Cities 124 (2022) 103608

Contents lists available at ScienceDirect

Cities
journal homepage: [Link]/locate/cities

Persistence of open-air markets in the food systems of Africa's

secondary cities
Corrie Hannah a, *, Julia Davies b, Rachel Green c, Andrew Zimmer b, Patrese Anderson c,
Jane Battersby d, e, Kathy Baylis c, Nupur Joshi b, Tom P. Evans b
a
Arizona Institutes for Resilience, University of Arizona, United States of America
b
School of Geography, Development & Environment, University of Arizona, United States of America
c
Department of Geography, University of California, Santa Barbara, United States of America
d
African Centre for Cities, University of Cape Town, South Africa
e
Department of Environmental and Geographical Science, University of Cape Town, South Africa

A R T I C L E I N F O A B S T R A C T

Keywords: In a rapidly urbanizing sub-Saharan Africa (SSA), changing urban food systems are expected to affect the ways in
Food systems which households access food. While significant research and policy efforts address food production to support
Food security food security in the context of city-region food systems, the complex nature of urban food systems has received
Urbanization
less attention, particularly in secondary cities across the globe. Using household survey data collected in 2019
Urban planning
Secondary cities
across 4 cities in Kenya and 14 cities in Zambia, we examined patterns of household food acquisition from key
Africa food retail locations. Households predominantly purchased food from open-air markets and vendors, and only a
subset of households bought food from supermarkets. Variations in urban population size, area, and connectivity
across the 18 cities presented a clustering of four city types that shape household food access and purchasing
patterns at food retail locations. Despite a growing narrative that supermarkets and western diets will dominate
SSA's urban food systems, our findings suggest that low- to middle-income households will continue to purchase
food at open-air markets in SSA's secondary cities in the coming years. Attention to these urban household
purchasing patterns can inform urban planning and governance priorities to ensure that SSA's food systems can
meet urban food security needs.

1. Introduction 2015).
Urban food systems planning becomes particularly crucial when
Cities across the globe have grown rapidly in recent decades, espe­ considering these demographic forecasts in the coming decades (Béné
cially in Asia and Africa. In sub-Saharan Africa (SSA), one billion people et al., 2020; Kraemer et al., 2016). Food systems broadly encompass the
will be living in urban areas by 2050, which is a substantial increase interactions between and within the biogeophysical and human envi­
from the region's current urban population of 350 million (UN-DESA, ronments, where a suite of activities span food production to food
2014). Although much of this urban growth has occurred in primary or consumption, including food processing, packaging, distribution, retail,
mega-cities, such as Lagos in Nigeria or Cairo in Egypt, secondary cities and waste management (Ericksen, 2008). At the city scale, urban food
now host a growing proportion of urban dwellers globally and represent systems comprise a hybrid mix of market and non-market food sourcing
an important component of the African urbanization process (Sat­ strategies, and formal and informal food retail environments (Battersby
terthwaite, 2017; Satterthwaite et al., 2010). Since definitions of sec­ & Watson, 2018; Tefft et al., 2017). As urbanization increases, super­
ondary and tertiary cities can vary, we use the term ‘secondary cities’ in markets across Africa, Asia, and Latin America are projected to increase
this paper to broadly encompass the range of tertiary and secondary in prevalence, particularly where there is growth in the number of
urban areas with populations between 5000 and 500,000 people. Sec­ moderate-income households (Pingali, 2007; Reardon et al., 2003;
ondary cities comprise 26.5% of the world's overall population (Chai & Reardon & Timmer, 2014).
Seto, 2019) and over half of the global urban population (Buettner, A common narrative tied to urbanization and the rise of

* Corresponding author at: Environment & Natural Resources, 2 Building, 1064 E Lowell Street, PO Box 210137, Tucson, AZ 85721-0137, United States of America.
E-mail address: corrieh@[Link] (C. Hannah).

[Link]
Received 19 March 2021; Received in revised form 17 January 2022; Accepted 19 January 2022
Available online 1 February 2022
0264-2751/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license ([Link]
C. Hannah et al. Cities 124 (2022) 103608

