*Article* **Food Swamps and Poor Dietary Diversity: Longwave Development Implications in Southern African Cities**

#### **Bruce Frayne \*and Cameron McCordic \***

School of Environment, Enterprise and Development (SEED), Faculty of Environment, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada

**\*** Correspondence: bfrayne@uwaterloo.ca (B.F.); c2mccordic@uwaterloo.ca (C.M.)

Received: 1 November 2018; Accepted: 24 November 2018; Published: 27 November 2018

**Abstract:** While the literature on food deserts focuses on limited availability of food in urban settings, 'food swamps' may better characterize the extensive prevalence and accessibility of cheap, highly processed foods. For urban populations, access to nutritionally inadequate poor-quality food has dire developmental consequences. The long-wave impacts of malnutrition at gestational and early childhood stages are negative and can be non-reversible. Moreover, those who survive into adulthood may face a lifetime of sub-optimal physical and mental development that undermines the second and third UN Sustainable Development Goals—to end hunger and to ensure healthy lives. This paper assesses the long-term health vulnerability of children with limited access to adequate and nutritious food in rapidly urbanizing cities. The analysis focuses on the African Urban Food Security Network (AFSUN) data drawn from 6453 household surveys in 11 cities and nine countries in Southern Africa. The results indicate that children in these households are consuming a limited diversity of food, have limited access to resources and have greater odds of experiencing both short-term and long-term food and nutrition insecurity. These findings demonstrate an underlying vulnerability to long-term health impacts stemming from nutritionally inadequate diets, with potentially significant costs to human capital.

**Keywords:** food deserts; food security; malnutrition; children; urbanization; Southern Africa

#### **1. Introduction**

The United Nations' 2030 Agenda poses two direct development challenges that pivot on Sustainable Development Goal (SDG) 11: Sustainable Cities. These challenges are captured under SDG 2: Zero Hunger, and SDG 3: Good Health and Well-Being. Cities of the Global South have become a "ground zero" for these compounding challenges. This investigation assesses the poverty, food security and food consumption characteristics of poor urban households in Southern Africa containing children aged 5 years old and younger. The findings identify household vulnerabilities to longwave nutrition-related health hazards faced by the children growing up in these environments.

The urban transition unfolding across the Global South has the potential to create great prosperity and provide the means by which these SDGs can be achieved. Cities provide economies of scale that make global sustainability possible [1]. However, accessible, nutritious food is a key component without which hunger cannot be eliminated, nor can good health and well-being be achieved [2]. Yet the promise of ending hunger and achieving good health for all is a major challenge in the context of the rapid urbanization of the Global South [3,4]. Sprawling informal settlements are now a common feature of the urban form alongside the rise of megacities (cities with more than one million residents) in the developing world [5–13]. Typically associated with rural populations, hunger and malnutrition are now increasingly associated with urban populations [2,14–16]. As a result, "malnutrition is in turn a major contributor to both mortality and morbidity and is consequently also a vexing development problem, the locus of which is increasingly urban" [16] (p. 119).

Notwithstanding the positive development potential of widespread urbanization in Sub-Saharan Africa, the continent remains beset by persistent hunger and malnutrition [14–17]. Research indicates that the long-wave demographic impacts of malnutrition at gestational and early childhood stages are negative and non-reversible [18–20]. Moreover, those who survive into adulthood many face a lifetime of sub-optimal physical and mental development that undermines the second and third UN Sustainable Development Goals—to end hunger and to ensure healthy lives [18,21–24].

Hunger and malnutrition are part of the epidemiological transition that is also underway in the Global South. The epidemiological transition describes a shift in the determinants of morbidity and mortality from predominantly communicable diseases (e.g., tuberculosis, influenza, hepatitis) towards predominantly non-communicable diseases (e.g., heart disease, cancer, diabetes) [25]. While the epidemiological transition has been a helpful conceptualization of changing disease prevalence, the theory has evolved as empirical evidence has come to light. As an example, Harper and Armelagos [26] note that new infectious diseases have begun to emerge and spread because of antimicrobial resistance and globalization. The theory has also expanded to include socio-economic factors that have been identified as drivers of the epidemiological transition [27]. Wilkinson [28] highlighted the role of socio-economic inequality in mortality trends linked to the epidemiological transition. In response, Santosa et al. [29] recommended further research into the socio-economic determinants of health to inform needed revisions in the evolving concept. Dye et al. [30] identified a specific interaction between the prevalence of tuberculosis infection rates, and diabetes in a study of India and South Korea. This study noted the role of urbanization (the urban transition) as well as nutrition as key drivers of the epidemiological transition in these countries. Uauy and Kain [31] highlighted the growing need to focus on obesity prevention, in addition to malnutrition, in nutrient programming. This point was reiterated by Broyle et al. [32] who identified a growing pandemic of childhood obesity, driven in part by socio-economic factors like household income.

As would be expected under these transitions, global human nutrition itself is in a state of flux and is described by a third shift: the nutrition transition. As outlined by Drewnowski and Popkin [33], the nutrition transition refers to the shift from the consumption of carbohydrates and fibers to sugars and saturated fats. This transition has been linked to the epidemiological transition through the health outcomes of this dietary shift. Shetty [34] notes the growing challenge of obesity and non-communicable diseases resulting from the nutrition transition. Popkin [35,36] noted that the speed of the nutrition transition appears to differ between the Global North and South. This observation has been conceptualized as the "dual burden" of nutrition where developing countries are faced with a high prevalence of diseases stemming from both under-weight and over-weight populations [37]. In other words, rather than proceeding through the nutrition transition, many developing countries are faced with the burden of both widespread hunger and obesity (or a dual burden). The urban poor are particularly at risk in these countries, where food systems have evolved to accommodate cheap processed food high in sugar and saturated fat [38]. Popkin [39] notes that the nutrition transition in the developing world may also be linked to the urban transition, with urban diets and activity levels becoming increasingly distinct from rural diets and activity levels [40].

Together, these transitions highlight a shift in the vulnerability profile of poorer communities that mirrors the transition from rural to urban livelihoods. These transitions indicate a growing public health threat to future urban residents in the Global South. Children growing up in an environment of limited access to nutritious food are at an increased risk of developing chronic diseases into adulthood [41]. The co-occurrence of stunting and obesity among poor urban neighborhoods is indicative of a food system where highly processed food are more easily accessed while nutritional food is often out of reach to poor families [42]. This situation is highlighted particularly in the context of Southern African cities [43–45].

