Agricultural commercialisation and food consumption: pathways and trade-offs across four African contexts

Abstract

The relationship between agricultural commercialisation and nutrition is empirically contested, with findings suggesting different trends for different indicators, and across different contexts. Using novel cross-sectional survey data from a study of commercialisation in four African countries, this paper aims to understand associations and trade-offs between agricultural commercialisation and food consumption using three different indicators: household perceptions of their own food security; the dietary diversity of adult women and men; and the consumption of unhealthy ultra-processed foods and drinks. We find that a higher level of commercialisation is associated with respondents perceiving the household as more food secure in three countries and hypothesise that this relationship only holds where food price inflation remains low. Perception of better food security is subsequently associated with better reported dietary diversity, but diversity remains low in all countries even in the most commercialised households, and men’s dietary diversity overtakes women’s, on average, at higher levels of commercialisation. Ultra-processed foods were not highly consumed in our samples but were also linked with increasing commercialisation. Through this analysis, we show that commercialization and food consumption outcomes are linked in both positive and negative ways – and that both, retaining own food production, and the food market context, plays a moderating role, as does household experience of uncertainty in transition to commercial agriculture. Findings for policy in the context of inevitable but highly varied forms of agricultural commercialisation in Africa therefore include ensuring that diverse nutritious foods are available and affordable in local markets, but also that household food security and diets are socially supported through the process of transition where commercialisation is pursued.

1 Introduction

Agricultural commercialisation refers to the process of farmers becoming more integrated with markets of different kinds. Though over half of livelihoods in Africa south of the Sahara are based to some extent in agriculture, very few farmers remain purely at subsistence level without any formal marketing of their produce (Davis et al., 2017). There are however myriad different ways that farmers might interact with markets, and the outcomes in terms of livelihoods, food security and diets through different levels and forms of commercialisation—and the trade-offs among these outcomes—are still unclear, including how these may also differ across contexts.

Researchers have been assessing the relationship between agricultural commercialisation and food security and nutrition since at least the 1970s, when globalisation of food systems and economic markets had already reached into the lives of some of the most remote communities. Early work in the 1970s and 80s concerned adverse effects on welfare and nutrition of commercialisation framed as a result of colonialist policies and increasing globalisation, but findings were varied and mechanisms relegated to a ‘black box’ (Carletto et al., 2017; Maxwell & Fernando, 1989). Much of this early work has been described as ideologically biased by recent researchers of the topic (Carletto et al., 2017), though this might say more about epistemological divides than it does about the quality of the early work.

A separate round of research in the 1980s and 1990s aimed to bring together epidemiological and micro-economic approaches to understand how agricultural commercialisation affects theoretical steps between production and nutrition, and therefore to open the ‘black box’ of earlier research (Pinstrup-Andersen 1981). This work – combining reviews of previous studies with econometric analysis of national-level datasets – found mixed impacts on child malnutrition based on local context, type of crop being commercialised, and type of outcome indicator assessed (Bouis & Haddad, 1990; Von Braun & Kennedy, 1986). Researchers looking at child malnutrition outcomes have found their studies confounded to some extent by child illness (Kennedy, 1994), which affects malnutrition irrespective of food consumed and is therefore not directly on what might be theorised as the causal pathway between agricultural commercialisation and nutrition outcomes (though some have argued that income from agricultural sales can also be spent on health and so impact nutrition that way – see conceptual frameworks such as Kadiyala et al. (2014)). Some authors during this time pointed out that context – including type of crop, and broader agricultural and social policy environment – appears to moderate overall nutrition outcomes of agricultural commercialisation (Bouis & Haddad, 1990; De Walt, 1993; Longhurst, 1988).

Recent papers using national datasets have again revisited the questions above, and have again found heterogeneous results: either no relationship between commercialisation and child malnutrition (Carletto et al., 2017); or that market participation is associated with better nutrition outcomes (Demeke & Haji, 2017; Fischer & Qaim, 2012); or that cash crop producers in some contexts are more exposed to economic shocks, which can be nutritionally devastating in a time of high food prices (Wood et al., 2013). While incomes have generally increased with commercialisation, who controls that income (men or women) has also been seen to make a difference to child nutrition in particular, in the commercialisation literature and other nutrition literature (Fischer & Qaim, 2012). These recent studies – all undertaken in Africa – seem to confirm that context matters in the links between agricultural commercialisation and nutrition.

