Secondary supermarket revolution: food sources and food security in Northern Namibia

Abstract

A central feature of the transformation of urban food systems in cities of the Global South is the growing presence of supermarkets and their supply chains, often termed supermarketization or a supermarket revolution. A key issue in the African context is whether supermarkets are a threat to other sources of food including informal sector vendors. Most research on the supermarket revolution and competition with other food retailers focuses on large urban conurbations with little attention paid to the role of supermarkets in smaller urban centres. This paper aims to rectify this situation through a case study of the role of supermarkets in three small urban centres in northern Namibia. The paper uses data from a representative household food security survey in 2018 which collected detailed information on household food consumption and food purchasing patterns. We show that supermarkets have established a dominant role in the local food system and are patronized by almost all households. However, the informal food sector displays considerable resilience and is patronized on a regular basis by low-income households. Future research on the impact of the secondary supermarket revolution should examine the experience and strategies of informal food vendors and whether the relationship with supermarkets is truly symbiotic or not.

Policy and Practice Recommendations

• Unregulated supermarket expansion is transforming the food system of Namibian secondary urban centres and needs greater monitoring of impacts on other food sources

• Traditional food sources such as urban agriculture are in decline as the food system transforms and may have to be abandoned as viable food security policies

• The role of easy access to highly processed and unhealthy foods from convenience and supermarket outlets needs to be factored into Namibia’s strategies to combat the growing burden of non-communicable disease

Introduction

Urbanization is unfolding at an unprecedented pace in Sub-Saharan Africa (SSA), fueled by natural population growth and rural–urban migration, with some projections estimating that the urbanized population of the continent will exceed 55% by 2050 (Lall et al., 2017; UN-Habitat, 2014). The urban population will likely triple by mid-century, increasing from less than 400 million in 2015 to 1,200 million in 2050. While the greatest numerical increases are occurring in large cities, urban growth is not confined to large primate cities. Virtually every urban centre up and down the continental urban hierarchy is experiencing significant growth (Christiansen and Kanbur, 2016; Roberts, 2014). The total number and proportion of the urbanized in small and medium size cities (of less than 500,000 inhabitants) now exceeds that of larger primate cities. As many as 162 million people (or 55% of the total urban population) live in secondary cities and towns of less than 500,000, compared with 60 million (20%) in cities of 500,000 to 2 million, and 73 million (25%) in cities of greater than 2 million (Sattherthwaite, 2017).

A recent UN-Habitat (2020) report concludes that the residents of secondary cities in SSA experience ‘multiple deprivations’ in relation to governance, economics, water and sanitation, the living environment, education, health and crime (UN-Habitat, 2020). Food is mentioned only once in a 56-page report and food insecurity and deprivation not at all. The absence of any discussion of secondary city food insecurity by the UN agency responsible for urban development and governance leaves the misleading impression that it is not an important challenge for residents and local government in Africa’s secondary cities. Part of the reason for this silence is the bias of the international food security agenda which tends to see food insecurity as a rural not urban challenge (Crush et al., 2012; Crush and Riley, 2019). However, a growing body of research by the African Food Security Urban Network (AFSUN) and others shows that food insecurity is a growing crisis in Africa’s secondary cities (Battersby and Watson, 2019; Crush and Battersby, 2016; Frayne et al., 2018; McCordic and Abrahamo, 2021; Mackay, 2019; Riley and Crush, 2022).

In major Southern African cities, urban food wholesaling and retailing and food environments are undergoing rapid change (Battersby and Haysom, 2019; Frayne and Crush, 2018; Tacoli, 2019; Hannah et al., 2022). In South Africa, the five major supermarket chains have consistently expanded their market share and account for three-quarters of the formal grocery retail segment overall and an increasing share of the market in low-income townships and informal settlements (CCSA, 2019). However, claims about the disruptive effect of supermarkets deprive informal businesses of any agency, adaptability and innovation. Recent research has demonstrated that the informal food sector in large African cities has displayed considerable resilience, adaptability and growth in meeting the daily food needs of low-income households, despite an often hostile regulatory environment (Crush et al., 2017). As Skinner (2019: 104) observes, “despite greater supermarket penetration in many urban areas, informal food retailers – street and market traders, ‘spaza’ or small informal shops – remain ubiquitous across the African continent.” Over the past decade, South African supermarkets have expanded to over 25 countries in SSA, aided by the accessibility of supermarket outlets to global and South African supply chains (das Nair, 2021). The proliferation and growing influence of South African as well as other supermarkets is seen in visible changes to the built environment as well as the food sourcing and consumption behaviour of urban residents (Battersby, 2017; das Nair, 2018, 2019). Conventional wisdom suggests that supermarkets in Africa target and serve the middle and upper-classes and higher income neigbourhoods in large cities (Reardon and Gulati, 2008).

