Crop diversity, sustainable food and nutritional security among smallholder farmers in Ghana

Abstract

Purpose

The nexus between sustainable agri-food production and food security outcomes of farm households in sub-Saharan Africa is attracting policy attention. This study analyzes the effects of crop diversity on the incidence of food scarcity, dietary diversity, and the sale and consumption of own crops.

Design/methodology/approach

The study uses panel data collected in 2015 and 2018 on a randomly selected sample of 2553 households from 49 villages in northern Ghana. The study employed a fixed effects modeling approach in the empirical analysis.

Findings

The study finds that crop diversity is positively associated with better dietary diversity, reduced hunger, lower food expenditure, and higher consumption of own produce. The results show positive effects of crop diversity on the total harvested output and sale of agricultural production. Whilst sales improved sustainable food and nutrition security by providing purchasing power to buy nutritional inputs in the market, consumption of own produce rather improved food availability by reducing food scarcity and malnutrition.

Practical implications

Crop diversity is one of the pathways for promoting sustainable agri-food production systems to ensure the food and nutritional security of vulnerable populations and promote biodiversity to achieve environmental goals in sub-Saharan Africa. Crop diversity reduces food expenditure and raises rural incomes through improved outputs and sales, which empowers farm households to diversify their dietary options to be able to overcome incidences of hunger and malnutrition in periods of food scarcity.

Originality/value

The present study improves the understanding of sustainable agri-food production through crop diversity and its implications on food and nutrition security outcomes. The panel data and fixed effects modelling approach address the endogeneity problem between crop diversity and household tastes and preferences.

Citation

Fremmpong, R.B., Gross, E. and Owusu, V. (2023), "Crop diversity, sustainable food and nutritional security among smallholder farmers in Ghana", British Food Journal, Vol. 125 No. 12, pp. 4372-4395. https://doi.org/10.1108/BFJ-12-2022-1060

1. Introduction

Approximately one in every nine individuals does not have enough food to eat. Many of the affected people live in developing countries, with sub-Saharan Africa having the highest prevalence of hunger (FAO et al., 2015, 2022). Notably, children and women tend to suffer disproportionately from hunger and malnutrition. The high rate of food insecurity and hunger in Africa contrasts with the widely-held view that agriculture is the backbone and predominant economic activity in the continent. This raises questions about the pathways through which agriculture affects sustainable household food and nutrition intake. Experience from previous agricultural policies and the rising effect of climate variability on smallholder sustainable agricultural production in sub-Saharan Africa call for an agricultural strategy that meets the food needs of the ever-increasing population. Solving food insecurity with an appropriate agricultural system is essential for the development of low-income countries since adequate nutrition tends to be a basic requirement for a healthy population and sustained economic development (Breda et al., 2020; Downs et al., 2017; Johnston et al., 2014). Current estimates show that in spite of the modest steady economic progress in low-income countries in sub-Saharan Africa, many people, especially children, are still vulnerable to hunger and food insecurity, with potentially adverse future welfare effects (FAO et al., 2022).

Because of climate change, the ability of smallholder agricultural production to meet current household food security needs is becoming a challenge (Shahbaz et al., 2022; Zsögön et al., 2022). Evidence show that crop diversity can be a panacea for improving households' food security (Amao et al., 2023; Zsögön et al., 2022) because of its positive income effect, poverty reduction and positive environmental effect by conserving soil and water resources (Joshi et al., 2021; Vernooy, 2022). The work by Pellegrini and Tasciotti (2014) also indicates that crop diversity could directly affect dietary quality and health status. This is particularly important for subsistence smallholder farmers.

There is a growing theoretical and empirical literature on the relationship between crop diversity on food security (Bellon et al., 2020; Dillon et al., 2015; Jones et al., 2014; Kumar et al., 2015; Pellegrini and Tasciotti, 2014; Snapp and Fisher, 2015). Recent empirical studies have sought to evaluate the effect of crop diversity on households' food security (Appiah-Twumasi and Asale, 2022; Balana et al., 2022; Owoputi et al., 2022; Bellon et al., 2020). Besides the positive socioeconomic effects of crop diversity on farmers, there is evidence that as part of an agrobiodiversity strategy it can mitigate climate-change impacts on agriculture and has positive environmental effects (Verooney, 2022; Shroff and Cortés, 2020). Therefore, crop diversity can lead to sustainable improvements in the livelihoods of people and of the land they use to nourish them. The results of this study support the existing findings on the positive effects of crop diversity on several economic indicators and food security by an example from northern Ghana.

Food insecurity is particularly prevalent in the northern regions of Ghana because of the high monetary poverty and low possession of productive assets (Issahaku et al., 2023) and crop harvest failure (Boansi et al., 2023). Evidence from the 2020 Comprehensive Food Security and Vulnerability Analysis indicates that approximately 11.7% (3.6 million) of Ghanaians, particularly in the three northern regions of Ghana, including the Upper East, Upper West and Northern regions, are food insecure (MoFA et al., 2022). Notably, these regions' livestock-keeping households and smallholder crop farmers are most vulnerable to poverty and resulting food insecurity (Boansi et al., 2023; Danso-Abbeam et al., 2023; Kuivanen et al., 2016). A defining feature of agriculture in the study area is that many farms cultivate below 12.3 acres (5 hectares). Many of the farmers work for subsistence, with few or no marketable surpluses. That smallholder farmers are food-insecure calls for a thorough understanding of the linkages between sustainable agricultural practices and food security in the area.

We analyzed the effect of crop diversity on a household's subjective and objective food security status. We hypothesized that households that grow different crops may consume more varied diets and suffer less from food insecurity. Our study contributes to the existing literature on the effect of crop diversity on food security and nutritional status by estimating a panel data that allows us to control for unobserved household fixed effects, such as tastes and preferences, which could potentially bias the estimates of the effects of crop diversity on food security and sustainable nutritional outcomes. While the primary purpose of this study is to estimate the effect of crop diversity on household food security and nutritional status, we further analyze the channels through which the effect is realized by examining its effects on households' food expenditure, the value of farm output and sales.

The remainder of this paper is organized as follows. Section 2 provides an overview of the related literature and recent findings on crop diversity and food security. Section 3 discusses the materials and methods used in this study. The results of the study are presented in Section 4. We discuss the key findings of the study in Section 5. In Section 6, we provide the conclusion and policy recommendations.

