Journal of Rural Studies 94 (2022) 385–398 Contents lists available at ScienceDirect Journal of Rural Studies journal homepage: www.elsevier.com/locate/jrurstud Implications of socioeconomic change for agrarian land and labour relations in rural Ghana Fred Mawunyo Dzanku a,*, Dzodzi Tsikata b a Institute of Statistical, Social and Economic Research, University of Ghana, P.O. Box LG 74, Legon, Accra, Ghana b Institute of African Studies, University of Ghana, P.O. Box LG 74, Legon, Accra, Ghana 1. Introduction high rural unemployment, and in elevated rates of tenure insecurities and land use conflicts. Therefore, as pointed out by Tsikata (2009), these This article is concerned with how socioeconomic changes repre- changes in land and labour relations need to be discussed together rather sented by increasing rates of youth schooling, declining rates of reci- than separately as is often the norm in the received literature. procity in agrarian labour relations, and reduced migrant labour flows Perceived returns to schooling influences both private and public have contributed to changes in land and labour relations in smallholder investment decisions. Lessons from the Asian Green Revolution commercial agricultural households in Ghana. Our central argument is demonstrate that farm households tend to invest in children’s education that while private and public investments in education are indispens- using income from increased productivity and commercialisation able, there are important household level labour market related op- (Estudillo and Otsuka, 1999; Estudillo et al., 2006). In Ghana, a number portunity costs associated with such investments, particularly for rural of policies and programmes have been implemented by the state to agrarian households in developing countries where agricultural pro- encourage school enrolment. Examples include the capitation grant, the duction is still largely unmechanized. Therefore, we argue that accel- School Feeding Programme, and since 2017 the Free Senior High School erating the creation of opportunities for increasing returns to schooling policy known as Free SHS. In rural areas of Ghana where agriculture is is vital for ensuring that private investments in education, including the main economic activity, children contribute significantly to both forgone farm labour contributions of youth household members, is farm and non-farm labour and income (Koomson and Asongu, 2016). worthwhile even in the short to medium term. One would thus expect the increase in demand for youth schooling to Labour is a fundamental input for both agricultural and non- affect household farm labour availability. Schooling and child work are agricultural production. This is more so in rural sub-Saharan Africa not always mutually exclusive, but Canagarajah and Coulombe (1999) (SSA) where the use of labour-saving and productivity-enhancing tech- found that increase in demand for schooling significantly lowers chil- nologies is low. In a rural economy with abundant land and labour as dren’s contribution to family labour. Indeed, one of the most important well as limited off-farm employment opportunities, smallholder house- reasons offered by parents for their children not being in school was that holds have relied on own labour supply with little or no hired labour they needed their labour (Ye and Canagarajah, 2002). In this article, we input. As land markets develop due to population growth, increased examine if and how increased demand for youth schooling has modified commercialisation of agriculture, among others, and as farming be- rural agricultural labour markets. comes more individualised, family labour becomes scarce and hired Exchange or reciprocal labour, the arrangement whereby farmers labour markets develop in rural areas (Amanor, 2010). provide non-cash remunerated agricultural labour services to each other Rural-urban migration and labour movement from the agricultural to on a rotating basis, has been an important source of agricultural labour the non-agricultural sector during the process of economic trans- in rural agrarian economies for centuries (Moore, 1975). While there are formation further boosts commercial agriculture, which leads to further social motives for labour exchange, the basic economic motivation arises land and labour commodification. At the same time, the less than opti- in the presence of thin labour (and credit) markets, that is, when hired mum earnings from smallholder commercial agriculture compromises labour markets, for instance, are characterised by a small number of the ability to pay the level of wages and land rents that would support participants due to prohibitive cost of transactions. It has long been the development of healthy agricultural labour and land markets. This documented that changes in the socioeconomic environment influences results in paradoxes such as agricultural labour shortages in contexts of the prevalence of labour exchange (Moore, 1975). In this article, we * Corresponding author. E-mail addresses: fdzanku@ug.edu.gh, fdzanku@gmail.com (F.M. Dzanku), dtsikata@ug.edu.gh, dzodzit@yahoo.co.uk (D. Tsikata). https://doi.org/10.1016/j.jrurstud.2022.07.010 Received 12 August 2020; Received in revised form 21 July 2022; Accepted 28 July 2022 Available online 6 August 2022 0743-0167/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 identify and analyse changes in agrarian labour exchange offerings and cocoyam serve a dual purpose of household consumption and cash in- implications for rural agricultural labour markets and outcomes. We ask come to varying degrees. In the northern study areas, commercial how farm households have been responding to the changes in agrarian agriculture is based on food crop production – yam, rice, maize and land and labour markets, distinguishing between labour-related re- cassava in East Gonja; maize, millet, sorghum, and vegetables (mainly sponses and non-labour-related ones. onions and okra) in Garu. We also examine gender, geography and migration status as sources Aside the north-south contrasts regarding agro-ecology, commerci- of difference in land and labour relations, livelihood outcomes and so- alisation, infrastructure, and economic welfare, there are also differ- cial change. This is in keeping with a body of work which has drawn ences with respect to kinship and descent rules, and socio-cultural attention to these differences (Agbosu et al., 2007; Beals and Menezes, norms. Asunafo North and Kwaebibirem are Akan areas that tradition- 1970; Britwum et al., 2014; Skinner, 2012). With regard to geography, ally operate under matrilineal descent rules where inheritance (partic- while Ghana is heterogeneous in several respects, there is a conspicuous ularly land) is from one’s maternal uncle rather than from father. The and well-researched north-south divide in social and economic devel- opposite is true in Garu and East Gonja where inheritance is through the opment due to several factors including differences in insertion in the patrilineage. While these inheritance rules are dynamic, whether as a colonial political economy, agro-ecology and infrastructure. Because result of policy reforms such as the Intestate Succession Law1 of 1985 or Ghana’s agricultural production systems are predominantly rainfed, the through increasing individualisation of descent rules or by dissent agro-ecological variation has resulted in agricultural production po- (Amanor, 2001), they have implications for land and labour relations tential differences that are conditioned by a unimodal rainfall pattern in because of the interrelationship between land and labour issues in the north compared with a bimodal pattern in the south. This dichotomy agriculture (Tsikata, 2009). precipitated colonial and post-independence economic policies that Table 1 provides a summary of selected district level indicators based treated the north as an area of migrant labour reserve for work in the on Ghana Statistical Service data. Asunafo North is the largest of the four south to boost the export economy that was centred mainly on cocoa and districts by population size. East Gonja has by far the lowest population mining (Hear, 1984; Plange, 1979b; Scully and Britwum, 2019; Shep- density among the four districts; population density is highest in herd, 1981; Thomas, 1973). This dichotomy fits neatly as a microcosm of Kwaebibirem. The differences in population density have implications Samir Amin’s description of Ghana as part of the coastal sub-region of for agricultural land availability and labour market conditions. Africa of the colonial trade economy, where, in our case, southern Ghana The large north-south gap in the proportion of households headed by can be viewed as the ‘rich’ area, and the north as the hinterland that females is suggestive of differences in socio-cultural norms. The Kwae- served as a labour pool for the south (Amin, 1972). The relative paucity bibirem and Asunafo North districts have relatively large migrant pop- of educational facilities and institutions in the north further fuelled la- ulation shares (31% and 39% respectively). This is, in part, due to bour migration southwards (Plange, 1979a). historical labour migration from northern to southern Ghana, which has Differentiation by gender is important because pervasive gender contributed substantially to the growth of the cash crop economy in the gaps persist against women with respect to access to and control of south (Beals and Menezes, 1970; Scully