Oil Palm Production and Educational Outcomes: Gender-Differentiated Evidence from Cameroon

Abstract Oil palm production continues to expand in many developing countries in the tropics. Its expansion has been associated with economic gains, but oil palm production could also have immense social implications, especially affecting human capital development with significant labour implications. We use a farm household dataset from a native but emerging oil palm production zone, Cameroon to examine the relationship between oil palm production and child educational outcomes such as enrollment rate, attendance rate and the number of school days missed. Using different analytical techniques, we show that oil palm production is positively associated with the enrollment rates of both boys and girls. We do not find any statistical relationship between oil palm production and attendance rates by gender. However, we find evidence of a strong negative association between oil palm production and the number of school days missed by boys. That is, oil palm production is associated with more school days attended by boys. Exploring the mechanism that could be explaining these results, we show that households may be investing the income gains from oil palm production in the human capital development of their children. Our results are robust over different regression estimators and alternative specifications. We also show that the results are unlikely to be driven by omitted variable bias. These findings have implications on whether oil palm production could stir integrated growth and human capital development, especially in rural areas.


Introduction
The global demand for palm oil has led to the continuous expansion of oil palm plantations in many developing countries (Qaim, Sibhatu, Siregar, & Grass, 2020).With emerging evidence of oil palm led developments in Southeast Asian countries, especially Indonesia and Malaysia, oil palm hotspots are re-emerging in some West African countries like Cameroon which happens to be a native production zone of oil palm (Tabe-Ojong, Ordway, Nkongho, & Molua, 2022a;Ordway, Naylor, Nkongho, & Lambin, 2017).The expansion of oil palm has been shown to contribute to poverty alleviation and rural economic progress in Indonesia and Malaysia which represent the biggest oil palm producers in the world (Krishna & Kubitza, 2021;Santika et al., 2019).However, these economic gains and benefits have been accompanied with immense environmental and social implications (Qaim et al., 2020;Pye, 2019;Abram et al., 2017;Sayer, Ghazoul, Nelson, & Klintuni Boedhihartono, 2012).
While millions of people depend on oil palm plantations for their livelihood, challenges exist around its problematic expansion and exacerbation of inequalities among smallholders, as well as other socioeconomic concerns (Ayompe, Schaafsma, & Egoh, 2021b;Krishna & Kubitza, 2021).From a social perspective, anecdotal evidence suggests that oil palm production may have significant child labour concerns and negative educational and schooling outcomes given that households may involve their children in the production of palm oil (Li, 2018;UNICEF., 2016).This may even be more intense in the context of Cameroon where oil palm production is labour intensive given the various processes involved in the value chain of palm oil (Tabe-Ojong et al., 2022a;Ordway et al., 2017;Nkongho, Feintrenie, & Levang, 2014).Besides, empirical evidence shows that agricultural households that are cash-strapped may not be able to enroll their kids in school or withdraw them from school when faced with production and income shocks (Dillon, 2013;Kruger, 2007).By contrast and beyond these negative implications, oil palm production could also relax the liquidity constraints of smallholder farmers and increase their incomes (Tabe-Ojong et al., 2023;Ayompe, Nkongho, Masso, & Egoh, 2021a;Santika et al., 2019;Krishna, Euler, Siregar, & Qaim, 2017;Euler, Krishna, Schwarze, Siregar, & Qaim, 2017).This income could then be invested in the human capital development of their children such that a link between oil palm production and education could exist (Rist, Feintrenie, & Levang, 2010).Whether this linkage exists is an important issue that warrants empirical answers, but little work explores the relationship between oil palm production and educational outcomes of children.This study attempts to answer this question and provide empirical insights on the oil palm education linkage.