2. Related literature

Conserving and improving biodiversity is part of Sustainable Development Goal (SDG) 15. Besides fighting hunger and poverty, crop diversity can thus contribute to global environmental goals. Agriculture is the main economic activity and source of livelihood for an estimated 78% of the world's rural population (World Bank, 2019). An expectation of the production-consumption-linkage of the non-separable agricultural household model is that higher consumption calls for higher production and vice versa. Thus, sustainable agricultural production can reduce hunger and malnutrition through the quantity and quality of food produced and consumed (Hoddinott, 2012). Given that smallholder agricultural production is sustainable, increased household income from crop sales could enhance nutritional status (Jones et al., 2014). Increased productivity could lead to higher income, but whether the increased income improves nutrition depends on the intra-household ownership and allocation of resources and institutional factors (Aziz et al., 2022; Chege et al., 2015; Hoddinott, 2012; Sibhatu et al., 2015).

The past decades have seen rapid growth in agricultural productivity, with a reduction in the proportion of people living in extreme hunger (United Nations Organization, 2015). However, we cannot draw the same conclusion for the number of undernourished individuals (FAO et al., 2015) [1]. This is because past efforts have concentrated on increasing calorie supply to reduce hunger with little emphasis on ensuring dietary quality (Béné et al., 2019; Haddad et al., 2016; Popkin and Reardon, 2018). Meeting the targets of SDGs of ending hunger (Target 2.1) and malnutrition (Target 2.2) by 2030 requires a comprehensive understanding of the linkage between agriculture and nutrition, as well as the linkage between agriculture and health in developing countries.

Myriad of studies have emphasized on the crucial roles that crop diversity play in effectively reducing production risks, ensuring sustainable production and providing a steady income for smallholder farmers who do not have access to irrigation and other productivity-enhancing technologies (Donfouet et al., 2017; Heady, 1952; Lin, 2011; Massawe et al., 2016; Mugi-Ngenga et al., 2016; Powell et al., 2015; Zsögön et al., 2022). For instance, Dillon et al. (2015) observed that farmers use crop diversity as a coping strategy against rising temperatures in Nigeria. Thus, beyond providing immediate direct benefits, crop diversity contributes to food stability, which is considered the fourth pillar of food security (Tendall et al., 2015). Additionally, crop diversity could improve nutritional status of household members in areas where food markets are generally inaccessible or do not offer a variety of food items (Caviglia-Harris and Sills, 2005; Kurosaki, 2003). Given the nutritional benefits of agricultural bio-diversity, some recent development initiatives have sought to promote crop diversity among smallholder farmers through the introduction of new crops and livestock species (Qaim et al., 2016).

Empirical studies on the relationship between crop diversity and food security have found a positive correlation between crop diversity and dietary diversity (Dillon et al., 2015; Jones et al., 2014; Kumar et al., 2015). At the individual level, Muller (2009) found that producing different staple and non-staple food crops improved the body mass index of household members between 18 and 50 years of age in Rwanda. In the Punjab province of Pakistan, rural households that adopted climate-smart agriculture and farm diversification strategies were food secure (Haq et al., 2021, 2022). However, the literature on this subject still remains sparse, requiring further studies to establish more conclusive evidence. Moreover, we expect the effects of crop diversity on food security and nutritional outcomes to differ depending on the local infrastructure and access to food and input markets. Sibhatu et al. (2015) found that, although the effect of crop diversity on dietary diversity is positive, the effect is relatively weaker and smaller than the effects on market access and market-based transactions.

In addition to socioeconomic endowment and behavioral practices, the market links farm income to household dietary diversity (Rajendran et al., 2017). The absence of a well-functioning market limits household consumption and sustainable food production (Feyisa, 2021; Ramos et al., 2021; Muller, 2009). A well-functioning input and output market thus allows farmers to specialize, which increases yields and crop sales (Ahmed, 2022; Barrett, 2008; Tesfay, 2020). Markets also allow households to buy supplementary food and health inputs. However, market access does not always lead to better nutritional outcomes (Moe, 2002; Nandi et al., 2021; Usman and Callo-Concha, 2021). Similarly, as established in the intra-household decision-making literature, control and ownership of household income and assets tend to influence crop sales and nutrition outcomes (Diamond-Smith et al., 2022; Kulkarni et al., 2021).

While numerous studies have found a positive association between crop diversity and various dimensions of food security outcomes, some literature found a negligible effect (Bellon et al., 2020) or no statistically significant effect (M'Kaibi et al., 2017; Nkonde et al., 2021) or a negative association between crop diversity and household food security status, specifically, dietary diversity (Rajendran et al., 2017). Using the number of edible plants and animals to measure agricultural biodiversity, M'Kaibi et al. (2017) found no statistically significant relationship between child stunting and agricultural biodiversity in rural Kenya. Using Zambian data, Nkonde et al. (2021) found that crop diversity did not exert significant effect on dietary diversity of children under 5 years. The study by Rajendran et al. (2017) further argued that simply increasing crop diversity may not necessarily increase the dietary diversity of farm households. These empirical evidence therefore suggest that the relationship between crop diversity and food security is context-specific and requires in-depth and rigorous analysis to be able to proffer appropriate policies. In this context, this study provides additional empirical evidence that extends the literature on the effects of crop diversity on household food security outcomes.

The positive environmental impacts of crop diversity go beyond the scope of this study. But diversifying cultivated crops means also to improve biodiversity and together with improving food security and income opportunities of smallholder farmers, crop diversity has the potential to make a sustainable contribution to living conditions of the rural population and planet health by conservation and expansion of plant biodiversity (Shroff and Cortés, 2020; Verooney, 2022).

3. Materials and methods

3.1 The study area

This study is based on a panel data set collected on farming households across four distinct districts in northern Ghana: West Mamprusi, Mamprugu-Moaduri, Gonja North and Builsa South (see Figure 1). We interviewed the same households and their members in 2015 and 2018.

The study area is part of Ghana's northern rural savannah area. Approximately 92.9% of the households in the districts engage in farming activities (GSS, 2014). The population in the area is predominantly smallholder farmers who cultivate maize, millet, sorghum and cowpeas as their major staple crops. Farming in the area is usually rain-fed and the average farmer uses simple farming inputs, such as hoes and cutlasses. Only large-scale farmers use chemical fertilizers on a minimum level. Farmers with marketable surplus usually sell at the local markets in Fumbisi, Walewale and Yagaba in the Northeast region and sometimes at the regional markets in Tamale in the Northern region.