The rapid growth of these cities has also strained the food systems supporting the urban populations in the Global South, leaving pockets of food deserts in many Southern African cities. In a study of Cape Town, Battersby, and Peyton [46] note that the geographic distribution of supermarkets across the city limits access for poor households. Those supermarkets that are in poor areas of the city often stock fewer healthy foods than are available among supermarkets in high-income areas. This practice may be interpreted as a form of retail redlining, where food retailers, often driven by profit margins, are unwilling to service certain vulnerable sectors of the population or provide inferior goods and services in those areas [47]. In response to this limited accessibility, poor households in the city often rely on informal food markets [48]. As a result, the urban food desert has been criticized as having too narrow a view of the urban food system in the Global South [49]. While traditionally defined by limited geographic availability of supermarkets [50], Crush and Battersby [51] note that the concept of food deserts in the African context often ignores the informal economy and the importance of food access rather than availability. Battersby [52] further highlights the importance of accounting for non-market food sources. This investigation posits that given that supermarkets are not the only indicator of the presence or absence of food availability, the idea of a food desert might be more usefully thought of as a food swamp—readily available, cheap, poor quality and nutritionally inadequate food [53,54]. The existence of this kind of food swamp (both in terms of food source availability and nutritional diversity) poses a significant threat to the long-term health of poor urban households in Southern Africa [43].

The health impacts of this food system on poor urban households are keenly felt among children. Popkin [35] notes that the regular intake of sugars and saturated fats during early childhood could have significant implications for the prevalence of non-communicable diseases later in life. Caesar et al. [55] further suggest that the food insecurity may be linked to communicable diseases (like HIV and TB) in Southern African cities through circuitous socio-economic poverty. Crush et al. [56] suggest that food insecurity and HIV may share a cyclical relationship via precarious and uncertain household income. Household members carrying these diseases often require greater nutritional diversity but are unable to afford it, further progressing the disease impacts.

It is within this broad developmental context that this paper assesses the odds of exposure to health risks for children, precipitated by limited access to adequate and nutritious food in rapidly urbanizing cities. This investigation has two research objectives: First, to determine the change in the odds of household food insecurity among poor urban households in Southern Africa based on whether those households contain children aged 5 years old and younger. This objective assesses the distribution of food insecurity to assess the positioning of these households in access-defined food deserts/swamps. Second, to describe the food security and poverty characteristics of poor urban households with children 5 years old and younger in Southern Africa. This objective identifies the vulnerability of these households according to their nutrition access and adaptive capacity. The analysis focuses on the African Urban Food Security Network (AFSUN) data drawn from household surveys in 11 cities and nine countries in Southern Africa. This investigation argues that nutrition-related health outcomes are less a consequence of food deserts as they are of highly constrained access to already available food in these cities by individuals and households, and that cheap, processed, and nutritionally poor foods dominate food affordability. This investigation provides novel insight into the experiences of urban food swamps among households with small children in Southern Africa by going beyond the spatial availability of food and directly assessing the food access patterns of these households to determine their nutrition-related vulnerabilities.

#### **2. Materials and Methods**

#### *2.1. Approach and Limitations*

To achieve this study's research objectives, this investigation relied on household survey data and nonparametric statistics. Household surveys are a common method for understanding the food consumption patterns of a large population where direct observation is often logistically unfeasible. That said, urban household surveys, particularly among poor households, are challenging given the rapidly changing nature of cities in the Global South. In response to this challenge, this study's

household survey was completed in partnership with local institutions and experts in each sampled city to ensure that the survey design and sampling was guided by the most up-to-date research, census data and maps. The use of nonparametric statistics in this investigation allowed for direct comparison of groups of household respondents categorized by variables of interest (e.g., food security, food consumption, household demographics). While neither open to causal interpretation or precise predictive modeling, these methods provide accurate and interpretable descriptions of the health vulnerabilities of poor households housing young children.

#### *2.2. Research Objectives*

Determine the change in the odds of household food insecurity among poor urban households in Southern Africa based on whether those households contain children aged 5 years old and younger.

Describe the food security and poverty characteristics of poor urban households with children 5 years old and younger in Southern Africa.

#### *2.3. Sample*

The sample for this investigation was drawn from a survey of 6453 poor households sampled using systematic and random sampling across 11 cities in 9 countries in Southern Africa. From that original survey sample, this investigation selected only those households containing children 5 years old and younger. This revised sample contained 2499 poor households distributed across the sampled cities. Both samples are demonstrated in Table 1.


**Table 1.** The Household Sample Distributed Across the Sampled Cities.

#### *2.4. Variable Descriptions*

This investigation measured household food consumption using the food item list in the Household Dietary Diversity Score (HDDS). The HDDS measures whether these food items were consumed by any member of the household in the last 24 h [57]. The food items included in the HDDS are: bread and other grains, potatoes and other foods made from roots or tubers, vegetables, fruits, red meats, eggs, fish or shellfish, beans or nuts, dairy, oils or other fats, sugars, and other foods such as condiments, coffee, or tea.

These food items were also used to measure the kinds of food items that household went without due to high food prices in the last six months. As a result, this food items list is used in this investigation to identify the foods that are commonly consumed and those food items that are vulnerable to in-access due to rising food prices. This investigation also measured the frequency with which households went without food due to rising food prices in the last six months. The investigation also measured the extent to which households had consistent or inconsistent access to water, medical care, and cooking fuel in the last year.

Finally, this study included the Household Food Insecurity Access Prevalence (HFIAP) that measured the frequency of household food access challenges in the month prior to the survey [58]. This scale is administered as a series of 9 sub-scale Likert questions measuring the frequency of social, physical, and economic experiences of limited food access by any member of the household. The score for this scale is derived using a weighted scoring algorithm that ranks household as: food secure, mildly food insecure, moderately food insecure, and severely food insecure. In addition, this investigation relied on the Months of Adequate Household Food Provisioning (MAHFP) scale to measure long-term household food access [59]. The scale measures the number of months in the last year during which a given household had access to adequate household food provisioning.

#### *2.5. Analysis*

In order to assess the change in the odds of household food security based on whether a household contains children aged 5 years and younger, this investigation will make use of odds ratios, Pearson's chi-square tests, and Fisher's exact tests. Odds ratios measure the change in odds of a household being categorized as food secure/insecure based on other characteristics (e.g., whether the household contains children aged 5 years old or younger). This investigation makes use of Pearson's chi-square tests and Fisher's exact tests to determine whether the distribution of households across any two variables is significantly non-random.

This investigation then provides descriptive statistics to indicate the experiences of food insecurity among poor urban households in the Southern African region that contain children aged 5 years old and younger. This analysis will highlight the potential impediments to attaining sufficient nutrition for these children and hypothesize the long-term ramifications of this experience for those children.