Evidence for the patronage of supermarkets by the urban poor is mixed. Wanyama et al. (2019), for example, argue that relatively few households in informal settlements in Kampala and Nairobi buy any of their food in supermarkets. In contrast, Owuor (2018) shows that most Nairobi households, including those in the lowest income brackets, purchase some of their foods in supermarkets. Other research confirms that the urban poor do patronize supermarkets, but primarily to purchase staple foods such as mealie meal and rice in bulk (Caesar and Crush, 2016; Crush and Frayne, 2018; Odunitan-Wayas et al., 2018). In turn, supermarket chains have responded by creating budget subsidiaries with limited product ranges and tapping into mass consumer markets in townships and informal settlements (Battersby and Peyton, 2016). In Namibia, the expansion of the supermarket sector has been rapid and dramatic, and a powerful exemplar of the ability of a supermarket revolution to fundamentally reshape the food system, purchasing patterns and dietary intake in the capital city, Windhoek (Kazembe et al., 2021; Nickanor et al., 2021). In Windhoek, a symbiotic rather than competitive relationship between supermarkets and informal food vendors has emerged, primarily because the latter have shown considerable ingenuity and agency in the face of local government policy to contain their spread (Crush et al., 2019; Kazembe et al., 2019; Nickanor et al., 2019a).

To date, research on the supermarket revolution in Africa has tended to focus on large cities. As a result, there has been limited assessment of whether a secondary supermarket revolution is also in progress, with supermarkets moving down the urban hierarchy to smaller cities and towns in search of new markets. Nor is there much research on their reception from municipal governments, informal sector retailers and more traditional food suppliers in urban and peri-urban food markets. In three small towns in Kenya, Rischke et al. (2015) found that supermarkets had established a strong presence, with over 80% of households shopping at supermarkets although the average share of total food expenditure was less than 20%. In another study in Msunduzi, South Africa, Caesar and Crush (2016: 54) found “extraordinarily high levels of reliance on supermarkets, especially compared with the informal food economy.” However, a recent project on the food system of three secondary African cities – Kisumu in Kenya, Kitwe in Zambia and Epworth in Harare – suggested that although supermarkets are present and expanding in number, more traditional sources such as small grocery shops, market vendors and the informal food sector still play the dominant role in household food provisioning (Fuseni et al., 2019; Opiyo and Ogindo, 2019; Tawodzera et al., 2019).

In this paper, we build on this literature by examining the case of small urban centres in northern Namibia. Although there is considerable debate about the definition of secondary cities or secondary urban centres, in Namibia all urban areas outside the capital Windhoek with their own municipal government are categorized as such (Haysom, 2022; Zimmer et al., 2020). The urban population of the country increased from 390,000 (28% of the total in 1991) to 886,000 (42% of the total in 2011), and is projected to increase to 2.3 million (two-thirds of the population) by 2040 (NSA, 2014). In 2016, Windhoek had an estimated population of 440,000 (or 43% of the urban population). The other 57% of the urban population resided in 27 secondary urban centres ranging in size from 2,000 to 260,000 (Ottolenghi and Watson, 2017: 22–23). This paper focuses on three of those centres located close together along the main road between Windhoek and the Angolan border. They increasingly form a single urban conurbation which we refer to here as the Oshakati-Ongwediva-Ondangwa Urban Corridor. The paper is based on data from a representative household food security survey conducted in 2018. The study aimed to examine the association between food security and food sourcing patterns among a representative sample of households in the Corridor. The paper addresses three main questions: first, is the Corridor undergoing a secondary supermarket revolution as the urban population grows in size? Second, are there regular or discernible patterns in the food sourcing strategies of households in the Corridor? And third, which households are most likely to patronize the various formal and informal food retail outlets? The next section of the paper discusses the methodology and data source for addressing these questions. The following section presents the results of the data analysis, which is followed by a discussion of the significance of the findings. The conclusion reflects on the implications of the findings for the notion of a secondary supermarket revolution in Africa.