In addition to agriculture, communities in the study area have limited economic activity because of their remoteness. Even though there are feeder roads that link communities to relatively larger cities such as Tamale and Walewale, the roads are mostly inaccessible during the rainy season, making it difficult for farmers to assess input and output markets. Farmers can sell their products to large corporations or rotating markets [2]. Farmers sometimes sell to intermediate traders who visit local communities during harvest season to purchase yields. As a result, they must sometimes choose between selling their harvest below the market price to traders or eventually suffering post-harvest losses.

3.2 Data

We employed a secondary panel data consisting of 2,553 agricultural households in the four districts interviewed over two years, 2015 and 2018. The first round of the data employed in the study was collected in March 2015 at the end of the farming season, when the farmers were relatively less busy and available for the agricultural survey, and the second round of data was collected in March 2018. The same households were interviewed in both rounds. The data was part of a broader agricultural survey to understand the impacts of commercial farming and irrigation project-dependent farming on the livelihood of smallholder farmers in the area. A two-stage cluster random sampling technique was used during the data collection. In the first stage, all villages in the four districts were listed, and their respective population data were obtained from the Ghana Statistical Service. Then 49 villages were selected where the main source of income stems from farming activities. All households in the 49 villages were listed and put into two strata – farm (90%) and non-farm (10%, petty traders, craftsmen and so on) households. Farm households reported that at least one member was engaged in crop farming, while in non-farm households' agriculture was not the main source of income.

The data employed in the study comprised of information on the demographics and economic characteristics of households and their members, such as age, education, farm size and farm input expenditure. Other variables in the data include crop sales, food expenditure, households' food scarcity/shortage incidence in the month preceding the survey, and information on food consumed two days before the survey interview, which assisted us to generate a dietary diversity score. The data entry was performed using CSPro software to ensure accuracy and consistency.

3.3 Conceptual framework

This study follows the non-separable agricultural household framework of Dillon et al. (2015) and Kumar et al. (2015), which assume a joint production and consumption decisions. Nonseparability arises due to missing or imperfect markets and high transaction costs. The geography of the study area and socioeconomic conditions satisfy the assumptions underpinning the non-separable agricultural household model. Notably, the area lacks a well-functioning input and output market, making household agricultural decisions dependent on its food preferences and labour supply. The remoteness of the communities in the study area also increases transaction costs, and there is a complete lack of an insurance market. The primary aim of the farmers is to provide food for their households but sells only when there was a surplus. The nonseparability assumption in our empirical model is, therefore, plausible and in line with the study by Boansi et al. (2021) using data on agricultural households in northern Ghana [3].

Following Kumar et al. (2015), we assume an agricultural household that derives utility from consumption of home-produced and market-purchased food, non-food commodities, health status and leisure. The households are assumed to maximize the following utility function:

(1)U=u(xf,xnf,H,L)

where xf is a vector of the total number of foods consumed by the household consisting of own-produced and purchased food items. xnf is the number of non-food commodities consumed by the household, and H and L are vectors of health and leisure enjoyed by household members. In the presence of market failure and transaction costs, the household allocates its production resources such that the total number of crops produced xfj=(xf1,xf2,xf3,..,xfj) maximizes profits from the production side and utility as in (1) from the consumption side (Kumar et al., 2015).

3.4 Empirical specification and identification strategy

Using the reduced form equation (1), the household's utility is expressed as a function of food security proxied by dietary diversity, the experience of food scarcity and food expenditure. Crop diversity as a driver of the components of food security is of major interest in this equation. Controlling for household demographics, wealth and farm characteristics, we formally specify the empirical model as:

(2)foodsecurityjit=ai+β+γcropdiveristyit+Householdit′δ+Wealthit′θ+Farmit′λ+ϵit

where foodsecurityj is a set comprising of three food security indicators-dietary diversity, food scarcity and food expenditure, for the ith household at time t. The variable cropdiveristy is a count of the different crops harvested by the household. We control for the age of the household head, the number of adults, the number of children below fifteen years and the ratio of non-farming to farming household members. These variables are contained in the set Household. The set Wealth contains the household total expenditure in the dietary diversity and food scarcity models and non-food expenditure when estimating the food expenditure model. The set also includes an index of the household's assets calculated using the principal component analysis (Filmer and Pritchett, 2001) [4]. The farm-level characteristics denoted as Farm, include farm size measured in acres, the expenditure on farm inputs and the number of crop shocks that affected the household in the previous farming season. The definitions of all the variables used in the regression models are provided in Table A2.

The function for the value of crops harvested was then specified using the same control variables as:

(3)outputjit=αi+β+θ0cropdiveristy+Householdit′θ1+Wealthit′θ2+Farmit′+υit

where output denotes the total sold and home-consumed harvest in the last farming season preceding the respective surveys.

To be able to interpret the coefficients γ and θ0 as causal effects of crop diversity on household food security or output, the cropdiversity variable is required to be exogenous in equations (2) and (3). However, in equation (1), the household's taste and food preference may be affected by what it grows and eats. We could directly control for this variable if it was observable. However, no variable in the data set can accurately capture the direct effect of taste and preferences. This endogeneity problem has not been treated in the empirical literature because of data constraints. We address this problem by estimating household fixed-effect models for equations (2) and (3). We assume that the households' tastes and preferences remain stable over the sample period. Hence these can be captured by household fixed effects. We further investigate how crop diversity affects the total output, sales and consumption of own harvest by estimating separate regression models for total output, total crop sales and value of own consumption.

Because we use panel data and none of the household's changed community between 2015 and 2018, the household fixed effects take care of community-level fixed effects, which could confound the relationship between crop diversity and the various indicators of crop diversity [5]. We employed the Poisson regression models and reported the incidence rate ratios for the dietary diversity and food scarcity functions and ordinary least square (OLS) regression models for the expenditure and sales functions. For robust estimates the standard errors of the individual observations within a group should not be correlated. If this assumption is violated, the precision of the estimator could be significantly overstated (Cameron and Miller, 2015). Given that our households are clustered within small communities with intra-household knowledge sharing of farming practices, we expect the standard errors in our model to defy this independence assumption. We therefore provide community cluster robust standards errors to satisfy this independence assumption in the linear panel models [6]. Clustering the standard errors also accounts for any potential serial correlation.

