New evidence regarding the effects of contract farming on agricultural labor use

Ruml, Anette; Qaim, Matin Article — Published Version New evidence regarding the effects of contract farming on agricultural labor use Agricultural Economics Provided in Cooperation with: John Wiley & Sons Suggested Citation: Ruml, Anette; Qaim, Matin (2021) : New evidence regarding the effects of contract farming on agricultural labor use, Agricultural Economics, ISSN 1574-0862, Wiley, Hoboken, NJ, Vol. 52, Iss. 1, pp. 51-66, https://doi.org/10.1111/agec.12606 This Version is available at: https://hdl.handle.net/10419/233750 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. http://creativecommons.org/licenses/by-nc/4.0/ Received: 18 December 2017 Revised: 6 November 2020 Accepted: 11 November 2020 DOI: 10.1111/agec.12606 ORIGINAL ARTICLE New evidence regarding the effects of contract farming on agricultural labor use Anette Ruml1,2Matin Qaim1 1Department of Agricultural Economics and Rural Development, University of Goettingen, Goettingen, Germany 2German Institute for Global and Area Studies, Hamburg, Germany Correspondence AnetteRuml,University of Goettingen, Heinrich-Dueker-Weg12, 37073Goettin- gen, Germany. Email:[email protected] Abstract Contractual agreements between smallholder farmers and agribusiness companies have gained in importance in many developing countries. While productivity and income effects of contracting in the small farm sector were analyzed in many previous studies, labor market and employment effects are not yet well understood. This is an important research gap, especially against the background of continued population growth and structural transformation. Here, we investigate the effects of two types of contractual agreements between large international processing companies and smallholder farmers on agricultural labor use, household labor allocation, and hired labor demand in Ghana’s palm oil sector. We use cross-sectional survey data and a willingness-to-pay approach to control for unobserved heterogeneity between farmers with and without contracts. We find that agricultural labor intensity is substantially reduced through the contracts, because contracting in Ghana is associated with the adoption of laborsaving procedures and technologies. Simple marketing contracts lead to reallocation of the saved household labor to off-farm employment, whereas resourceproviding contracts lead to a stronger reallocation of labor within the farming enterprise. Household labor is more affected by labor savings than hired labor. KEYWORDS agricultural labor use, child labor, contract farming, gender, oil palm, rural employment JEL CLASSIFICATION J23, J43, O13, Q12 1 INTRODUCTION Contract farming has gained in importance in many developing countries, with agribusiness companies contracting small- and medium-scale farmers (Bellemare, 2018; Meemken & Bellemare, 2020;Otsuka,Nakano, & Takahashi, 2016; Ton, Vellema, Desiere, Weituschat, & D’Haese, 2018). Contract farming has positively This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2020 The Authors. Agricultural Economics published by Wiley Periodicals LLC on behalf of International Association of Agricultural Economists contributed to income gains in the small farm sector and to broader agricultural development in many situations (Otsuka et al., 2016). However, the effects of contract farming on agricultural labor markets and employment are not yet sufficiently understood. Depending on the situation, contract farming can lead to higher farm labor demand and more labor-intensive agricultural production, or it can also contribute to on-farm labor savings, and Agricultural Economics. 2021;52:51–66. wileyonlinelibrary.com/journal/agec 51 52 RUML and QAIM thus promote structural transformation with labor shifts from agriculture to other sectors. Here, we investigate the effects of contracts between large processing companies and smallholder farmers on agricultural labor use, including household labor and hired labor, in Ghana’s palm oil sector. This is an interesting empirical example, because international palm oil companies are increasingly investing in Africa, so that the search for socially-inclusive business models is important from a rural development policy perspective (Byerlee, Falcon, & Naylor, 2017;Qaim, Sibhatu, Siregar, & Grass, 2020). Many studies analyzed the effects of contracts on agricultural productivity and income in the small-farm sector (e.g., Ashraf, Giné, & Karlan, 2009; Barrett et al., 2012; Khan, Nakano, & Kurosaki, 2019; Mishra, Kumar, Joshi, & D’Souza, 2016; Ragasa, Lambrecht, & Kufoalor, 2018;Rao, Brümmer, & Qaim, 2012; Ruml & Qaim, 2020). Possible effects of contracts on agricultural labor use received much less attention in the empirical literature. This is surprising, because agricultural labor use, household labor allocation, and hired labor demand are all important aspects of household welfare, rural development, and broader structural transformation. The few available studies that analyzed labor market effects suggest that contracting leads to additional labor use in farm production, harvesting, and postharvest handling (Benali, Brümmer, & Afari-Sefa, 2018; Meemken & Bellemare, 2020; Neven, Odera, Reardon, & Wang, 2009; Rao & Qaim, 2013). However, we argue that these results cannot be generalized, because contracting can also involve the adoption of labor-saving technologies and procedures. Labor-reducing effects through contracts were not shown previously in a small-farm context. Here, we show