#### **3. Results**

#### *3.1. Research Objective One*

Across the board, households with children aged 5 years old and younger had greater representation among food insecure households when compared to households that did not contain children aged 5 years old and younger. That said, for each measure of food insecurity, most of the sampled poor urban households were categorized as food insecure on each measure (indicating the widespread prevalence of food insecurity among the sampled households) (Table 2). These findings demonstrate the widespread nature of food insecurity faced by the sampled poor households in cities across Southern Africa. These frequency distributions also indicate that the majority of the sampled households did not contain children aged 5 years old and younger.


**Table 2.** Descriptive statistics of household food security by households with children.

\* Household Food Insecurity Access Prevalence. \*\* Months of Adequate Household Food Provisioning.

The odds ratio calculations of household food insecurity and the age of children contained in households validate many of the observations made via descriptive statistics. First, all of the food insecurity variables indicated a statistically significant relationship with households categorized based on the age of the children in the household. Second, households with children aged 5 years old and younger had greater odds of experiencing food insecurity according to each of the food insecurity measures included in this investigation when compared to households without children aged 5 years and younger (Table 3). While no other demographics were controlled in these findings, the results indicate a broad positioning of the sampled households with children in vulnerable conditions. Despite these findings, the odds ratios do not indicate a large increase in the odds of food insecurity based on the child membership in the sampled households. As a result, these findings are better interpreted as descriptive rather than predictive.


**Table 3.** Odds ratio calculations of household food security by households with children.

\*\* *p* < 0.01 on both the Pearson-Chi-Square Test and Fisher's Exact Test (2-Sided).

#### *3.2. Research Objective Two*

The sampled households with children 5 years old and younger demonstrated limited dietary diversity. The household survey data indicated that these households consumed just over 5 food groups, on average, in the last 24 h. Among the foods consumed in the previous 24 h, the most commonly consumed food types were bread, condiments, sugar, vegetables and oils (Table 4). 70% of the surveyed households with young children did not consume dairy in the previous 24 h. Meats and beans were only consumed by a minority of these households. The surveyed households with young children also favored refined sugars over fruit consumption.

**Table 4.** Food Items Consumed by Households with Children 5 Years Old and Younger. HDDS: Household Dietary Diversity Score.


The sampled households also demonstrated a high degree of food insecurity. Only 12% of the sampled households were categorized as food secure on the HFIAP (Household Food Insecurity Access Prevalence), while almost 60% were categorized as severely food insecure. These statistics indicate that food access challenges were a common experience among the sampled households. These findings were confirmed by the frequency with which the households went without food due to food prices in the last 6 months. Only 16% of the sampled households did not go without food due to rising food price in the last 6 months (Table 5). Finally, over 70% of the sampled households with young

children experienced at least one month of insufficient food provisioning the previous year and 8% indicated that they did not have a single month of adequate food provisioning the previous year.


**Table 5.** Food Security and Food Price Impacts Among Households with Children 5 Years Old and Under.

Some key nutritional food items were not accessed by households in the last 6 months due to food prices. The sampled households with young children identified meats, dairy, eggs, fish, and fruits as largely inaccessible in the previous 6 months due to rising food prices (Table 6). This table may provide an explanation for some of the dietary trends observed among these sampled households so far in this investigation. Sugars and condiments were identified by these households as more affordable than meats or fruits. That said, it is likely that other factors like preference or availability may be at play. While vegetables were ranked by these households as the most affordable, this food item was not the most commonly consumed by these households in the previous 24 h.

Those sampled households with children 5 years old and younger also demonstrated challenged access to key infrastructure resources. As an example, approximately 40% of the sampled households indicated inconsistent access to water and medical care in the last year. Of importance to household food security, about 60% of the sampled households also went without consistent access to cooking fuel in the last year. Limited access to cooking fuel limits the potential food items that a household can consume (Table 7). The limited access to these infrastructure services indicate the marginal coping capacity of these households to manage the onset of diseases. As a result, these findings demonstrate the limited adaptive capacity of these households in the face of long-wave nutrition-related health impacts.


**Table 6.** Unaffordable food types for households with children aged 5 years and younger.

**Table 7.** Poverty characteristics of households with children aged 5 years and younger.


#### **4. Discussion**

The findings from this investigation indicate that, among the sampled urban households in Southern Africa, households with children aged 5 years old and younger had increased odds of experiencing food insecurity when compared to households that did not contain children in this age bracket. When those households with children aged 5 years old and younger are assessed further, they demonstrated limited dietary diversity, widespread food insecurity, and vulnerability to food price increases with limited access to key urban infrastructure services. As a result, this investigation found that the children living in these households are susceptible to the long-term health implications of limited dietary diversity and inconsistent food access. In addition, the findings from this study identified the limited capacity of these households to manage the nutrition-related health outcomes of their current consumption.

These findings describe the vulnerability context of poor households with young children living in an access-based food swamp [60,61]. While further research will be needed to identify the long-wave health-outcomes of the nutritional patterns observed here, this investigation identified that the current dietary diversity of children growing up in poor urban households across Southern Africa suggests that they are positioned for sub-optimal physical and cognitive development (in addition to long-term nutrient-related diseases) [62–64]. Future longitudinal research should also investigate how the vulnerability context observed here might relate to the onset of communicable diseases as well.

These findings highlight the precarious position of many poor households with young children in Southern African cities and indicate a looming public health threat [60,61]. As Popkin et al. [38] noted, the widespread intake of sugars and saturated fats during infancy has the potential to instigate the onset of non-communicable diseases later in life and speed the epidemiological challenges predicted by both the nutrition transition and the epidemiological transition. Furthermore, the limited capacity of these households to maintain food security increases their vulnerability to communicable diseases like HIV, TB and the new disease-scape of antimicrobial resistant pathogens [26,56]. Given the urban transition underway in Africa over the coming decades [8,9], these vulnerabilities are likely to become exacerbated by poorly planned and implemented urbanization [10,15].

#### **5. Conclusions**

This study is not alone in suggesting that children are not all receiving sufficient food to develop fully, from conception through to adulthood [60,61]. The issue is not simply one of food availability [16,62], characterized in this paper as urban food swamps; nor is it only the distribution of that food, characterized more broadly in the literature as urban food deserts [51]. At the heart of the urban nutrition discussion is a more complex interplay of economic, social, political, infrastructural and environmental factors that together underpin the vulnerability of children (and adults) to hunger and malnutrition [21].

To end hunger (SDG 2) and to ensure health and well-being for all people (SDG 3), the international development agenda has to focus on food and nutrition security in urban areas, where the majority of people already live—or in the case of Africa—will live within the coming decade. Yet, with high levels of food and nutrition poverty in cities of the Global South (and Sub-Saharan Africa in particular), and with the long-term, negative impacts of poor-quality diets and resultant malnutrition on children, the very basis of much needed human capital to achieve sustainable development is undermined. On this specific point, Ogundaria and Awokuseb [63] argue that health is an even more important determinant than education in human capital development in Sub-Saharan Africa and is a crucial component of economic growth.