3.5 Measurements of food and nutritional security indicators

1. Crop diversity

Crop diversity has been measured in a variety of ways in the empirical literature. Some studies have used indices based on the number of crops planted and the proportion of land devoted to each crop (Cavatassi et al., 2012). The study by Kumar et al. (2015) used three indicators: (1) the total number of crops, (2) the total number of agricultural activities (production of field crops, production of fruits and fruits/vegetables, rearing animals and production of animal source foods) and (3) the number of food groups out of a total of seven that is cultivated by the household. The work by Jones et al. (2014) also used the number of crop species, count of crops and livestock, and Simpson index as three different proxies for farm production diversity. Given our data and the nature of the outcome variable we evaluated, crop diversity was measured as the number of crops harvested during the last farming season. We used the number of crops harvested instead of the number of crops planted/cultivated because in this study, the primary outcome variables we analyzed (dietary diversity, food scarcity, food expenditure and output value) tend to be affected by previous harvests instead of current production and planting decisions [7]. The crop diversity indicator was therefore computed as the count of all food crops harvested by the households in the previous farming season.

1. Food scarcity

The food scarcity outcome variable counts the number out of eight symptoms of food scarcity experienced by the household in the last month preceding the survey. Following similar empirical studies (Owusu et al., 2011; Chagomoka et al., 2016), the food scarcity variable employed in this study was constructed using the following questions – Did the households do any of the following activities because there was no food available to the household in the last month: (1) sell livestock to buy food? (2) sell any assets to buy food for the family (apart from livestock)? (3) borrow money from family/friends to buy food? (4) borrow food from family/friends? (5) skipped meals for a whole day because no food was available? (6) reduce the quantity of food served to men in this household? (7) reduce the quantity of food served to women in this household? (8) reduce the quantity of food served to children in this household? The last three questions highlight the intensity of and vulnerability to food security since women and children tend to be the most vulnerable to food shortages (Harris-Fry et al., 2017; Richard and Messner, 2022). It is possible that some of the proxies for food shortage were used for other reasons than food shortage. For example, small-scale farmers sometimes keep livestock as strategic or liquid assets, which they sell in times of financial challenges (FAO, 2012; Herrero et al., 2013; Taruvinga et al., 2022).

1. Dietary diversity

The household dietary diversity score employed in this study was constructed using seven food groups: (1) cereals (e.g. maize, millet and rice); (2) dairy products (e.g. cheese, milk and yoghurt); (3) fruits; (4) starchy staples (cassava, cocoyam, plantain and potatoes); (5) protein (meat and fish); (6) vegetables; and (7) oil and fats. All the food items consumed during the last two days were assigned to one of the seven food groups. A count of the groups consumed out of these food items was used to measure the dietary diversity. The two-day recall test was used to decrease recall bias.

1. Household food expenditure

For household food expenditure, we relied on detailed household information on food expenditure of 48 different food items including different grains, vegetables, fruits, meet, oil and infant nutrition in the week preceding the survey.

4. Results

4.1 Descriptive results

Table 1 summarizes the indicators of food scarcity, dietary diversity and household food expenditure examined in this study. The first eight variables represent various indicators of food scarcity. About 15% of the sampled households that harvested up to three crops had to sell livestock in the last month to buy food, 8% had to borrow money to buy food and 7% had to borrow food. However, no farm household that cultivated more than six crops borrowed food or money to feed their households. About 13% of the households that harvested three crops or less had to reduce food intake for men and women in the last month, compared to those harvesting six or more crops. Thus, food scarcity tended to intensify when households cultivated fewer crops.

The average dietary diversity score for a household that harvested up to three crops was 4.43 which is lower than the sample average of 4.49. However, the dietary diversity score for households that harvested more than six crops tended to be higher than the sample average. Similarly, household food expenditure tends to increase with crop diversity. Table 1 further show that the number of crops harvested tends to increase with farm size in the study area.

Table 2 presents summary statistics of the variables used in the regression models. The average number of crops harvested by households in the sample was approximately three. In contrast, an average of five and six crops were cultivated at the community level in 2015 and 2018, respectively. Dietary diversity in our sample increased from 4 food items in 2015 to 5 in 2018. Similarly, the households experienced less food scarcity in 2018 than in 2015. However, we found a decline in total household expenditure, food expenditure and non-food expenditure between the two periods. Compared to 2015, the total output and amount sold and consumed at home were higher in 2018. Specifically, while the estimated average harvest was GH₵2972.3 ($782) in 2015, the corresponding average for 2018 was GH₵4552.5 ($944.50). These values show that the households consumed and sold approximately 40% of their output. The mean weekly expenditure on food-related items was approximately GH₵133 ($24.50), while GH₵7 ($1.04) was spent on non-food items. These results show that a significant proportion of household production and income is spent on food and consumer goods.

Regarding farming practices, Table 2 shows that the farmers spent about GH₵532 ($140) on pesticides, herbicides, improved seeds and fertilizer in 2015 and GH₵1145 ($237.55) in 2018. The households cultivated 11 acres in both years, with a marginal increase from 10.3acres in 2015 to 12 acres in 2018. Farmers reported having suffered three different crop shocks during the survey period. The average household households has five adult members, three children and headed by 46-year old person.

4.2 Empirical results

4.2.1 Effects of crop diversity on output, sales and consumption

In this section, we present the empirical results on the effects of crop diversity on the total household crop output, sales and consumption. For smallholder farmers, home-consumed produce and revenue from output sales are the two main channels for household food access and food security status. While, sales improve sustainable food and nutrition security by providing purchasing power to buy nutritional inputs in the market, consuming own produce directly improves food availability, thereby reducing food scarcity and malnutrition. Hence, the linkages between consumption, sales and crop diversity highlights the effect of crop diversity on a household's sustainable food security status.

The estimated results for the models explaining the effects of crop diversity on total output, sales and home-consumed output are presented in Table 3. All the models in Table 3 were estimated with household fixed effects, which also indirectly capture the community and district-level fixed effects. The estimates show that, for an additional crop harvested, the total output and sales tend to increase by GH₵506.84 ($105.15) and GH₵150.81 ($31), respectively. Similarly, the household will consume an extra GH₵309.92 ($64.3) worth of food from an additional crop harvested. The findings thus suggest that increasing crop diversity may increase food and nutrition security in the four districts when sales and own consumption are used as proxies for access to food (see columns (2) and (3) in Table 3).

The findings also show that expenditure on farm inputs improves all the three outcome variables. Thus, investing in productivity-enhancing inputs is associated with increased output and farm revenues. Some of the household demographics also exhibit significant explaining power in the estimated models. Notably, the ratio of non-farming members to farmers (Non-farmers/farmers in household) reduces the amount of output sold (see column (2) in Table 3). The results show that household output increased in 2018 compared to their 2015 levels. The random effect estimated in Table A1 shows that the results are similar to what we observe in Table 3.