Methods

Study area

All of Namibia’s secondary urban centres are growing rapidly through rural to urban migration, including those in the north of the country close to the border with Angola (Table 1). The three towns of Oshakati, Ongwediva and Ondangwa are relatively close to one another along a major road transportation route (C46) and increasingly constitute a single urban corridor (Fig. 1). Oshakati and Ongwediva are 5 km apart and Ondangwa is a further 30 km away from Ongwediva. The Corridor had a combined population of nearly 80,000 in 2011 (an increase of 122% since independence from South Africa in 1991) and numbers an estimated 120,000 at the present time resident in 26,800 households (Table 1). The Corridor has been a major focus of post-independence economic development in the north of the country and the hub of trans-border trade with Angola, only 60 km away (Fig. 1). The rural population around the towns and further to the south live in scattered villages and primarily engages in communal cultivation of staple crops such as pearl millet (mahangu), livestock-rearing and the harvesting of wild foods.

Table 1 Urbanization in Namibia, 1981–2021

Fig. 1

Location of the secondary urban corridor in northern Namibia

Survey data

The data on which this paper is based comes from a representative household survey conducted by the African Food Security Urban Network (AFSUN) in 2018. A two-step cluster sampling design was adopted. In the first step, primary sampling units [PSU] (clusters) were selected from a list of all PSUs from the 2011 Population and Housing Census using the probability proportional to size method. The second step involved a random selection of a fixed number of households within each PSU using a map created by NSA that positioned each PSU using GPS. The final sample size was 853 households of an estimated 23,550 households in 2018 (or 3.5% of the total) (Table 2). The AFSUN household food security survey instrument was mounted on tablets using ODK Collect and administered to the selected household heads or their spouse/partner by a team of student fieldworkers from the University of Namibia. The survey instrument collected a wide range of demographic, economic, food consumption, and food sourcing behaviour at the household level.

Table 2 Corridor households and sample size

Household food sourcing was captured by the inclusion of the Hungry Cities Food Purchases Matrix (HCFPM) in the survey instrument. Crush and McCordic (2017) argue that the HCFPM is a unique tool for tracking patterns of household procurement of 30 or more staple, fresh, frozen and processed food items. The HCFMP collects information on the number of households that obtain each item, how often they do so, where they normally obtain it, and the spatial location of the source, with a one-year recall period. Crush and McCordic (2017) themselves use HCFPM data from Maputo in Mozambique to demonstrate its potential in the descriptive identification of food procurement patterns. This is the first paper to make systematic use of the HCFPM in a study of household food procurement in secondary cities and identify sourcing patterns by applying bivariate and multivariate statistical analysis to data generated by the HCFPM.

In order to measure the prevalence of household food insecurity in the corridor, validated assessment tools were used (Haysom and Tawodzera, 2018; Leroy et al., 2015). Food access was measured using the Household Food Insecurity Access Scale (HFIAS) which is based on the idea that the experience of food insecurity causes predictable reactions and responses at the household level that can be quantified through a summary score for each household (Coates et al., 2007; Jones et al., 2013). The Household Food Insecurity Access Prevalence (HFIAP) categorizes households into four levels of food insecurity based on their HFIAS scores: food secure, and mild, moderately, and severely food insecure. Households are categorized as increasingly food insecure as they respond affirmatively to more severe conditions and/or experience those conditions more frequently. In this paper, the four HFIAP categories were binned into two variables: food secure and food insecure.