The urgency of childhood malnutrition cannot be overstated within the broader sustainable development debate [64] and further research that considers the prevalence of urban malnutrition in the context of adequate aggregate food supply is important within the broader food and nutrition security policy arena. Finally, as argued in this paper, while access to food affects nutrition outcomes (as in the case of food deserts), the ubiquitous presence of cheap, industrially manufactured food products (referred to as food swamps in this paper) has serious negative health implications for all people, but especially for children. Both research and policy must focus on food quality and not just availability and access to food.

**Author Contributions:** Conceptualization, B.F.; Methodology, B.F.; Validation, C.M.; Formal Analysis, B.F. and C.M.; Investigation, B.F.; Data Curation, B.F. and C.M.; Writing—Original Draft Preparation, B.F. and C.M.; Writing—Review & Editing, B.F. and C.M.; Supervision, B.F.; Project Administration, B.F. and C.M.; Funding Acquisition, B.F.

**Funding:** This research was funded by the Canadian International Development Agency (CIDA Agreement No. S63441).

**Acknowledgments:** The research used in this paper was funded by the Canadian Government through the Canadian International Development Agency (CIDA) under its University Partners in Cooperation and Development (UPCD) Tier One Program. We wish to thank our colleagues in AFSUN for their assistance and wish to wish to thank the following for their assistance with research planning and implementation: Ben Acquah, Jane Battersby, Eugenio Bras, Asiyati Chiweza, David Coetzee, Bronwen Dachs, David Dorey, Scott Drimie, Miriam Grant, Gareth Haysom, Trevor Hill, Krista House, Florian Kroll, Clement Leduka, George Matovu, Chileshe Mulenga, Peter Mvula, Ndeyapo Nickanor, Sue Parnell, Wade Pendleton, Akiser Pomuti, Ines Raimundo, Michael Rudolph, Maria Salamone, Christa Schier, Nomcebo Simelane, Godfrey Tawodzera, Percy Toriro, Maxton Tsoka, Daniel Warshawsky and Lazarus Zanamwe.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Mapping Obesogenic Food Environments in South Africa and Ghana: Correlations and Contradictions**

**Florian Kroll 1,2,\*,†, Elizabeth Catherina Swart 1,3,† , Reginald Adjetey Annan 2,†,**

**Anne Marie Thow 4,†, David Neves 5,† , Charles Apprey 2,†, Linda Nana Esi Aduku 2,†, Nana Ama Frimpomaa Agyapong 2,†, Jean-Claude Moubarac 6,†, Andries du Toit 5,†, Robert Aidoo 7,†and David Sanders 1,†**


Received: 4 April 2019; Accepted: 13 May 2019; Published: 18 July 2019

**Abstract:** In sub-Saharan Africa, urbanisation and food systems change contribute to rapid dietary transitions promoting obesity. It is unclear to what extent these changes are mediated by neighbourhood food environments or other factors. This paper correlates neighbourhood food provision with household consumption and poverty in Khayelitsha, South Africa and Ahodwo, Ghana. Georeferenced survey data of food consumption and provision were classified by obesity risk and protection. Outlets were mapped, and density and distribution correlated with risk classes. In Khayelitsha, 71% of households exceeded dietary obesity risk thresholds while 16% consumed protective diets. Obesogenic profiles were less (26%) and protective more prevalent (23%) in Ahodwo despite greater income poverty in Khayelitsha. Here, income-deprived households consumed significantly (*p* < 0.005) less obesogenic and protective diets. Small informal food outlets dominated numerically but supermarkets were key household food sources in Khayelitsha. Although density of food provision in Ahodwo was higher (76/km2), Khayelitsha outlets (61/km2) provided greater access to obesogenic (57% Khayelitsha; 39% Ahodwo) and protective (43% Khayelitsha; 16% Ahodwo) foods. Consumption and provision profiles correlate more strongly in Ahodwo than Khayelitsha (rKhayelitsha = 0.624; rAhodwo = 0.862). Higher obesogenic food consumption in Khayelitsha suggests that risky food environments and poverty together promote obesogenic diets.

**Keywords:** obesity; food environments; urban; mapping; nutrition; South Africa; Ghana; governance; supermarkets; ultra-processed

*Sustainability* **2019**, *11*, 3924; doi:10.3390/su11143924 www.mdpi.com/journal/sustainability

#### **1. Introduction**

Urbanisation, poverty, globalisation, industrialisation, climate change and the emergence of a concentrated corporate food regime [1–3] are converging and mutually reinforcing global transitions [4]. Cities in the Global South and their large populations are at the epicentre of these transitions. These have significant implications for sustainability in terms of the health of populations, in terms of the ecosystem disruptions caused by food system activities and in terms of the economic exclusion of large segments of populations from food system activity. Although the United Nations sustainable development goals (UN SDGs 2 and 12) make reference to food security and poverty, they do not recognise the urbanisation of food insecurity nor the dietary transition [5]. This entails increasing consumption of ultra-processed, energy-dense and micronutrient-poor foods, compounded by sedentary lifestyles with reduced physical activity. Consequently, rates of obesity and non-communicable diseases (NCDs) are rising rapidly [6–8]. This shift is particularly severe in low- and middle-income countries (LMICs) and is characterised in some by an increasing "double burden" of simultaneous obesity and persistent childhood stunting [9–11]. Obesity is a major public health concern as it promotes the development of NCDs such as diabetes, hypertension, circulatory disorders and some cancers [12].

This transition has been particularly rapid in Sub-Saharan Africa [13,14], with nutritional change, obesity and NCDs in South Africa and Ghana especially advanced [15,16]. Thus, in 2016, 43% of Ghanaian adults were overweight or obese. Urban (48%) populations were more obese than rural (25%) and rates were higher among women (50%) than men (28%). Forty-three per cent of all adults in the Ashanti region were obese [17]. Similarly, in South Africa, 68.5% of women in urban areas were overweight or obese in 2015. The highest prevalence of female obesity (73%) was recorded in the Western Cape province. The corresponding statistics for men are 34.2% (SA urban) and 43.7% (men, Western Cape) [18]. In Ghana, in 2011, 23% of children were stunted nationally, and approximately 21.7% of children in the Ashanti region [19,20]. The 2014 Ghana Demographic and Health Survey reported that nationally 18.8% of children under five were stunted, while 16.1% in the Ashanti region. In South Africa, 26.5% of children under five were stunted in 2013 [21], and, in 2016, 27.4% of children under five in South Africa and 22.9% in the Western Cape were stunted [18]. Stunting in South Africa persists [22], with obesity often affecting other individuals in the same household [23].