that they exist in Ghana’s palm oil sector. In particular, using data from a survey of farm households, we investigate the effects of two types of contracts— namely, marketing and resource-providing contracts—on labor use in oil palm production. While farmers without a contract do some of the postharvest handling themselves, farmers with a contract sell the oil palm fruit bunches to the buying company immediately after harvest. Some of the contracted farmers also use labor-saving chemical inputs, such as herbicides, thus further reducing labor intensity. We quantify the effects of contracting on total labor use per unit of land and investigate the resulting implications for household labor allocation and hired labor demand. In addition, we differentiate between male, female, child, and youth labor. Differentiation is useful to better understand possible broader social implications. Endogeneity issues in the evaluation of effects are addressed through including farmers’ willingness-to- pay (WTP) for certain contract features as an additional explanatory variable in the regressions, which is a useful approach to control for possible unobserved heterogeneity (Bellemare & Novak, 2017). Contract farming in Ghana’s palm oil sector is not a peculiar case. Many smallholders in Africa have traditionally produced palm oil for home consumption and local markets. However, demand for palm oil from domestic and international markets is growing, so that modern supply chains with new actors and smallholder contract schemes are increasingly emerging in Africa (Byerlee et al., 2017). Similar trends are also observed in other crops traditionally grown by smallholders. Against this background, better understanding the labor market implications of contract farming is particularly important. The rest of this article is structured as follows. The next section presents further details of trends in Africa’s palm oil sector, including a description of traditional and modern supply chains. Section 3describes the data collection and the statistical methods, Section 4presents and discusses the empirical results, while Section 5concludes. 2BACKGROUND Over the last few decades, international demand for palm oil increased tremendously. This led to a substantial rise in the area under oil palm cultivation, particularly in South East Asia (Byerlee et al., 2017). In West Africa, where oil palm actually originates, production levels stagnated in recent decades (Huddleston & Tonts, 2007). This situation is now gradually changing. In South East Asia, the land for future oil palm expansion is limited and production growth increasingly conflicts with tropical rainforest conservation objectives (Qaim et al., 2020). Hence, to meet the further rising international demand, palm oil companies have also started to invest in Africa. In Ghana, the area under oil palm increased from 160,000 hectares in the year 2000 to over 370,000 hectares in 2018. During the same period, national production volumes rose from 1 million tons to 2.6 million tons of fresh fruit bunches (FAO, 2019). Similar trends are also observed in other West African countries. Oil palm is already one of the most important cash crops produced in West Africa and substantial further growth is expected in the future (Byerlee et al., 2017; Rhebergen et al., 2016). The transformation of oil palm from a local semisubsistence crop, which it was for centuries in Africa, to a major cash crop is associated with supply chain modernization and the entry of large processing companies. In Ghana, the location of company-owned palm oil plantations and processing facilities is primarily determined by land concessions that the companies obtain from the Ghanaian government. Some of the palm oil that the companies process is produced on these company-owned plantations. In RUML and QAIM 53 TABLE 1 Production and marketing characteristics in oil palm with and without contract Traditional, without contract Marketing contract Resource-providing contract Buyer Local customers, small processing mills Processing company Processing company Product sold Oil palm fruits, palm oil Oil palm fruit bunches Oil palm fruit bunches Production assistance None None Inputs, technologies, technical support on credit Labor operations Plot maintenance (m) Plot maintenance (m) Plot maintenance (m) Input application (m) Input application (m) Input application (m) Harvesting (piecemeal) (m, f) Harvesting (at once) (m, f) Harvesting (at once) (m, f) Picking of fruits (m, f, c, y) Processing (m, f) Marketing (m, f) Notes: (m) indicates that the operation is typically performed by adult males. (f) indicates that the operation is typically performed by adult females. (c)and(y) indicate that children and youths are also involved occasionally. addition, the companies procure oil palm fruit bunches from surrounding smallholder farmers through contractual agreements. Smallholder farmers continue to be the main producers of oil palm in West Africa. In Ghana, smallholder production accounts for 75% of total palm oil supply (Byerlee et al., 2017). Smallholder palm oil producers are also an important employer in Ghana, providing jobs for several hundred-thousand farm workers (Manley & Leynseele, 2019; Ministry of Food and Agriculture, 2011). To this point, five large national and international palm oil processing companies procure their supply from contracted smallholders in Ghana, but the sector is evolving. Many smallholders still produce palm oil for traditional local markets without any contracts, whereas new