and Britwum, 2019). Aside the productive and reproductive resources and livelihood outcomes north-south movement of agricultural labour, there is also a south-south (Dzanku et al., 2021; Lambrecht et al., 2018). Women have long been movement driven primarily by the demand for cocoa and oil palm land reported to have more difficulty organizing unpaid labour for work on their farms than men (Okali, 1983). Such labour constraints faced by women have been reported in relation to access to family labour (Doss Table 1 and Morris, 2001), particularly male labour (Andersson Djurfeldt, 2018; Selected district level characteristics. Hill and Vigneri, 2014). Indeed, such inequalities exist within farm Indicators (1) (2) (3) (4) households in Ghana where the provision of labour on the husband’s Kwaebibirem Asunafo East Garu farm is considered part of the woman’s conjugal responsibilities (Apu- North Gonja sigah, 2009; Tsikata, 2009). Given the above, one would expect the la- Population 121,698 150,198 117,755 71,774 bour deficit created by increased demand for youth schooling and the Female (%) 50.6 49.2 48.9 52.0 declining availability of exchange labour to affect men and women Population density 151.3 105.2 27.7 106.2 2 differently. (persons/km ) Female headed households 34.1 31.4 14.0 16.3 The rest of the article is structured as follows. The second and third (%) sections describe the study context and data collection procedures, Migrant population (%)⸷ 31.1 39.4 14.5 4.7 respectively. The observed socioeconomic changes in the study areas Youth in school (%)⸷ 21.0 17.2 11.7 11.7 and associated labour commodification issues are discussed in the fourth Female (%) 18.4 14.5 9.2 8.2 section. This is followed in the fifth section by analyses of farmers’ re- Male (%) 23.8 20.0 14.0 15.0 Economically active 72.2 75.1 72.4 77.2 sponses to household and community level changes in labour avail- population (%) ability, implications for farm labour markets, and changing labour Female (%) 70.8 73.1 67.9 77.6 arrangements. The penultimate section deals with gendered heteroge- Male (%) 73.6 77.1 76.9 76.7 neity in farmers’ response to labour and land scarcity and agricultural Employed in agriculture 39.6 60.3 77.3 85.2 (%) labour relations. The last section concludes. Female (%) 33.7 57.7 66.5 82.8 Male (%) 45.6 62.8 86.4 88.4 2. The study context Source: The population and population density figures are from the 2021 pop- ulation and housing census. The rest were compiled from various district In order to capture the geography and gender of social change and analytical reports of the 2010 census because those are not yet available from the the effects on agricultural labour availability and labour relations, we 2021 census. ⸷This is based on the definition of crossing and living outside the situated our case studies in two districts in southern Ghana (Asunafo administrative boundary of one’s birth. North and Kwaebibirem) and two in northern Ghana (Garu and East Gonja). Kwaebibirem and Asunafo are in the semi-deciduous forest zone, which has a bimodal annual rainfall regime. By contrast, East Gonja and Garu are in the Savannah zone where annual rainfall distribution is unimodal. The main commercial crops in the southern districts are 1 This law mandated that a significant share of a man’s property should be cocoa, oil palm, cashew and citrus; maize, cassava, plantain and inherited by his children, which was in conflict with the matriliny. 386 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 as well as to offer labour in richer agricultural areas. 4.1. Youth schooling The GSS data shows that secondary and higher education attendance was higher in the south than in the north, and there were gender gaps in Education is widely considered an investment in human capital, favour of males in all districts (Table 1). Economic activity was which is expected to yield returns that include improved well-being in marginally higher among males in all districts except Garu. Since such general. Indeed, governments, households, and individuals are widely statistics do not include reproductive work, women’s labour contribu- believed to expect educational investments to yield high rates of return tions to social and economic well-being are underreported. Except in (Rolleston and Oketch, 2008; Schultz, 1961). Yet, there are important Kwaebibirem, agriculture employed far more than half (60–85%) of the trade-offs between youth schooling and labour market participation that economically active population in all the districts. Agriculture is more need to be recognised. Aside from state policies aimed at reducing important for employment in the north (and among males) than in the cost-related school dropouts, our qualitative interviews revealed high south (and among females). These gender gaps must be considered private demand and investments in children’s education. This demand within contexts where women make substantial labour contributions to for schooling comes partly from sensitisation campaigns by various Civil agricultural and home production, most of which are unremunerated Society Organizations (CSOs) and Non-Governmental Organizations (Raney et al., 2011). (NGOs), some of which have been targeted at eliminating child labour in cocoa production. However, while children’s education was reported to 3. Data collection have become a higher priority for many families than it was in the past, it has also become a source of financial pressures because of the seasonal This paper draws on qualitative and two rounds of household panel character of agricultural earnings. An assemblyman and farmer from a data collected in 10 villages (Fig. 1) across the four districts described in community in the Kwaebibirem district observed2 section 2. Between January and March 2016, we carried out district and People are spending a lot on school fees. If you come here again and community level qualitative interviews using three data collection see that I have still not completed my new house, it is because school fees techniques: key informant interviews, in-depth interviews and focus is killing me. group discussions (separately for men and women). The key informant The paradox here is a growing sense that increasingly high levels of interviews involved institutions at the district capitals as well as com- education does not deliver the expected levels of entry into work outside munity leaders. The in-depth interviews and focus group discussions agriculture. This is creating disenchantment with education as a source involved men, women, and youth. In all, we conducted 27 interviews at of alternative livelihoods for the next generation, although this has not the district level and 138 at the community level (24 focus group dis- translated into a reduction in school attendance and an increased cussions and 114 in-depth and key informant interviews). The sampling availability of household labour for agriculture. Educational studies scheme ensured that diverse demographic groups based on gender, age, have found that young people with some schooling, no matter the length ethnicity, and migration status were represented. About 44% of the of time spent in education are not likely to take up agricultural as per- qualitative interview respondents were female. manent work (Sumberg et al., 2017). In March 2017, we carried out a household survey in all the com- How prevalent is school attendance by household members who munities after a census of all households to generate a sampling frame. could contribute to family farm labour? The survey collected informa- Depending on village population, we randomly sampled between 22 and tion on schooling for those aged 15 years and older. Here, we focus on 75 households from each community to obtain a representative sample, school attendance by youth (15–24-year-olds) – the age at which they yielding a total survey sample of 484 farm households containing 2,586 are normally expected to be in second cycle and higher educational individuals of which 53% were female. Based on the household census in institutions. There was a gender balance in the proportion of youth in each community, we ensured that female-headed households and male- the sample (50.3% females and 49.7% males). Nationally representative headed households were adequately represented. In March 2020, we rural survey samples show that youth schooling increased from almost carried out a follow-up survey to the same households and successfully 24% in 2017 to about 31% in 2019 (Table 2). About 29% of youth reinterviewed 87% of the original households. We also sampled 348 new belonging to households in our pooled survey sample were attending households for the purpose of studying contract farming, bringing the secondary school and above; the rate of youth schooling was seven total sample for the 2020 survey was 770 households and 4,269 in- percentage points less in the north than in the south (p-value = 0.020). dividuals. Thus the pooled sample contains 1,254 and 6,855 household There was a statistically significant increase of approximately 12 per- and individual observations, respectively. centage points in youth schooling between 2017 and 2020 in our survey sample (Table 2). This increase could be associated with the Free SHS 4. Socioeconomic changes and rising labour commercialisation policy. Table 2 shows statistically significant gender gaps at the 5% level in youth schooling in favour of boys in both the nationally representa- We focus on education as a