supermarkets suggests that supermarkets can offer products of consis­ removal of street vendors and the upgrading or relocation of open-air
tent quality at low costs due to economies of scale. Some scholars have markets has caused conflict between vendors and local government
argued that these modern retailers may soon replace traditional food authorities in cities in Zambia and Ghana (Asante, 2020; Asante &
retail outlets in SSA, such as open-air markets and vendors, which is a Helbrecht, 2019; Battersby & Muwowo, 2018).
trend that has occurred in Latin America and East and Southeast Asia While the Sustainable Development Goals (SDGs) include both “zero
(Balsevich et al., 2003; Reardon et al., 2005). However, much of the hunger” and “sustainable cities and communities” (Goals 2 and 11,
supermarket development that has occurred in SSA has taken place in respectively), neither has targets or indicators that explicitly address
larger cities and in specific countries, most notably in South Africa, urban food systems (Battersby, 2017b; Crush et al., 2020). The United
Kenya, Zambia, and Zimbabwe (Reardon et al., 2003; Weatherspoon & Nations' 2017 New Urban Agenda prioritizes food planning and gover­
Reardon, 2003). Yet, supermarkets may have less influence in secondary nance dimensions that predominantly focus on regional scales and rural-
cities in SSA. For example, households in two secondary cities in Nigeria urban linkages, but does not explicitly engage with the role that urban
predominantly procure food through open-air markets and informal food systems planning can play in facilitating food security and nutrition
means (Resnick et al., 2019). in smaller cities (Battersby & Watson, 2019). These efforts fail to
Understanding how and why urban residents obtain food from su­ adequately reflect the challenges and realities of contemporary urban
permarkets, open-air markets, and other retail options is important for food systems and have paid little attention to spatial planning, retail
the study of food security in an urban planning context. Pursuing sus­ logistics, and governance of the urban components of food systems
tainable urban food systems as an urban planning strategy can facilitate (Battersby & Watson, 2019). Several translocal networks, such as the
urban food security, environmental security, and social welfare out­ Milan Urban Food Policy Pact, C40 Cities, and ICLEI-Local Governments
comes (Ericksen, 2008; Melesse et al., 2020). However, achieving this for Sustainability, are now directing the focus of local governments to­
level of urban food systems sustainability will require integrated policies wards urban food systems (Moragues-Faus, 2021). However, few
and urban planning initiatives based on empirical data and analyses of research initiatives have investigated how the structural dynamics of
urban food systems (Battersby & Watson, 2019; Tacoli, 2019). One key cities shape food provisioning and household food procurement patterns
motive for urban food systems planning and the study of urban food in urban areas (Milan Urban Food Policy Pact, 2015).
systems is to address growing challenges with urban food insecurity, Our study contributes to this growing literature and global policy
which has been exacerbated by the COVID-19 pandemic and the asso­ dialogue on urban food systems planning with an analysis of how
ciated economic crisis (Béné, 2020; Devereux et al., 2020; Moseley & household food purchasing patterns vary across different retail locations
Battersby, 2020). Food security exists “when all people, at all times, in SSA's secondary cities. A fundamental challenge experienced by urban
have physical, social and economic access to sufficient, safe and nutri­ governments in SSA and in developing countries globally is a lack of data
tious food to meet their dietary needs and food preferences for an active and standardized methodologies for characterizing trends in urban food
and healthy life” (FAO, 2009). Yet, the context and challenge of food systems, which results in missed opportunities to articulate strategies for
security differs between rural and urban settings (Haysom & Tawodzera, integrated urban food systems planning. Thus, with a large-N sample of
2018). For example, the Food and Agricultural Organization's (FAO) urban households in Kenya and Zambia, we present the first multi-city
survey of 146 countries in 2014–2015 found that 50% of urban house­ analysis of urban food systems in secondary cities in SSA. This system­
holds in the least developed countries were food-insecure, compared atic data collection approach offers a baseline for understanding the
with 43% in rural areas. The situation is even worse in informal urban nature of household acquisition of food in smaller cities more generally.
settlements, with rates of food insecurity reaching 70–95% of the pop­ The motive of our research is to promote the development of a stan­
ulation (Tefft et al., 2017). dardized methodology that scholars and practitioners can use to study
Yet, from an urban planning perspective, the problems of hunger and and compare urban food systems in cities of all sizes.
malnutrition in cities are often neglected relative to “more urgently Our analysis also addresses a knowledge gap regarding how
visible problems” such as housing or sanitation (Maxwell, 1999). Low- household-level food acquisition varies across cities, especially in SSA.
income households also tend to employ diverse strategies to cope with While some analyses have found inter-city variation with respect to food
food insecurity, which can render the urban food security problem consumption patterns and associated environmental impacts across nine
invisible to local government officials (Battersby, 2017a; Maxwell, large Indian cities (Boyer et al., 2019), less is known about how
1999). Moreover, local governments in smaller cities often do not have household-level food acquisition varies across urban food systems in
the resources and governance capacities needed to address urban food secondary cities. By comparing household food acquisition across
systems challenges (Sonnino et al., 2019). Mandates for addressing food several cities, our quantitative, multi-city approach also complements
security typically sit with the national departments that are responsible the qualitative, case study-based approach of the Hungry Cities Part­
for ensuring that rural agricultural production meets the demands for nership and the Africa Food Security Urban Network (Haysom &
food availability and supply. With this national focus, the challenges Fuseini, 2019). We categorized the secondary cities in our sample into
surrounding food access at the urban scale are often overlooked in food four types based on their size and/or connectivity, namely: 1) smaller
systems planning and governance (Battersby & Marshak, 2017; Crush & cities, 2) remote mid-sized cities, 3) connected mid-sized cities, and 4)
Frayne, 2010; Crush & Riley, 2018). larger cities. We then assessed the relationship between these secondary
There is also a global need for studying urban food systems for urban city types and the prevalence of household food purchases made at
planning purposes that is not just attributed to a set of cities or a region. different food retailers, including open-air markets, vendors, local
International attention to urban food systems planning has thus far shops, and supermarkets.
largely been partial, contingent, and, at times, poorly aligned to local Finally, we identified some reasons for why households in SSA's
realities. For instance, in a systematic survey of 170 Asian cities in 21 secondary cities purchase food from certain retailers over other retailers
countries, Acharya et al. (2021) found that only 8% of cities intervene in and discuss the household- and urban-scale factors that shape these
their urban food systems in forward-looking, holistic, and inclusive behaviors. While scholars of urban food systems have tended to focus on
ways. Moreover, due to a lack of food systems policy and governance, the ultimate outcome of household-level food security, either measured
the majority of these cities are only in the early stages of developing with food insecurity indices (Blekking et al., 2020; Legwegoh & Riley,
urban planning policies, and many respond to urban food systems 2014; Tuholske et al., 2020) or biometric indicators (Demmler et al.,
problems reactively (Acharya et al., 2021). In the SSA context, urban 2017; Khonje et al., 2020), we take an alternative approach by identi­
food systems planning has largely been shaped by a modernization fying household- and city-level factors that explain where and why
agenda in which governments favor formal food retail, leading to ten­ households tend to access food in an urban food system. We found that
sions with the traditional and informal food sectors. For example, the most households in our sample source their food from open-air markets.