High prevalence of obesity and stunting are linked to multiple dimensions of poverty and deprivation, including water, fuel, medicine, food and income [24]. De-agrarianisation is leading to increased urbanisation of poverty and food insecurity in Southern African [25–27] and West African cities [16,28]. Food insecurity contributes to nutrition transitions in low- and middle-income countries (LMICs), as food insecure households reduce dietary diversity and substitute cheaper foods such as starchy staples, sugar and oils. The level of national food insecurity in Ghana appears on the rise, especially in northern Ghana [29–31]. State estimates of food insecurity in Ghana suggest that about 5% of the population are food insecure with another two million people vulnerable [30]. There seem to be few analyses of urban food insecurity for Ghanaian cities. Evidence from major cities such as Accra suggests that large percentages of the urban poor in Ghana experience food insecurity [32–34]. Accra households spend an average of 54% of income on food, meaning that they are vulnerable to food price fluctuations [34]. Kumasi has experienced significant poverty linked to challenges in governance and service provision, particularly the provision of water, sanitation, and markets critical for food provision [35]. Food insecure city dwellers in Ghana cope by reducing the amount of food consumed, consuming fewer portions, or substituting nutritionally inferior foods such as gari for rice [20] or cutting back on supplementary foods such as rice to enable consumption of primary staples such as maize or yam [36].

In middle-income but high inequality South Africa, food insecurity affects 54.3% of the population, with 28.3% at risk of hunger and 26.0% experiencing hunger. This is especially severe in urban informal

areas (shantytowns), where 68.5% experience hunger or are at risk of hunger. In the Western Cape, 42% of households are food insecure [21]. Particularly high levels of urban food insecurity affect poor city-dwellers [37,38]. Levels of food insecurity in the low-income "black" African township of Khayelitsha are especially high, with 89% moderately or severely food insecure in a 2008 survey [39].

Both urban food insecurity and the nutrition transition are attributed to transformations of global food systems [40], understood as the web of processes, actors and infrastructure by which food is produced, processed, distributed and sold [41,42]. These transformations entail trade liberalisation, concentration in agrofood value chains, the dominance of corporate agribusiness, the rise of Big Food—large food manufacturing corporations which frequently operate at regional or global scales [43]. This has been accompanied by the expansion of supermarket retail and corporate fast food chains. Market liberalisation has facilitated food imports, enabling foreign direct investment into the expansion of formal retail and service outlets and supply chains, consolidation of value chains, and exposing domestic industry to increasing global competition. These dynamics accelerate the nutrition transition by making ultra-processed foods and animal protein products more available and affordable [9,44–46].

#### *1.1. Obesogenic Food*

Several recent epidemiological studies and meta-analyses correlate dietary composition with obesity and NCD risk in the United States. These studies indicate that the consumption of ultraprocessed crisps, fried potatoes, sugar-sweetened beverages, processed meat, unprocessed red meats, sweets and desserts, butter or margarine, and refined grains (in descending order) are associated with increased risk of obesity and associated NCDs [47–50].

Consensus appears to be emerging on several key points. Firstly, there is increased emphasis on synergistic effects of nutrients, foods and patterns of food combinations, and reduced emphasis on individual nutrients of concern [51,52]. Secondly, total fat consumption itself is less problematic than long thought [51–54]. Instead, it is the type of fats which is critical: saturated fats derived from red meat and industrial trans-fats seem particularly risky, especially in combination with refined carbohydrates [49,50,52,54]. Deep-frying foods, especially starchy foods, tends to increase the trans-fat content of such foods making these particularly problematic in terms of obesity and cardiovascular health [51,52]. Neutral association was found for most dairy foods, except consumption of butter. Instead, thirdly, consumption of refined starches with a high glycaemic index (GI) and low fibre content, appear to be key factors promoting obesity and related NCDs [48,55,56]. Fourthly, and conversely, consumption of certain foods appears to reduce the risk of obesity and associated NCDs. These foods include whole grains, nuts, seeds, and fish high in polyunsaturated fats, as well as yoghurt, vegetables and fruit [51]. A fifth convergence entails the recognition that ultra-processed foods, cheaply mass-produced using multiple industrially-refined ingredients and additives, are consistently obesogenic and risky. These foods are being made increasingly available, accessible and desirable by Big Food corporations extending their reach into Africa [43,57]. The NOVA framework classifies food according to four types depending on the nature, purpose and extent of processing. Type one foods are whole and minimally-processed foods; Type two are ingredients such as oil, butter or sugar derived from whole foods; Type three are combinations of Types one and two, frequently used to preserve food; and Type four, ultra-processed foods, are typically industrial mass-products composed of multiple refined ingredients including sugar, cheap starches and oils, salt and various other additives which increase shelf life or alter the flavour, texture or colour of food. Examples include processed cheese, processed meats, confectionery, instant noodles, most breakfast cereals, and sugar-sweetened beverages. Consumption of ultra-processed food decreases intake of fibre, protein and various health-promoting micro-nutrients, while typically increasing the intake of free sugar, sodium and problematic fats. Detailed studies using this framework have been conducted in several countries. These

studies have revealed a direct association between consumption of ultra-processed products and weight gain and increased risk for various NCDs [58–62]. The transformation of systems of food production towards industrial mass-production of ultra-processed food has been mirrored by other structural changes including the transformation of food retail.

#### *1.2. Changing Food Retail Environments*

Changes in retail, especially linked to the expansion of supermarkets in Africa, have transformed food environments. These changes have included centralised and consolidated procurement and distribution systems of increasingly regional scale, more direct contractual relationships with large-scale producers and suppliers who are able to meet demanding formal-sector standards. These upstream trends are matched by a downstream diffusion and penetration of "supermarkets for the poor" into areas previously dominated by traditional and informal markets [46,63].

Ultra-processed foods are becoming increasingly dominant in the global food system [64]. In Ghana, not only has there been a rise in supermarkets but ultra-processed foods have also penetrated the traditional food retail outlets and are widely available. Although about 6% of processed foods in Ghana are imported from South Africa, they derive predominantly from continents other than Africa [65]. In South Africa, the food retail transition has unfolded extensively—shopping malls and supermarkets are rapidly expanding into erstwhile underserviced, impoverished neighbourhoods, as documented for Cape Town [66–69]. This expansion and the entry of more vertically-integrated networks of informal shops which employ more competitive business practices appears to be transforming the local informal economy [70]. This represents a hybrid system with a highly consolidated formal core and an informal periphery closely linked to the formal economy and to transnational networks of people, goods and finance [5,67,68].