key socioeconomic variable and examine tive samples and in our own survey sample in 2017. However, the how changes in educational investments and expected returns by gender differences narrowed by 2019/20 in all cases and became sta- households influence land and labour relations in contrasting rural tistically not different from zero except in the northern sample where the agrarian societies of northern and southern Ghana. In particular, we gender gap more than doubled. focus on school attendance by youth (15–24-year-old) household members who could otherwise be an important source of agrarian la- 4.2. Youth schooling, declining migrant labour flows and rising labour bour and the implications for various land and labour arrangements. Our costs argument here is not that schooling is the cause of rural agricultural labour market imperfections, particularly given that unemployment and Labour migration, including seasonal labour movements from the underemployment are still serious problems in rural Ghana (Adeniran north to the south for work on cocoa farms, has historically been an et al., 2020; Dzanku and Hodey, 2022) due, in part, to the persistent important resource for growth in the cocoa sector (Amanor, 2010; Beals seasonality and uncertainties in agricultural production under rainfed and Menezes, 1970; Hill, 1997; Miracle and Berry, 1970). Migrant la- conditions (Charlton et al., 2021). However, our aim is to analyse the bour from northern Ghana has been particularly important in Asunafo opportunity costs of schooling and implications for family farm labour where commercial agriculture is synonymous with cocoa production. relations in order to draw attention to why there is the need to raise marginal returns to youth schooling under the present conditions of high youth unemployment across the African continent. 2 It must be noted that this was before the introduction of the Free Senior High School education policy in 2017. 387 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Fig. 1. Map of Ghana showing the study communities. However, the narrative from our qualitative interviews suggests that the argued that the family labour deficit created by absent youth, the increasing rate of youth schooling has had an impact on labour move- reduced flow of migrant labour from the north, and the concomitant ments from the north: labour scarcities have led to rising cost of hired labour. During FGDs, In the past, people could hire vehicles, go to the north, and bring young however, while some interpreted the scarcity of hired labour to mean men here to work as labourers. Now such things can no longer be done shortage of labourers, others argue that it is more an issue of escalating because parents are making sure that their children stay in school. This has wage rates than availability: affected labour availability in this community. (63-year-old widow and Labourers are there; it is only their price that has increased. If I need one retired farmer, Asunafo North). right now, I will get. (53-year-old migrant farmer, Asunafo North, However, we noted that the youth schooling effect on labour avail- Southern Ghana). ability is not a phenomenon only in the southern cocoa growing area; we If you have money, getting your work done is very simple, but if there is no also found similar narratives of increasing labour scarcity due to money, they [labourers] will be there but you cannot hire them. (27-year- increasing rates of youth schooling in the northern study areas: old male farmer, East Gonja, Northern Ghana). The labour situation in this area is tough because most children are in With lower levels of family labour and higher commercialisation school and do not get enough time to help their parents on the farm. (Elite male farmer, Garu). There is a general rise in land-labour ratio in Ghana (Diao et al., 2014). In theory, this should lead to rising labour cost. Some households 388 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Table 2 the pods and carrying the beans. Average school attendance by 15–24-year-olds (percent). With the higher rates of input (land and labour) and output com- Demeter sample mercialisation in the south (driven mainly by cocoa and oil palm pro- duction) than in the north, one would expect exchange labour (1) (2) (3) (4) (5) availability to be higher in the north. Our qualitative interviews show National Overall South North South – North that, indeed, this is the case. For example, a 60-year-old female farmer in 2017 Grushie Zongo (East Gonja) who has nine grown children living outside Overall 23.5 22.7 26.1 20.9 5.2 her community explains: Boys 26.4 25.5 30.7 23.0 7.7 Sometimes, I use hired labour when my children send me money, but Girls 20.1 20.0 22.2 18.6 3.6 Sex gap 6.3 5.5 8.5 4.4 4.1 I do not have money most of the time so I use exchange labour. With 2020 exchange labour, I only prepare food for the workers. Overall 31.2 34.6 44.2 31.0 13.2 We found that even with the observed decline in exchange labour use Boys 31.8 38.5 48.0 35.8 12.2 in the southern communities, the phenomenon is more common among Girls 30.6 30.5 41.4 25.0 16.4 Sex gap 1.2 8.0 6.6 10.8 4.2 migrant farmers from the north than among natives. A 45-year-old − second-generation migrant farmer in Asunafo North whose parents Note: The national average figures for 2017 are from the rural sample of the 7th migrated from Navrongo in the Upper East Region explains: round (2016/17) of the Ghana Living Standards Survey (GLSS7); those for 2020 Exchange labour is not common now as it was when I was a little boy, are from the 3rd round of Ghana Socioeconomic Panel Survey (2019). Source: Household survey, 2017 & 2020. but for us the northerners, we know ourselves, so we sometimes engage in exchange labour. 3 It must be noted, however, that there were also narratives of rates in the south, we expected the employment of and expenditure on declining availability of exchange labour in the northern study areas, hired labour to be higher in the south, which is what Fig. 2 shows based although to a lesser extent than the south. This was pointed out in some on our survey data.4 The north-south difference in hired labour demand of our in-depth interviews as indicated by a 51-year old married female is reflected in the cost thereof (Table 3). For example, although the by- farmer and trader in Garu: day (daily-paid labour wage) rates in all the districts were above the The labour situation has changed. At first, there was a lot of exchange statutory national daily minimum wage of US$1.86, a daily-paid farm labour, but now by-day (daily-paid labour) is in the lead. worker in the north only received between 28% and 43% of what was The north-south gap in exchange labour availability indicated by the received in the south. qualitative data is observed in the survey data too (Fig. 3); even then, only about a third of all households reported using some exchange la- 4.3. Dwindling reciprocity? bour in the north, compared with about 9% in the south. With increased demand for schooling, and changing agricultural la- Exchange labour has been an important source of labour in Ghana’s bour markets and arrangements induced by the increasing commodifi- agriculture and was even adopted by the state as a model of rural cation of land and labour relations, how have households responded to development in the 1970s (Dadson, 1988). Takane (2000) distinguished the real or apparent labour shortages? It is to this question that we now between two kinds of exchange labour. The first is when a farmer works turn. on another’s farm ad hoc without strict obligations; these are often task-specific such as the breaking of cocoa pods in cocoa growing areas. 5. Households’ responses to social change and labour scarcity The second is when well-organized exchange groups are formed with a specified number of members who offer labour services to each other on Here, we ask how households are responding to the family labour a rotating basis. scarcity due to the increasing demand for child schooling, the scarcity or With declining availability of family labour, exchange labour could rising cost of hired labour, and the declining reciprocal labour avail- be a substitute, but exchange labour availability could also be affected ability exacerbated by increasing individualisation and commodifica- by the absence of youth that could contribute to their household’s ex- tion of land and labour relations. The observed responses can be grouped change labour pool. The increasing commercialisation of land and la- into labour and non-labour responses. bour relations has led to the altering of reciprocal labour services into cash-based labour arrangements Amanor (2010), which explains the rise in hired labour and sharecropping. A middle-aged cocoa farmer in 5.1. Labour-related responses the Kwaebibirem district explains: Most people are not involved in exchange labour anymore because There have been a number of labour-related responses to the re- people are now conscious of money and they want to get money from ported scarcity of labour. By labour-related, we mean responses asso- every activity. ciated with farm labour decisions that adjust for the family labour Even tasks such as breaking cocoa pods and conveying cocoa beans deficit. These include a more intensive use of available household la- to the drying point, which were done using exchange labour, have bour, employing more hired labour, uptake of labour-saving technolo- become commercialised in some areas as indicated by a 47-year old gies, and substituting less labour intensive crops for labour demanding farmer in Asunafo North: crops. We examine some of these options by using the survey data to My father had a large cocoa farm but other farmers helped with har- analyse differences in external labour resources and labour-saving vesting and cracking the [cocoa] pods, but now, we hire labour for cracking technology use between those with and without youth in school. First, we test the null hypothesis that the availability of family farm labour is identical between households with and without youth in 3 Male adult equivalent labour available was higher in the north (2.5) than in school. Consistent with our a priori expectation, the results (Table 4) 5 the south (1.8). Using the ratio of the gross value of crop sales to the gross value show that male adult equivalent (MAE) family labour was about 18% of crops produced as a measure of level of commercial agriculture, our pooled less for households with youth in school than those without youth survey data shows that the mean level of commercialisation in the north was only 46% of the level in the south. 