4.2.2 Effects of crop diversity on food scarcity and dietary diversity

In Table 4, we examine the effects of crop diversity on household's experience of food scarcity, dietary diversity [8] and food expenditure using household fixed effects modelling approach. Columns (1), (2), (3) and (4) show the incidence rate ratios from the Poisson estimates of the effects of crop diversity on household food scarcity and dietary diversity whilst columns (5) and (6) present the linear panel estimates of the effect of crop diversity on household food expenditure. We test the robustness of the crop diversity estimates by including community-level crop diversity as an additional control variable in columns (2), (4) and (6). Intra-community trade and food exchange could significantly influence household food scarcity and dietary diversity. Hence, we included the community-level production diversity, specifically, the number of crops harvested in the community as a control variable to identify the effect of household-level crop diversity.

The results indicate that the probability of a household experiencing any form of food security increases with higher crop diversity. We find that an extra crop harvested tend to reduce the household's rate of experiencing food scarcity by a factor of 0.87 (see column (1) in Table 4). When the community-level crop diversity is included, the estimated incidence of food scarcity reduced slightly to a factor of 0.86. The results thus suggest that households may be more food secure when they cultivate/harvest more crops. Within the conceptual expectations, we find a higher positive effect of crop diversity on household's dietary diversity. In columns (3) and (4), the expected rate of dietary diversity increased by a factor of 1.03 for an additional crop harvested. Consistent with Kumar et al. (2015), the community-level crop diversity exerts a positive effect on the rate of household dietary diversity.

Table 4 shows that household crop diversity positively correlates with total food expenditure. The effect size suggests that for an additional food crop harvested, the household's food-related expenditure tend to increase by about GH₵7.5 ($1.60). Given the study setting, the observed positive effects of crop diversity on output sold and home-consumed output in columns (5) and (6) are not surprising. This effect may have come from the fact that higher revenues from output sold could be expended on supplementary food and condiments to accompany staple food.

The results also indicate that households with more adults tend to spend more on food. Notably, an additional adult member in the household raises household food expenditure by GH₵11 ($2.30). As the number of non-farming household members to farmer members increases, we find that the household's diet tends to be less diversified (see columns (3) and (4) in Table 4). The relationship between farm size and food security is also found to be positive. Households that cultivate an extra acre of land tend to reduce their rate of food scarcity by a factor of 0.97 but increase their rate of dietary diversity only marginally.

The relationship between wealth and food security suggests that a household with more assets tends to have less experience with food shortage. The implication is that it is more likely to consume a more diversified diet. We find no significant association between the asset score and total food expenditure. However, total household expenditure tends to increase the expected rate of diet diversity, as well as the likelihood of experiencing food scarcity. We further observe that the incidence of crop shocks tends to reduce the rate of household crop diversity. Table A2 provides. The random effect estimates of the model presented in Table A2 are similar to the estimates in Table 4.

5. Discussion of results

The incidence of food insecurity is high in developing countries. About 20% of the population in Africa suffers from food scarcity. The most affected are usually rural households that suffer from several socioeconomic deprivations. Crop diversity can improve food security by providing a consistent food supply, steady income, a buffer against agricultural shocks and environmental sustainability. This paper has focused on the effect of crop diversity on dietary diversity, food scarcity and household food expenditure as proxies for food security.

The results from the study have shown that crop diversity does not only improve dietary diversity but increases output and farm revenue as well. We further find that the incidence of food scarcity declines with higher crop diversity. The papers' findings confirm earlier studies in the literature (Bellon et al., 2020; Dillon et al., 2015; Jones et al., 2014; Kumar et al., 2015). The findings from the study further highlight the importance of market access and agricultural investment (expenditure on farm inputs) as drivers of food security. Given the relationship between market access and farm revenue on the one hand and revenue and food security on the other, we argue that the impact of crop diversity on food security could be higher if farmers in the study districts had access to reliable output markets. Thus, improving accessibility to markets and infrastructure for farmers should be considered a complementary instrument to a crop diversity strategy in the area. The results have relevance for the so-called New Green Revolution for Africa (Denning et al., 2009), which has seen several African countries tackle food insecurity with input subsidies. Incorporating crop diversity into such policies will ensure food self-sufficiency and dietary quality in those countries.

We do not directly analyze the effect of crop diversity on environmental sustainability. However, the reviewed literature shows that crop diversity could yield environmental benefits by improving biodiversity. It is important to note that biodiversity is essential to our food system's long-term sustainability and stability (Verooney, 2022; Shroff and Cortés, 2020). Thus, promoting crop diversity will have positive (indirect) environmental benefits. This is extremely important for the communities in the study area, which depend on a fragile ecosystem threatened by desertification.

The findings from our study agree with Bellon et al. (2020) that crop diversity is a viable strategy to tackle environmental heterogeneities for smallholder farmers in northern regions of Ghana. However, policies geared toward addressing food insecurity in the region must not focus only on the promotion of crop diversity, as this might not be enough to reduce smallholder household food insecurity. Policies designed to promote crop diversity should be within a broader policy framework that provides other needed correlates, like a functioning market for inputs and outputs. In this regard, recent agricultural interventions like government input subsidies to small-scale farmers could augment benefits by including crop diversity in input subsidy packages. While it may be an effective strategy in rural areas on a micro-scale, agricultural interventions should also aim at generating enough marketable surplus to achieve Target 2 of the SDGs. Improving the productivity of small-scale farmers is also essential to meet the growing food demand because of rapid urbanization.

While the study's findings provide valuable information on the effects of crop diversity on food and nutrition security, the data employed in this study is limited in geographical scope to provide a concrete basis for generalization on which crops should be included into a diversified cultivation package. Second, using panel data and fixed effects modelling approach in this study improve the causal inference. However, there is a possibility that measurement errors could affect the actual effect of crop diversity on the outcome variables we examined in the various model specifications. The lack of a suitable instrument made it impossible to deal with this issue in the present study. Hence, we suggest future studies employ methods that address this problem. The R2 values of the models are generally low, probably due to the low variations in the time-varying explanatory variables in our fixed effects model.

6. Conclusion and policy recommendations

Despite the moderate agricultural productivity increases over the years, a sizable proportion of the world's population, notably smallholder farmers in developing countries, lives in hunger or is malnourished. There is therefore an urgent need for the design and implementation of succinct agricultural policies and interventions that address food shortage and hunger in developing countries to achieve the SDGs of ending hunger and food insecurity. This paper has analyzed the effects of crop diversity on household food and nutritional security using a panel data collected in northern Ghana and a fixed-effects modelling approach. First, we have demonstrated that promoting crop diversity among smallholder farmers in rural areas of northern Ghana increases their total harvest, sales and overall food consumption. Second, we have shown that crop diversity directly affects various indicators of food and nutritional security such as nutritional intake, food expenditure and food scarcity.