We examine the relationship between oil palm production and educational outcomes.Specifically, we look at enrollment rates, attendance rates and the number of school days missed by children.In the interest of understanding gender differences and implications, we estimate separate models for girls and boys.We use a farm household data from a native oil palm production zone, Cameroon and employ different regression techniques including linear and non-linear models such as ordinary least squares, probit, and logit models as well as the poisson regression model.Our results show that oil palm production is positively associated with enrollment rates for both boys and girls.We do not find any statistically significant relationship between oil palm production and attendance rates for either boys or girls.However, we show that oil palm production is associated with a reduction in the number of school days missed by boys.Exploring the mechanism that could explain these results, we find suggestive evidence that income increases may well explain this relationship.Oil palm farmers may be investing the income gains from their production into the education of their children.These findings are robust to several empirical specifications and estimators.Besides this, we also show that our estimated regression coefficients are stable and unlikely to be driven by omitted variable bias following the procedures of Oster (2019).
We offer three contributions to different aspects in the empirical literature.In the first place, we contribute to the literature on the welfare implications of oil palm production.Many studies have reported the income and welfare effects of oil palm production (Ayompe et al., 2021a;Santika et al., 2019;Krishna et al., 2017;Euler et al., 2017).We add to this strand in the literature by confirming that oil palm production is associated with income increases.Beyond this direct implication, our novel contribution comes from showing that this income may be invested to build the human capital development of children.Related to the first, our second contribution relates more to context.Our study is one of the first to assess the income and educational implications of oil palm in an emerging but native production zone of oil palm.Many previous studies are based on insights from the oil palm led growth and development in Indonesia.However, context matters as the situation in Indonesia is not the same in emerging oil palm producing countries in West and Central Africa 1 (Ruml et al., 2022;Tabe-Ojong, Alamsyah, & Sibhatu, 2023).
The third and most important contribution comes from the gender disaggregation analysis where we examine the educational implications of oil palm production for both boys and girls.To our knowledge, we are one of the first to perform such gender disaggregated analysis along the lines of children.Previous analysis on the gender and time implications of oil palm production exists (Rowland, Zanello, Waliyo, & Ickowitz, 2022;Mehraban, Debela, Kalsum, & Qaim, 2022;Chrisendo, Krishna, Siregar, & Qaim, 2020;Elmhirst, Siscawati, Basnett, & Ekowati, 2017;Tabe-Ojong, 2023).However, all these studies only focused on the spouse of the primary oil palm producer.We take a step ahead to understand the educational implications for children along gender lines.Chrisendo, Siregar, and Qaim (2022) examined the association between oil palm cultivation and education (including school education and higher education) in Indonesia with several rounds of panel data and found that oil palm production is associated with increases in educational expenditures.The authors also differentiate between boys and girls attending school in households with and without oil palm and show that boys in oil palm producing households are more likely to be enrolled and less like to drop out from school than boys in non oil palm producing households, although the difference is not statistically significant.For children of school age, time-savings are important, for reinvestment into school participation and attendance.There may exists significant heterogeneity in schooling outcomes along gender lines arising from cash crop cultivation (Nkamleu & Kielland, 2006).Given that child educational outcomes are related to child labour, though inversely (Putnick & Bornstein, 2015;Beegle, Dehejia, & Gatti, 2009;Guarcello, Lyon, & Rosati, 2006), our findings add some empirical insights to the child labour literature.In the wake of growing calls on child labour, studying the incidence of child labour in commercial agriculture, examining the educational implications of oil palm production remains relevant, and highly justified.Studying the educational implications of Oil palm production could be seen as addressing the fundamentals of comprehensive human development which cuts across most of the SDGs, such as SDG 1 (no poverty), 2 (zero hunger), 3 (good health and wellbeing), 4 (quality education), and 5 (gender equality).Given this, our findings have important implications for rural policy, child development and human capacity development.
The rest of the article is structured as follows.Section two presents the background about oil palm production in Cameroon and establishes a conceptual framework where the specific study hypotheses are discussed.The research design is presented in section three.This is followed by a discussion on the measurement of the outcome variables and the empirical strategy used in the paper.Both descriptive and regression results are presented and discussed in section four and the research concludes in section five with some policy recommendations.