4 Here, and elsewhere, the error bars on the graphs are indicative that the 5 In calculating male-equivalent farm labour, a teen and an adult female difference between the groups is statistically significant at the 5% level if the farmer labour was estimated to be equivalent to 0 0.75 of an adult male’s la- bars do not overlap. bour based on FAO (1999). 389 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Fig. 2. Hired labour use is significantly higher in the south than in the north. schooling (p-value = 0). This labour deficit is, however, only statistically Table 3 significant in the southern sample where the difference is about 21%. Cost of hired farm labour. Next, we examine whether those with youth in school used their District Contract labour (US By-day (daily-paid) casual labour available household labour more intensively and relied more on external $/acre) (US$) sources of labour (exchange and hired) for maintaining their farms Kwaebibirem 16.28 5.81 owing to the shortfall in household labour supply. We find that the re- Asunafo North 13.37 4.07 sults depend on the geographic location of the household. For those East Gonja 7.56 1.74 Garu 7.56 1.63 located in the north youth schooling is not significantly correlated with any of the outcomes except the likelihood of using hired labour, which Source: Demeter qualitative interviews, 2016 was significantly higher in 2020 for households with youth in school. The fact that most of the outcomes show no significant difference in the north could be expected because farming is generally less intensive in Fig. 3. The use of exchange labour is significantly higher in the north than in the south. 390 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Table 4 Differences in various labour responses between households with and without youth in school. Indicator 2017 2020 Overall South North Diff. Overall South North Diff. Family labour available (mae) Overall 2.2 1.9 2.5 ¡0.5 2.1 1.7 2.6 ¡0.9 In school 2.0 1.7 2.5 − 0.8 2.0 1.6 2.6 − 1.1 Not in school 2.3 2.1 2.5 − 0.4 2.4 1.9 2.6 − 0.8 Diff. ¡0.3 ¡0.4 0.1 ¡0.5 ¡0.4 ¡0.3 − 0.0 ¡0.3 Family labour (person-days/mae) Overall 119.4 162.8 78.4 84.4 150.3 218.3 86.0 132.4 In school 138.7 177.1 78.5 98.6 179.9 236.4 86.8 149.7 Not in school 110.7 153.7 78.4 75.3 118.9 183.4 85.5 97.9 Diff. 28.0 23.4 0.1 23.3 60.9 53.0 1.3 51.7 Exchange labour use (%) Overall 20.9 7.8 33.2 ¡25.4 29.9 24.4 35.0 ¡10.6 In school 13.0 3.8 27.5 − 23.7 26.3 22.2 32.9 − 10.7 Not in school 24.4 10.4 34.9 − 24.5 33.7 28.6 36.3 − 7.7 Diff. ¡11.4 ¡6.7 − 7.5 0.8 ¡7.4 − 6.3 − 3.4 − 3.0 Hired labour use (%) Overall 60.0 70.7 49.8 21.0 63.0 82.0 45.2 36.8 In school 68.7 76.3 56.9 19.4 76.5 90.4 53.7 36.7 Not in school 56.0 67.2 47.6 19.6 48.8 65.7 40.0 25.7 Diff. 12.7 9.0 9.3 − 0.2 27.7 24.7 13.7 11.0 Hired labour expenditure (US$/ha) Overall 81.9 132.3 34.4 97.8 135.3 216.1 59.1 157.0 In school 40.8 56.9 15.5 41.4 58.6 72.4 35.8 36.7 Not in school 26.1 38.8 16.5 22.3 33.1 48.4 25.1 23.3 Diff. 14.7 18.1 − 1.0 19.1 25.5 24.0 10.7 13.3 Weedicides expenditure (US$) Overall 17.6 25.0 10.6 14.3 35.7 46.6 25.5 21.1 In school 24.2 32.5 11.3 21.2 41.9 52.7 24.0 28.7 Not in school 14.6 20.2 10.4 9.8 29.3 34.7 26.4 8.3 Diff. 9.6 12.3 0.9 11.4 12.6 18.0 − 2.5 20.4 Note: The numbers in bold font means that the mean difference is statistically significant at the 5% level. A positive difference of the differences means that the gender gap is wider in the south. Source: Household survey, 2017 & 2020. northern Ghana because of the longer dry spell under rainfed conditions. expenditures grew faster in the north between 2017 and 2020 – real In the southern pooled sample, however, consistent with our proposi- expenditures increased by about 86% in the south compared with 140% tion, we find that, on average, households with youth in school used in the north over the period. Table 4 shows a statistically significant their available labour more intensively (about 27% higher) and had difference in weedicide use between households with and without youth access to significantly less exchange labour (7 percentage points less in in school in the southern sample, and this difference increased by about 2017 but not 2020) than households without youth schooling. Exchange 46% between 2017 and 2018 (from about US$12 to US$18). labour use may be less common among households with youth in school The associations between schooling and the various indicators pre- due to the reciprocity involved. sented in Table 4 could be spurious because we did not adjust the results With other conditions remaining the same, households with youth in for factors other than youth schooling that could drive the differences. school are expected to hire in more labour since they have less access to Therefore, we evaluate the associations using regression analysis, which exchange and family labour than their counterparts do. Indeed, the allows us to account for factors such as household socioeconomic and incidence of farm labour hiring was about nine percentage points higher demographic characteristics, commercialisation rates, unobserved in 2017 and increased to about 25 percentage points in 2020 among household-specific heterogeneity or village fixed effects that capture household with youth in school in the south than among those without unmeasured community-specific characteristics.6 such members in school. The difference in average hired labour The marginal effects associated with youth schooling (after adjusting expenditure per hectare between households with and without youth for confounding factors) are shown in Table 5. In the overall sample, the schooling is significant and increased in real terms from about US$18 in effect of youth schooling had a statistically significant effect on three 2017 to US$24. outcomes: about 27% less available household labour, about 29% higher What about the use of labour-saving technologies in response to la- family labour use intensity, and about US$/ha expenditure on hired bour scarcity? A common narrative during our qualitative interviews labour. For two of the outcomes (family labour availability and use in- was the substitution of weedicides for labour during land clearing in tensity), unlike the results reported in Table 4, the youth schooling response to family labour scarcity and the rising cost of hired labour. In the qualitative study, the use of weedicides in response to labour scarcity and the rising cost of labour was mentioned in all study locations: ′ In the past, I had to pay so much to hire labour to clear the land with 6 The general form of the regressions is yit,j = α+ δschoolit + β xit + ci + εit, cutlass, but now it is easier and cheaper to clear the land by spraying with where yit,j is the value of the jth indicator of interest for household i in time t; chemicals. (40-year-old male farmer, Kwaebibirem). schoolit is the indicator variable that takes on the value 1 if the household has a With the introduction of the chemicals, farmers and their children can youth in school and 0 if not; δ is the estimate of interest which shows the effect spray the farm themselves, they do not need to hire labour for land clearing. of youth schooling on y; xit is a vector that contains other factors that could (Middle-aged female farmer, East Gonja). influence the value of y besides school attendance; the corresponding param-eter vector associated with xit is represented by β; ci is the household-specific The survey data shows that, although weedicide expenditures were unobserved heterogeneity, which in the absence of panel data constitutes an significantly higher in the south where labour was scarcer (Fig. 4), omitted variable; εit, is the idiosyncratic error term. 391 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Fig. 4. The use of weedicides as a labour-saving technology is higher in the south. exiting agriculture altogether. Of course, there are constraints associated Table 5 with these responses too; non-farm jobs, for example, may not be Marginal effects of youth absence due to schooling on selected outcomes. available, may be precarious, or household members may not have the Dependent variable (1) (2) (3) required skills. Overall South North On farm size, data from the qualitative interviews provided mixed Family labour available (mae) − 0.269*** − 0.156** − 0.317*** results about