The nutritional deficit dimension of food insecurity was captured using the household dietary diversity score (HDDS) (Swindale and Bilinksky, 2006). The HDDS is defined as the number of unique food groups consumed within the household in the previous 24 hours, and is seen as a validated proxy for both the quality and quantity of food consumption (Jones et al. 2013; McCordic and Frayne, 2018; Ruel 2003). The HDDS scores were binned into two categorical variables: low diversity (HDDS = 0–5) and higher diversity (HDDS = 6 +). Household economic status was measured using two variables: household income quintiles and the Lived Poverty Index (LPI). The LPI measures on a scale of 0 to 4 how frequently a household went without five separate essential goods and services in the previous year (Mattes et al., 2016).

Data analysis

For the analysis of HCFPM data, we first generated frequency distributions for household variables and descriptive summaries of food sourcing by type of outlet and frequency of patronage. Second, we used K-means cluster analysis to group like sources (Everitt et al. 2011; Dolnicar et al. 2018; Pedraza et al. 2021). The main advantage of this clustering technique is that it allows us to move beyond the more common formal-informal food source dichotomy. For example, two of the three clusters identified by cluster analysis actually contain a mix of formal and informal food outlets. The initial iteration for the k-means clustering analysis started with seven food sources. However, the relative change in AIC between clusters of 4 or more was sub-optimal (Santos-Pereira and Pires 2013). Instead, three clusters were found to be optimal: 1 = Convenience; 2 = Supermarkets; and 3 = Traditional. Third, all households were allocated to only one of the three clusters, based on where they predominantly purchased a 33-item food basket. This metric was computed based on the probabilities derived from the average distance of dissimilarity of each household belonging to a particular food source. Finally, we modelled cluster membership as the dependent variable using bivariate analysis and a multinomial logistic regression with household factors as explanatory variables. Variables from the bivariate correlations (p < 0.10) were simultaneously entered into regression models. All statistical analyses were conducted using R statistical software (R Core Team, 2020).

Results

Household and food security variables

Table 3 provides various key characteristics of the sampled households. Household size was relatively evenly distributed around the national average of 4.0 with 52% at or above this average. The most common household type was female-centred (i.e. households with a female head and no male spouse or partner) at 40%. Of the male-headed households, more were extended and male-centred than nuclear. Nearly 40% of the households live in informal housing (corrugated iron or tin shacks). Only 20% of households said they receive any type of social grant (mainly child, old age or disability grants). Table 3 also shows income quintiles for the 687 households who answered the question. These range from a low of less than N$1,100 (USD73) per month to a high of more than N$12,000 (USD795) per month. The majority of households (60%) reported incomes below N$4,200 (USD278 per month). The HFIAP and HDDS categories show that only 23% of households were completely food secure and only 34% had a diverse dietary intake.

Table 3 Household variables (n = 853)

Types of food source

Most households in the Corridor source different foods from multiple outlets over the course of a month. Figure 2 shows the wide variety of formal and informal, cash and non-cash sources. Food purchase is the dominant form of food procurement with only 20% growing any of their food in the rural areas and less than 5% engaged in urban agriculture. The predominant non-cash form of food consumption is food sent by relatives in rural areas which benefits around 35% of households. Figure 2 highlights that the dominant food retail sources include supermarkets, open markets, small shops (such as grocers, butcheries and bakeries), street vendors, fast food outlets and tuck shops (small informal outlets in informal settlements). Significantly, as many as 97% of households obtain food from supermarkets compared with 50% from open markets, 29% from street vendors and 19% from tuck shops. Food secure and food insecure households patronize supermarkets at roughly equally rates. Food insecure households are more likely than food secure households to patronize open markets, small shops, street vendors and tuck shops.

Fig. 2

Food sources by food security status

Figure 3 shows that there are major differences in the frequency of patronage for four of the main food retail outlets. While nearly all households purchase some of their food at supermarkets, almost 80% only shop there once per month and another 17% once per week. Both street vendors and tuck shops are patronized almost daily while open markets are predominantly patronized weekly or monthly.