2
C. Hannah et al. Cities 124 (2022) 103608

Hence, we discuss the relative persistence of open-air markets in new store investments. However, even in SSA's larger sized secondary
different types of secondary cities and the implications of this for urban cities, open-air markets are an important food source for many house­
food systems planning in SSA and globally. holds (Crush & Frayne, 2010; Smit, 2018).
Similar to studies of the ‘supermarketization’ trend in the Global
2. Background: urban food systems South, urban agriculture has received significant attention in the urban
food systems literature (De Zeeuw et al., 2011; Frayne et al., 2014; Prain
Understanding the nature and diversity of the urban food retail & Lee-Smith, 2010; Thornton, 2020; Zezza & Tasciotti, 2010). Some
environment and household food sourcing strategies is important for benefits of urban agriculture include its contribution to household food
dissecting the systemic challenges of urban food security and for consumption, income generation, urban greening, and social cohesion
developing urban planning strategies that support sustainable urban (Lee-Smith, 2010; Nkrumah, 2018; Thornbush, 2015; Toriro, 2018).
food systems. To begin, the urban food retail space is a critical compo­ Urban agriculture can also provide a buffer for households that are
nent in the study of urban food systems, because urban households vulnerable to food insecurity in the event of economic stress or shocks
overwhelmingly purchase, rather than produce, their food (Crush et al., (Lee-Smith, 2010; Nkrumah, 2018; Smart et al., 2015; Thornbush, 2015;
2012; Haysom & Fuseini, 2019; Seto & Ramankutty, 2016; Smit, 2016). Toriro, 2018). However, multiple barriers (e.g., available land, pro­
Major urban food retailers in SSA generally include open-air markets, ductive resources, and tenure security) prevent low-income households
street vendors, local shops, and supermarkets, and households typically from engaging in urban agriculture in SSA (Davies et al., 2020).
obtain food from a mix of these retailers. For example, a household may Beyond the retail sector and urban agriculture, some urban house­
buy vegetables from an open-air market, meat from a local butcher, and holds also have farm holdings outside of the city, which may be
milk and bread from a supermarket. Restaurants and fast-food outlets managed by themselves or by relatives (Cattaneo et al., 2021). Further
can also play an increasing role in household food sourcing strategies in rural-urban linkages of food acquisitions occur through cash remittances
some cities. In other cities, households may supplement purchased food and food transfers between rural and urban households, where urban
with food obtained from urban or peri-urban agriculture, food sharing households can collect food produced in peri-urban and rural locations
and remittances, and civil society organizations (e.g., non-government from familial or social connections (Crush & Caesar, 2017; Crush &
organizations or religious groups). Caesar, 2018; Crush & Ceasar, 2020; Mabrouk & Mekni, 2018). Food
One line of scholarly work on urban food systems has examined the transfers play a particularly important role in contributing to food
tensions and competition between the traditional or informal food sec­ supply among migrant households (Frayne & Crush, 2017; Tawodzera &
tors (i.e. street vending, open-air markets, and locally owned shops, Crush, 2016), and social networks can be a crucial strategy for
called ‘Spazas’ in South Africa), which have long dominated food retail improving food security within urban communities (Claasen & Lemke,
in African cities, and the rise of formal shops, particularly large, modern 2019). The urban food system shapes the culmination of these house­
supermarkets (Berger & van Helvoirt, 2018; Crush & Frayne, 2011; hold food acquisition strategies that range from retail food purchases to
Owusu-Sekyere et al., 2016; Reardon et al., 2021). As supermarkets have urban agriculture to food transfers. Key attention to the study of the
increasingly spread throughout SSA (das Nair, 2020), they tend to be nature of how urban residents navigate these urban food systems across
favored by urban governments over informal food retail because they cities is warranted for urban food systems planning.
bring in tax revenue and are thought to stimulate local economies
(Skinner, 2016). These larger supermarkets are often envisioned as 3. Materials & methods
symbols of modernity, and are associated with cleanliness, quality, and
economies of scale (Abrahams, 2009; Battersby, 2017a; Zhong et al., 3.1. Research setting
2019).
Despite these purported benefits of supermarkets, their role in SSA's We conducted our study in 18 secondary cities in Kenya and Zambia
urban food systems has been met with some reproval by some local (Fig. 1). Situated on the coast of eastern Africa, Kenya has a population
urban populations. For example, supermarkets tend to target wealthy of approximately 52 million, with 27% of people living in urban areas.
consumers, which does little to alleviate food insecurity among the Zambia, in contrast, is a landlocked country with a population of nearly
urban poor (Battersby, 2019; Battersby & Watson, 2018; Berger & van 18 million and roughly 45% of people living in urban areas (World Bank,
Helvoirt, 2018; Nickanor & Kazembe, 2017; Peyton et al., 2015). Su­ 2020). Agriculture and livestock dominate livelihoods in both countries
permarkets can create unfair competition with small, locally-owned (FEWSNET, 2011; FEWSNET, 2014), and rural-urban and urban-urban
shops as well as the informal sector as they are able to undercut pri­ food transfers are common, especially among households most vulner­
ces, particularly when it comes to non-perishable food items but able to food insecurity (Andersson Djurfeldt, 2015; Frayne, 2010).
increasingly also for fresh produce (Minten et al., 2010; Minten & Studying urban food systems across secondary cities in Kenya and
Reardon, 2008). This monopoly on non-perishable items has suggested Zambia presents an opportunity to compare how households are navi­
that supermarket expansion is contributing to a transition in African gating urban food systems in two SSA countries that have undergone
cities towards a more westernized diet, which introduces greater risks of various urban food system transformations. For example, Karatina in
obesity and malnutrition due to higher consumption of low-nutrient, Kenya is home to one of the largest open-air markets in Central and East
highly processed foods (Battersby, 2019; Battersby & Peyton, 2016; Africa, serving as a central node in facilitating economic exchange be­
Demmler et al., 2017). This is especially true when households are tween urban and rural areas, and as a crossroad for trade across the
located close to supermarkets and fast-food restaurants (Otterbach et al., entire region (Mbataru, 2017). Retail modernization in Kenya has been
2021). viewed as a market tool to combat food insecurity. However, the pro­
Beyond modern retail locations, the informal food retail economy motion of supermarket development at the expense of existing open-air
continues to be a key source of household food procurement, particu­ markets has been challenged for primarily serving higher-income con­
larly for low-income households (Giroux et al., 2020; Skinner, 2016). sumers (Berger & van Helvoirt, 2018). While traditional markets and
Several studies from SSA have found that a large proportion of the urban informal traders remain the main sources of fresh produce, some
poor rely substantially on open-air markets to access food (White et al., scholars have observed that the development of supermarkets in Kenyan
2018). For example, a survey of 11 southern African cities found that cities has led to a nutrition transition, as households are more readily
70% of households frequently obtained food from open-air markets able to buy packaged and processed food (Battersby & Watson, 2018;
(Frayne et al., 2010). This may be attributed to the fact that many sec­ Opiyo & Agong, 2020).
ondary cities in SSA have not yet seen the introduction of supermarkets, Secondary cities form important transportation and market links
who naturally target areas with higher market densities and incomes for between local agricultural producers and urban consumers. However,