companies and contract schemes are emerging, largely depending on where the government provides additional land concessions. The production and marketing conditions between traditional supply chains without contracts and modern supply chains with contracts differ remarkably. In traditional supply chains, farmers have no secure sales market. They harvest the fruit bunches and then pick the individual fruits out of the bunches, in order to sell to local customers or home process to palm oil. Picking, processing, and finding a buyer are time-intensive operations. As the quantities traded in local markets are small and the fruits are perishable, harvesting in traditional supply chains typically takes place in a piecemeal fashion. In contrast, farmers in modern supply chains with a contract have a secure sales market where prices are fixed annually. Contracted farmers harvest the bunches, but instead of picking the individual fruits out of the bunches and processing themselves, they sell the bunches to the buying companies at the farm gate. In other words, the labor-intensive postharvest operations are no longer carried out on the farm. The companies have large mechanized mills where the fruit bunches are processed. Compared to the home processing of palm oil, and the small local mills that continue to use manual techniques (Byerlee et al., 2017), larger mills produce at higher processing capacities of 20–30 tons per hour. This means that farmers in modern supply chains with a company contract can harvest and sell larger quantities of fruit bunches at once. In Ghana, two types of contracts exist in the palm oil sector, namely, marketing and resource-providing contracts, as shown in Table 1. For both types of contracts, the harvest and sales conditions are as described above. However, the contracts differ in terms of the additional assistance provided for production inputs and technologies. While farmers with a marketing contract do not receive production assistance, farmers with a resource-providing contract can obtain planting material, chemical inputs, other production tools, and technical support on credit from the contracting company. This credit is paid back through a share of the harvest and the commitment to sell to the contracting company. Thus, in addition to providing a secure sales market, the resource-providing contract addresses farmers’ financial constraints through interlinking output, input, and credit markets. Farmers producing under marketing contracts and farmers without contracts are not involved in such market interlinkages. The described differences between traditional and modern supply chains lead to the expectation that contract farming has a labor-saving effect on oil palm production at the smallholder level. Whether this is really observed empirically is analyzed below. The expected reduction in agricultural labor intensity raises additional questions. Farmers could either use the labor saved per unit of oil palm land to expand the area cultivated, thus keeping the total agricultural labor use constant, or they could reallocate the labor saved to off-farm activities. Obviously, 54 RUML and QAIM FIGURE 1 Map of study area in Ghana Source: Authors’ own presentation using tools provided in Kahle and Wickham (2013). expansion of the area cultivated would require access to additional land and capital. In our study region in Ghana, land is not the major limiting factor. In fact, most farmers have more land than they actually cultivate. However, farmers typically face financial constraints to expand the oil palm area, as new oil palm plantations are costly to establish and only start bearing fruits after several years. Against this background, it is likely that marketing contracts and resource-providing contracts lead to different types of labor reallocation. As mentioned, farmers with resource-providing contracts have access to credits for the establishment and maintenance of oil palm plantations, whereas farmers with simple marketing contracts do not. Indeed, another recent study using the same survey data showed that resource-providing contracts contribute to higher smallholder production investments and larger areas cultivated with oil palm, whereas simple marketing contracts do not have such effects (Ruml & Qaim, 2020). Effects on labor use and labor reallocation were not analyzed previously but will be evaluated here. In particular, we investigate the effects of both types of contracts on (a) total agricultural labor use per acre of oil palm, (b) household labor use, (c) hired labor use, and (d) the time worked in off-farm employment. In addition, we disaggregate the labor use effects by gender, separating between male and female adults, and also analyze possible implications for child and youth labor. Table 1suggests that both types of contracts lead to a reduction or omission of farm operations that often also involve women, children, and youths, so that examining effects by gender and age can provide useful additional insights. 