Fig. 3

Frequency of sourcing from main food retailers

Food source clusters

Thirty three of the 35 food items in the HCFPM are available for purchase in each cluster (Table 4). However, the Cluster Mean Square and F Statistic values show that the purchase pattern for each of the items differs across the sample (with the exception of fresh fish [F = 0.002] and dried fish [F = 0.61]) (Table 3). For ease of reference, the dominant cluster for each food item is bolded in Table 4. Overall, the Supermarket Cluster is dominant for 18 of the food items, the Convenience Cluster for another 16 and the Traditional Cluster for only 1. The Supermarket Cluster is dominant for all of the cereal staples, 60% of the fresh produce items and 50% of the processed foods. The Convenience Cluster is dominant for all of the cooked foods, fruit, vegetables, eggs, offal and frozen meat, as well as tinned produce and sweets/chocolate. While the Traditional Cluster is a minority source for all foods except vegetables, the HCPMF did not include wild or indigenous foods which are commonly obtained from these sources.

Table 4 HCFPM food clusters

Table 5 shows how often the households in each cluster procure food from seven types of food source. The average Silhouette coefficient was 0.54, indicating fair-to-good cohesion and separation. The ratio between the largest and smallest cluster was 1.513, indicating relatively balanced cluster sizes. Within all three clusters, food shopping in supermarkets is primarily an infrequent (monthly) occurrence. However, Convenience Cluster households (and to a lesser extent Traditional Cluster households) do shop more frequently at supermarkets (36% and 25% respectively at least once per week). Both also shop more frequently at small shops, although Supermarket Cluster households do patronize these outlets on a more frequent basis too (53% at least once a week). All three Clusters show a similar pattern of more infrequent patronage of fast-food outlets and restaurants. They also share a common approach to food shopping at open markets with monthly visits more the norm. By contrast, patronage of street vendors differs considerably between clusters with Convenience Cluster households purchasing food far more frequently from street vendors. Food transfers from rural areas are infrequent in all three Clusters.

Table 5 Food sourcing frequency by food cluster

Bivariate analysis

Table 6 shows the association between the three clusters and the household variables. Household size and household type have a weak relationship with cluster type as the proportional distribution is relatively consistent across the three clusters. Female-centred households, the dominant type overall, are almost equally important in each cluster. The relationship between housing type and cluster membership appears stronger. Traditional Cluster households are most likely to be in informal housing and Convenience Cluster households least likely. In the Convenience Cluster, membership increases with increased income as 10% of households fall in the lowest quintile and 41% in the upper. This may seem counterintuitive since poor households are more likely to frequently patronize convenience outlets such as street sellers. However, the Convenience Cluster also includes fast-food and restaurant patronage which are far more likely to be patronized by higher-income, less poor households. The opposite is true with the Traditional Cluster households that have a greater chance of being in the lowest two income quintiles. Supermarket Cluster households are most evenly distributed across the income quintiles, evidence that patronage is not strongly related to household income but that supermarkets appeal to all income groups.

Table 6 Bivariate association between food source clusters and household variables

The variation in the relationship between income and procurement in the three clusters is reflected in the lived poverty, food security and dietary diversity variables. In addition to lower incomes, households in the Traditional Cluster are most likely to have higher lived poverty (50% > = 1.00), to be food insecure (76%) and to have lower dietary diversity (86%). Households in the Convenience Cluster are most likely to have low lived poverty (86% <  = 1.00), to be food secure (50%) and to have higher dietary diversity (71%). The Supermarket Cluster households fall between the two in relation to all three indicators with intermediate lived poverty (65% <  = 1.00), food security (29%) and dietary diversity (64% less diverse).

Multinomial logistic regression

Multinomial logistic regression was used to model cluster membership as the response variable with household characteristics as explanatory variables. As the response variable has K categories, k-1 logit models were required with the kth category as the reference. Traditional Cluster was set as the reference category which gives two logit models, one for Traditional versus Convenience and the other for Traditional versus Supermarket. Table 7 presents the results of the modelling exercise. Household size and housing type did not significantly increase the odds of food shopping at either convenience outlets or supermarkets as opposed to traditional food sources. Household type did have a more significant effect. For example, nuclear households had 13.236 times the odds of shopping at convenience outlets and 9.093 the odds of shopping at supermarkets as opposed to traditional food sources when compared to single person households, controlling for the other independent variables in the model. Similarly, female-centred households had 7.873 and 10.577 times the odds of shopping at convenience outlets and supermarkets respectively as opposed to traditional food sources when compared to single person households, controlling for the other independent variables in the model.