3
C. Hannah et al. Cities 124 (2022) 103608

Fig. 1. Study site locations in Kenya and Zambia with an example presentation from Choma, Zambia, of how households were sampled across residential areas in
each city.

Zambia relies more heavily on these in-country food flows, importing household spends the greatest proportion of their food budget, what
fewer food products (i.e., 7% of merchandise products in 2019) types of food are purchased from these retailers, how much money was
compared to Kenya (i.e., 15% of merchandise products in 2019) (World spent across all retail locations in the past week, and how often their
Bank, 2020). Zambia has also been transitioning to a more formalized household purchased food at these retail locations in the past two weeks.
approach to urban food systems governance by banning informal street Responses for currencies were recorded in Kenyan Shillings and Zam­
trading and attempting to confine street vendors to open-air markets bian Kwacha, and then converted into USD based on the currency ex­
(Blekking et al., 2017). However, street vendors continue to play an change rate at the time of data collection in 2019. We included an
important role in helping to overcome spatial disparities in food access additional subset of questions aimed at understanding how households
and sourcing (Blekking et al., 2017; Giroux et al., 2020). were embedded in the broader food system. These questions included
the number of times a household obtained food from a rural source in
the past year, the distance to the location where they purchase the most
3.2. Household data collection food, whether they were a recipient of food remittances from a rural or
an urban source, whether they engaged in urban agriculture, and
Our dataset consists of 3020 household surveys that were conducted whether they owned a farm outside of the city.
in four Kenyan cities (n = 1012) and 14 Zambian cities (n = 2008) from We sampled secondary cities in each country, using the Global
May to August 2019, in partnership with Mpala Research Centre in Human Settlement Population (GHS-POP) dataset for reference (Schia­
Kenya and the Zambia Agriculture Research Institute in Zambia. The vina et al., 2019). We selected cities within our study sites that have
four cities in Central Kenya included Nanyuki, Karatina, Timau, and populations ranging from 5000 to 200,000 people, which represents
Naro Moru. The fourteen cities in Zambia include Batoka, Choma, smaller secondary cities according to our definition. We focus on these
Chongwe, Itezhi-Tezhi, Kapiri Mposhi, Maamba, Mazabuka, Mbabala, smaller secondary cities, because nearly half of Africa's urban popula­
Mkushi, Mpongwe, Namwala, Nyimba, Pemba, and Petauke. We tion is concentrated in urban settlements with fewer than 300,000 in­
sampled more cities in Zambia than Kenya because the size of our field habitants (United Nations and Social Affairs, 2018). Setting our upper
teams differed in the two countries. However, our selection of towns in sampling threshold at 200,000 people allows us to systematically focus
both Kenya and Zambia constituted a census of towns for the regions on the nature of food systems in these smaller secondary cities, where
sampled. In Kenya, we sampled within one region on the west side of Mt. little to no empirical data collection has been pursued, especially across
Kenya, and in Zambia, we sampled across the Southern, Eastern, Central, multiple cities.
and Copperbelt Provinces. Local enumerators who were fluent in local We also selected cities that represent the closest market town for
languages administered the survey using tablet computers. We obtained surrounding rural communities. In the field, we opportunistically sur­
permission from municipal-level authorities and community leaders in veyed additional towns not represented in the GHS-POP dataset. These
both countries to conduct research activities in each urban area. We also ranged from micro-urban places of less than 1000 households (i.e.,
obtained Institutional Review Board approval from the first author's approximately 5000 people) up to secondary towns that are regional
home institution to conduct human subjects-based research via house­ centers of commerce. Some of these cities also serve as administrative
hold surveys in Kenya and Zambia. decision-making centers, providing governance oversight for the local
The survey included questions about where the respondent's