Oil palm production in Cameroon
Cameroon is an archetypical lower middle income country (LMIC) in the ecological zone of the Congo Basin of Central African subregion.It constitutes a huge part of the Congo Basin Forest (CBF) which is home to pristine and evergreen forests (Tabe-Ojong et al., 2022a).Although being very relevant and important for countries in the CBF, deforestation is actively taking place to enable the cultivation of some food and cash crops such as cocoa, banana, rubber, and oil palm (Ordway, Naylor, Nkongho, & Lambin, 2021).Oil palm is a perennial tree crop whose production cuts across time with its supply responding to market diktats for which Cameroon is critical in the global market.Not only does oil palm define the country's export basket, but Cameroon also remains a significant player in the export of other tropical cash-crops such as cocoa, coffee and rubber.Despite their economic importance, export crops from most developing countries are tainted with ecological and social challenges ranging from deforestation to labour-related exploitation which has become an Achilles heel to their market supply.
Cameroon is one of the countries besides Nigeria, Ghana and Congo that are native producers of oil palm.The crop has always been cultivated as a rich source of red palm oil which constitutes an important part of the country's diet (Tabe-Ojong et al., 2022a).Some of the biproducts from oil palm such as palm kernel oil and palm wine also explain the relevance of the crop for many farming households but also domestic consumers (Nkongho et al., 2014).Oil palm is cultivated in Cameroon by both agro-industries as well as smallholder farmers (Tabe-Ojong et al., 2022a).Most of the smallholder farmers are operating on an independent basis where they heavily rely on on-farm processing of oil using artisanal mills.Between 1970 and 1990, there have been some attempts by the government to support smallholder farmers in the country.For instance, the defunct FONADER (Fonds National de Developpement) sponsored smallholder scheme (1978 to 1991) sought to provide technical and financial support to smallholder farmers and link them to agro-industries but they were arguably unsuccessful as the scheme collapsed in 1991 (Nkongho et al., 2015).
In some oil palm producing districts, especially around Eseka in the Littoral region of the country, some farmers are connected to large agro-industries such as SOCAPALM (Soci et e Camerounaise de Palmeraies) which is one of the main oil palm producing agro-industries in the country.These contracts are usually marketing contracts that are mostly concerned with purchasing fresh fruit bunches (FFBs) from farmers, with little or no provision of inputs which is a critical constraint in oil palm production in the country (Tabe-Ojong et al., 2022a).Moreover, most of these contracts are very weak, informal, and opportunistic with farmers not benefiting as expected.Farmers are usually offered fixed prices for the FFBs no matter the season and most contracts are usually not transparent.Similar contracting issues have been reported in the oil palm sector in Ghana by Ruml and Qaim (2021a).Given this, farmers prefer to remain independent, especially as they can harness some of the value added from their FFBs despite the seeming inefficiency.
Besides being a critical part of the consumption bundle of most farm households, oil palm is also a profitable venture and has been shown to be associated with income increases in Cameroon (Tabe-Ojong et al., 2023;Ayompe et al., 2021a).Despite this, farmers face a myriad of constraints including but not limited to lack of capital, labour, technical know-how and little institutional and policy support from the government (Tabe-Ojong et al., 2022a;Ordway et al., 2017;Nkongho et al., 2015).Land is arguably not a binding constraint for farmers as there exist possibilities for land acquisition via customary inheritance, leases, and outright purchases to expand cultivation.However, given that land expansion entails huge investments in labour and capital, area under oil palm is still low as compared to the area owned by farmers 2 .Farmers in the oil palm producing systems in Cameroon also cultivate other cash crops such as cocoa and food crops such as bananas, plantains, yams, as well as fruits and vegetables (Tabe-Ojong, 2023).