Some recent work has started to move away from looking at child malnutrition indicators, and towards looking at effects on food security (largely in terms of calories consumed) and diet quality (largely in terms of the particular types of foods consumed) (Fischer & Qaim, 2012). These indicators seem a more logical (even if not more important) proximal outcome measure of agricultural commercialisation, so more work is required to understand these links, provided these indicators are rightly used and not misinterpreted.

Additionally, few papers have yet looked at the links between commercialisation and how diets are changing in the context of the nutrition transition (Popkin, 1999; Popkin et al., 2020), and the role of increased incomes in changing access to ultra-processed foods (Monteiro et al., 2019). An ethnography of rapidly commercialising farming systems in rural Peru documented increased incomes from cash crops being spent on snack foods and sugar-sweetened beverages (McDonell, 2016), and this aspect of dietary change also merits further investigation in other contexts.

The debate on commercialisation and nutrition (or its proximate determinants) is therefore far from resolved, and is intertwined with normative ideas about the role of agricultural intensification in national economic development and global environmental sustainability (Béné et al., 2019). Extending previous work, this paper uses novel cross-sectional survey data from a study of agricultural commercialisation in four African countries (Ghana, Nigeria, Zimbabwe, and the United Republic of Tanzania, hereafter Tanzania) to take a direct look at three indicators in relation to commercialisation: household perceptions of their own food security; the dietary diversity of adult women and men; and the consumption of unhealthy ultra-processed foods and drinks by women and men. In this way, the research is able to illustrate trade-offs among these outcomes in different commercialisation contexts for a more nuanced understanding of the relationship between aspects of commercialisation and aspects of nutrition in Africa.

The remainder of the paper is structured as follows. Section 2 describes the methodology in more detail, including the data available to us, and our econometric approach. Section 3 presents the results, including the descriptive analysis, the basic results and the role of moderating factors for the relationship between agricultural commercialisation and food consumption; followed by Sect. 4 that discusses the findings in detail and reflects on study limitations. Finally, Sect. 5 provides brief concluding remarks.

2 Methods

2.1 Research aims and approach

The aim of this study is to understand the differential outcomes from agricultural commercialisation. The research question is: How does agricultural commercialisation, through different routes and in different crops, relate with food security and food consumption outcomes in different contexts? The hypotheses for this question are laid out in the text and box 1 below.

• 1. Higher commercialisation may be associated with less household production of food for own consumption (both staple and non-staple) and more for sale for income. Both of these might mean greater reliance on food markets to acquire food. Depending on the market context for food purchases (different foods available, different prices of foods, and physical/geographic access to the food market), this might be associated with differential outcomes:

• 2. Food security (considered as adequate quantity of food) might be positively linked with enough food being available and affordable at the right times in the markets; might be maintained through concurrent own-production of staple foods; or might be negatively associated with markets not meeting demand and reduced or discontinued own-production. Despite objective measures of food security, a household’s own perceptions of their food security may increase or decrease as changes in perceived risk along with commercialisation bring security or uncertainty.

• 3. Dietary quality (considered as adequate diversity of food) might be positively related to enough healthy (nutrient-dense) food being available and affordable at the right times in markets, and if households choose to purchase this food; might be maintained through concurrent own-production of nutrient-dense (plant and animal) foods; or might be negatively related with markets not meeting demand for healthy foods and own-production being reduced or discontinued, or if more unhealthy (ultra-processed, nutrient-poor, salt/sugar/fat-rich) foods are available and affordable and chosen for purchase.

• 4. It is likely that men and women experience agricultural commercialisation differently in terms of livelihood and food consumption outcomes, depending on the social context. It is also likely that different households experience commercialisation differently, depending on their own socio-economic position within different social contexts.

• 5. Overall, it is likely that there are trade-offs between income, food security, dietary diversity, and ultra-processed food intake. These are moderated by market context (what is available and affordable to buy in food markets, how accessible food markets are), by household socio-economic position, and by household preferences and choice of which foods are purchased.

  

  Box 1: Summary of research hypotheses

2.2 Data and contexts

Our analysis begins with descriptive evidence on commercialisation and dietary patterns for the four countries. We then turn to multiple regression analysis to examine the relationships between dietary diversity and various explanatory variables of interest.