We further explore the pathways through which the observed effects are achieved. We find that total output and the proportions of harvest sold and consumed at home increase with higher crop diversity while food expenditure decreases. This indicates the potential of crop diversity to reduce hunger, improve small-scale farmers' nutritional status and economic situation. Crop diversification could be an approach to provide additional welfare gains for poor households with limited access to food markets. An agriculture strategy as crop diversity could effectively improve food security and nutritional status of households in remote rural areas in developing countries. Through higher, diversified output, independence from food markets by home consumption of own farm output, increases food security.

From a policy perspective, our study re-enforces existing evidence on the effectiveness of crop diversity in ending food insecurity and hunger of smallholder subsistence farmers. One way to make this strategy more effective is to consider the nutritional content of the crops cultivated. The current policy landscape regarding agriculture shows that several countries (Ghana, Malawi, Zambia, Nigeria and others) have already put in place input subsidy programmes that purposely target smallholders. A critical component of these policies is the provision of subsidized inputs. In addition to offering fertilizers and improved seeds/seedlings, we recommend including nutrition-sensitive components and offer input subsidy packages to promote crop diversity and good nutritional content varieties.

Crop diversification can be a sustainable combining strategy for cash crops like cotton, cocoa, or coffee. Relying on single cash crops can be riskier for farmers in case of shocks. Therefore, a strategy that combines crop diversification for achieving food security, in addition with cash crops, which have a clear income producing motivation, could fight hunger and malnutrition and give farmers chances to raise some income. Strategies which focus on single crops put farmers at a larger risk in case of shocks. Necessary is a need-based and pro-poor approach that tackles the most urgent problems of food insecurity and gives farmers a chance to solve it independently by diversifying production risks of farmers facing challenging climate conditions.

Fostering crop diversity can thus be a viable instrument for policymakers to achieve sustainability of actions in various dimensions: improve farmers’ lives in rural areas to decrease hunger and poverty and promoting biodiversity to preserve the environment.

Figures

Figure 1

Map of study area

Table 1

Distributions of categories of crop diversity and households with food scarcity and food security experiences

	2015				2018				Pooled sample
	Crops harvested				Crops harvested				Crops harvested
Indicator	1–3	4–6	>6	Total	1–3	4–6	>6	Total	1–3	4–6	>6	Total
	(n = 900)	(n = 334)	(n = 14)	(n = 1,248)	(n = 964)	(n = 323)	(n = 14)	(n = 1,305)	(n = 1,864)	(n = 657)	(n = 32)	(n = 2,553)
% Of households that experienced food scarcity	33.48	31.74	14.29	32.80	22.25	27.64	27.78	23.66	27.67	29.73	21.88	28.13
% Of households that sold livestock for food	18.35	19.46	0.00	18.44	12.68	14.91	11.11	13.21	15.42	17.23	6.25	15.77
% Of households that sold an asset for food	5.67	4.79	0.00	5.37	7.48	4.35	5.56	6.68	6.61	4.57	3.13	6.04
% Of households that borrowed to buy food	7.45	8.68	0.00	7.70	8.11	7.14	0.00	7.76	7.79	7.93	0.00	7.73
% Of households that borrowed food	5.78	5.69	0.00	5.69	7.38	4.35	0.00	6.53	6.61	5.03	0.00	6.12
% of households that reduced food for men	14.91	12.87	14.29	14.35	12.27	10.56	16.67	11.90	13.54	11.74	15.63	13.10
% of households that reduced food for women	14.02	12.28	7.14	13.47	12.06	9.94	16.67	11.60	13.00	11.13	12.50	12.51
% of households that reduced food for children	8.23	11.38	7.14	9.06	11.02	9.01	11.11	10.52	9.67	10.21	9.38	9.81
% of households that sipped meals for a whole day	7.45	7.49	7.14	7.46	10.81	12.11	11.11	11.14	9.19	9.76	9.38	9.34
Household dietary diversity score	3.97	4.41	4.57	4.10	4.86	4.86	4.94	4.86	4.43	4.63	4.78	4.49
Household food expenditure	126.89	145.38	215.64	132.84	96.74	125.04	134.81	104.26	111.30	135.39	170.17	118.24
Size of farmland	9.32	12.87	13.42	10.32	11.76	13.82	15.23	12.32	10.58	13.34	14.44	11.34

Source(s): Authors' computation, 2022

Table 2

Descriptive statistics of variables used in the regression models

Variables	Variable definitions and measurement	2015 (n = 1,248)		2018 (n = 1,305)		Total (n = 2,553)
		Mean	SD	Mean	SD	Mean	SD
Crops harvested by household	Number of crops harvested by the household in the last harvest season	2.96	1.22	2.98	1.22	2.97	1.22
Crops harvested in the community	Number of unique crops harvested in the community/village	5.44	1.78	6.45	1.89	5.96	1.90
Household dietary diversity score	The number consumed out of seven food groups – meat and fish, cereals, starch staples, dairy, vegetables, oils and fats and fruits	4.10	1.41	4.86	1.21	4.49	1.37
Household food scarcity score	Count of household experience of eight food scarcity indicators in Table 1	0.82	1.53	0.79	1.86	0.80	1.71
Value of output per acre	Total values of crops harvested per acres (GH₵/acre)	2972.31	4903.92	4552.49	5065.89	3778.83	5048.48
Value of output sold per acre	The market value of sold output total household total harvest per acre of farmland (GH₵/acre)	1183.23	1660.36	1895.42	2978.26	1546.73	2449.80
Value of output consumed per acre	The market value of harvest consumed in the house harvest per acre of farmland (GH₵/acre)	1216.74	3213.87	1747.21	2004.63	1487.49	2678.74
Household total expenditure last week	Household total expenditure in the past week before the survey (GH₵)	168.80	196.10	44.96	56.96	105.59	155.92
Household food expenditure	Household total expenditure on food in the past week before the survey (GH₵)	133.13	163.14	104.29	130.00	118.41	147.83
Household nonfood expenditure	Household total expenditure on non-food items the past week before the survey (GH₵)	7.88	15.45	2.62	25.88	5.19	21.57
Cost of other inputs	The household's expenditure on agricultural inputs (GH₵)	531.91	667.04	1146.12	1451.98	845.40	1178.00
Age of household head	Age of the household head in years	43.94	14.96	47.95	14.56	45.99	14.89
No. of adults in household	Number of adult household members older than 15 years	3.83	2.06	6.11	2.99	5.00	2.82
Number of children under 15 yrs in Household	Number of household members up to 15 years	3.41	2.52	2.37	1.95	2.88	2.31
Household head has been to school	The household head has received at least one year of primary education	0.18	0.38	0.15	0.35	0.16	0.37
Non-farmers/farmers in household	Number of household members who are active farmers	4.43	3.33	3.05	1.94	3.72	2.80
Size of farmland	Size of land cultivated by the household in acres	10.32	5.84	12.33	6.34	11.34	6.18
Number of crop shocks suffered	Number of adverse crop shock experienced by the household	2.54	1.83	2.46	1.52	2.50	1.68
Household asset index	PCA score of household asset ownership	0.23	0.12	0.16	0.08	0.19	0.11