changing farm sizes and why. In the south, land scarcity (0.049) (0.067) (0.070) seemed relatively more important than labour shortages in driving Family labour use intensity (log) 0.293*** 0.182** 0.386*** declining farm sizes. This view is exemplified by a 45-year old male (0.065) (0.091) (0.096) migrant farmer and trader in a Kwaebibirem community who held 4.5 Use of exchange labour 0.004 0.014 0.011 (0.060) (0.082) (0.053) ha under cocoa, oil palm and rice: Hired labour expenditure (US$/ha) 8.537** 22.148*** 3.132 Now, land has become very scarce in this community and that is why (4.154) (7.464) (3.672) my farm is small; if land were available, I would have a bigger farm. Weedicide expenditure (US$/ha) 0.733 5.519*** − 1.644* Even the land I have now I got it because I am hardworking and people (0.994) (1.792) (0.880) have given good testimonies about me. Observations 844 410 434 Similar views were expressed during focus group discussions in the Note: The results are from household fixed effects regressions that allow us to Asunafo North communities where land was scarcer than in Kwaebi- model household-specific unobserved heterogeneity. Cluster robust standard birem because of the relatively older cocoa farms and land exhaustion: errors are in parentheses. ***; **, and * denote statistical significance at the 1%, Our fathers came to acquire the lands and they have now divided it among 5% and 10% levels, respectively. All the estimates adjust for household de- their children, so now lands belonging to individuals are smaller than what mographic characteristics (sex of household head, age, level of education, child dependency ratio), wealth, off-farm employment, scale of production, and people had in the past. All the lands that were acquired by our fathers are sources of labour. finished, the sizes of farms belonging to individuals are becoming smaller. (Male FGD, Asunafo North). effects were statistically significant in both the north and south sub- Using the survey data, we examine whether there are significant samples (columns 2 and 3), which shows the importance of adjusting for differences in farm size and non-farm employment between households other confounders using regression analysis. with and without youth in school. The results (Table 6) show no evi- As we found earlier, however, the schooling effects on hired labour dence of youth schooling associated reduction in farm sizes. In fact, we and labour-saving technology were only statistically significant at the observe in the overall sample that mean farm size was significantly 1% level in the south subsample (column 2) – youth schooling was larger for households with youth members in school, a result that is associated with an average of about US$22/ha extra expenditure on driven by the southern subsample, but is not surprising because of hired labour, and almost about US$5/ha more on weedicides as sub- specialization in the production of perennial crops (cocoa and oil palm) stitute for labour. in the south, which is less likely to be affected by current education of young household members. In the north where annual crops dominate households’ crop portfolio, mean farm size is smaller for households 5.2. Non-labour-related responses with youth in school, but the differences in both panel years are statis- tically insignificant. Aside labour-related responses, farm households could adopt non- We also observe that off-farm participation and income are signifi- labour responses in the face of labour scarcity, and by non-labour we cantly higher among households with youth in school. The results show mean responses that are not directly associated with farm labour use, some temporal and spatial nuances – in 2017 the off-farm income dif- including farm size reduction, increased non-farm participation, and ference was driven primarily by the north subsample and by the south 392 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Table 6 Differences in farm size and off-farm income among households with and without youth in school. Indicators 2017 2020 Overall South North Diff. Overall South North Diff. Farm size (ha) Overall 3.2 4.1 2.4 1.6 3.5 4.2 2.9 1.3 In school 3.8 4.8 2.4 2.4 3.8 4.4 2.8 1.7 Not in school 3.0 3.7 2.5 1.2 3.3 3.8 3.0 0.8 Diff. 0.9 1.1 − 0.1 1.2 0.6 0.6 − 0.2 0.8 Off-farm income participation (%) Overall 72.7 87.8 58.5 29.3 71.3 91.2 52.5 38.7 In school 78.6 90.0 60.8 29.2 78.8 92.6 56.1 36.5 Not in school 70.1 86.4 57.8 28.6 63.4 88.6 50.4 38.2 Diff. 8.5 3.6 3.0 0.6 15.4 4.0 5.7 − 1.7 Off-farm income (US$) Overall 715.3 1030.8 417.3 613.4 704.0 1114.1 316.5 797.7 In school 933.6 1064.9 727.7 337.1 892.4 1197.9 389.4 808.5 Not in school 617.1 1009.0 322.0 687.0 504.5 952.6 272.2 680.4 Diff. 316.5 55.9 405.7 − 349.8 387.9 245.3 117.2 128.1 Note: The numbers in bold font means that the mean difference is statistically significant at the 5% level. A positive difference of the differences means that the gender gap is wider in the south. Source: Household survey, 2017 & 2020. subsample in 2020 (Table 6). were unemployed (Ghana Statistical Service, 2022) – this rate of un- It must be noted that these results do not establish causality since employment was higher than the rate in the general population, which both observed and unobserved factors that drive off-farm income could was 12% for males and 16% for females. be correlated with youth schooling. We, in part, mitigate against the risk It is important to emphasize that youth schooling is not the only of spurious correlations by regressing farm size and off-farm income on cause of labour shortages, and that there is some contradiction in the youth schooling and other covariates that may be correlated with the fact that while households complain about labour shortages and the cost outcomes using the fixed effects regression, which allows correlation thereof, the 2021 census shows that a sizable proportion of the rural between household specific heterogeneity and youth schooling (and population are unemployed (about 12% of the population 15 years and other covariates). The results (Table 7) show that although youth older and about 19% of 15–35 year-olds). schooling seemed associated with smaller mean farm size, the associa- In sum, the issue of returns to schooling is vital as Ghanaian parents tion is not strong enough to achieve statistical significance. The effects of generally expect to reap future benefits in the form of remittances, for youth schooling on off-farm participation and income are only signifi- example, from private investments in their children’s education cant in the south subsample where youth schooling is associated with a (Ahiakpor and Swaray, 2015; Lloyd and Gage-Brandon, 1994). How- 6-percentage point higher off-farm participation and about 52% larger ever, our qualitative interviews show that private investments in child off-farm income. In sum, we observe trade-offs associated with in- education does not always yield the expected private returns in the short vestments in youth schooling but, as with labour-related responses, most to medium term, as lamented by a 50-year-old single mother, trader and of the strong effects are observed in the southern study areas where farmer in Kwaebibirem: agricultural production is more intensive. The above results show the presence of household level farm-related I spent so much on my daughter’s school fees in nursing training. I had to opportunity costs associated with the absence of youth household stop by building along the way to concentrate on her school fees, but she is members due to schooling. However, investments in education are ex- still unemployed and at home two years after graduation … she still de- pected to yield benefits that outweigh these costs (Raju and Younger, pends on me [financially]. 2022). Nonetheless, the high rates of unemployment, even among educated youth (Baah-Boateng, 2015), could reduce expected economic 6. Gendered heterogeneity returns to education. The 2021 population census shows that about 17% of male and 22% of female youth (aged 15–35 years) in the labour force Here, we assess whether the articulated social change in terms of the burden of child schooling and the attendant farm labour deficits are different for men and women. We also examine gender gaps, if any, in Table 7 farm labour availability and demand for hired labour as well as explore Marginal effects of youth schooling on farm size and off-farm employment. gender differences in farmers’ response to labour scarcity. Where the Dependent variable (1) (2) (3) survey data allows, we examine both intra– and inter-household gender Overall South North differences, with the later distinguished by household headship. Farm size (log) − 0.084 − 0.137 − 0.065 (0.064) (0.106) (0.084) 6.1. Gendered land and labour arrangements Off-farm participation 0.077 0.064** 0.027 (0.055) (0.032) (0.093) Gender roles in agriculture are dynamic, with some evidence from Off-farm income (log) 0.308 0.522** 0.164 (0.235) (0.261) (0.341) Ghana (Lambrecht et al., 2018) suggesting that some changes have Observations 844 410 434 occurred. For instance, both men and women are venturing into the Note: The results are from household fixed effects regressions that allow us to production of crops that were hitherto considered men or women’s model household-specific unobserved heterogeneity. Cluster robust standard crops. Here, we first examine gender differences in farm size and the errors are in parentheses. ***; **, and * denote statistical significance at the 1%, various farm labour arrangements (Table 8). As could be expected, 5% and 10% levels, respectively. The estimates adjust for household de- female-headed households (FHHs) cultivated less land than mographic characteristics (sex of household head, age, level of education, male-headed households (MHHs) – the gender gap was about 59% in household composition), wealth, sources of labour, and time effect. 