3 MATERIALS AND METHODS 3.1 Sampling strategy and farm household survey We conducted a survey of oil palm-producing farm households in Ghana between April and July 2018. When we sampled regions and households for the survey, there were a total of five large palm oil processing companies with company plantations and smallholder contract schemes (Figure 1). All five companies were located in the Southern parts of Ghana. Out of the five companies, we RUML and QAIM 55 purposively selected two that were located in neighboring regions quite close to each other, namely, Benso Oil Palm Plantation owned by Wilmar International in the Western Region and Twifo Oil Palm Plantation owned by Unilever in the Central Region. Benso has simple marketing contracts with farmers and started the contract scheme in the Western Region already in the 1990s. In contrast, Twifo uses resource-providing contracts and started to work with smallholders in the Central Region in 2008. From both company schemes, contracted oil palm farmers were selected randomly based on complete lists of villages and farmers involved. Both companies stated that they offer the contracts to all oil palm farmers in the selected contract villages, provided that farmers agree to the contract conditions. This was also confirmed in focus group discussions, which we carried out in nonsampled villages prior to the actual survey.1In contract villages, most oil palm-producing farm households were contracted, and the few that were not sometimes still sold parts of their harvest to the company through informal arrangements with their neighbors. Given that most farmers in the contract villages selfselected into a contract, we could not select comparison farmers without contract in the same villages without the risk of serious selection bias. Nor could we select comparison farmers in neighboring or nearby noncontract villages because these villages had not been selected by the companies for their contract schemes, probably due to different village or farmer characteristics. Our alternative was to select comparison villages and farmers in a different region located outside of the current contract area but otherwise sufficiently similar to the contract villages and farmers. This was possible because—as discussed above— Ghana’s palm oil sector is evolving, and new company plantations and contract schemes are being planned and implemented.2 1Focus group discussions were carried out separately with village officials (village chiefs, assembly men, lead farmers, and elder councils) and farmers. Village officials were asked about their perception of the contract farming scheme and the opportunities and challenges faced with the contracts, whereas farmers were asked about their production methods, access to land, labor, and inputs, and personal experiences with the contract scheme. The discussions were informal and often lasted for several hours of debate in local languages with English translation by a local interpreter. In total, eight focus group discussions were held: four in villages with resource-providing contracts, two in villages with marketing contracts, and two in villages without any contracts. 2As mentioned above, the regions and locations where company plantations are established cannot be freely chosen by the companies but are determined by where the government provides land concessions. Many of the recent concessions to palm oil companies were provided in the southern parts of Ghana, where our study is also located. Around the concession land for the company plantations, companies can select villages where they want to contract smallholder farmers. Villages are typically With the help of the Ministry of Food and Agriculture (MoFA), we identified a suitable comparison area in the Ashanti Region also in the southern part of Ghana, where the conditions are very similar, no contracting existed at the time of the survey, but a new contract scheme was about to start.3MoFA provided us with a list of villages in the Ashanti Region that had already been selected for the upcoming contract scheme. From this list, we randomly sampled villages and oil palm-producing farm households. The focus group discussions confirmed that farmers in these comparison villages were not aware of the upcoming contract scheme at the time of the survey, which was advantageous for us to collect comparable data from oil palm farmers in traditional supply chains without company contracts. Our strategy to sample farmers with and without contracts in different (neighboring) regions (Figure 1)haspros and cons. On the pro side, it reduces or avoids selection issues within each region. On the contra side, there is perfect correlation between contract schemes and regions, so if the regions differ systematically, it would be difficult to know whether differences in the outcomes are due to differences in contracts or regional characteristics. We tried to minimize the possibility of systematic regional differences. All three regions—Western, Central, and Ashanti— are located in Ghana’s green belt, which is classified as suitable for oil palm cultivation (Rhebergen et al., 2016). The three regions have very similar rainfall and temperature conditions (Table A1 in the online Appendix). Nor are there systematic regional differences in terms of soil quality and irrigation (Ruml & Qaim, 2020). In addition to agroecological factors, we also compared the three regions in terms of various economic and social indicators, such as mean income levels, human development, and employment rates, which could possibly influence the labor market effects of contract farming. For all these indicators, we do not observe systematic regional differences (Table A1). This is also in line with our microlevel observations and the focus group discussions in all three regions. Of course, we cannot rule out completely that certain unobserved regional differences exist, so that some caution is warranted. We explain below how we try to control for observed and unobserved heterogeneity in our regression models. In total, we randomly selected 463 oil palm-producing farm households from 31 villages in the three regions: 193 from the Western Region with a marketing contract, 164 chosen based on distance, infrastructure conditions, and the number of oil palm-cultivating households. 