4
C. Hannah et al. Cities 124 (2022) 103608

administrative units of counties in Kenya and districts in Zambia. Within 3.4. Analytical approach
each city, we sampled households in low- to middle-income residential
areas to capture the characteristics of households that are most at risk of To address the research questions of where and why urban house­
experiencing food insecurity. Most sampled cities did not have large holds in SSA shop at certain food retailers, we presented a series of
wealthy residential areas that are typical of primary cities (e.g., Nairobi, quantitative analyses from the survey data. We supplemented these
Kenya and Lusaka, Zambia). Although moderate sized urban areas (with survey data with open-ended responses and observations from the field
populations of 100,000+) had some stratification of residential areas by to clarify and contextualize our findings. Following common principles
income, these wealthier areas make up a small proportion of the total and practices in mixed-methods analysis (Fetters et al., 2013), we in­
population. tegrated these different types of information to elucidate coherence in
Finally, we used a systematic random sampling approach in each the trends we identified in our findings.
residential area to select households. We used a combination of satellite We first presented descriptive statistics to illustrate where house­
imagery on GPS-enabled tablet computers and guidance from research holds across different cities in Zambia and Kenya purchase food, what
partners to sample households along roads. We skipped a set number of types of commodities are purchased at food retail locations, and how
households relative to the size of the sampling area to ensure a repre­ often households purchase these commodities. Then, using a hierar­
sentative and spatially distributed sample within each residential area. chical cluster analysis, we identified four city types based on city size
Apartment buildings located in the central business districts in Kenyan and connectivity. We used these city types to identify household char­
urban areas were treated similarly. We estimated the number of acteristics that are associated with food purchasing patterns at food
households living in one apartment complex, chose a random starting retail locations. We performed an additional set of regressions to
point and then surveyed every three to five apartments, depending on examine where households spent most of their food expenditures rela­
the size of the apartment complex. If an enumerator arrived at a house or tive to where they have recently purchased food. This allowed us to
apartment with no available or eligible respondent, then the enumerator gauge which food retail locations are likely substitutes or complements
continued to each successive house until they were able to successfully to one another. We concluded our analysis with results that provide
conduct a survey, and then again followed the procedure of skipping further insight for why households shop at specific retail locations. We
three to five households. In most residential areas the number of surveys presented information regarding household perceptions of retail loca­
conducted ranged from five to thirty, depending on the size and density tions, the relative demand for a retail location based on a willingness to
of the residential area. For example, some smaller cities essentially had pay exercise, and household preferences for food products with a hy­
only one residential area. This yielded a sample size of 30 to 300 pothetical increase and decrease in food expenditures. The culmination
households sampled in each town. This data collection approach of these findings provides a greater understanding for how households in
allowed for a spatially stratified random sample within the residential SSA's secondary cities navigate their urban food systems.
areas to get a spatially represented sample of households, where the
household numbers sampled are estimated as proportional to city size. 4. Analysis & results

3.3. City-level data 4.1. Overall trends

To understand how city-level characteristics (namely urban area, First, we asked household respondents to specify where their
high density area, and connectivity) shape urban food systems and household typically spends the most money (i.e., the largest proportion
household food purchasing patterns in secondary cities, we incorporated of their food expenditure budget) on food purchases. Options in the
additional city-level data from secondary geospatial datasets into our survey included open-air markets, street vendors, local shops, super­
analyses. Following Tuholske et al. (2019), we derived urban area and markets, or other locations (e.g., wholesale retailers, directly from
population density using a two-step approach for 12 of the 18 cities. farmers). Open-air markets are designated centers where traders sell
First, for each city, we overlaid urban area polygons derived from goods to consumers. Municipal councils often oversee and regulate these
OpenStreetMap (OSM) with the 2015 GHS-POP gridded population markets via collection of daily levies, and many markets use formal
database (Haklay & Weber, 2008; Schiavina et al., 2019). Then, we governing mechanisms, such as a constitution or by-laws, to organize
selected all grid cells that intersected with the OSM data with a popu­ market activities. Vendors are the individuals who sell goods on tarps or
lation density exceeding 1500 people per km2. Adjacent grid cells with makeshift kiosks alongside streets or roadsides. In contrast to open-air
population density values greater than 300 people per km2 were added markets, vendor activities are largely unregulated by an external au­
to the initial grid cell selection to account for suburban and peri-urban thority. Local shops are independently owned and operated retail loca­
areas. These combined grid cells created an urban footprint for each tions that may be situated in a small, designated retail space in a city's
city to represent the urban area in km2. We approximated population central business district, a residential neighborhood, or directly attached
density with a high density area variable by summing the number of to a household. Supermarkets are large, commercial, transnational, or
pixels associated with the GHS-POP dataset for the year of 2015 in the regional chain outlets such as Shoprite, Choppies, Nakumatt, or Ukwala.
urban footprint. OSM data was unavailable for six of the cities in our Across all cities, food purchases made at open-air markets and ven­
sample. Hence, we manually digitized the urban area extent in these dors dominated household-level purchases in terms of food expenditures
cities based on satellite imagery and field notes to create the urban (Fig. 2). Kenyan household respondents reported spending the most
footprint, which was then used to generate the variables for urban area money on food at vendors (30%), open-air markets (28% of re­
and high density area. spondents), supermarkets (22%), and local shops (11%). In comparison,
To approximate the connectivity of a city, we derived two proximity most household respondents in Zambia spend the most money on food
measures from the open-source gROADS database (CIESIN & ITOS, purchases at open-air markets (78% of respondents), followed by local
2013): 1) the distance from the city to the closest main road, and 2) the shops (12%), vendors (5%), and supermarkets (4%). Example locations
average road distance from the centroid of the city to the centroids of the in the “Other” category included wholesale markets, canteens, restau­
three closest urban centers. The neighboring urban centers came from a rants, prepared street food, mobile sellers, and traders.
database of urban areas ranging from secondary cities to large primary Food purchasing patterns also varied by city size (Fig. 2), which
cities across SSA (Zimmer et al., 2020). we've approximated using a measure of high density area (i.e., the
number of pixels in a city with population densities exceeding 1500
people per km2 in combination with adjacent pixels of 300 people per
km2 to account for suburban and peri-urban areas). Households in larger