Oil palm production and education
To structure our conceptual understanding of the relationship between oil palm production and educational outcomes, we briefly describe some pathways that may help explain this relationship.Oil palm production and child education can be linked through two main ways.First, oil palm production may be associated with education through income increases.As oil palm has been highlighted to be income increasing in Cameroon (Ayompe et al., 2021a;Tabe-Ojong et al., 2023), it is intuitive to expect that these income increases could translate to increased educational expenses for children.Farm households could use these incomes to cater for all education related costs such as school fees, purchase of uniforms and books and feeding costs during school hours.
The second pathway linking oil palm production and child educational outcomes may be child labour.Children in many farm families participate in agricultural production which may well prevent them from attending school (Li, 2018;UNICEF., 2016;Chrisendo et al., 2022).The oil palm sector in Cameroon is labour demanding given that processing is performed onfarm.Given this, the value chain seems to be long as farm households are involved in the entire production process from production through processing to commercialisation.Processing is one activity that is highly resource demanding.Previous analysis from Ghana showed that the provision of marketing contracts can help absorb some of this labour since contracting is associated with the adoption of labour-saving procedures and technologies (Ruml & Qaim, 2021b).The saved labour could then be reallocated to other uses such as educational efforts and offfarm employment.Marketing contracts are not prevalent in the study area in Cameroon, so households rely on family labour (including children) for production but also processing and commercialisation tasks.Children in oil palm producing households are usually involved in fruit picking, harvesting, and processing.Thus, it may be the case that children do not go to school to support oil palm production.
Talking about the participation of children in oil palm production, one would expect some differences along gender lines, especially as most of the tasks involved are labour-intensive.While young boys in oil palm producing households may assist in all production and processing activities including harvesting, young girls may only participate in less labour-intensive activities such as fruit picking and some processing activities.They may be excluded from harvesting which is a more difficult activity.Also, girls, unlike boys, may not even participate in oil palm production as they are normally involved in other household non-farm activities which may even be more time demanding.We thus expect boys to be more involved in oil palm production than girls.However, things may still go the other direction depending on the labour needs of households.For instance in Ethiopia, boys have been shown to be more likley to attend school than girls (Haile & Haile, 2012).

Farm household survey
This analysis is based on a farm household survey conducted between August and September 2021 in the Littoral region of Cameroon.The Littoral region is one of the regions where oil palm production is expanding.It is home to the Ngwei forest which is a crucial hotspot for oil palm production in the country.The sampling framework followed a two-staged procedure where in the first step, 39 villages were randomly selected using the probability proportional to size sampling strategy.These villages are found in both the Sanaga Maritime and Nkam divisions of the Littoral region.About 13-17 households were randomly selected from these villages using the random walk sampling approach.A total of 582 households were surveyed in these 39 villages with the assistance of village chiefs.
Interviews were conducted using survey-based tablets enabling real-time data checks.The questionnaire was programmed on the Survey solutions platform of the World Bank and was administered by 7 enumerators who were trained prior to this activity.Interviews were mostly conducted in French and is some cases the native language of the area.The survey gathered information from households both at the plot and household level.At the plot level, information was collected on input use, management practices, outputs, and other farm characteristics.Most of the other information was collected at the household level.Some of these include socio-economic and demographic characteristics, access to institutional characteristics, wealth and income characteristics, food security and dietary diversity, off-farm income, and labour characteristics.Information was also collected on enrollment and attendance rates of children between the ages of 6 and 15.Beyond these farm and household level interviews, some interviews and discussions were held with village chiefs and some opinion leaders in the village to obtain qualitative and anecdotal insights.Given our interest in the relationship between oil palm production and educational outcomes, we rely mostly on the household survey and collapse the plot level data to the household level when needed.

Measurement of outcomes
We have three main outcomes of interest: enrollment rate, attendance rate and the number of school days missed.The enrollment rate is measured as a dummy that takes the value of 1 for children that are enrolled and 0 otherwise.We asked the enrollment rate for children and youngsters between 6 and 15 years to capture impacts on human capital development.Given that we also look at the attendance rate, we could also infer some insights into child labour.To measure attendance rate, we use a binary scale that takes the value of 1 for children in households that missed classes and 0 otherwise.Besides this, we also computed the number of school days missed per academic year by the children, as a count variable.All these three proxy outcomes are gender-separated to understand implications for both boys and girls.