However, these transformations in economic regimes and food systems take place at a scale which is far removed from the everyday lived experience of the many poor who are affected. Paradigmatic food systems models emphasise the global and national scale [41,42,71], or discuss household food security outcomes while neglecting intermediate scales of analysis [4]. Consequently, there is a theoretical disjuncture between macro-scale transitions and shifts in household purchasing and consumption patterns. Structural determinism emphasising systemic transitions fails to show how food system change translates into micro-level dietary changes and neglects how the the poor respond to structural change and how these responses in turn might influence systemic transitions. Finally, while trade and investment policy could influence food environments at a national scale by limiting the import and raising costs of problematic foods, global and regional processes are beyond the reach and remit of local governance. This raises the question how local governance processes can engage with these macro-level drivers of food systems transitions.

#### *1.3. Food Environments in South Africa and Ghana*

The concept of food environments introduces an intermediate scale of analysis which may bridge this gap [72–74]. Conceptually, food environments enable, constrain and shape people's food purchase and consumption patterns in several ways related to food availability, accessibility and affordability, and desirability. The notion of food environments still suggests an environmental determinism wherein external environmental factors decisively shape people's purchasing and consumption behaviour. However, this should not lead to neglect of the specific role played by poverty and its various implications. For this reason, we also draw attention to systemic disadvantage faced by poor people. The poor inhabit food geographies differently from wealthier populations, and analysis needs to be sensitive to that and to how large-scale poverty itself shapes the nature of markets. In this sense, the nature of the retail environment

and the strategic decisions of food retailers themselves are shaped by the constraints on aggregate demand in poor food geographies.

A multi-scale perspective on food environments suggests distinctions between personal and external food environments composed of various factors emanating from different scales, e.g., local, national, regional, and global [73]. For this paper, we do not consider the personal scale, but two intermediate scales instead: the household and neighbourhood (i.e., within walking distance of a household) food environment, and the degree to which broader structural drivers such as poverty and food systems transitions influence these. This perspective reveals how systemic transformations change the nature of the local environments which shape the availability and accessibility of food. Conversely, household food purchasing and consumption patterns aggregate to reinforce structural shifts, suggesting multiple cross-scale feedback loops. The food environment concept also has implications for food governance because it frames food consumption drivers at a territorial scale that is amenable to local policy, planning and design interventions [75,76] while also making explicit linkages to higher levels of scale such as regional and global flows of goods and finance.

Food environments in Ghana and South Africa are characterised by a mix of formal and informal food outlets. Informality entails various economic activities which operate without formal registration for tax, licensing, or providing employee benefits such as retirement or paid leave [77]. The informal sector is a key source of food for the poor [39,78,79] but this role is not adequately considered in policy debates [80] and is still poorly researched [81]. Supermarkets and the informal food economy provide different and complementary retail sources of food [66,68,82]. Informal food trade presents advantages to the urban poor: affordable unit sizes, convenient locations, long opening hours, credit, daily re-stocking of fresh produce, which is often cheaper than at supermarkets, and meat cuts that cater to cultural preferences. However, there are disadvantages, too, such as higher unit costs for non-perishables, a limited range of goods of perceived lower quality and shorter shelf life due to the lack of a cold chain, and perceived food safety risks particularly in meat retail. The informal economy thus offers food sources which respond to key needs of poor consumers. This is particularly relevant in cities where unemployment and poverty are often concentrated in fragmented and remote peri-urban informal settlements [83,84].

Food environments in Ghana have traditionally been characterised by large markets and ubiquitous informal roadside stalls and shops. Although research on urban food environments in Ghana is limited, it appears that they are undergoing considerable changes, with increasing prevalence of supermarkets commonly frequented by wealthier city-dwellers, while open-air markets and hawkers are more usually patronised by the poor [85,86]. Fruit and vegetables are scarce in some areas [33], while there is an abundance of cooked foods and convenience stores selling processed and ultra-processed foods that require little or no cooking before consumption. In Accra, informally-sourced food comprised mainly polished rice, vegetable oil, frozen chicken and frozen fish. Increased density of convenience stores appears associated with increases in BMI [87], but food safety issues seem of primary policy concern in Ghanaian and South African cities. Research has therefore emphasised the risks of microbial contamination [34,88,89] and pesticide residues [81,90] in the informal economy.

In South Africa, consumption of street foods correlates with low dietary diversity. Despite widespread consumption of fruit from street trade, there is also a high consumption of ultra-processed foods such as sugar-sweetened beverages and savoury snacks [91]. In poorer areas of Cape Town, food environments promote unhealthy choices [92]. In addition to a challenging food environment, a lack of public safety discourages physical activity, converging with psychosocial stress to promote NCDs, particularly in settlements such as Khayelitsha [93,94].

#### *1.4. Framing Food Environments as Objects of Governance*

Key elements of African food systems are clustered in cities, including retail, distribution and processing, presenting opportunities for improved governance [95]. Given the importance of food environments, it is clear that they warrant more effective governance. Governance typically involves various combinations of hierarchical state power, indirect governance through markets, and "adaptive governance" through transversal multi-stakeholder networks and alliances [94,96]. A growing food systems governance narrative emphasises adaptive governance eliciting multi-stakeholder participation [97–100]. There is a growing movement in cities of the Global North towards the democratisation of food systems governance at the urban scale through mechanisms such as departments of food, food policy councils, local food charters, and extensive processes of public participation [4]. Proponents also advocate the incorporation of food issues within urban planning, design and management, for example through food sensitive planning guides or incorporation within local ordinances [101]. However, this requires the development of suitable indices, measurement technologies, and data collection to set benchmarks, design interventions and assess change [102,103]. For governance purposes, complex data need to be simplified and aggregated to render them accessible and amenable to interpretation and intervention by non-academics such as officials, activists and designers. Charts and maps [104] are among the media used to promote discursive and policy aims in ideologically loaded and contested environments [104–106]. These media can draw on novel sources of data by leveraging technical innovations that facilitate the rapid and simple collection of geo-referenced data. These approaches permit the spatial representation of food-related datasets, thus making them more relevant for spatial planning and urban design, which are key local governance competences. However, such media can be obscure and exclusionary. This paper therefore builds on previous approaches [82,107] to develop ways to make the hybrid and diverse food environments of African cities visible, accessible and legible.

#### *1.5. Research Question*

This paper explores correlations and linkages between neighbourhood food environments and household food environments, with particular emphasis on the risk they pose for obesity. At the household scale, we are concerned with household purchasing and consumption patterns as elements of household food environments, rather than as a reflection of individual consumption. We are thus interested to know household consumption levels of obesogenic foods and diverse healthier foods, and the aggregate patterns of consumption. At the neighbourhood level, we are interested to describe the density of food outlets, their variety, and the types of food available. The underlying interest is to understand to what extent food outlets in the local food geography facilitate the consumption of obesogenic foods or of healthier food options. The second line of inquiry considers the policy, planning and governance implications of these findings.