2.2.1 Data

We use multi-country household survey data from the first round of the panel studies under the Agricultural Policy Research in Africa (APRA) Programme of the Future Agricultures Consortium. Overall, the data we use are available for 3,993 households in four countries – 726 in Ghana, 537 in Tanzania, 2110 in Nigeria and 620 in Zimbabwe.¹ While different sampling strategies were used, stratified sampling was adopted across all four contexts. What is common across our four samples is that the data is representative of increasingly commercialised areas in specific parts of these countries (study areas), even if in four very different contexts (Saha et al., 2022). Hence, this gives us an opportunity to study trade-offs between commercialisation and food security and food consumption outcomes for sample households that are progressively engaging with markets. Given the varied commercialisation contexts, we present the regression results for each country separately.

Data for Ghana were collected from November–December 2017 for 20 village communities in Ahanta West and Mpohor districts situated in the oil palm belt of South-Western Ghana across three broad oil palm commercialisation models, classified as: (a) Formal: offered by Norpalm Ghana Ltd² and Benso Oil Palm Plantation³; (b) Incentive: offered by the medium-scale oil palm processor B-BOVID⁴ (i.e., B-BOVID) and (c) Independent: independent smallholders with no formal contracts with buyers. A random sample of households was drawn for each of the 20 village communities.

Tanzanian data is for a sample of farm households across ten villages (Njage, Mkusi, Mchombe, Nakaguru, Ijia, Luvukila, Chita, Mngeta, Makutano and Itongoa) within 30 km area from the Kilombero Plantation Limited (KPL)—a large-scale rice farm in Mngeta division. Stratified sampling was done by classifying villages in three categories according to their electricity status—had electricity by 2016/17 production season; gaining electricity between 2017/18 and 2018/19; and without electricity by 2019. Data collection was done in September–October 2017, with a proportionate random sample of households from the villages, consisting of small-scale farmers, medium-scale farmers and those that were members of farmer groups practicing System of Sustainable Rice Intensification (SRI) technologies.

The Nigeria survey was done across the state of Kaduna from the North and Ogun from the South -purposively selected based on their rapid strides in development of commercial agriculture (primarily maize). Three Local Government Areas (LGAs) were selected from three senatorial districts in Kaduna (Kaduna South, Kaduna Central and Kaduna North) and Ogun (Ogun East, Ogun Central and Ogun West) respectively—in Kaduna: Kachia LGA (Kaduna South), Chikun LGA (Kaduna Central), and Soba LGA (Kaduna North); in Ogun: Ijebu East LGA (Ogun East), Imeko – Afon LGA (Ogun West), and Obafemi Owondo LGA (Ogun Central). In each State, a proportional random sampling was apportioned equally between the medium/large (500) and smallholder farms (500), yielding a total of 2,110 households.

Zimbabwe data were collected from March–April 2018 with A1 resettlement farmers⁵ drawn from Mvurwi and Concession areas of Mazowe District as beneficiaries of the country’s fast track land reform. Prior to 2007, these farmers were engaged in maize production for sale to traders in nearby Mvurwi and Concession towns, and after 2007, there has been rapid growth in tobacco production for sale through Harare-based auction floors. While Mvurwi has seen increased participation of smallholder farmers in non-food commercialisation with rise in tobacco production, farmers in Concession are cultivating maize for food and sale.

While these data are representative of the commercialised study areas, they are not regionally or nationally representative. However, the levels of commercialisation in our sample are comparable to those of households with increasing commercialisation observed in the nationally representative Living Standards Measurement Surveys (LSMS) of the World Bank.

2.2.2 Variables

Our main outcome of interest is food consumption as measured by dietary diversity metrics for an adult woman and an adult man in the household using the Minimum Dietary Diversity (MDD) score (FAO & FHI, 2016). One adult woman and man in the household were asked a series of yes/no questions about foods consumed in the previous twenty-four hours. Food was grouped into ten groups (Table 1). Summing answers to these questions gives the dietary diversity score for women and men, which ranges in value from zero to ten, with a cut-off of more than four groups used to define minimum dietary diversity (note: This cut-off is validated for women, but not yet for men).

Table 1 Key variables

Commercialisation is our key explanatory variable, measured as the (log) total value of crop sales. In addition, we differentiate households using quartiles based on this measure, and divide the sample into those ranked by commercialisation levels: (1) Least, (2) Low, (3) Moderate, and (4) Most.

We classify crops by starchy staples (defined as the major carbohydrate-based energy foods consumed by a population), non-staples and non-food crops (Table 2). Using this classification, we identify if a household is producing starchy staples and non-staples. Additionally, we also measure the quantity of food retained for consumption, calculated as the difference between quantity of staple or non-staple foods produced and sold. It is possible that retained production is used for other purposes, but for this study, we use it to proxy household’s reliance on own production for food.

