types of contract households (displayed in columns 1-3). All variables are household-level variables.
In line with descriptive statistics (Tables 4 and 5), results displayed in Tables 6 and 7 show that crop choices and production decisions are significantly correlated with participation in contract farming, irrespective of the gender of the contract holder. For example, contract households are less likely to grow food crops. More detailed information on households' most important income-generating crops is displayed in the Supplementary Appendix (Tables A1-A12). Country-specific regressions are displayed in Tables A13-A22.

Household welfare
Figure 2 displays descriptive statistics for male-headed households (panel a) and female-headed households (panel b) and the share of households below the international poverty line. We find that in most countries male-headed households (panel a) with a male contract holder are somewhat less likely to be poor than households without contract holders (in all countries but Côte d'Ivoire). Households with only female contract holders are less likely to be poor in Nigeria and Uganda. Conversely, households with both male and female contract holders are more likely to be poor (in all the countries except Côte d'Ivoire). For female-headed households, households with only female contract holders are less likely (or equally likely) to be poor compared with their counterparts without contract holders. As before, given small sample sizes for female-headed households with only male or male and female contract holders, these figures should be interpreted with caution. Descriptive statistics displayed in Figure 2 do, however, not account for the fact that households with and without contract farmers vary in several ways (see above). In Tables 8  and 9 we display CEM matching results, which account for such differences, for the male and female-headed households, respectively. As our outcome variable indicates whether the household is living below the international poverty line, a negative coefficient implies lower poverty levels.
We find that for male-headed households, households with only male contract holders are less likely to be poor, and these results hold also when we exclude Tanzania. For female-headed households, households with only male or only female contract holders are more likely to be poor. These results do, however, not hold once we exclude Tanzania. Overall, there is HH is an abbreviation for household. Standard errors in parentheses. Ã p < 0.1, ÃÃ p < 0.05, ÃÃÃ p < 0.01. a Base category: Staple foods and others (not specified). These variables refer to the most important crops grown by the household. b Base category: Côte d'Ivoire. c The outcome variables in all columns are dummy variables, indicating whether any male member (column 1), female member (column 2), or both male and female members (column 3) hold a contract. The reference groups in all the 3 columns are households with no contract holders. The regression in column (1) for example compares households with only male contract holders against households with no contract holders and excludes households with only female contract holders and those with both male and female contract holders. substantial variation across countries, implying that context matters. Previous studies have highlighted, that the type and characteristics of the contract crop (Oya, 2012;Verhofstadt & Maertens, 2014), type of contract (Bellemare & Lim, 2018;Ruml & Qaim, 2020), and overall institutional environment (Sulle & Dancer, 2020) can matter for outcomes among participating farmers. Additionally, gender can interact with other social categories (for example, ethnicity and religion), which also matters in the context of contract farming (Sulle & Dancer, 2020). This might explain why we cannot detect clear patterns across countries. Overall, our findings speak against the hypothesis that women might hold contracts with less beneficial features than menand against the hypothesis that households with more than one household member (women and men) holding contracts benefit more.

Conclusion
This study provides novel evidence on women's participation in contract farming, using nationally representative data from five African countries. We find that womenespecially in maleheaded householdsparticipate less often in contract farming than men, supporting concerns that women face higher barriers to participation in contract farming than men (Dolan, 2001;Maertens & Swinnen, 2012;Raynolds, 2002). Relatedly, the finding that women in male-headed households are less likely to participate than those in female-headed households supports the   notion that it is often easier to successfully integrate female household heads in higher-value agrifood supply chains than female spouses of male household heads (Oduol et al., 2017;Oduol & Mith€ ofer, 2014). Successfully integrating womenboth in male-and female-headed householdsinto contract farming and other supply chains requires gender-sensitive approaches to value chain development (Fischer et al., 2020). Such approaches carefully assess and take into account the different resources and barriers faced by women. Identifying and evaluating gender-sensitive approaches to contract farming is beyond the scope of this paper and will require more research.
Our results regarding the importance of the gender of the contract holder for household living standards are inconclusive, for both male-headed and female-headed households, and there is great heterogeneity across countries. It is known that the type and characteristics of the contract crop (Oya, 2012;Verhofstadt & Maertens, 2014), type of contract (Bellemare & Lim, 2018;Ruml & Qaim, 2020), and overall institutional environment (Sulle & Dancer, 2020) can matter in important ways, which we are unable to explore. Relatedly and regarding overall welfare effects, our results are in line with a previous study that used the same data (Meemken & Bellemare, 2020). The study concludes that previous case studies might have overestimated the effect of contract farming, for example, because they have focused on well-functioning contract schemes. The sample and data they and we use, in turn, capture various and potentially also illfunctioning contractual arrangements. Thus, it might be little surprising that when income effects can hardly be detected (Meemken & Bellemare, 2020), we also do not find pronounced gendered differences. Thus, more research will be needed, using datasets that capture diverse contract schemes but also more detailed information on contract features as well as individuallevel data. To date, such data are scarce, and future research teams could account for these issues in their study and questionnaire designs.
Finally, we highlight other issues that merit further exploration as they could not be fully answered based on the data and methods used here. We use cross-sectional, observational data, so our options to deal with selection bias at different levels (individual, household, region) are limited. Like many other studies, we use matching techniques, which do not control for unobserved heterogeneity. One option to improve internal validity would be to conduct a randomized control trial (RCT) in collaboration with a company offering contracts and to randomize who within the household is offered a contract. A handful of such experiments were conducted in the context of contract farming but without a specific gender focus (Arouna, Michler, & Lokossou, 2021;Casaburi & Willis, 2018). Often, RCTs are associated with high costs, might raise ethical concerns, and might not be feasible in all situations (for example when women and men grow different types of crops). Another option that would allow for more advanced econometric analyses is the use of secondary (panel) data. As mentioned above, such data are scarce, but increasing interest in post-farm gate value chains might change this (Yi et al., 2021). The realization of and funding for the aforementioned data collection and research projects would be highly policy-relevant as contract farming is often supported by policymakerswho could prioritize gender-sensitive contract schemes. This paper provides novel, quantitative evidence but should be understood as an initial step towards a better understanding of the gendered patterns, drivers, and implications of contract farming. Note 1. We used the CPI to adjust for inflation for Uganda and Mozambique. Unlike in the other countries, data were collected in 2015.