Summary statistics
Table 1 presents the summary statistics of some of the model variables.To understand differences between oil palm and non-oil palm producers, we also statistically compared these two groups of farmers.As shown on this table, enrollment rates for both boys and girls are similar, although it is higher for girls (87%) than for boys (84%).The school enrollment rates seem to be rather low, even though schooling is compulsory and subsidized by the State.This could be due to little enforcements and follow up especially in many rural and forested areas such as in the study area.Also, the attendance rate seems to be very high as just 4 percent of boys and 6 percent of girls do not attend schools.In terms of the number of school days missed, girls miss 15 school days while boys miss 11 school days.Although enrollment is higher among girls, their attendance is lower than boys, and they also miss more school days than boys.About 56 percent of households in the study areas are cultivating oil palm.We observed significant differences between oil palm and non-oil palm farmers in terms of school enrollment for both boys and girls.We do not find any significant difference in the attendance rate for both boys and girls.However, we find that the number of school days missed by boys are statistically different between oil palm farmers and non-oil palm farmers.Specifically, boys in oil palm producing households are less likely to miss school days than their counterparts in non-oil palm producing households.
Most household heads (73%) are males and are middle-aged with an average age of about 50 years.Household heads also have about 9 years of education which is equivalent to above primary level certificate attainment.Households have sizes of about 5 members and make use of both hired and farm labour for production purposes.Access to institutional services such as extension, credits, and cooperative membership is low (20%).Households have huge land parcels where they cultivate other cash and food crops such as maize, vegetables, legumes as well as root and tuber crops (Molua, Tabe-Ojong, Meliko, Nkenglefac, & Akamin, 2020;Tabe-Ojong, Molua, Ngoh, & Beteck, 2021).

Empirical estimation
Given our interest in examining the relationship between oil palm production and various educational outcomes, we specify a linear regression model of the type: (1) Y i represents the educational outcomes (enrollment rates, attendance rates, and school days missed) for child j in household i. OP is a dummy variable that takes the value of 1 for oil palm production and 0 otherwise.d represents the vector of additional controls and d is the vector of parameter estimates for these controls.e i is the random error term.The parameter of interest is b which shows the relationship between oil palm production and the educational outcomes.A positive b coefficient speaks to a positive association between oil palm production and enrollment and attendance rates (including the number of school days missed), while a negative b coefficient implies oil palm production may be associated with a reduction in both enrollment and attendance rates.Although the outcome variable is a dummy, we estimate ordinary least squares (OLS) regression in the framework of the linear probability model.Since we are interested in the relationship between oil palm production and education, this model avoids identification by functional form that is common with probit and logit models (Angrist & Pischke, 2008).Notwithstanding, we still show the results of these models for robustness purposes.Given that we also have the number of school days missed by children, we also estimate a poisson regression model.
There exist some potential endogeneity threats that may affect b: We discuss some of these threats and highlight how they do (not) affect the analysis and how we controlled for them in the analysis.Endogeneity may arise from reverse causality, measurement error, and unobserved heterogeneity.For the case of reverse causality, oil palm production may increase or reduce enrollment and attendance rates, but it is hard to think of how enrollment and attendance rates can determine oil palm production given that oil palm is a perennial tree, and most plantations are already established.Although labour seems to be a key constraint in establishing new plantations, anecdotal evidence obtained from talks with the farmers show that capital seems to be the main binding constraint.Capital investments in starting new plantations are usually very huge.One could argue that the financial returns from their human capital investments in the children may be used to start new plantations, but this is unlikely to be the case given that most of these investments only pay off in the long run.Measurement error does not seem to matter for the analysis given that we could correctly observe the cultivation of oil palm by households.However, for enrollment and attendance rates, there may be misreporting since school attendance is compulsory.In this case, the bias may be upwards as households may over-report school enrollment and attendance to be in line with the norms and regulations imposed by the State.
For the case of unobserved heterogeneity, it is possible that unobserved factors such as risk, managerial abilities, preferences, and skills may trigger oil palm production and be associated with both oil palm production and educational outcomes.For instance, it is possible that wealthier households grow more oil palm and send their children more often to school because they are rich, and not only because they grow oil palm.Since we only have a one period crosssectional data, it is difficult to satisfactorily control unobserved heterogeneity.However, we do two things to see whether omitted variable bias matters in the analysis: (1) add many control variables and observe coefficient stability with and without these control variables and (2) estimate bounds on which omitted variable bias may explain away the estimated relationships (Oster, 2019).
Despite these, we still use instrumental variable estimators 3 to control for any residual endogeneity.We estimate the following 2SLS regression model: Some of the variables and notations are maintained from above.Following Smale and Mason ( 2014) and Tabe-Ojong, Mausch, Woldeyohanes, and Heckelei (2022b), we use the village oil palm production rate as an instrumental variable (IV) 4 .The choice of this variable is based on the literature highlighting the role of social networks and neighborhood effects in farm production (Di Falco, Doku, & Mahajan, 2020).This IV could serve as a good proxy for information access and flow which is relevant for production.However, we really cannot claim our IV is strong especially as we are using cross-sectional data.Moreover, our IV could suffer from some limitations such as only capturing between and not within variation, omitted variable bias concerns as well as the reflection problem (Kubitza & Krishna, 2020;Tabe-Ojong et al., 2023).Also, the village oil palm production rate may be determined by infrastructure conditions in the village that can also affect economic development and thus child schooling through mechanisms other than own oil palm cultivation.To reduce some of the concerns with the IV, we also use the Lewbel (2012) instrumental variable approach which produces consistent and unbiased estimates in the presence of weak or no instruments.It is a heteroskedasticity based estimator that exploits heteroskedastic covariance restrictions (Baum & Lewbel, 2019;Lewbel, 2012).We estimate an additional regression here where we allow the model to generate internal instruments in addition to our standard IV.At this point, it is important to mention that eliminating all forms of endogeneity is not trivial with cross-sectional data, but as we employ different robust strategies, our analysis may be in order.Nevertheless, we refrain from implying any causality from our results and refer to the resulting estimates as correlations that are suggestive of the relationship between oil palm production and education.