2017 and increased to 84% in 2020. The gender gaps are significant in 393 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Table 8 Comparison of selected indicators between male– and female-headed households. Indicators 2017 2020 (1) (2) (3) (4) (5) (6) (7) (8) Overall South North Diff. Overall South North Diff. Farm size (ha) Male 3.7 5.0 2.8 2.2 4.1 5.2 3.4 1.8 Female 2.3 2.9 1.3 1.6 2.2 2.8 1.3 1.4 Diff 1.4 2.1 1.5 0.6 1.9 2.4 2.0 0.4 Family labour available (mae) Male 2.5 2.1 2.7 ¡0.6 2.4 1.8 2.8 ¡1.0 Female 1.7 1.7 1.6 0.1 1.7 1.5 2.1 ¡0.7 Diff 0.8 0.5 1.1 ¡0.7 0.7 0.3 0.7 − 0.3 Exchange labour use (%) Male 23.9 9.2 34.1 ¡24.9 32.6 26.8 36.9 ¡10.1 Female 14.3 5.8 29.8 ¡24.0 23.7 20.7 28.6 − 7.8 Diff 9.6 3.4 4.3 − 0.9 9.0 6.1 8.3 − 2.2 Hired labour expenditure (US$/ha) Male 31.6 53.8 15.9 37.9 42.7 62.2 28.4 33.8 Female 28.6 34.8 17.3 17.4 54.0 67.3 31.7 35.7 Diff. 2.9 19.1 − 1.4 20.5 − 11.3 − 5.2 − 3.3 − 1.9 Weedicide expenditure (US$/ha) Male 7.9 13.3 4.2 9.1 11.9 15.3 9.5 5.8 Female 7.7 9.2 4.9 4.2 7.8 7.3 8.7 − 1.3 Diff 0.3 4.1 − 0.8 4.9 4.1 8.0 0.8 7.2 Note: Boldface numbers indicates statistically significant difference at the 5% level or less. A positive difference of the differences means that the gender gap is wider in the south. both the north and south subsamples. gender gaps to be wider in the north than in the south because of more Table 8 compares mean values of various types of labour input and binding sociocultural norms against women in the north (Ragsdale et al., labour-saving technology use between MHHs and FHHs. In general, fe- 2018) as well the confluence between the scarcity and cost of labour and male farmers in Ghana tend to have less access to family farm labour the enforcement of sociocultural norms given that the recourse to norms than men (Doss and Morris, 2001), particularly male family labour that exclude some social groups particularly women tend to be higher (Andersson Djurfeldt, 2018; Hill and Vigneri, 2014). Thus, factors that when resources are scarce, (Hazarika et al., 2015). However, we did not reduce the availability of family labour such as youth schooling could find any significant consistent pattern of north-south differences in the either increase women’s family labour deficits or affect them less than gender gaps – some of the gender gaps are significantly wider in the men, supposing they have equal access to substitute labour. In the north (male labour in 2017) while others are wider in the south (hired overall sample, FHHs had less family farm labour available, participated labour in 2017 and labour-saving technology in 2020). less in exchange labour, spent about US$19/ha less on hired labour in Looking within the household (Table 9), we observe, as could be the south in 2017, and spent about US$8/ha less on weedicides as a expected, that women’s plots are smaller than men’s are. Comparing the labour-saving technology in the south in 2020. gender gaps in farm size (Tables 8 and 9), we observe that the between Table 8 also shows geography-specific differences. We expected the household gender gaps are wider than the within household gaps. As Hill Table 9 Mean intra-household gender differences. Indicators 2017 2020 (1) (2) (3) (4) (5) (6) (7) (8) Overall South North Diff. Overall South North Diff. Plot size (ha) Men’s plot 1.8 2.2 1.5 0.7 1.7 2.0 1.4 0.6 Women’s plot 1.3 1.5 1.0 0.5 1.1 1.4 0.8 0.6 Diff. 0.6 0.7 0.5 0.2 0.6 0.7 0.6 0.0 Person-days of family labour Men’s plot 78.6 101.3 56.6 44.8 82.3 114.5 54.6 59.9 Women’s plot 65.7 74.4 52.2 22.2 71.2 89.1 49.0 40.1 Diff. 12.9 27.0 4.4 22.6 11.2 25.4 5.6 19.8 Exchange labour use (%) Men’s plot 15.5 6.0 24.6 − 18.6 21.7 18.7 24.3 − 5.6 Women’s plot 11.7 4.1 23.4 − 19.3 17.8 12.7 24.2 − 11.6 Diff. 3.8 1.9 1.2 0.7 3.9 6.0 0.1 6.0 Hired labour expenditure (US$/ha) Men’s plot 30.8 45.9 16.1 29.9 41.9 65.7 21.4 44.3 Women’s plot 50.6 70.3 19.9 50.5 36.9 47.7 23.5 24.3 Diff. − 19.8 − 24.4 − 3.8 − 20.6 5.1 18.0 − 2.0 20.1 Using weedicide (%) Men’s plot 46.2 60.4 32.4 28.0 60.8 51.1 69.2 − 18.1 Women’s plot 40.4 44.5 34.0 10.5 44.1 38.0 51.6 − 13.6 Diff. 5.8 15.9 − 1.6 17.5 16.8 13.2 17.6 − 4.5 Note: Boldface numbers indicates statistically significant difference at the 5% level of less. A positive difference of the differences means that the gender gap is wider in the south. 394 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 and Vigneri (2014), we find that women have access to smaller quan- tities of family labour than men but that the gender gap is statistically significant only in the southern subsample where labour is scarcer. On external sources of farm labour, we found that exchange labour was more commonly used on men’s plots than women’s plots in the overall sample. In the region-specific samples, however, the gender difference is only statistically significant in the south in 2020 (column 6 of Table 9). In addition, while average hired labour expenditure on women’s plot was higher by about US$20 in the overall sample in 2017, the intra-household gender differences are not statistically significant. In 2020, however, hired labour expenditures on men’s plot were higher than on women’s plots, particularly in the southern subsample where there was a statistically significant US$18/ha more expenses on hired labour on men’s plots, on average. The use of labour-saving technology was also significantly higher by about 16 percentage points on men’s plot in 2017; in 2020 the gaps were significantly larger on men’s plots in both the south (about 13 per- centage points) and north (almost 18 percentage points) subsamples. As others (Andersson Djurfeldt, 2018; Doss and Morris, 2001; Hill and Fig. 5. Percent of cropland owned and used by women in heterosexual Vigneri, 2014), our results so far suggest that women have fewer claims couple households. over household and exchange labour, which triggers their need to hire in more labour than men did in 2017. However, it seems that women could areas, and that within each region, the shares were lower in districts not sustain the use of more hired labour on their plots three years hence. with relatively more severe land and labour scarcity. Thus, the quali- Conventional wisdom suggests that women’s participation in paid tative and quantitative findings told a similar story, that women were agricultural labour markets are limited. Our qualitative data suggest hiring-out their labour more in areas where they had less control over that women’s participation in hired farm labour markets could be con- land, other factors remaining similar. strained by factors that include preference for male hired labour. This is We used the survey data to seek further evidence of association be- the case even in situations of reported increasing labour scarcity. A tween control over land and female participation in hired labour market common response from the in-depth interviews and FGDs to the ques- as labourers. Although we cannot show causality, we can show whether tion of why male hired labour is generally preferred was that ‘women a correlation exists or not. One would expect more female hired labour cannot weed because they are not strong enough’. Yaro et al. (2017) service in the north, particularly in Garu where women’s ownership and reported similar views in relation to agricultural related tasks in western control over land was lowest. It is striking that indeed the highest Ghana. A 46-year-old male migrant farmer in Asunafo North whose incidence of female hired labour services occurs in Garu even though the household cultivated 4 ha to cocoa, plantain and citrus stated: same was not true in East Gonja (Fig. 6), because farm sizes were rela- I do not work with women on my farm. Women can weed but they tively larger in East Gonja and women supply relatively more labour to are not that strong and they take more days to work. their husband’s plots than in Garu. Our qualitative studies found that in We found such sentiments more prevalent in the south than in the East Gonja, many women were using their husband’s fallow lands to north. In fact, we found in some villages in the north that the most grow their own crops. common hired labour groups are constituted by women, but this was reported as a recent development. In terms of by-day labour [daily-paid labour], women are in the ma- 6.2. Are the opportunity costs of youth schooling gendered? jority, from sowing to harvesting. I know the terrain very well so I know what is happening. (Assemblyman, Garu). Because of the relatively lower levels of family labour available to There is an underlying reason for this; our in-depth interviews show FHHs and on plots operated by women, one could conjecture that the large gender gaps in land ownership and control in favour of men in social change related to youth schooling could affect women and men communities where female farm labourers are common. If women have differently. As we have shown, the opportunity costs of youth absence access to land at all, their portions tend to be much smaller than men’s. This suggests that women are ‘forced’ to make a living by hiring out their labour on contracts. An elite commercial farmer in Garu stated: Apart from family labour, women are the most commonly used labourers because they are readily available and ready to work for money. They are common because most of them do not have lands and only make income from hiring out their labour on by-day basis. In keeping with the evidence from the qualitative survey, we found gender gaps in plot ownership and size to be widest in Garu where fe- male hired labourers are most common. The survey collected data on whether plots were operated by husbands, wives or jointly. We found that, in the overall sample, only 11% of the 672 plots cultivated within couple households were operated by wives (74% were operated by husbands and 14% jointly).7 Fig. 5 plots the share of cultivated land operated by wives within the household. We observed that the proportions were lowest in the northern study 7 A plot operator is considered the person who ‘owns’ the crops on the plot and is in charge of managing the plot. Fig. 6. Incidence of females hired farm labour services. 