Table 7 Multiple logistic regression of food source clusters

Household income did not significantly increase the odds of food shopping at convenience outlets as opposed to traditional food sources controlling for the other independent variables in the model. However, there were increased odds of food shopping at supermarkets for almost all income classes as opposed to traditional sources when compared to the highest income quintile controlling for the other independent variables in the model. Of the two food security metrics, the most significant finding was that households that shopped at convenience outlets had 1.598 the odds of being food secure as opposed to traditional food sources when compared to food insecure households.

Discussion

Since independence in 1990, Namibia has experienced three decades of uninterrupted urban growth including of the capital, Windhoek, and the country’s numerous secondary urban centres. Windhoek has also experienced a fundamental transformation of its food system with supermarkets playing an increasingly dominant role (Nickanor et al., 2021). A key question is whether secondary urban centres in Namibia have experienced a similar transformation of their food system and whether the supermarket revolution has diffused down the urban hierarchy. This seems indisputable; an inventory of supermarkets in the urban corridor of northern Namibia, for example, indicates that there are now as many as 26 supermarkets in the three towns including 12 foreign-owned and 14 Namibian-owned outlets (Table 8). Given the heavy supermarket footprint in the corridor, the two subsidiary issues addressed in this paper are, first, the relationship between supermarkets and the food sourcing behaviour of households in the corridor, and second, the nature of the relationship between supermarkets and other formal and informal food sources.

Table 8 Supermarkets in urban corridor of Northern Namibia

The representative household survey clearly shows that virtually all of the households surveyed (over 95%) patronize supermarkets, but that supermarkets are far from being their only food source. Many households engage in what McCordic et al. (2018) call cross-platform shopping, procuring food from different types of food outlet during the course of the average month. The Hungry Cities Food Purchasing Matrix (HCFPM) and cluster analysis provide new tools for understanding these kinds of household food purchasing patterns, including in secondary urban centres. Using cluster analysis, the paper reduces a significant number of different food sources into three main types: convenience, supermarkets, and traditional. This classification contrasts with the more standard formal-informal food retail dichotomy (Crush and Frayne, 2011), as it allows the grouping together of like formal and informal sources, as well as market and non-market sources, in the same thematic cluster. The HCFPM shows the purchasing patterns for over 30 different types of staples, fresh produce, frozen and cooked food, and processed foodstuffs. Supermarkets are the most important source for all five staple foods, five fresh foods, two of three frozen foods, and five of ten processed foods. Almost all of the foods in the HCFPM matrix are also obtained by a minority of households from outlets in the Convenience and Traditional clusters, a point to which we return below.

The bivariate analysis of household income and cluster membership revealed clear differences between the clusters. For example, 10% of households in the Convenience Cluster are in the lowest income quintile and 41% are in the upper quintile. The Traditional Cluster equivalents are 35% (lowest quintile) and 10% (upper quintile). In the case of the Supermarket Cluster, by contrast, not only are the lower and upper quintile very similar (19% and 16% respectively) but it is clear that household income is not a significant determinant of whether a household shops at supermarkets. Studies in the capital Windhoek have shown that there is a distinct gender component to household food insecurity with female-centred households being especially vulnerable (Nickanor, 2013). In northern Namibia, however, female-centred households are no more food insecure than other types of household (Nickanor et al., 2019b). However, their dietary diversity is slightly lower on the HDDS scale which is likely reflected in food sourcing strategies. In each cluster, around 40% of households were female-centred. However, the odds of being in the convenience cluster were lower for female-centred than nuclear households. They were higher, however, for the supermarket cluster. The reason for this difference could reflect the need on the part of low-income female-centred households to conserve irregular income flow for monthly bulk purchase of staples at supermarkets.