3In other parts of the Ashanti Region, contract schemes of palm oil companies already existed at the time of the survey (Figure 1). 56 RUML and QAIM from the Central Region with a resource-providing contract, and 106 from the Ashanti Region without any contract. For the structured survey, personal interviews were carried out with the head of each of the sampled households in the local language, using a questionnaire developed for this purpose and programmed into tablet computers. The questionnaire captured information on the household structure, all income sources, the time spent by household members in various economic activities, and other socioeconomic details. Input–output details for oil palm production were captured at the plot level for all plots managed by the sample household. We use complete data for 524 oil palm plots, after excluding those that did not yet bear any fruits. In addition to the household interviews, we also conducted shorter structured interviews with the chief in each of the villages, capturing information on villagelevel characteristics. 3.2 Regression models In a first step, we estimate the effects of contract farming on labor use in oil palm farming with a regression model of the following type: 𝑌𝑖ℎ𝑗 =𝛽 0+𝛽 1𝑀𝐶𝑖ℎ𝑗 +𝛽 2𝑅𝑃𝐶𝑖ℎ𝑗 +𝛽 3𝑋𝑖ℎ𝑗 +𝑢 𝑖ℎ𝑗, (1) where 𝑌𝑖ℎ𝑗 is total labor use per acre of oil palm on plot 𝑖, in household ℎ, and village 𝑗.𝑀𝐶 represents the marketing contract and 𝑅𝑃𝐶 the resource-providing contract; these are dummy variables taking a value of 1 if the household and plot are part of the respective contract scheme and 0 otherwise.4Thus, 𝛽1measures the effect of the marketing contract and 𝛽2the effect of the resource-providing contract. A negative and statistically significant coefficient would indicate that the respective contract reduces total labor use per acre of oil palm. We also control for other factors that may influence labor use in oil palm farming through the vector 𝑋𝑖ℎ𝑗, which includes plot, household, and village characteristics. 𝑢𝑖ℎ𝑗 is a random error term that we cluster at the village level. In a second step, we estimate disaggregated models using household labor and hired labor per acre of oil palm as separate dependent variables. As there are some farmers who do not use both types of labor, the dependent variables in these models include zero observations leading to corner solutions. This is accounted for by modeling two decisions for each type of labor as follows: 4MC and RPC are possibly endogenous, which could lead to biased estimates. We discuss endogeneity issues and how we address them further below. 𝐷𝑖ℎ𝑗 =∝ 1𝑀𝐶𝑖ℎ𝑗 +∝ 2𝑅𝑃𝐶𝑖ℎ𝑗 +∝ 3𝑋𝑖ℎ𝑗 +𝜇 𝑖ℎ𝑗 𝜇𝑖ℎ𝑗 ∼𝑁 (0, 1),(2) 𝑄𝑖ℎ𝑗 =𝛾 1𝑀𝐶𝑖ℎ𝑗 +𝛾 2𝑅𝑃𝐶𝑖ℎ𝑗 +𝛾 3𝑋𝑖ℎ𝑗 +𝜀 𝑖ℎ𝑗 𝜀𝑖ℎ𝑗 ∼𝑁 (0, 𝜎2),(3) where Equation (2) models the binary decision whether or not to use household (hired) labor on oil palm plot i, and Equation (3) models the decision of how much household (hired) labor to use on this plot, conditional on the first decision being positive. Hence, 𝐷𝑖ℎ𝑗 is a dummy and 𝑄𝑖ℎ𝑗 a continuous variable. The other variables are defined as above. We estimate Equations (2)and(3) separately for household labor and family labor. In a third step, we test whether contract farming leads to reallocation of household labor from farm to off-farm activities. This is tested with the following equations, which are estimated at the household level: 𝑉ℎ𝑗 =𝜋 1𝑀𝐶ℎ𝑗 +𝜋 2𝑅𝑃𝐶ℎ𝑗 +𝜋 3𝑋ℎ𝑗 +𝜏 ℎ𝑗 𝜏ℎ𝑗 ∼𝑁 (0, 1),(4) 𝑊ℎ𝑗 =𝜑 1𝑀𝐶ℎ𝑗 +𝜑 2𝑅𝑃𝐶ℎ𝑗 +𝜑 3𝑋ℎ𝑗 +𝛿 ℎ𝑗 𝛿ℎ𝑗 ∼𝑁 (0, 𝜎2),(5) where 𝑉ℎ𝑗 is a dummy variable taking a value of 1 if at least one member of household hworks in off-farm employment, and 0 otherwise, whereas 𝑊ℎ𝑗 is a continuous variable measuring the number of labor days worked in off-farm employment by all household members. Household labor is reallocated to off-farm activities if the coefficients 𝜋1,𝜋2and/or 𝜑1,𝜑2are positive and statistically significant. As discussed above, differences in labor reallocation effects between marketing and resourceproviding contracts can be expected. Moreover, we examine whether the effects are different for male and female household and hired laborers. This is tested by running the models in Equations (2)–(5)separately for male and female labor and comparing the coefficients. Finally, we investigate the effects on child and youth labor participation in the production of oil palm by reestimating the models in Equations (2)and(3)withchildand youth labor as dependent variables. We use double-hurdle specifications to estimate the models in Equations (2)–(3)and(4)–(5). The doublehurdle specification is suitable to estimate corner solution models with a binary first-stage decision and a continuous variable in the second stage. As such, it estimates two interlinked choices: the decision to employ the RUML and QAIM 57 particular type of labor (Equations (2)and(4)), and the choice on the quantity of the type of labor (Equations (3) and (5)) (Burke, 2009; Cragg, 1971; García, 2013). Doublehurdle models were used recently in the agricultural economics literature to estimate labor market effects (Benali et al., 2018; Rao & Qaim, 2013). The continous outcome variables in Equations (3) and (5) are not normally distributed, which we tested prior to estimation. Therefore, we use an exponentional double-hurdle model, which is suitable for our variable distributions.5Alternatively, a tobit model using hyperbolic sine transformations could be employed. Hyperbolic sine transformations are suitable transformations if the variable includes meaningful zeros (Bellemare & Wichman, 2020). We test the double-hurdle specification against the more specific tobit alternative using a likelihood ratio test. The results reject the hypothesis that the tobit is a suitable specificastion in all cases, meaning that the double-hurdle model is preferred (Table A2 in the online Appendix). 