Baseline estimation (OLS model)
We begin this section by presenting the results of the linear probability model that estimates the relationship between oil palm production and educational outcomes.As shown in Figure 1, oil palm production is positively associated with the enrollment of both boys and girls.The results are robust when we include and exclude various controls.Oil palm production is associated with an increase in the enrollment rate of boys by about 12 percentage points.The magnitudes are a little higher for girls where oil palm production increases the probability of enrolling girls by about 13 percentage points.This finding corroborates and provides empirical support to Gehrke and Kubitza ( 2021) who demonstrated the positive externalities of the oil palm boom in increasing returns to investments and higher school enrollment in Indonesia.Increase in returns to education may trigger households to increase the enrollment of school-age children.
Beyond school enrollment, we also examine the attendance rate.As shown in Figure 2, we find little or no statistical relationship between oil palm production and the probability of a child to not attending schools.However, as shown in the fourth column, we establish a positive association between oil palm production and the probability of a female to miss school.Here, girls in households that cultivate oil palm are more likely to miss schools.This result may be because some young girls may be charged with various activities along the value chain of palm oil.This may be because of its long value chain which has many steps from production through harvesting and processing and commercialization.Moreover, oil processing is mainly mechanical using artisanal mills (Ordway et al., 2021;Nkongho et al., 2014).It may be the case that women and girls participate more in the long chains after harvesting like preparing and boiling the nuts for artisanal pressing and packaging after pressing to make ready for the market.Beyond this, it could also be the case that girls are charged with other househld chores and activities which are usually time demenading and may make girls to not go to school .

Oil palm production, enrollment, and attendance rates
In Figures 3 and 4 we present the results of the association of school enrollment and attendance with oil palm production, using instrumental variable (IV) estimators.Beginning with the relationship between oil palm production and enrollment rates, we find consistent and similar results like in the case of the linear probability model.The magnitudes are somewhat slightly larger 5 than in the case of the linear probability model.As these two models estimate average treatment effects (ATE) and late average treatment effects (LATE) respectively, they  A2.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.
speak to the robustness of the estimated relationships.As observed, the LATE estimates represent an upper bound to the ATE estimates and given that we are interested in understanding the link between oil palm production and education, LATE seems to be the more relevant coefficient here.Oil palm production is associated with increases in enrollment rates for both and girls by about 13 and 14 percentage points, respectively.These increases in enrollment rates because of oil palm production, although just correlations, is an important policy finding given the many anecdotal claims that oil palm cultivation could be associated with child labour and the potential negative educational outcomes (Li, 2018; UNICEF., 2016).3. Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.A4.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.
Moving to the attendance rates, we again find similar findings like in the case of the linear probability model that oil palm production is not associated with attendance rates for both boys and girls respectively (Figure 6).