Table 2 Crops produced

Men and women were also asked about consumption of ultra-processed foods, as a proxy for ‘unhealthy’ food consumption (Monteiro et al., 2019). Both the adult woman and man in the household were asked yes/no questions about the three food groups as: sweets, snacks, and sugar-sweetened beverages. We construct a dummy variable that takes the value 1 if a woman or man consumes any of these ultra-processed foods in a household.

Household food security is measured by the Food Insecurity Experience Scale (FIES) (Ballard et al., 2013). The FIES is an experience-based metric and captures a household’s subjective perception about their food security. It is constructed based on eight questions about facing increasingly constrained access to food, from worrying about not having enough food, to going without eating for a whole day. A raw score is calculated by summing the number of affirmative responses (0–8) and food security status is based on the raw score. A higher score denotes greater perceived food insecurity.

We also have a variable on adult equivalent income available to the household from all sources. Using 12-month data on farm income (calculated as the value of all agricultural output sold minus production costs) and off-farm income (income from all employed and self-employed activities and transfers), we construct household income on a per capita basis, where the total net incomes are equivalised using an adult equivalence scale factor.

We do not have a variable available from our dataset that describes the final aspect of our hypothesis: Access to markets to buy food. This limits to some extent the modelling we can undertake to examine pathways to changes in food security and food consumptions, but we instead use the amount of own-production retained by the household to proxy the alternative scenario of households producing a proportion of what they consume.

2.3 Econometric approach

The econometric analysis focuses on the relationship between consumption of a range of food groups and commercialisation. Commercialisation is measured using the household’s total value of crop sales, as well as the household’s status under commercialisation quartiles based on the sample for each country. Other explanatory variables of interest include production of starchy staples and non-staple foods; an exception is Tanzania where all households produce rice and we do not include this variable. Additionally, we examine household perception of their food security measured as the FIES, consumption of ultra-processed foods by men and women, and a proxy for own production retained for consumption. We recognise that this relationship may be affected by unobserved characteristics that are correlated with household’s food consumption outcomes, and control for household-level factors that could be related with these: age of head, female headed household, schooling of head, household size, land size, household income, assets, housing characteristics, and access to basic services. In addition, we control for individual-level factors related to women and men’s food consumption: individual woman’s or man’s age, their relation to household head and main job.⁶

Means and standard deviations for all variables are reported in Table 3,⁷ discussed below and in later sections. In terms of household characteristics, we note that across the four country contexts, on average, our samples are commercialised at high monetary levels, with specific differences by commercialisation quartile. The next observation about the production of starchy staples also reflects each country context, as we see across the countries, about 16%, 99%, 74% and 68% households produce starchy staples in Ghana, Tanzania, Nigeria and Zimbabwe respectively. Across the same four countries, in terms of non-staples, the situation is quite different, as only about 5%, less than 1%,13% and 6% households produce these. On average, the FIES score across the countries is between 1.5–3 reflecting mild perceived food insecurity, and consumption of ultra-processed food is low. There is a wide variation in average per capita incomes across countries⁸– such that we use the adult equivalent income in our analysis; and generally, about 40% of the households are below the poverty line in all four countries. Households produce very varied amounts for own consumption, often in combination with production for sales: average production of all crops retained for consumption in Kg across countries is as follows: 231 in Ghana; 2,328 in Tanzania; 2,810 in Nigeria; and 2,204 in Zimbabwe. Household heads are on average 45–50 years of age with 7–9 years of schooling, and there are few female headed households (between 6%-20%) across countries.

Table 3 Summary statistics

Our primary dependent variables are the number of food groups consumed by an adult woman and an adult man in the household. We estimate these models using the approach of Headey et al. (2019) – beginning with ordinary least square (OLS) regression methods, we also estimate a Poisson estimator to account for the count nature of our data.

We model the number of food groups consumed by an adult woman or an adult man i from household h located in village v (DDihv) as a function of the commercialisation level of the household (Chv):

$${DD }\;_{ih\nu }= exp( \beta\;{ C}_{h\nu\;}\;+\;\delta\;{ x}_{h\nu\;}+\;\gamma\;{\;z}_{ih\nu\;}\;+\vartheta\;{k}_{hv} )\;+\;{\varepsilon}_{ihv}$$

where ${x}_{hv}$ is a vector of household-level characteristics including the age of head, female headed household, schooling of head (years), household size, (log) land size in hectares, (log) adult equivalent income, (log) value of assets, index of housing characteristics, and index of access to basic services. Individual specific controls ${z}_{ihv}$ include age, relation to household head, and main job. The other characteristics include village variables (${k}_{hv}$) that attempt to capture the unobserved locality level characteristics that may be related with dietary diversity and known participation in commercialisation models. ${\varepsilon }_{ihv}$ is the error term.