Note(s): PCA denotes Principal Component Analysis

Exchange rate: US$1 = GH₵ 3.79 in 2015; US$1 = 4.82GH₵ in 2018 (Bank of Ghana, 2022)

Source(s): Authors' computation, 2022

Table 3

Fixed effects estimates of the effect of crop diversity on values of output, sales and own consumption

	(1)	(2)	(3)
Variables	Value of output per acre	Value of output sold per acre	Value of output consumed per acre
No. of crops harvested by household	509.41***	154.26**	310.17***
	(179.75)	(65.17)	(51.77)
Age of household head	5.66	−9.12	4.91
	(19.37)	(8.52)	(7.76)
No. of adults in the household (years)	−139.26	−116.61*	23.86
	(94.28)	(58.74)	(62.50)
Number of children under 15 yrs in household	−89.08	−27.94	−7.27
	(126.21)	(57.33)	(100.50)
Non-farmers/farmers in household	−99.16	−99.12**	−27.88
	(96.69)	(37.11)	(22.74)
Size of farmland	81.04**	36.27	9.12
	(33.34)	(23.07)	(17.23)
Expenditure on farm inputs (GH₵)	0.48**	0.17**	0.21**
	(0.22)	(0.08)	(0.10)
Household asset index	2215.82	−873.90	2647.37**
	(1887.69)	(1146.37)	(1041.54)
The number of crop shocks suffered	−69.18	−11.38	19.82
	(90.97)	(44.32)	(30.86)
Year of survey = 2018 (base = 2015)	1036.01***	500.74***	426.44***
	(239.91)	(154.26)	(139.65)
Constant	1211.99	1926.30**	−717.00
	(1229.42)	(849.74)	(550.63)
N	2,553	2,553	2,553
Log likelihood	−23819.37	−21732.11	−22649.53
F-statistic	25.15	20.88	12.70
Prob > F	(0.00)	(0.00)	(0.00)
Within R2	0.08	0.10	0.04
Between R2	0.14	0.00	0.09
Overall R2	0.11	0.02	0.06
AIC	47090.76	42936.88	44780.68
BIC	47149.09	42995.21	44839.01

Note(s): Standard errors are in parentheses

*denotes significance at 10%; ** denotes significance at 5% and *** denotes significance at 1%

Source(s): Authors' computation, 2022

Table 4

Fixed effects estimates of the relationship between household crop diversity and food security

Variables	Poisson estimates				OLS estimates
	(1)	(2)	(3)	(4)	(5)	(6)
	Household food scarcity score	Household food scarcity score	Household dietary diversity score	Household dietary diversity score	Household food expenditure	Household food expenditure
No. of crops harvested by household	0.87**	0.86**	1.03***	1.03***	7.67**	7.66**
	(0.06)	(0.05)	(0.01)	(0.01)	(3.46)	(3.49)
Age of household head (years)	1.00	1.00	1.00	1.00	−0.52	−0.53
	(0.01)	(0.01)	(0.00)	(0.00)	(0.64)	(0.64)
No. of adults in the household	0.95	0.95	0.99	0.99	10.64**	10.58**
	(0.06)	(0.06)	(0.01)	(0.01)	(4.62)	(4.70)
Number of children under 15 yrs	1.09	1.09	1.00	1.00	6.51	6.47
	(0.08)	(0.08)	(0.01)	(0.01)	(5.00)	(5.04)
Non-farmers/farmers in household	1.00	1.00	0.99***	0.99***	1.61	1.57
	(0.04)	(0.04)	(0.00)	(0.00)	(1.79)	(1.76)
Size of farmland	0.97**	0.97*	1.00**	1.00*	−0.36	−0.33
	(0.01)	(0.01)	(0.00)	(0.00)	(0.79)	(0.78)
Household asset index	0.18*	0.20*	1.44***	1.49***	39.89	38.25
	(0.16)	(0.17)	(0.16)	(0.17)	(34.33)	(33.81)
The number of crop shocks suffered	1.04	1.04	0.98***	0.98***	0.17	0.14
	(0.04)	(0.04)	(0.01)	(0.01)	(2.00)	(2.00)
Year of survey = 2018 (base = 2015)	1.79***	1.82***	1.29***	1.27***	−16.62*	−15.75
	(0.34)	(0.34)	(0.03)	(0.03)	(8.62)	(9.43)
Log household expenditure (GH₵)	1.44***	1.44***	1.03**	1.03**
	(0.10)	(0.10)	(0.01)	(0.01)
Household nonfood expenditure (GH₵)					2.63***	2.63***
					(0.34)	(0.34)
No. of crops harvested in the community		0.96		1.02***		−0.94
		(0.03)		(0.00)		(2.16)
Constant					23.53	29.46
					(45.17)	(54.28)
Number of observations	966	966	2030	2030	2,522	2,522
Log likelihood	−825.11	−822.60	−1517.21	−1513.83	−14036.09	−14035.68
Wald test (chi-square)	40.06	46.27	272.60	303.82
Prob > chi	(0.00)	(0.00)	(0.00)	(0.00)
F-statistics					11.47	10.54
Prob > F					(0.00)	(0.00)
Within R2					0.12	0.12
Between R2					0.13	0.13
Overall R2					0.12	0.12
BIC	1678.07	1680.93	3023.70	3025.53	28150.51	28157.52
AIC	1629.72	1627.74	2967.81	2964.04	28092.18	28093.36

Note(s): Standard errors in parentheses

*Denotes significance at 10%; ** denotes significance at 5%, and *** denotes significance at 1%

Source(s): Authors' computation, 2022

Table A1

Random effects estimates of the effect of crop diversity on values of output, sales and own consumption