Just because almost all households patronize supermarkets, it does not necessarily follow that they all do so for the same reasons or with the same frequency. Supermarkets tend to be located along the main road or one of the side roads leading out of the Corridor. They are not located in residential areas or suburban shopping malls (as in the capital Windhoek) but most households are able to reach at least one supermarket on foot given the small distances involved. Even then there are distinct patterns of patronage. Overall, there are two broad types of supermarket patron. This distinction emerged in the multinomial logistic regression which found the odds of supermarket patronage were highest for lowest and highest income households. Another survey question added further probative value to the bimodal odds ratio distribution.

All households that shopped at supermarkets were asked to rate various reasons for shopping at these outlets on a three-point scale (agree, disagree and neither). Middle and higher-income households valued the fact that supermarkets have a greater variety of foods (81% in agreement) and better quality food (80%). Lower-income households placed higher value on the fact that foods could be bought in bulk at supermarkets (67% in agreement). The question is whether the bulk purchasers only buy in bulk and, if so, when and why. This raises the issue of the frequency of purchase and the survey data makes clear that 83% of households in the Supermarket Cluster only shop at supermarkets once per month and that they do so primarily to purchase staples in bulk for household use over the course of the month. The heavy dependence on cereal staples is consistent with the low level of dietary diversity in the Cluster, but it is notable that the presence of supermarkets and supermarket shopping do not ensure overall food security for 70% of households in the cluster.

Conclusion

Early iterations of the global supermarket revolution model tended to see the increasing dominance of food systems by global and regional agribusiness corporations as an inexorable process which, according to Reardon and Gulati (2008), “has progressed far and will continue apace for years to come in developing countries.” Although the revolution had come late to Africa compared with other regions, supermarkets would eventually commandeer food production, distribution, marketing and retailing in African cities. Hitherto, supermarkets had been seen as niche players targeting economic elites in large cities. Proponents of the revolution model argued that supermarkets would eventually tap emerging mass markets accompanying rapid urbanization and become the primary food procurement source for lower-income households and the urban poor as well. As Reardon and Gulati (2008) noted, supermarkets have “now gone beyond the initial upper- and middle-class clientele in many countries to reach the mass market.” They also suggested that supermarketization would fundamentally challenge, disrupt and even replace pre-existing food supply chains and smaller formal and informal producers and retailers “who are not equipped to meet the new competition and requirements from supermarkets.”

In Namibia, both South African and locally-owned supermarkets have closely followed the South African supermarket revolution model in the capital, Windhoek. Not only do supermarkets command a large share of food retail, they have systematically expanded their geographical reach towards the low-income and informal settlements on the north side of the city. However, studies of the informal food sector in Windhoek also suggest that supermarket penetration and domination does not automatically mean the dissolution of other forms of food retail. Despite an, at times, unfriendly regulatory environment that contrasts with the enabling conditions for supermarkets, the informal food sector has continued to grow, adapting to supermarket expansion, and entering into a more symbiotic relationship with the supermarket sector (Crush et al., 2019; Nickanor et al., 2019a).

The supermarket revolution model does not address the issue central to this paper; that is, is there such a thing as a secondary supermarket revolution in smaller urban centres and, if so, is it distinctive in some way or does it follow the path of the region’s big city revolution. By examining the urban food system and food sourcing behaviour of households in the northern Namibian urban corridor, this paper seeks to address this question in the Southern African context. The HCFPM and cluster analysis show that supermarkets have established a dominant role in the food system and are patronized by almost all households. The fact that many labels on supermarket products are in Portuguese suggests that proximity to Angola may account for some of the market share. The primary ‘loser’ in Namibia’s secondary supermarket revolution are more traditional food sources including a potential decline in informal food transfers. However, lower-income households engage in monthly target shopping at supermarkets and daily food shopping from convenience outlets. Higher-income households shop for most foods at supermarkets but also regularly patronize convenience outlets such as fast-food and restaurants. Future research on the impact of the secondary supermarket revolution on the urban food system needs to examine the experience and strategies of informal food vendors and whether the symbiotic relationship with supermarkets observed in Windhoek holds in smaller towns as well. In addition, the impact of the supermarket revolution and food system transformation on other formal sector convenience outlets such as small family-owned shops and corner stores needs further study.