The coefficient $\beta$ captures the main relationship of interest: between commercialisation (log of value of all crops sales of the household) and food consumption (the number of food groups consumed by the adult woman and man). We also estimate models where DD is binary, obtaining a value of one if the adult woman or man consumed at least 5 different food groups during the past twenty-four hours, and zero otherwise. In this case, we use probit estimators. We report our estimates based on both the linear and Poisson models.

We also examine the trade-offs as a result of commercialisation between (perceived) household food insecurity, and dietary diversity of women and men in the household (Adubra et al. 2019); and trade-offs (if any) with ultra-processed food consumption. As a secondary objective, we examine the role of own production of food crops retained for consumption for the relationship between commercialisation and food consumption (Hirvonen & Hoddinott, 2017). Finally, we introduce an interaction term for commercialisation and adult equivalent income available to the household from all sources—to examine the extent to which a household’s per capita income plays a role between commercialisation and food consumption.

3 Results

Our analysis begins with descriptive evidence on commercialisation and food consumption patterns for the four countries. We then turn to multiple regression analysis to examine the relationship between commercialisation, dietary diversity and other explanatory variables of interest.

3.1 Descriptive Analysis

Table 4 provides overall dietary diversity outcomes for women and men. Overall, the average woman and man consumes 3 food groups each, and only 18% of both meet the minimum diet diversity of more than four food groups. These percentages are slightly lower in Nigeria, and slightly higher in Zimbabwe. When we look at dietary outcomes by commercialisation levels using quartiles, we note a marginal improvement at higher levels of commercial incomes across all countries, both for women and men.

Table 4 Descriptive Statistics for Dietary Diversity by country

Table 5⁹ outlines more detail on diets, in terms of women and men reporting consumption of each food group. Starchy staple foods including grains, white roots and tubers, and plantains are the most commonly consumed food group for women and men across all four countries (81%-98%), followed by meat, poultry and fish (51%-76%), and other fruit and vegetables. Of ultra-processed foods, beverages (11%-55%) are the most commonly consumed group reported by men and women.

Table 5 Proportion of women and men consuming from different food groups, by country

Figure 1¹⁰ gives the distribution of perceived food insecurity according to the FIES score, by commercialisation levels using quartiles. In three of the four countries, there is a general trend of perceptions of better food security with increasing commercialisation; the exception is Nigeria, where the trend is largely flat. Farming households in Tanzania and Zimbabwe perceive themselves as particularly food insecure at lower levels of commercialisation.

Fig. 1

FIES score – by commercialisation quartiles and country

FIES score by the four commercialisation quartiles, by country. (1) Least, (2) Low, (3) Moderate, and (4) Most.

Figure 2 summarizes food production retained for consumption – starchy staples and non-staples. We note that in Ghana, households retain minimal proportions of starchy staples and non-staple foods. In the other three countries, households are retaining starchy staples (rice or maize, depending on the context), and this is increasing with levels of commercialisation, most strikingly in Tanzania. Households in Nigeria and Zimbabwe also retain a small amount of non-staple foods. Average household production of all crops retained for consumption in Kg across countries is 231 in Ghana; 2,328 in Tanzania; 2,810 in Nigeria; and 2,204 in Zimbabwe.

Fig. 2

Starchy staples and non-staples retained for consumption – by commercialisation quartiles and country

Starchy staples and non-staples retained for consumption by households shown as a cumulative by quartiles and country. (1) Least, (2) Low, (3) Moderate, and (4) Most.

3.2 Basic Results

Table 6 reports the regression results for all four countries – Panel A for Ghana, Panel B for Tanzania, Panel C for Nigeria, and Panel D for Zimbabwe.¹¹ Our results examine associations between commercialisation and food consumption for the four different contexts. For each panel, columns (1)-(4) are results for dietary diversity for women, and columns (5)-(8) are results for dietary diversity for men. Columns (1)-(2) and (5)-(6) provide the OLS and Poisson coefficients using commercialisation levels measured as (log) total value of crop sales; columns (3)-(4) and (7)-(8) provide the OLS and Poisson coefficients using commercialisation quartiles as a categorical variable. We present a summary of results for the sake of brevity, and full results are provided in the supplementary material. The discussion below attempts to capture the relationship between commercialisation and food consumption – with an additional focus on differences for women and men’s food consumption within households.