3.3 Definition of key variables The dependent variables in the different regression models are total labor use per acre of oil palm, as well as labor use by different categeries of laborers, including household and hired labor, male and female labor, and child and youth labor. All these variables are measured in labor days worked per acre of oil palm during the 12 months prior to the survey. Laborers are considered adult if they are 18 years or older. Youth labor includes persons between 15 and 17 years of age, and child labor refers to individuals that are 14 years or younger. Child and youth participation is only counted as labor when the individuals were actively involved in any of the agricultural operations. Activities such as delivering food or water to other laborers or simply accompanying family members without own active involvement are not counted as labor. Collecting data on child labor can be difficult as employing child labor is forbidden and farmers may be hesitant to provide this information. However, the ban on child labor applies primarily to hired child labor, which is uncommon in oil palm farming in the study area. The use of hired youth labor is also rare in oil palm production. Therefore, child and youth labor in our context refers to children and youths belonging to the farm family, for which farm- 5The exponential double-hurdle model uses the exponential value of the independent variables on the right-hand side instead of taking the logarithm of the dependent variable on the left-hand side of the equation. The left-hand side cannot be log-transformed, due to meaningful zeros in the data. ers openly provided details during the interviews. Nevertheless, we cannot completely rule out a certain reporting bias, which should be kept in mind when interpreting the results. The key explanatory variables are the two dummies for particiation in marketing and resource-providing contracts, which were already explained above. In addition, we include a set of control variables. At the plot level, we control for soil quality, irrigation, the number of palms per acre, the age of the palms, and the distance from the plot to the closest road that is accessible with a truck, measured in walking minutes. At the household level, we control for the number of adult household members, which is a proxy for the availability of household labor, and the total land size. As the current land size can be influenced by contracts, we use land availability in 2008, which is before most of the farmers had any oil palm contracts.6Total land size includes all plots available to the household for cultivation, regardless of whether or not the plots were actually cultivated in 2008. Furthermore, we control for socioeconomic characteristics of the oil palm farmer (age, sex, education, and farming experience). In the household-level models, we control for the characteristics of the household head, who is not necessarily the same person as the oil palm farmer. Finally, we control for distance to the closest market measured in km as a village-level variable. 3.4 Dealing with potential endogeneity We use the regression models explained above to evaluate the impact of marketing contracts and resource-providing contracts on labor use. However, farmers self-select into contract participation, so that the exposure variables may be endogenous. Some of the variables that influence contract participation are observed and controlled for. But there may also be unobserved factors that are jointly correlated with contract participation and labor use decisions. Such type of endogeneity could lead to correlation of the contract dummy variables with the error terms and thus bias the estimation results. Our sampling framework helps to reduce issues of farmer self-selection, because farmers with and without contracts were chosen in different regions, namely, regions that are very similar in terms of regional characteristics but differ in terms of contract availability (see above). However, unobserved heterogeneity may still exist. To control 6Some of the farmers with marketing contracts were already contracted before 2008, but the marketing contracts did not affect farm investments and the scale of production, as another recent study with the same data showed (Ruml & Qaim, 2020). The resource-providing contracts, however, affect investments and the scale of production, and these were not available before 2008. 