5
C. Hannah et al. Cities 124 (2022) 103608

Fig. 2. Food retail locations where house­

hold respondents spend the most money on
food purchases by city in Kenya and Zambia.
Although we did not scale our sample size
directly to the city size (i.e., high density
area) in our sampling design, the number of
sampled households is generally larger in
cities that are larger in population. Cities
with one or more operating supermarkets at
the time of data collection are noted with a
starred symbol, and we assume that each
city has one or more open-air markets
alongside several vendors and local shops.

cities spent the most money at supermarkets compared to households in markets for vegetables and fruits, but purchase maize, meat, and fish
smaller cities. Some smaller cities did not have any supermarkets, which at local shops. While some respondents in Zambia also use local shops
may also explain why a high number of households in these cities pur­ for their maize, meat, and fish purchases, many respondents purchase all
chased food at open-air markets and local shops. food commodities at open-air markets. Fig. 3B illustrates that most
The location of household food purchases also varied by commodity. households in both Kenya and Zambia purchase vegetables frequently
As shown in Fig. 3A, respondents in Kenya primarily shop at open-air over the course of the month. Conversely, respondents in both countries

Fig. 3. Country-level plots for (A) the retail locations where households most often purchase different food commodities (i.e., vegetables, fruit, maize, meat, and fish)
and (B) how many times households purchased these food commodities in the past month. ‘Not Applicable’ refers to households who do not purchase those products.

6
C. Hannah et al. Cities 124 (2022) 103608

reported purchasing fruit, maize, meat, and fish less than ten times per For some cities that do not have a supermarket, some households
month (Fig. 3). indicated that they nevertheless made purchases at a supermarket
Several factors may contribute to the greater proportion of house­ (Table 1). While enumerators were trained to use the term “supermar­
holds in Kenya purchasing maize meal in supermarkets. Kenya has a kets” for large, commercial, transnational, or regional chain outlets,
more developed retail sector than Zambia both in terms of supply chain survey respondents held a variety of perceived notions regarding what
networks and formal food retailers (supermarkets and retail shops). On constitutes a supermarket. For example, only six of our sampled 18 cities
average urban households in Kenya also have higher income levels than had at least one supermarket, as per our definition thereof. Yet one or
households in Zambia. Households that have the financial means buy more households in 16 of the 18 cities reported that there indeed was a
maize in large quantities because the price per kilogram is lower than supermarket in their city. These definitions of supermarkets among
when purchasing in smaller quantities. In both countries, large quantity residents are subjective, and purchases made at a so-called “supermar­
bags of maize meal are more available in supermarkets, whereas vendors ket” could have included any formal food retail store ranging from a
at open-air markets and local shops typically sell a wider range of sizes of locally owned mid-sized outlet to a large-scale commercial supermarket.
maize meal, ranging from 2 kg to 25 kg to accommodate the range of An additional reason for these incongruencies may be attributed to cases
income earners. Since Zambia has a greater proportion of lower income whereby a smaller city is located near to a larger city with relatively
households, there is a greater proportion of households that purchase easier access to a supermarket, such as in the case of Batoka, which is
maize meal more frequently in smaller quantities from open-air markets. close to Choma in Zambia.