Table 6 Basic results – summary

The results for dietary diversity are fairly varied across the four country contexts. For Ghana, we note that even when controlling for various household, individual and village-level factors, the association of dietary diversity with total value of crop sales is not statistically significant—both for women’s and men’s food consumption. When we replace the continuous variable of commercialisation with a categorical variable that differentiates between households in different commercialisation quartiles, the same result persists. The association is stronger for Ghanaian households producing starchy staples in addition to oil palm.

For Tanzania, the associations between total value of crop sales and dietary diversity differ for women and men – being negative and statistically significant for women’s diets, and positive but statistically insignificant for men’s diets. The result for women’s diets is further supported when we use the commercialisation quartiles – being highly commercialised is clearly associated with declining diversity of women’s diet in the household. On the other hand, we note a positive and significant coefficient for men’s diets when households are moderately commercialised. These results suggest that even though these are rice producing households (99%), there are differences in intra-household allocation of food that may be at the expense of diets for women in the household.

For Nigeria, we find no gender differences in dietary outcomes. Increasing crop sales are negatively and significantly associated with dietary diversity both for women and men. This result is robust also to using the commercialisation quartiles. However, we find that producing starchy foods such as maize and non-staple foods such as soybean are associated with more diverse diets, both for women and men.

Finally, for Zimbabwe, we note no association between commercialisation and diets for women and men. Production of starchy staples such as maize is negatively associated with dietary diversity – significantly for women’s diets. However, commercialisation has no significant association with women or men’s diets.

The coefficients on the control variables are a priori correct across all four contexts.¹² Wealthier and more educated households enjoy better diets – a relationship that appears more consistently for men than women across countries. We note that while there may be concerns of correlation between our variables of interest and the disturbance term, especially for commercialisation, we try to address this by controlling for various characteristics that may be related with commercialisation and village fixed effects.

3.3 Commercialisation, food security and ultra-processed food consumption

We now examine trade-offs between commercialisation, household food insecurity and diets. Table 7 reports the Poisson regression results for all four countries, using the commercialisation measure (log) Total value of crop sales: Panel A for Ghana, Panel B for Tanzania, Panel C for Nigeria, and Panel D for Zimbabwe.¹³ Earlier results for commercialisation (Table 5) are robust to including FIES. We find that household (perceived) food insecurity exhibits an inverse relationship with dietary diversity – both for women and men. This result is consistent across all four countries, suggesting that improved subjective well-being in terms of food security is associated with more diverse dietary outcomes, both for women and men despite that levels of food insecurity are quite different across countries.

Table 7 Accounting for Food security

We found above that very few people consume ultra-processed foods in the surveyed households in most countries (Table 4). When we examine this consumption by commercialisation quartiles, we see that the picture is quite different across the four countries (Fig. 3). There is some increase in the proportion of households where individuals report consuming snacks, sweets and beverages at higher levels of commercialisation – this is true in Ghana and Zimbabwe. We also find that households able to consume ultra-processed foods such as snacks, sweets or beverages are also able to afford basic food groups which are reflected in the dietary diversity outcomes for women and men in the household.

Fig. 3

Consumption of ultra-processed foods – by commercialisation quartiles and country

Proportion of HHs where individuals are consuming ultra-processed foods by the four commercialisation quartiles, by country. (1) Least, (2) Low, (3) Moderate, and (4) Most.

3.4 Substitutions between production and income

Table 8 reports the Poisson regression results, examining the moderating results for all four countries, using the commercialisation measure (log) Total value of crop sales: Panel A for Ghana, Panel B for Tanzania, Panel C for Nigeria, and Panel D for Zimbabwe.¹⁴

Table 8 Moderating role of own production and incomes

First, we assess the role of own production of food crops retained for consumption in the commercialisation and food consumption relationship. The results differ across the four countries. We find no significant moderating role for production retained for consumption for the commercialisation and food consumption relationship in Ghana and Nigeria. This likely reflects the fact that these households are either not producing many food crops (the case in Ghana, where households produce oil palm as the primary crop) or are selling most of their food crops in the market (possibly the case for Nigeria). In Tanzania we note a positive and significant interaction for men’s dietary diversity, and a negative and significant one for women’s diets. In the latter case, the coefficient for household production retained for consumption is positive, suggesting an inverse relation such that if the household retains lower amounts for consumption, increasing commercialisation will worsen dietary diversity for women. This also ties in with our basic results where we find a negative and significant association between commercialisation and food consumption for women in Tanzania. We note the same result for women’s food consumption in Zimbabwe.