58 RUML and QAIM for possible unobserved heterogeneity, we include an individual WTP measure as an additional covariate in the regression models. Details of this approach are explained in the following. The WTP measure captures the farmer’s subjective preference for producing under contract, which is likely correlated with a number of farmer and locational characteristics, including unobserved ones such as risk aversion, time preferences, entrepreneurial skills, and individual market access. Hence, controlling for WTP in the models will reduce possible issues caused by unobserved heterogeneity. Using WTP measures to address endogeneity is an approach that was also recently used in other studies evaluating the impacts of contracts and related marketing institutions (Bellemare & Novak, 2017; Meemken & Qaim, 2018; Verhofstadt & Maertens, 2014). We derived the farmer’s WTP for contracts through a simple experiment that was implemented as part of the survey. In particular, we offered each farmer a set of hypothetical contract offers requiring varying amounts of initial investments. Respondents were asked: “Would you be willing to enter a contract agreement with a company for the establishment of one acre of oil palm that would increase your income but would necessitate an initial investment of ZGhanaian Cedis (GHS)?” For each respondent, Zstarted at a low value and gradually increased in follow-up questions.7The highest value of Zfor which the answer was “yes” represents the individual WTP, which we include as an additional control variable in our regressions. It should be noted that the WTP variable itself is also endogenous and was derived at a time when many farmers in our sample already had a contract. However, we do not use this variable to estimate the effect of WTP on labor use but only to control for unobserved heterogeneity when estimating the effects of contracting. 4RESULTS 4.1 Descriptive statistics Table 2shows descriptive statistics and mean difference tests for all outcome variables used. The upper part of Table 2shows labor use at the plot level. As expected, farmers with a contract use significantly less agricultural labor in oil palm production than farmers without a contract. This is true for both types of contracts, but the difference is especially large for the resource-providing contract. Farmers with a marketing contract use less than half, 7We included eight initial investment amounts, ranging from 500 GHS to 4000 GHS, in steps of 500 GHS. The average initial investment amount required for the resource-providing contract is between 3000 and 4000 GHS. However, this amount can vary substantially, depending on the type and quantity of support the individual farmer requests on credit. and farmers with a resource-providing contract only use about one-third of the labor that farmers without a contract use per acre of oil palm. Differences are primarily observed for household labor, including male and female, as well as child and youth labor. For hired labor, differences between plots with and without contracts are not statistically significant. This provides a first indication that both contracts are associated with lower agricultural labor use at the plot level, especially lower household labor use. The lower part of Table 2shows the number of days worked in off-farm employment at the household level. For the total number of days worked in off-farm activities, no significant differences between households with and without contract are observed. However, gender disaggregation reveals that households with a marketing contract have more female off-farm labor days than households without any contract. The differences in Table 2can- not be interpreted as effects of contracts, as the plots and households also differ in terms of several other characteristics (Table A3 in the online Appendix). The regression results presented below control for differences in plot and household characteristics and possible other confounding factors. Table 3provides additional descriptive statistics on the type of labor used in each production step. Male adults are more involved than other household members in plot maintenance and harvesting, while female adults are more involved in fruit picking and processing. The overall contribution of child and youth labor is relatively small and mostly concentrated on fruit picking and to a lesser extent harvesting and processing. As discussed above, fruit picking and processing are operations that are no longer carried out on-fam in the modern supply chains with company contracts. In terms of hired labor, male laborers are involved in all operations, except for fruit picking where their contribution is small. Female-hired laborers are mostly involved in fruit picking and harvesting. Noteworthy in Table 3are also the large standard deviations, indicating that a large variation in the use of household and hired labor exists across oil palm farms. 4.2 Effects of contracts on agricultural labor use Table 4shows the estimated effects of contract farming on total agricultural labor use. The results clearly suggest that contract farming reduces total labor use under both types of contracts. The marketing contract leads to a reduction of 43 labor days per acre of oil palm, which is equivalent to a 55% decrease when compared to the mean labor use of 78 days on oil palm plots without any contract. The resource-providing contract leads to a reduction of 48 RUML and QAIM 65 of decent agricultural and nonagricultural employment is key for sustainable rural development and structural transformation, especially in Africa where rural population growth is still substantial. ACKNOWLEDGMENTS This research was financially supported by the German Research Foundation (DFG), grant number RTG1666 (GlobalFood). The authors thank the Ministry of Food and Agriculture in Obuasi, Ghana, for the support during the field work. The authors also thank the survey enumerators and respondents for their willingness to participate in this study. Very valuable comments received from two anonymous reviewers of this journal are gratefully acknowledged. REFERENCES Ashraf, N., Giné, X., & Karlan, D. (2009). Finding