395 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 due to schooling could manifest in the form of smaller farm sizes, higher erogeneity and other covariates, we find evidence of neither inter– nor expenditures on hired labour, and higher dependence on labour-saving intra-household gender differences in the opportunity costs of youth technologies. The link with farm size is harder to establish even with schooling (Table 10). That is, we cannot reject the null hypothesis that panel data, particularly for perennial crops, since it is possible that the the effect of youth schooling on farm size, hired labour and labour- farms were established before school enrolment of youth in our sample. saving technology expenditures are not different for women and men. Our hypothesis is that women bear a higher cost of absent household In sum, these results show that the opportunity cost of youth schooling, members due to schooling, conditional on scale of production, wealth, in terms of potential farm labour forgone, is not significantly different and the availability of household and external labour. We test this hy- for women than for men. pothesis by estimating regression models with farm size, hired labour expenditure per hectare, and labour-saving technology use as dependent 7. Conclusion variables. The explanatory variable of interest is an interaction term between youth schooling and gender represented by either sex of In this article, we have used qualitative and quantitative methods to household head or sex of plot operator.8 This specification allows us to show important social and economic changes in the structure and compare the average marginal effects across groups with and without conduct of agrarian land and labour markets. We show that some of the youth members in school. observed changes and implications are associated with farm household Table 10 reports the results that evaluate the presence of gender investments in child schooling beyond the basic level, which is precip- differences in the effect of youth schooling on farm size and expendi- itated by expected private returns and state programmes. We have tures on hired labour and labour-saving technology (weedicides). Col- demonstrated that, as expected, the increased demand for youth umn 1 shows the estimated differences in effect between FHHs and schooling has affected farm labour availability for smallholder farming MHHs; column 2 reports the corresponding differences between female because post-basic education and youth contribution to family labour managed plots (FMP) and male managed plots (MMP) within the same are competitors. In effect, the increased demand for schooling has households. If gender gaps are present, the estimate of δ should be significantly lowered youth family labour contributions. significantly different from zero, if not then we conclude that there is no Moreover, the demand for schooling in northern Ghana has also evidence of a gender gap. After adjusting for household-specific het- contributed to a decline in the flow of migrant labour from northern Ghana, a region that has long served as a labour reserve for commercial Table 10 agricultural growth in the south. At the same time, reciprocal labour Gender gaps in the opportunity cost of youth schooling. exchange arrangements have all but dried up, particularly in southern Inter-household Intra-household Ghana, due to increasing individualisation and commodification of land and labour relations. These changes have led to rising cost of hired la- (1) (2) (3) (4) bour, triggering responses from agricultural households. South North South North Youth schooling was associated with smaller mean farm sizes, higher Panel A: Farm size (log) intensity of family labour use, lower likelihood of exchange labour use, Female (γ1) − 0.299** − 0.368*** − 0.448*** − 0.510*** higher mean expenditures on hired labour, and higher likelihood of (0.150) (0.104) (0.125) (0.119) labour-saving technology uptake. Thus, we have shown that youth School (γ2) − 0.188 − 0.089 − 0.099 0.033 (0.118) (0.074) (0.068) (0.088) schooling (beyond the basic level) is associated with significant Female x School (δ) 0.032 0.214 0.094 − 0.033 household-level opportunity costs. We show that most of the findings (0.155) (0.142) (0.100) (0.155) exhibit difference across female and male headed households, between Panel B: Hired labour (US$) wives and husbands in the same household, and between the north and Female (γ1) − 3.614 − 0.223 7.853 4.223 south. Our qualitative evidence showed that the opportunity cost of (8.670) (6.366) (11.372) (4.091) School (γ2) 18.586** 7.264 23.291*** 3.848 youth schooling was higher for women than for men; that where there (8.739) (5.016) (8.177) (3.433) are gender gaps with respect to the opportunity cost of youth schooling, Female x School (δ) 9.409 − 3.766 − 2.792 − 2.685 land and labour scarcity tended to exacerbate them. However, the (12.860) (8.781) (19.756) (6.058) qualitative evidence of gendered opportunity cost of youth schooling did Panel C: Weedicides (US$) Female (γ1) − 3.405 − 7.222*** 6.009 1.821 not hold up in the quantitative analysis after adjusting for household − − (2.365) (1.240) (4.505) (1.776) specific heterogeneity, suggesting that both men and women bear School (γ2) 6.911*** − 1.701 4.211 − 1.423 similar opportunity costs of the absence of youth household members (2.301) (1.076) (4.229) (1.624) due to post-basic education. Female x School (δ) − 3.314 0.727 − 2.590 0.996 A number of lessons could be distilled from our study for agricultural (2.828) (1.391) (5.579) (2.897) Observations 410 434 582 712 and rural development policy and practice. The first is the need for accelerating Ghana’s policy agenda of agricultural modernisation, Note: The standard errors (in parentheses) are robust to household clustering in particularly labour-saving technology availability and access as agri- the intra-household level analysis. ***; **, and * denote statistical significance at cultural labour becomes scarcer. While the observed family labour the 1%, 5% and 10% levels, respectively. All the estimates adjust for household demographic characteristics (age of household head/plot holder, level of edu- scarcity has led to increased agricultural wage rates, as would be ex- cation, child dependency ratio), wealth, off-farm employment, scale of pro- pected, this has not necessarily increased the supply of agricultural duction, sources of labour. wageworkers as many young people migrate or prefer non-farm work, including artisanal and small-scale mining, where available. Besides, some evidence shows that youth do not see small-scale farming in its current form as compatible with ‘modern life’ because they are better educated than their parents (Sumberg et al., 2017). 8 The general form of the model is: Second, systematic efforts such as expanding labour intensive sectors y α ′γ female γ school δ female school β X c ε , where y of the economy (including the rural economy) through prudent capital it = + 1 it + 2 it + ( it × it) + i + i + i it is the response variable of interest (i.e., farm size, per hectare hired labour investments in rural infrastructure and agroindustry could help increase expenditure, and the use of labour-saving technology) either at the household private returns to schooling by reducing youth unemployment. Indeed, or plot levels, female is the gender indicator captured as sex of household head if agriculture is sufficiently modernized to remove the drudgery, in the inter-household analysis or sex of plot user in the intra-household anal- educated young people could see the sector as one that is compatible ysis. If there is a gender gap δ should be significantly different from zero. with ‘modern life’ (Sumberg et al., 2017). 