Second, overall, we do not find any strong significant evidence of per capita incomes playing a moderating role for the association between commercialisation and food consumption across the countries. However, in some instances, we find that while commercialisation per se has no correlation with food consumption, higher sales made along with low per capita incomes can worsen food consumption – we find this for men’s diets in Ghana, and women’s diets in Tanzania and Zimbabwe. This likely suggests that with higher sales, households are exposed to greater risks and if per capita incomes are lower, then dietary diversity outcomes are expected to worsen.

4 Discussion

Recent work has found little evidence of a positive relationship between agricultural commercialisation and nutrition, and calls for a better understanding of the transmission channels between these issues (Carletto et al., 2017). We have started to address this gap through several methodological aspects of the current study: 1. Looking at food security and two complementary food consumption outcomes as key steps on the pathway between agriculture and nutrition; and 2. Looking at income and the role of markets versus own production as key moderating factors on this pathway.

Our study samples in the different countries covered a range of levels of commercialisation (from around 30% of crops sold, to almost 100%) and of different commercial crops produced, across food and non-food crops. Our study confirms previous findings that context in terms of foods produced and sold, and what is kept for consumption and what is available to buy in markets, clearly plays a role in how commercialisation is related with food security and diets (Bouis & Haddad, 1990; Carletto et al., 2017).

When we split our sample into quartiles of commercialisation (least commercialised, low, moderate, and most commercialised), we find that a higher level of commercialisation is associated with respondents perceiving the household as more food secure in three countries, though less steeply in Ghana, and with no association in Nigeria. We further hypothesise therefore that this relationship only holds where income goes up but where food price inflation remains low; in the years leading up to our survey, Nigeria saw high food price inflation of up to 20% whereas the other countries saw much lower levels (Fig. 4).

Fig. 4

Food price inflation 2016–2017 in the study countries

Perception of better food security is subsequently associated with better reported dietary diversity. Dietary diversity was low overall in every country, with the average score well below the 5 food groups that denotes a minimally diverse diet, and no more than a quarter of respondents achieving a minimally diverse diet in any context. There is no national data on adult dietary diversity scores with which to compare our findings, but consumption of the different food groups that make up a diverse diet is low in each of these contexts (Afshin et al., 2019). On dietary diversity for both men and women, we find that in Ghana scores rise steadily with increased commercialisation; but in Tanzania both the most and least commercialised quartiles have more diverse diets, suggesting a difficult transition period for farm households as they commercialise and move away from producing diverse diets at home. Diets in Nigeria and Zimbabwe, where all households surveyed were at higher levels of commercialisation, were not particularly diverse. Overall, men’s dietary diversity overtakes women’s, on average, at higher levels of commercialisation. Work using similar metrics in Tanzania and Ghana has found consistently lower dietary diversity in households that feel less food secure, outside of the context of commercialisation (Box 9, FAO (2020)); and recent work has shown that healthy diets are not affordable to a majority of populations in these countries (Herforth et al., 2020). Surveys of commercial farmers vs rural farming counterparts in other countries have found little difference between these groups in terms of dietary diversity (Hamann & Rosen, 2011).

Ultra-processed foods were not highly consumed in our samples, which were rural and agricultural and might be expected to have lower availability of these products, though availability is rising in rural areas in many low-income countries (Baker et al., 2020). Sugar-sweetened beverages were consumed in all countries, particularly in Zimbabwe, and their consumption rose steadily with commercialisation except in Nigeria where it remained low. This is despite the fact that sugar-sweetened beverages are more affordable in Nigeria than in Zimbabwe over time (Blecher et al., 2017), so we hypothesise that the local market context makes these drinks more or less available depending on context.