#### **2. Materials and Methods**

The study gathered quantitative data linking household consumption patterns and local food geographies. Two analytical approaches are combined in this paper. Firstly, descriptive statistics are employed to reflect key features of household and neighbourhood food environments in the research sites. This includes the number and density of food outlets and the relative availability of obesogenic foods in diverse food outlets. Secondly, geo-referenced indices of provision and consumption of obesogenic food are mapped to reveal potential spatial correlations between food consumption and food environments. Ethics clearance was obtained in South Africa from the University of the Western Cape research ethics council (BM17/8/20) and in Ghana from the Centre for Scientific and Industrial Research (CSI: RPN 011/CSIR-IRB/2017).

#### *2.1. Sites, Instruments and Analysis*

Two urban research areas in Ghana and South Africa, namely Khayelitsha (Cape Town) and Ahodwo (Kumasi), were chosen. Khayelitsha was chosen as an urban South African site due to previous research experience in the neighbourhood, its large population of urban poor, and its peripheral geographic location. It is located on the far outskirts of the Cape Town metropolis, a major port city with 3.4 million inhabitants. Ahodwo is a central suburb of Kumasi, a metropolis of approximately two million people located in the central Ashanti region of Ghana. Kumasi was chosen as an urban area due to the high level of development and socio-economic status of residents. The spatial dislocation of Khayelitsha as a remote peri-urban dormitory settlement, a legacy of apartheid-era spatial planning, reinforces poverty as it traps the urban poor in areas far from job opportunities [83,84,108]. By comparison, Ahodwo is located fairly centrally in the Southeast of Kumasi metropolis, which reflects a ubiquitous and vibrant street economy.

Digital survey instruments utilising the ODK smartphone survey app were developed in a consultative process involving the entire interdisciplinary research team. The instrument incorporated standardised survey instruments such as the Lived Poverty Index [24] as well as the adapted PURE food frequency questionnaire incorporating the NOVA classification framework. Surveys were additionally reviewed and validated by enumerators as part of an iterative training process including several workshops to ensure clarity and consistency of comprehension. A link to the survey instruments is in the appendix. The instrument was pilot-tested with a small sub-sample of respondents to further assess comprehension and time required. The survey recorded: (i) key aspects of household socioeconomic status and foodways including consumption, sourcing, and preferences; and (ii) the types of outlet, the variety of food sold, key aspects of business practice such as operating times, upstream sources, and modes of provisioning. Georeferences were recorded for each survey response. The research teams in Ahodwo selected a six-area sample frame demarcated by a roundabout and two exit roads which border the township (Figure 1). Two survey areas were selected in Khayelitsha using a transect principle (Figure 2). The one area is Site B, the other is Enkanini-Makhaza. The transects focused on a central zone defined by the railway stations, adjacent minibus taxi ranks and nearby shopping mall complexes. Data were gathered between September and November 2017. Enumerators walked along the roads identified in the sample frame and selected residential properties in the Khayelitsha sampling frame on a 1 in 7 ratio, and in Ahodwo on a 1 in 5 ratio. Enumerators interviewed the household member most knowledgeable about food purchasing and consumption. Enumerators sought to interview all food outlets within the sample frame. Because the sampling rate in Khayelitsha was higher and properties smaller, the sample frame area in Khayelitsha was also smaller (1.32 km<sup>2</sup> as compared with 5.62 km2).

**Figure 1.** Ahodwo Sample Frame, 5.62 km2.

**Figure 2.** Site B (far left) and Makhaza-Enkanini (far right) sample frames 1.32 km2.

#### *2.2. Analysis*

Data were anonymised and cleaned by checking for inconsistencies, identifying outliers, and correcting obvious errors contradicting known observations (especially for fast food outlets). In some cases, georeferences could not be collected due to technical errors or because enumerators considered the use of smartphones unsafe. In these cases, georeferences were randomly allocated based on the approximate location of the sample.

Customised analysis frameworks for both household and neighbourhood scales were developed on google sheets. Standard spreadsheet functions were used to recode data to generate composite indices of obesogenic and protective food provision and consumption. These were analysed using descriptive statistics such as counts and frequency distributions. Crosstabulations were done to compare dietary risk class distribution between Ahodwo and Khayelitsha samples, internally between income-deprived and non-deprived households in each area sample, and between outlet risk distribution in both sites. Significance of distribution patterns were tested using Pearson's chi-square test. Finally, the correlation (r) between provision and consumption risk classes was tested for both sites.

#### *2.3. Food Outlet Typology*

Food retail and service providers were categorised according to a typology based on whether the outlet was located in public or private space, their trading history or temporal persistence and the degree to which permanent or fixed structures have been built. This typology draws on previous work done by AFSUN [37,82]. The food provider typology included the food retail and service categories shown in Table 1.


It is important to note that there are significant differences between South African and Ghanaian supermarkets. In South Africa, this usually means a large floor space (200 m <sup>2</sup> or more), multiple aisles, trolleys, large-scale refrigeration, multiple (5 or more) electronic tills with card payment facilities, barcode scanning systems used for electronic inventory management, and typically also corporate ownership or formal franchise operations with standardised corporate image and branding. By contrast, in Ghana, supermarkets are often smaller operations with floor space of 100 msq or greater. Where the supermarket is owned by a multinational company such as Shoprite, its characteristics are similar to that described for South Africa. Supermarkets may have some features such as trolleys, multiple aisles, barcode scanning systems, etc., but on smaller scales. Additionally, cash is the typically accepted mode of payment in most Ghanaian supermarkets.