396 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Third, as others (e.g., Bryceson, 2019) have shown, the persistent Baah-Boateng, W., 2015. Unemployment in Ghana: a cross sectional analysis from inter– and intra-household gender gaps with respect to access, owner- demand and supply perspectives. African Journal of Economic and Management Studies 6, 402–415. ship and control over productive resources including labour point to the Beals, R.E., Menezes, C.F., 1970. Migrant labour and agricultural output in Ghana. Oxf. need for relentless campaigns that discourage retrogressive gender Econ. Pap. 22, 109–127. norms and practices. The practice whereby available household labour Britwum, A.O., Tsikata, D., Akorsu, A.D., Ako, M.A., 2014. Gender and Land Tenure in Ghana: A Synthesis of the Literature. ISSER, Institute of Statistical, Social & is disproportionally distributed within the household with a focus on the Economic Research, University of Ghana. husband’s plot although women contribute significantly to child Bryceson, D.F., 2019. Gender and generational patterns of African deagrarianization: schooling means that women face the rising prices in hired labour evolving labour and land allocation in smallholder peasant household farming, markets disproportionally. Aside raising awareness about the regressive 1980–2015. World Dev. 113, 60–72. Canagarajah, S., Coulombe, H., 1999. Child Labor and Schooling in Ghana. The World nature of such gender norms, empowering women by boosting their Bank. capacity for gainful employment in non-farm rural enterprises is Charlton, D., Rutledge, Z., Taylor, J.E., 2021. Chapter 77 - evolving agricultural labor essential as non-farm earnings help moderate the gender gaps. The fact markets. In: Barrett, C.B., Just, D.R. (Eds.), Handbook of Agricultural Economics. Elsevier, pp. 4075–4133. that gender inequality in child education within households has nar- Dadson, J., 1988. The need for cooperative reorientation: the Ghanaian case. In: rowed considerably could help create an environment for more equi- Hedlund, H. (Ed.), Cooperatives Revisited. Scandinavian Institute of African Studies, table gender norms. Uppsala, p. 173. Diao, X., Cossar, F., Houssou, N., Kolavalli, S., 2014. Mechanization in Ghana: emerging demand, and the search for alternative supply models. Food Pol. 48, 168–181. Credit author statement Doss, C.R., Morris, M.L., 2001. How does gender affect the adoption of agricultural innovations? The case of improved maize technology in Ghana. Agric. Econ. 25, 27–39. Fred Mawunyo Dzanku: Conceptualization, Methodology, Formal Dzanku, F.M., Hodey, L.S., 2022. Livelihood Outcomes of Agricultural analysis, First draft preparation.; Dzodzi Tsikata: Conceptualization, Commercialisation, Women’s Empowerment and Rural Employment. APRA Working Reviewing and Editing Paper 92. Future Agricultures Consortium, Brighton, UK. Dzanku, F.M., Tsikata, D., Ankrah, D.A., 2021. The gender and geography of agricultural commercialisation: what implications for the food security of Ghana’s smallholder Data statement farmers? J. Peasant Stud. 1–30. Estudillo, J.P., Otsuka, K., 1999. Green revolution, human capital, and off-farm employment: changing sources of income among farm households in central Luzon, The data employed for this article is available from the authors upon 1966-1994. Econ. Dev. Cult. Change 47, 497–523. request. Estudillo, J.P., Sawada, Y., Otsuka, K., 2006. The Green Revolution, development of labor markets, and poverty reduction in the rural Philippines, 1985–2004. Agric. The codes for replicating the results contained in the article are Econ. 35, 399–407. available from fdzanku@ug.edu.gh. FAO, 1999. Agricultural Censuses and Gender Considerations - Concept and Methodology. Food and Agriculture Organization of the United Nations, Rome, Italy. Ghana Statistical Service, 2022. Ghana 2021 Population and Housing Census General Declaration of competing interest Report Volume 3E: Economic Activity. Ghana Statistical Service, Accra, Ghana. Hazarika, G., Jha, C.K., Sarangi, S., 2015. The Role of Historical Resource Scarcity in Modern Gender Inequality. Louisiana State University Baton Rouge, LA. None. Hear, N.V., 1984. ‘By-day’boys and dariga Men: casual labour versus Agrarian capital in Northern Ghana. Rev. Afr. Polit. Econ. 11, 44–56. Hill, P., 1997. The Migrant Cocoa-Farmers of Southern Ghana: A Study in Rural Data availability Capitalism. LIT Verlag Münster. Hill, R.V., Vigneri, M., 2014. Mainstreaming gender sensitivity in cash crop market Data will be made available on request. supply chains. In: Quisumbing, A.R., Meinzen-Dick, R., Raney, T.L., Croppenstedt, A., Behrman, J.A., Peterman, A. (Eds.), Gender in Agriculture: Closing the Knowledge Gap. Springer, Dordrecht, The Netherlands, pp. 315–342. Acknowledgements Koomson, I., Asongu, S.A., 2016. Relative contribution of child labour to household farm and non-farm income in Ghana: simulation with child’s education. Afr. Dev. Rev. 28, 104–115. The authors are grateful to the Ghana Demeter team made up of Lambrecht, I., Schuster, M., Samwini, S.A., Pelleriaux, L., 2018. Changing gender roles in Martha Awo, Daniel Ankrah, Dzifa Torvikey, Promise Eweh, Peter agriculture? Evidence from 20 years of data in Ghana. Agric. Econ. 49, 691–710. Atupare, Irene Tagbor, and all the research assistants who helped during Lloyd, C.B., Gage-Brandon, A.J., 1994. High fertility and children’s schooling in Ghana: sex differences in parental contributions and educational outcomes. Popul. Stud. 48, the fieldwork and transcription. The Demeter project was funded 293–306. generously by the Swiss Programme for Research on Global Issues for Miracle, M.P., Berry, S.S., 1970. Migrant labour and economic development. Oxf. Econ. Development (r4d programme). Pap. 22, 86–108. Moore, M., 1975. Co-operative labour in peasant agriculture. J. Peasant Stud. 2, 270–291. References Okali, C., 1983. Cocoa and Kinship in Ghana: the Matrilineal Akan of Ghana. Routledge. Plange, N.K., 1979a. ‘Opportunity cost’and labour migration: a misinterpretation of Adeniran, A., Ishaku, J., Yusuf, A., 2020. Youth employment and labor market proletarianisation in northern Ghana. J. Mod. Afr. Stud. 17, 655–676. vulnerability in Ghana: aggregate trends and determinants. In: McLean, M.L. (Ed.), Plange, N.K., 1979b. Underdevelopment in northern Ghana: natural causes or colonial West African Youth Challenges and Opportunity Pathways. Palgrave Macmillan, capitalism? Rev. Afr. Polit. Econ. 6, 4–14. Cham, Switzerland, pp. 187–211. Ragsdale, K., Read-Wahidi, M.R., Wei, T., Martey, E., Goldsmith, P., 2018. Using the Agbosu, L., Awumbila, M., Dowuona-Hammond, C., Tsikata, D., 2007. Customary and WEAI+ to explore gender equity and agricultural empowerment: baseline evidence Statutory Land Tenure and Land Policy in Ghana. Institute of Statistical Social and among men and women smallholder farmers in Ghana’s Northern Region. J. Rural Economic Research (ISSER), University of Ghana, Accra technical publication No. Stud. 64, 123–134. 70. Raju, D., Younger, S.D., 2022. Benefits and Costs of Public Schooling in Ghana. Policy Ahiakpor, F., Swaray, R., 2015. Parental expectations and school enrolment decisions: Research Working Paper;10017. World Bank, Washington, DC. evidence from rural Ghana. Rev. Dev. Econ. 19, 132–142. Raney, T., Anríquez, G., Croppenstedt, A., Gerosa, S., Lowder, S.K., Matuschke, I., Amanor, K., 2001. Land, Labour and the Family in Southern Ghana: a Critique of Land Skoet, J., 2011. The Role of Women in Agriculture. Food and Agriculture Policy under Neo-Liberalisation. Nordic Africa Institute. Organization of the United Nations. Amanor, K.S., 2010. Family values, land sales and agricultural commodification in south- Rolleston, C., Oketch, M., 2008. Educational expansion in Ghana: economic assumptions eastern Ghana. Africa. Journal of the International African Institute 80, 104–125. and expectations. Int. J. Educ. Dev. 28, 320–339. Amin, S., 1972. Underdevelopment and dependence in black Africa: historical origin. Schultz, T.W., 1961. Investment in Human Capital. The American economic review, J. Peace Res. 9, 105–119. pp. 1–17. Andersson Djurfeldt, A., 2018. Assets, gender, and rural livelihoods. In: Andersson Scully, B., Britwum, A.O., 2019. Labour reserves and surplus populations: northern Djurfeldt, A., Dzanku, F.M., Isinika, A. (Eds.), Agriculture, Diversification, and Ghana and the eastern cape of South Africa. J. Agrar. Change 19, 407–426. Gender in Rural Africa: Longitudinal Perspectives from Six Countries. Oxford Shepherd, A., 1981. Agrarian change in northern Ghana: public investment, capitalist University Press, Oxford, UK, pp. 54–80. farming and famine. J., H., P., R., G., W. In: Rural Development in Tropical Africa. Apusigah, A.A., 2009. The gendered politics of farm household production and the Palgrave Macmillan, London, pp. 168–192. shaping of women’s livelihoods in Northern Ghana. Feminist Africa 12, 51–67. 397 F.M. Dzanku and D. Tsikata J o u r n a l o f R u r a l S t u d i e s 94 (2022) 385–398 Skinner, E.P., 2012. Labour migration and its relationship to socio-cultural change in Tsikata, D., 2009. Gender, land and labour relations and livelihoods in sub-Saharan mossi society. Africa 30, 375–401. Africa in the era of economic liberalisation: towards a research agenda. Feminist Sumberg, J., Yeboah, T., Flynn, J., Anyidoho, N.A., 2017. Young people’s perspectives on Africa 12, 11–30. farming in Ghana: a Q study. Food Secur. 1–11. Yaro, J.A., Teye, J.K., Torvikey, G.D., 2017. Agricultural commercialisation models, Takane, T., 2000. Incentives embedded in institutions: the case of share contracts in agrarian dynamics and local development in Ghana. J. Peasant Stud. 1–17. Ghanaian cocoa production. Develop. Econ. 38, 374–397. Ye, X., Canagarajah, S., 2002. Efficiency of Public Expenditure Distribution and beyond: Thomas, R.G., 1973. Forced labour in British west Africa: the case of the northern a Report on Ghana’s 2000 Public Expenditure Tracking Survey in the Sectors of territories of the gold coast 1906–1927. J. Afr. Hist. 14, 79–103. Primary Health and Education. World Bank. 398