	(1)	(2)	(3)
Variables	Value of output per acre	Value of output sold per acre	Value of output consumed per acre
No. of crops harvested by household	393.58***	94.83*	242.90***
	(138.65)	(52.87)	(55.69)
Age of household head	−11.60**	−5.32**	−3.39*
	(4.55)	(2.70)	(2.05)
No. of adults in household	65.91**	41.31**	40.55
	(32.22)	(19.75)	(26.20)
Number of children under 15 yrs in Household	122.07**	54.65*	81.04**
	(58.88)	(30.12)	(38.37)
Household head has been to school	−347.30	11.58	−202.24*
	(318.15)	(143.81)	(118.91)
Non-farmers/farmers in household	−43.94	−10.32	−72.39**
	(68.62)	(19.35)	(35.99)
Size of farmland	104.48***	48.83***	34.48**
	(23.45)	(14.66)	(14.48)
Expenditure on farm inputs (GH₵)	1.17***	0.51***	0.19***
	(0.35)	(0.16)	(0.05)
Household asset index	4965.07***	1006.82*	2798.58***
	(1164.65)	(536.00)	(632.87)
Number of crop shocks suffered	−162.33***	−71.37***	−5.85
	(60.79)	(25.99)	(22.55)
Year =  = 2018	940.45***	344.77**	444.51***
	(307.57)	(153.19)	(140.42)
Constant	−531.76	−1.71	−517.02***
	(377.71)	(184.87)	(137.50)
N	2,549	2,549	2,549
Wald test (chi-square)	453.57	256.26	465.06
Prob > chi	(0.00)	(0.00)	(0.00)
Within R2	0.07	0.06	0.04
Between R2	0.28	0.20	0.12
Overall R2	0.19	0.15	0.08

Note(s): Standard errors in parentheses

*Denotes significance at 10%; ** denotes significance at 5%, and *** denotes significance at 1%

Source(s): Authors' computation, 2022

Table A2

Random effects estimates of the relationship between household crop diversity and food security

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	Household food scarcity score	Household food scarcity score	Household dietary diversity score	Household dietary diversity score	Household food expenditure	Household food expenditure
No. of crops harvested by household	0.92*	0.92**	1.02***	1.02***	4.34	4.35
	(0.04)	(0.04)	(0.00)	(0.00)	(2.75)	(2.76)
Age of household head	1.01*	1.01	1.00	1.00	−0.13	−0.13
	(0.00)	(0.00)	(0.00)	(0.00)	(0.16)	(0.16)
No. of adults in household	0.94**	0.94**	1.00*	1.00*	5.67***	5.68***
	(0.02)	(0.02)	(0.00)	(0.00)	(0.79)	(0.79)
Number of children under 15 yrs in Household	1.07***	1.07***	1.01***	1.01***	1.41	1.41
	(0.03)	(0.03)	(0.00)	(0.00)	(1.12)	(1.12)
Non-farmers/farmers in household	0.99	0.99	0.99***	0.99***	2.50*	2.51*
	(0.02)	(0.02)	(0.00)	(0.00)	(1.40)	(1.41)
Size of farmland	0.97***	0.97***	1.00***	1.00**	−0.06	−0.04
	(0.01)	(0.01)	(0.00)	(0.00)	(0.43)	(0.40)
Household asset index	0.09***	0.10***	1.45***	1.48***	63.99***	63.41***
	(0.06)	(0.06)	(0.09)	(0.09)	(20.01)	(19.55)
Number of crop shocks suffered	1.05*	1.05*	0.99***	0.99***	1.14	1.10
	(0.03)	(0.03)	(0.00)	(0.00)	(1.26)	(1.28)
Year =  = 2018	1.57***	1.58***	1.25***	1.23***	−10.56*	−10.15*
	(0.21)	(0.21)	(0.02)	(0.02)	(5.91)	(6.11)
Household head has been to school	0.88	0.88	0.98	0.99	9.18	9.07
	(0.11)	(0.11)	(0.02)	(0.02)	(5.63)	(5.70)
No. of crops harvested in the community		0.98		1.01***		−0.46
		(0.03)		(0.00)		(1.32)
Log household expenditure	1.42***	1.42***	1.03***	1.04***
	(0.07)	(0.07)	(0.01)	(0.01)
Household nonfood expenditure					2.47***	2.47***
					(0.24)	(0.24)
Constant	0.32***	0.38***	3.12***	2.88***	36.94***	39.48***
	(0.09)	(0.13)	(0.11)	(0.11)	(10.85)	(13.16)
N	2,553	2,549	2,553	2,549	2,518	2,518
Log likelihood	−3090.44	−3081.55	−4731.95	−4719.96
Wald test (chi-square)	171.30	177.43	104335.71	615977.32	302.63	323.44
Prob > chi	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Within R2					0.12	0.12
Between R2					0.15	0.15
Overall R2					0.14	0.14
AIC	6164.28	6149.85	9377.79	9358.68
BIC	6234.28	6231.49	9447.78	9440.32

Note(s): Standard errors in parentheses

*denotes significance at 10%; ** denotes significance at 5%, and *** denotes significance at 1%

Source(s): Authors' computation, 2022

Notes

1. The report deems target 1C of the MDG of reducing hunger as achieved in some countries (FAO et al., 2015).

2. The relatively large communities take turns to host local markets every week. During these days market women and middlemen come from major cities like Accra and Kumasi to purchase from local farmers. The rotating markets also serve as the main sources of fish and other dietary supplements in these areas.

3. The area is sometimes referred to as the “overseas” region of Ghana due to its remoteness and distance from the major market centers.

4. We computed the asset index from the following assets owned by the households: bed, bicycle, boat, car, cats and dogs, cattle, chairs, chicken, cooking stove, cutlasses, donkey, bullock carts, ducks, fan, gas, cooker, generator, goat, hifi, hoes, modem, jewelry, knapsack, mattress, mobile phone, mosquito net, motorbike, other animals, personal computers, pigs, plough, radio, refrigerator, sewing machine, sheep, tv, table, tractor etc. The index was standardized to have zero mean and a standard deviation of one, where higher values denoted more asset endowments.

5. Migrants in our sample tend to move to urban areas and larger cities (Accra, Tamale, Kumasi, etc), not to other rural areas.

6. Otherwise, we present robust heteroskedastic standard errors for the fixed effects Poisson models.

7. The harvest is stored and consumed until the next harvest. In Ghana's northern region, farming is usually rain-fed. Farming is, therefore, done in the rainy season, usually, between May to November. In the dry season, fields lay fallow because of lack of water and irrigation.

8. Food scarcity is a count variable which ranges from zero to eight.