Oil palm production and the number of school days missed
Beyond the implications on enrollment and attendance rates, we also examine how oil palm production affects the number of school days missed.Here, we use the Poisson regression model.As shown in Figure 5, we observe an inverse relationship between oil palm production and the number of school days missed.Boys in households that cultivate oil palm are associated with a reduction in the expected number of missed school days by about 2. For girls in these households, we obtain contrasting results of a positive association with the number of missed school days again by about 2 days.This is an important result given that we observe contrasting correlation results for both boys and girls.As highlighted earlier, this result can be explained by the fact that girls may be more involved in the long value chains of producing, processing and marketing palm oil but also in other off-farm household chores.These findings are in line with earlier studies from Ethiopia that show that male children are more likely to attend school than female children (Haile & Haile, 2012).
Gender differentiated educational outcomes between boy and girls' is common in LDCs and rural households (Baten, Haas, de, Kempter, & Meier zu Selhausen, 2021).This may further perpetuate intergenerational cycles of hardship and inequality that undermine the stability of societies (Elmhirst et al., 2017).Closing the gender gap in education is important for several reasons (Grant & Behrman, 2010), particularly as expected with SDG 3, where both at the national and global levels, economic development is generally associated with a lower gender differential.Children's school performance weaves across most of the UN SDGs, as educational outcomes are captured in the goals and targets of SDGs 1,2,3,4, 5, and 10.The philosophy couched in these goals is aimed to reduce the inequities which may harm the rights and imperil the futures of individual children.Grant and Behrman (2010) show that differences exist in boy  A5.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.
and girl child school enrollment and progress, interestingly, with gender gaps that increasingly favor, rather than discriminate against, females.Among children who have ever attended school, girls younger than 16 years of age are observed to have equal or greater schooling progress than boys of the same age in all regions of the developing world (Grant & Behrman, 2010).

Impact pathway
As we argued in the conceptual framework, one pathway through which oil palm production may be associated with educational enrollment and attendance rates is through income.Previous analysis from Cameroon has shown that oil palm production is profitable (Ayompe et al., 2021a).We estimate some linear models where we regress oil palm cultivation and income to confirm this insight.After that, we also run models to test if income matters for school enrollment and in the final case, we interact income and oil palm production to see the joint association with enrollment.Our income measure includes all sources of household income including farm and non-farm income.To understand aspects of standard of living, we use per capita income.As shown in Table 2, column (1) oil palm production is associated with per capita income.Income is also a significant determinant of enrollment as shown in column (2).Interacting income and oil palm production, we find similar insights pointing to the important and possibly mediating role of income in boosting enrollment and attendance rates of children in oil palm production systems in Cameroon.Previous analysis from India highlights that high income households are less likely to involve their kids in agricultural production (Kruger, 2007;Self and Grabowski 2009) which supports our insights on the positive association between income increases and educational outcomes given that both schooling outcomes and child labour are inversely correlated (Guarcello et al., 2006;Beegle et al., 2009;Putnick & Bornstein, 2015).To the extent that income may explain the relationship between oil palm production and educational outcomes, any negative shock to income may reduce schooling outcomes (Dillon, 2013).Unlike other cash crops, the profitability of oil palm systems, within the domestic and foreign markets, deters the direct link to abusive child labour in Cameroon.These results thus indicate some of the opportunities surrounding oil palm investments in native and forested environments.608 M. P. Jr. Tabe-Ojong and E. L. Molua

Additional results and robustness checks
As described in the empirical strategy, we use Lewbel instrumental variable approach to establish robustness and confirm the use of the IV estimator.The results of the Lewbel based  A7.Additional controls include age and educational level, gender, household size, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.Notes: In specifications (1) and (3), the degree of omitted variable bias is calculated after setting the estimated coefficients to zero.In specifications (2) and (4), the omitted variable bias is assumed to equal the additional controls used, from which bias-adjusted beta coefficients are estimated.
IV estimator confirms the original IV estimates as shown in Figure 6.Here, the magnitudes, and statistical significance of the main regression estimates are maintained.These results give credence to the robustness of the identification strategy.
We also perform Oster bounds regressions where we evaluate coefficient stability and estimate how large unobserved selection has to be to rule out the established relationship between oil palm production and education.To show that omitted variables are not an issue in this analysis, we use the enrollment rates and run controlled and uncontrolled regressions.We then calculate beta and delta values.While the delta value shows how large unobserved selection must be relative to observed selection to explain away the estimated coefficients, the beta value on the other hand estimates how the coefficients will change if unobserved selection is as strong as observed selection (Oster, 2019).Table 3 shows the bias adjusted coefficients and delta values.The bias adjusted coefficients are only slightly lower than the true estimated coefficients, enabling us to already confirm that unobserved selection is not an issue in the analysis.For the delta values, we observe that unobserved selection would have to be three-time observed selection to explain away the estimated coefficients.