4.2. Understanding food system trends by city type 4.3. Household characteristics related to food purchasing behaviors

Secondary cities are often described as the intermediate urban ag­ To explore why households frequented open-air markets more often
glomerations located along a gradient from rural villages to large pri­ than other food retailers, we regressed a range of household character­
mary cities (Andreasen et al., 2017; Vernon Henderson & Kriticos, istics on the type of retail location where households in each country
2018). However, we found diversity across different types of secondary spent the most money on food purchases. Bigsten et al. (2003) and
cities. To explore this diversity by city type, we performed a hierarchical Demeke et al. (2011) adopted similar approaches to identify features of
clustering analysis of our 18 cities and then explored whether those households that correlate with poverty and food insecurity. At the city-
clusters correlate with household food purchasing patterns. We identi­ level, environmental factors that influence food choices include varia­
fied four unique city clusters based on four variables that capture: (1) tion in food prices by geography, store type, or nutritional quality of
city area, (2) high density area, and connectivity, which was measured food items. Additional disparities exist in the ability of consumers to
by (3) distance to the main road, and (4) average distance to the three access food outlets that sell healthier options such as fruits and vege­
closest cities (Table 1). The combination of these four variables explains tables, and which are also located along public transportation routes
86% of the variation in the cluster types. Appendix A in the Supple­ (Caswell et al., 2013). Our approach offers another perspective by
mentary Materials provides a full description of the methods that we focusing on household food procurement patterns with respect to both
used in the clustering analysis. retail and non-retail sources in the urban food system.
Cities in the first cluster were small in geographic size and urban area Our dependent variable for all regressions was the categorical vari­
density, within walking distance to the main road, and are well con­ able of the locations where households spend the most money on food
nected to other urban areas. Cities in the second cluster were of mod­ purchases (Fig. 2). We estimate two sets of regression models. The first
erate geographical size and population density, however they were set correlates where households generally spend the most money rela­
limited in their connectivity to other cities and the main road. Cities in tive to the frequency of their food purchases at other food retail locations
the second and third cluster shared many similarities regarding in the last two weeks. The second set of regression models identify how
geographical area and population size, however cities in the third cluster independent household characteristics are related to where households
had better access to the main roads. Cities in the fourth cluster were spend most of their money on food purchases. Since open-air markets
densely populated cities close to the main road and were within average were the predominant food retail location where households purchase
driving distance to the three nearest cities. food, we performed a multinomial logit model with open-air markets as
Consumer shopping patterns varied by city cluster type (Table 1, the reference category. This approach allows us to understand why
Fig. 4). Across all city cluster types in both countries, households spent a households would purchase food at vendors, local shops, supermarkets,
higher amount of money at supermarkets than other food retail loca­ and other food retail locations relative to open-air markets. We then
tions, even though frequencies of purchases made at supermarkets were developed a logistic regression model to understand why households
significantly less (Table 1). We found that in Kenya, a higher percentage would shop at open-air markets relative to all other food retail locations.
of households in larger cities (City Cluster Type 4) purchased food from Since 93% of households in Kenya and 79% of households in Zambia
supermarkets than in smaller cities (City Cluster Type 1), with a rela­ purchase food at more than one type of food retailer, we first explored
tively similar percentage of households purchasing from vendors and which retail types are substitutes or complements to one another. The
open-air markets in these two cluster types. In Zambia, a higher per­ first set of models serves as a correlation analysis to assess how likely
centage of consumers residing in smaller cities (City Cluster Type 1) households would be to spend most of their food budget at a given
purchased food from open-air markets compared to households in the location relative to the frequency of purchases at other food retail lo­
other three city types. Similarly, the proportion of households in cations (Table 2). From both the logistic and multinomial models in
disconnected cities and cities close to the main road (City Cluster Types Kenya and Zambia, we find that households are significantly less likely
2 and 3) purchased food primarily from open-air markets, followed by to spend most of their money at other food retail locations if they have
local shops and vendors. A higher percentage of our sample spent the frequently purchased food at open-air markets in the past two weeks.
most of their food expenditures at supermarkets in Zambia's largest Additionally, households in both Kenya and Zambia were significantly
cities compared to other city types, but even so this percentage was less likely to go to local shops than open-air markets if they had
nominal relative to the proportion of households purchasing food at frequently purchased food at vendors in the past two weeks – a result
open-air markets. The largest percentage of consumers spending money that suggests local shops and vendors may be substitutes. Moreover,
in local shops were in Kenya's small, connected cities (City Cluster Type households who purchased food at local shops in the past two weeks
1) and Zambia's mid-sized cities, and the largest percentage of con­ were also more likely to spend most of their money at both local shops
sumers spending money in supermarkets were in large densely popu­ and supermarkets relative to open-air markets and vendors, suggesting
lated cities (City Cluster Type 4). that a subset of our sample size tended to purchase food more regularly

7
C. Hannah et al. Cities 124 (2022) 103608

Table 1
Aggregated descriptive statistics for the four city cluster types based on the four variables that define the city cluster types (i.e., city area, urban area density, distance to
a main road, and the average distance to the closest three urban centers). We also included variables that describe the food system from our survey data and estimated
population sizes of these cities from the 2018 LandScan gridded population dataset. Cities with an asterisk have at least one supermarket.
City Cluster Type 1 (n = 485) – smaller cities Min Max Mean Median Std. dev.
Timau, Naro Moru, Mbabala, Batoka, Pemba

City area (sq. km) 0.82 3.51 2.29 2.32 0.96

High density area (number of pixels) 4 18 9.6 7 5.77
Distance to main road (km) 0.7 29.36 9.2 1.46 12.49
Avg. distance to next three cities (km) 22.16 39.72 28.88 24.33 8.34
Estimated population from LandScan 2018 231 2670 1384.4 1075 1169.95
Food expenditures in the past week (USD) – all retail locations
Open-air market 0 49.02 6.03 3.92 6.9
Vendor 0 53.92 3.62 1.96 5.34
Local shop 0 58.82 3.41 1.54 6.49
Supermarket 0 186.27 2.35 0 10.85
Number of food purchases in the past two weeks
Open-air market 0 14 7.71 4 5.21
Vendor 0 14 7.64 7 5.89
Local shop 0 14 3.16 2 4
Supermarket 0 14 0.46 0 1.67

City Cluster Type 2 (n = 450) – remote mid-sized cities Min Max Mean Median Std. dev.
Mpongwe, Namwala, Maamba, Itezhi-Tezhi*

City area (sq. km) 3.77 8.72 6.31 6.37 2.02

High density area (number of pixels) 5 34 17.25 15 13.77
Distance to main road (km) 42.69 88.74 70.71 75.7 22
Avg. distance to next three cities (km) 79.07 130.77 96.63 88.33 23.19
Estimated population from LandScan 2018 1122 4966 3807.25 4570.5 1826.19
Food expenditures in the past week (USD) – all retail locations
Open-air market 0 76.92 6.67 3.85 9.01
Vendor 0 9.23 0.51 0 1.12
Local shop 0 30.77 2.2 0.77 3.57
Supermarket 0 115.38 0.45 0 5.96
Number of food purchases in the past two weeks
Open-air market 0 14 6.58 5 4.99
Vendor 0 14 1.89 0 3.59
Local shop 0 14 2.07 1 2.97
Supermarket 0 2 0.04 0 0.24