Keeping own-production for (presumed) household consumption increases markedly with commercialization in 3 contexts (not in Ghana, because households surveyed here produce oil palm which is not consumed as a food in the same way as rice and maize). This is surprising in the context of general hypotheses that commercialisation creates income that can be spent on food rather than producing it. The production of starchy staple foods is both positively and negatively associated with dietary diversity in our data, depending on country context, agreeing with recent findings from Uganda that some households require a buffer against food insecurity risks even in the context of increasing commercialisation and that the process of moving from largely subsistence to increasingly commercial agriculture is an uncertain one for many households (Ntakyo & van den Berg, 2019). We found no significant moderating role of production retained for consumption for the commercialization and diets relationship in Ghana and Nigeria; positive and significant interaction for men’s dietary diversity, and a negative and significant one for women’s diets in Tanzania; negative and significant interaction for women’s diets in Zimbabwe, again illustrating the importance of food market context.

Increased income is negatively associated with dietary diversity in two contexts (Tanzania and Nigeria) and not associated in the other two (Ghana and Zimbabwe). There is no significant evidence of per capita incomes playing a moderating role for the association between commercialisation and diets across the countries. This is similar to other African contexts, where weak transmission from agricultural incomes to higher food expenditures and better diet is seen (Radchenko & Corral, 2018). Overall though, wealthier households have better dietary diversity in our data, more consistently for men.

Overall, we find that there are trade-offs between income, food security, dietary diversity, and ultra-processed food intake with increasing commercialisation. Incomes and perceptions of food security in general increased with higher levels of commercialisation, suggesting that commercialisation can help households to access sufficient quantities of food. Our analysis finds that there is a less positive association when it comes to metrics of food quality, with dietary diversity not consistently increased with commercialisation (and in fact decreased in some contexts) and increases in access to sugar-sweetened beverages detrimental to health in most contexts.

We suggest that these findings are moderated by market context (what is available and affordable to buy in food markets, and how accessible food markets are), as has been shown in Malawi (Koppmair et al., 2017) and by household preferences and choice of which foods are purchased. In our study countries, 55–65% of dietary energy supply is from starchy staple foods, and the cost of meeting nutrient requirements ranges from 57–92% of food expenditure (www.foodsystemsdashboard.org), suggesting that diversifying diets is difficult even outside of the context of agricultural livelihood change.

While our data across four countries allow us to look in some depth at interactions between agricultural commercialisation and various food security and dietary outcomes in different contexts, there are still some limitations of this study. Food markets (in terms of what is available and affordable to buy) and household food preferences are both ‘black boxes’ in this research, as we don’t have quantitative data on these. The dietary diversity metric we use is only validated for use with women (and therefore only recommended for use with women), though there is no reason it wouldn’t provide valid information on the diversity of men’s diets also. We know that some nutrition outcomes appear via longer timescales through inter-generational transmission of inequity, and conversely might be addressed over time through, for instance, accumulated education and opportunity, which is it not possible to assess here. Also not assessed here is the importance of commercialisation for women’s time, work, and caring resources, which may also be related with dietary choices or family food environments. Each of these are areas for further research, and future research looking at commercialisation and different aspects of the diet should ensure that adequate data is available on every step in the hypotheses laid out in Box 1.

5 Conclusion

Echoing the literature stretching back for decades, our work makes clear that context really does matter for the relationship between agriculture commercialisation and food security and diets: in particular, the food market context shapes the food environments in which additional income from commercialisation can be spent on diverse foods (Turner et al., 2018). Studies of the political economy of agriculture have found commercialisation to be seen commonly as efficient and poverty-reducing, but with contestation over whether it should also be redistributive, and there is a need to understand better the winners and losers from processes of commercialisation in different contexts (Shonhe, 2018). Analysing change outside of local context leaves too many questions remaining, so econometric studies such as ours need to be complemented with in-depth case studies to understand social and economic implications of change. Our study found fewer trade-offs among food security and diet quality than anticipated, and future work should consider how these relevant contextual factors shape the food security and diet outcomes of commercialisation.

Further, our work suggests that uncertainty may play a part in the decisions of farming households, who do not become perfect market players overnight but have to navigate livelihood change in a context of maintaining adequate food acquisition (Scoones et al., 2020). Findings for policy in the context of inevitable but highly varied forms of agricultural commercialisation in Africa therefore include ensuring that diverse nutritious foods are available and affordable in local markets, but also that household food security and diets are supported (for instance by social protection measures) through the process of transition where commercialisation is pursued.

Appendix

Table 9 Summary Statistics – by commercialisation quartiles

Table 10 Proportion of women and men reporting consumption by food groups—Dietary MDD by commercialisation quartiles

Table 11 Summary statistics for FIES questions by commercialisation quartiles