Conclusion
An important social welfare goal is getting school-age children to enroll, attend school and stay in class as part of the human capital building process.Based on these motivations, this research sought to assess the relationship between oil palm production and educational outcomes, an overlooked aspect in the literature, yet with significant policy relevance within the purview of the UN SDGs.To better understand gender differences and implications, we estimate separate econometric models for girls and boys.The results are instructive of the dynamics in rural household economies.Oil palm production is positively associated with enrollment rates for both boys and girls, with oil palm production shown to increase enrollment rates of boys and girls, with a possibility to reduce the number of classes missed by boys.These results are important and insightful to the role of oil palm production in inducing economic but also social gains.These results can be explained by increases in income as this could potentially reduce the need for child work and incentivise the human capital development of children.
Given these results, we offer some policy reflections and discussions.Although oil palm is associated with income increases in Cameroon, there still exists many constraints in exploiting and tapping the full potentials of the sector.Some of these constraints relate to on-farm processing, lack of contracting to large agro-industries and commercialization through local (traditional) marketing chains and little institution support from the government.These constraints could even explain the difference between the sector in Cameroon and Indonesia more broadly but also between Cameroon and some African production frontiers such as Ghana where contracting is increasingly taking central stage in oil palm production.Farmers in Cameroon like in other West and Central African countries rely on on-farm processing of FFBs using artisanal mills which are usually technologically inefficient in terms of the extraction of palm oil.Contracting farmers to large agro-industries for processing may help reduce such inefficiencies but may even more be beneficial in saving labour from oil palm and reallocating it to other activities which may improve the welfare of farm families.These welfare improvements could then be invested in promoting school participation as well as meeting other costs related to school registration and attendance.Such positive implications are further indicative of intergenerational human capacity development and progress.Improving the technical capacity of artisanal millers as well as enhancing the quality of on-farm processing of oil palm should be an important policy agenda for development plans in Cameroon given the rapid changes in global demand of palm oil.An improved cooperation and linkage of artisanal processors with large industrialists via different schemes will enhance their access to both modern technology and lucrative markets.This will allow Cameroon to take advantage of these growing demand and

Figure 1 .
Figure 1.OLS estimates of oil palm production and enrollment.Notes: Full model results are presented in the supplementary material, TableA1.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Figure 2 .
Figure 2. OLS estimates of oil palm production and missing classes.Notes: Full model results are presented in the supplementary material, TableA2.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Figure 3 .
Figure 3. IV Estimates of oil palm production and enrollment.Notes: Full model results are presented in the supplementary material, Table3.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Figure 4 .
Figure 4. IV estimates of oil palm production and school days missed.Notes: Full model results are presented in the supplementary material, TableA4.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Figure 5 .
Figure 5. Poisson estimates of oil palm production and number of missed classes.Notes: Full model results are presented in the supplementary material, TableA5.Additional controls include age and educational level, gender, household size, experience in crop production, years lived in the village, extension access, cooperative membership, credit access, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, family labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Figure 6 .
Figure 6.Lewbel estimates of oil palm production and missing classes.Notes: Full model results are presented in the supplementary material, TableA7.Additional controls include age and educational level, gender, household size, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, and food crop production.ÃÃÃ p < 0.01, ÃÃ p < 0.05, Ã p < 0.1.

Table 1 .
Summary statistics of regression variables Notes:The table presents the mean and standard error (parentheses) of the model variables.The value displayed for t-tests are p-values.ÃÃÃ , ÃÃ , and Ã indicate significance at the 1, 5, and 10 percent critical level.

Table 2 .
Oil palm production and income Full model results are presented in the supplementary material, TableA6.Robust standard errors are in parentheses.Additional controls include age and educational level, gender, household size, livestock ownership, farm size, asset index, number of crops cultivated, hired labour, and food crop production.

Table 3 .
Coefficient stability and unobserved selection