#### *2.4. Obesogenic Food Consumption and Provision Indices*

To operationalise the concept of obesogenic food environments, a nutritional classification framework was developed for the household and neighbourhood scales of analysis. This interpretive framework was informed primarily by relevant literature and deliberation among the authors. Although it requires further validation [109–111], in the absence of suitable alternatives, it represents a transparent and evidence-based framework which renders a high degree of complexity visible and legible. Food groups were allocated to two classes of obesity risk—risky and protective—based on their composition and their classification in terms of the NOVA system. The household questionnaire recorded the frequency of consumption of various foods but not the quantity. Consequently, this framework considers both frequency and diversity, but cannot predict dietary adequacy without further validation. Table 2 reflects the different foods considered part of the obesogenic risk index. Only foods where the evidence for impacts on obesity was clear and compelling were considered. Other foods, where the evidence is more ambiguous (e.g., maize meal, red meat, and chicken), were not counted towards the obesity risk index. Both white and brown bread in both sites were classified as ultra-processed food due to the typically high-volume, industrial Chorleywood production process involved [112], which results in just slightly more fibre in the brown bread, although otherwise almost identical with white bread. Nuts and seeds were not included as review of brand information showed that these typically had high salt and added oil content, which arguably offset possible nutritional benefits. A frequency cut-off was set to establish whether a given food is consumed frequently enough to contribute to obesity risk or prevention. Two or more occasions of consumption per week was selected as reflecting frequent consumption for risky foods, five per week for protective foods, as these need to be consumed at a high frequency in order to provide protective benefit. The obesogenic cutoff is set low in order to ensure the index is sensitive to the aggregate effects of occasional consumption of different obesogenic foods. A second cutoff was set to test for the number of foods exceeding the first (frequency) cutoff. As there is a larger number of food types in the risky category than in the protective category, this class has a higher cutoff (4) for number of foods. For example, if a household reported consuming processed meat three times a week, industrial bread five times a week, cookies twice, and sugar-sweetened beverages every day, they would have reached the cutoff for intake of obesogenic foods. If they also ate fruit only once or twice a week, cooked vegetables and legumes five times a week, and no other protective foods, the diet would not reach the minimum cutoff to be classed protective. A binary index was computed based on the number of foods that exceeded frequency cutoffs in either of these categories. This was used to compute a two-category "risk index" and a "protective index". Above the cutoff, that category was scored with 1, where it was below the limit, the category score is 0 (Table 2).



The same logic was followed to develop a synoptic indicator of the degree to which food outlets provide access to foods promoting obesity. A list of food types traded was assessed in terms of the obesity risk these foods pose. A threshold of two foods traded was set for each category. An outlet was scored high-risk if the outlet stocked two or more risky food types, low-risk if it stocked only one or less. Similarly, an outlet was scored based on the provision of foods known to mitigate obesity risk (Table 3). Low risk foods include mainly NOVA Class one (whole) staples such as maize meal, rice, potatoes, meat, eggs or Class two (refined ingredients such as oil). Where details of shop stock were not available due to refusal to participate in the survey, a score was interpolated based on average scoring of similar outlets and description of type of outlet (e.g., unsurveyed "fruit and veg stalls" were scored as "low risk and protective" as this was the dominant scoring for similar stalls).

Combinations of these two binary indices were used to calculate an aggregate risk index based on four mutually-exclusive categories of obesity risk and vulnerability of household and neighbourhood food environments (Table 4 below).


#### **Table 4.** Food environment risk and vulnerability matrix.


In the household example just mentioned, the household would be considered high-risk and vulnerable. Outlets marked as low-risk would stock few ultra-processed food types, but mostly the low-risk, minimally-processed foods mentioned above. Outlets stocking a variety of foods known to protect against obesity were classed as protective. Thus, supermarkets would generally be scored as "high risk and protective" as they provide access to both types of foods, while fruit and vegetable stalls

would generally be scored as "low risk and protective" because they do not stock a variety of obesogenic foods but offer various protective foods. The statistical distribution of households and food outlets belonging to these different risk classes was calculated at both the household and neighbourhood level. At the neighbourhood level, this distribution was calculated with reference to the number of food outlets per square kilometre. Household and neighbourhood food environment categories were mapped.

#### **3. Results**

In Khayelitsha, 327 households participated in the survey and 309 households in Ahodwo. Response rates were 91% and 97%, respectively. In total, 407 outlets were surveyed in Ahodwo and 83 in Khayelitsha. However, the outlet response rate in Khayelitsha was low (53%), possibly due to mistrust related to recent xenophobic persecution of foreign-owned informal outlets in South Africa [113] and absence of shop-owners. By contrast, 100% of Ahodwo retailers agreed to participate in the study. Descriptive statistics and maps reflecting survey findings are presented in the following two sections. The first section presents the distribution of the aggregate household risk index and profiles the key obesogenic and protective foods consumed. The second section first presents the diversity of food outlets in the local food environment, then maps the aggregate food provision index, juxtaposing this with aggregate food consumption.

#### *3.1. Household Food Environment*

The findings reflected in Figure 3 show that 71% of Khayelitsha respondents reported household diets that met or exceeded the cutoffs for obesogenic foods, while only 16% reported consuming diets meeting or exceeding cutoffs for protective food. In Ahodwo, the pattern is different, with approximately one quarter of households surveyed (26%) consuming diets high in obesogenic foods, and only 23% consuming high amounts and varieties of protective food. Here, two thirds of respondent households consumed low-risk diets composed largely of minimally-processed staples but lacking in protective food intakes. In both samples, there is a notable dearth of protective food consumption. The difference in the risk class distribution between the two sites is statistically highly significant (*p* < 0.000).

,SYWILSPHGSRWYQTXMSRVMWOERHZYPRIVEFMPMX]

**Figure 3.** Household consumption risk and vulnerability classes—numbers on the bars reflect class and aggregate totals.

Figure 4 shows which of the high-risk foods most commonly exceeded the thresholds. In Khayelitsha, commercial bread, sugary drinks, processed meat and sugar were the most prevalent obesogenic foods. In Ahodwo, commercial bread was followed by sugar, sugar-sweetened beverages, and confectionery as the top four high-risk foods exceeding threshold.

In Khayelitsha, the top three protective foods consumed were fruit, fresh and cooked vegetables (Figure 5). More than a third of households met the threshold for "protective" fruit consumption, just more than a third for cooked vegetables but just more than one in ten met the threshold for fresh vegetables. In Ahodwo, fish and fruit (38% each) were the most common obesity-mitigating food consumed at or above threshold, followed closely by cooked vegetables (28%). Only about one in five households (23%) exceeded the frequency threshold for fresh vegetables.

In Khayelitsha, supermarkets and formal retail outlets clearly dominate as sources for most foods except fruit, bread and sugary drinks (Figure 6). Informal and small shops dominate as sources of bread and sugary drinks, but play a minor role in the provision of protective food options. Informal stalls play an important role in the provision of fresh fruit and vegetables.

In Ahodwo, small shops play a greater role in providing access to the key foods than in Khayelitsha (Figure 7). Formal retailers are also a key source of sugary drinks, confectionery and sugar. Fruit, vegetables, legumes and fish are provided primarily by stalls and by "other", which mostly refers to open markets, table top (stationary) food vendors as well as mobile food vendors.

#### 'SRWYQTXMSRSJLMKLSFIWMX]VMWOJSSHWEFSZIXLVIWLSPH

63\*)%LSH[SERH/LE]IPMWLE

**Figure 4.** High-risk household food consumption above threshold.

### ,SYWILSPHGSRWYQTXMSRSJLIEPXLTVSQSXMRKJSSHWEFSZIXLVIWLSPH

63\*)%LSH[SERH/LE]IPMXWLE