missing markets (and a disturbing epilogue): Evidence from an export crop adoption and marketing intervention in Kenya. American Journal of Agricultural Economics,91(4), 973–990. Barrett, C. B., Bachke, M. E., Bellemare, M. F., Michelson, H. C., Narayanan, S., & Walker, T. F. (2012). Smallholder participation in contract farming: Comparative evidence from five countries. World Development,40(4), 715–730. Bellemare, M. F. (2018). Contract farming: Opportunity cost and trade-offs. Agricultural Economics,49(3), 279–288. Bellemare, M. F., & Novak, L. (2017). Contract farming and food security. American Journal of Agricultural Economics,99,357–378. Bellemare, M. F., & Wichman, C. J. (2020). Elasticities and the inverse hyperbolic sine transformation. Oxford Bulletin of Economics and Statistics,82(1), 50–61. Benali, M., Brümmer, B., & Afari-Sefa, V. (2018). Smallholder participation in vegetable exports and age-disaggregated labor allocation in Northern Tanzania. Agricultural Economics,49(5), 549–562. Burke, W. J. (2009). Fitting and interpreting Cragg’s tobit alternative using Stata. The Stata Journal,9(4), 584–592. Byerlee, D., Falcon, W. P., & Naylor, R. L. (2017). The tropical oil crop revolution: food, feed, fuel, and forests. Oxford: Oxford University Press. Cragg, J. G. (1971). Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica,39, 829–844. Davis, B., Di Giuseppe, S., & Zezza, A. (2017). Are African households (not) leaving agriculture? Patterns of households’ income sources in rural Sub-Saharan Africa. Food Policy,67,153– 174. Debela, B. L., Gehrke, E., & Qaim, M. (2020). Links between maternal employment and child nutrition in rural Tanzania. American Journal of Agricultural Economics,https://doi.org/10.1111/ajae.12113 FAO. (2019). FAOSTAT Statistical Database. Retrieved from http:// www.fao.org/faostat/en/#data/QC García, B. (2013). Implementation of a double-hurdle model. The Stata Journal,13(4), 776–794. Huddleston, P., & Tonts, M. (2007). Agricultural development, contract farming and Ghana’s oil palm industry. Geography,92, 266– 278. Kahle, D., & Wickham, H. (2013). ggmap: Spatial visualization with ggplot2. The R Journal,5(1), 144–161. Khan, M. F., Nakano, Y., & Kurosaki, T. (2019). Impact of contract farming on land productivity and income of maize and potato growers in Pakistan. Food Policy,85,28–39. Manley, R., & Leynseele, Y. V. (2019). Peasant agency in Ghana’s oil palm sector: The impact of multiple markets on food sovereignty. Journal of Agrarian Change,19,654–670. Meemken, E. M., & Bellemare, M. F. (2020). Smallholder farmers and contract farming in developing countries. Proceedings of the National Academy of Sciences USA,117,259– 264. Meemken, E. M., & Qaim, M. (2018). Can private food standards promote gender equality in the small farm sector? Journal of Rural Studies,58, 39–51. Ministry of Food and Agriculture (2011). Masterplan study for the oil palm industry in Ghana. Accra, Ghana: Ministry of Food and Agriculture. Mishra, A. K., Kumar, A., Joshi, P. K., & D’Souza, A. (2016). Impact of contracts in high yielding varieties seed production on profits and yield: The case of Nepal. Food Policy,62,110–121. Narayanan, S. (2014). Profits from participation in high-value agriculture: Evidence of heterogeneous benefits in contract farming schemes in Southern India. Food Policy,44,142–157. Neven, D., Odera, M. M., Reardon, T., & Wang, H. (2009). Kenyan supermarkets, emerging middle-class horticultural farmers, and employment impacts on the rural poor. World Development,37(11), 1802–1811. Otsuka, K., Nakano, Y., & Takahashi, K. (2016). Contract farming in developed and developing countries. Annual Review of Resource Economics,8,353–376. Qaim, M., Sibhatu, K. T., Siregar, H., & Grass, I. (2020). Environmental, economic, and social consequences of the oil palm boom. Annual Review of Resource Economics,12, 321–344. Ragasa, C., Lambrecht, I., & Kufoalor, D. S. (2018). Limitations of contract farming as a pro-poor strategy: The case of maize outgrower schemes in upper West Ghana. World Development,102, 30–56. Rao, E. J., & Qaim, M. (2013). Supermarkets and agricultural labor demand in Kenya: A gendered perspective. Food Policy,38,165– 176. Rao, E. J., Brümmer, B., & Qaim, M. (2012). Farmer participation in supermarket channels, production technology, and efficiency: The case of vegetables in Kenya. American Journal of Agricultural Economics,94(4), 891–912. Rhebergen, T., Fairhurst, T., Zingore, S., Fisher, M., Oberthür, T., & Whitbread, A. (2016). Climate, soil and land-use based land suitability evaluation for oil palm production in Ghana. European Journal of Agronomy,81,1–14. Ruml, A., & Qaim, M. (2020). Effects of marketing contracts and resource-providing contracts in the African small farm sector: Insights from oil palm production in Ghana. World Development, 136,105110. 66 RUML and QAIM Ton, G., Vellema, W., Desiere, S., Weituschat, S., & D’Haese, M. (2018). Contract farming for improving smallholder incomes: What can we learn from effectiveness studies? World Development, 104, 46–64. Verhofstadt, E., & Maertens, M. (2014). Smallholder cooperatives and agricultural performance in Rwanda: Do organizational differences matter? Agricultural Economics,45(S1), 39–52. SUPPORTING INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of the article. How to cite this article: Ruml A, Qaim M. New evidence regarding the effects of contract farming on agricultural labor use. Agricultural Economics. 2021;52:51–66. https://doi.org/10.1111/agec.12606