City Cluster Type 3 (n = 675) – connected mid-sized cities Min Max Mean Median Std. dev.
Nyimba, Petauke, Chongwe, Mkushi

City area (sq. km) 6.1 14.44 11.7 13.13 3.86

High density area (number of pixels) 8 45 28.5 30.5 15.29
Distance to main road (km) 0.27 13.64 5.7 4.43 5.82
Avg. distance to next three cities (km) 59.64 124.39 86.28 80.55 27.72
Estimated population from LandScan 2018 900 8170 4229.25 3923.5 3188.36
Food expenditures in the past week (USD) – all retail locations
Open-air market 0 115.38 5.51 3.85 5.96
Vendor 0 15.38 0.94 0 1.54
Local shop 0 46.15 2.49 1.15 4.29
Supermarket 0 15.38 0.07 0 0.88
Number of food purchases in the past two weeks
Open-air market 0 14 4.99 4 4.09
Vendor 0 14 2.78 1 4.05
Local shop 0 14 1.96 1 2.78
Supermarket 0 5 0.02 0 0.01

City Cluster Type 4 (n = 1417) – larger cities Min Max Mean Median Std. dev.
Karatina*, Nanyuki*, Choma*, Kapiri Mposhi*, Mazabuka*

City area (sq. km) 4.16 23.55 15.2 15.21 7.28

High density area (number of pixels) 29 137 89.2 91 41.51
Distance to main road (km) 0.89 16.11 8.33 4.43 5.82
Avg. distance to next three cities (km) 20.7 89.45 43.72 29.75 27.69
Estimated population from LandScan 2018 13,107 56,902 27,467.6 20,841 17,576
Food expenditures in the past week (USD) – all retail locations
Open-air market 0 98.04 6.13 3.92 7.47
Vendor 0 44.12 3.07 1.15 4.95
Local shop 0 68.63 3.29 1.76 5.35
Supermarket 0 98.04 4.92 0 10.92
Number of food purchases in the past two weeks
Open-air market 0 14 4.71 3 4.53
Vendor 0 14 6.19 4 5.97
Local shop 0 14 3.33 2 4.2
Supermarket 0 14 1.1 0 2.27

8
C. Hannah et al. Cities 124 (2022) 103608

Fig. 4. Food retail locations where household respondents spend the most money on food purchases (i.e., the largest portion of their food expenditure budget) by city
cluster type in Kenya and Zambia.

Table 2
Endogenous multinomial logit with open-air markets as the reference category and logit model comparing open-air markets to all other categories. Relative risk ratios
are presented with the significance values and standard errors are shown in parentheses. Holding all other variables constant, risk ratios greater than one indicate that
households are more likely to shop at a given food retail location relative to the reference location (i.e., all other locations in the logistic regression model and open-air
markets in the multinomial logit model). Conversely, risk ratios less than one suggest that households are less likely to purchase food at a given food retail location. Risk
ratio coefficients that are higher or lower than one indicate a stronger degree of risk or likelihood in either direction.
Dependent variable Kenya Zambia
Percent of sampled households
Open-air Vendor Local Supermarket Other Open-air Vendor Local Supermarket Other
spending the most money
market shop market shop

28 30 16 22 3 78 5 12 4 2

Logistic Multinomial logit Logistic Multinomial logit

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Open-air market purchases in 1.109*** 0.890*** 0.882*** 0.916*** 0.909* 1.194*** 0.530*** 0.894*** 0.740*** 0.858***
past two weeks (0.016) (0.021) (0.030) (0.021) (0.050) (0.017) (0.082) (0.019) (0.055) (0.041)
Vendor purchases in past two 0.980 1.155*** 0.922*** 0.975 0.950 1.005 1.202*** 0.884*** 0.921* 0.922
weeks (0.013) (0.020) (0.020) (0.016) (0.032) (0.016) (0.024) (0.027) (0.048) (0.050)
Local shop purchases in past two 0.951*** 0.938** 1.239*** 1.063*** 1.015 0.805*** 1.019 1.304*** 1.060 1.264***
weeks (0.018) (0.026) (0.025) (0.022) (0.049) (0.020) (0.050) (0.022) (0.059) (0.040)
Supermarket purchases in past 1.006 0.615*** 0.525*** 1.162*** 1.002 0.738*** 0.489** 1.100 2.120*** 1.197
two weeks (0.033) (0.092) (0.129) (0.040) (0.101) (0.045) (0.345) (0.084) (0.079) (0.147)
Constant 0.371*** 0.603** 0.805 0.936 0.247*** 2.576*** 0.177*** 0.178*** 0.087*** 0.041***
(0.169) (0.253) (0.245) (0.201) (0.381) (0.099) (0.209) (0.130) (0.228) (0.264)
Observations 970 1883
Log Likelihood − 554 − 879
Akaike Inf. Crit. 1118 2395 2395 2395 2395 1767 2473 2473 2473 2473
*
p < 0.1.
**
p < 0.05.
***
p < 0.01.

at formal retail locations. (people per room), tenure status (owning or renting), whether the
In the second set of regression models (Table 3), we included household was headed by a female, and monthly transportation
household-level variables that are likely determinants of where people expenditure, but our results remained unchanged. We also included
spend most of their money on food purchases (Caswell et al., 2013). The interactions with the household characteristics and city clusters, but our
variables are summarized in Appendix B. To test the model's robustness, results remained qualitatively unchanged. Hence, our final model in­
we included several household-level demographic characteristics, cludes the city clusters as binary control variables with the largest cities
namely: household size (number of members), household density (i.e., cluster four) as the reference city for which to interpret how

9
C. Hannah et al. Cities 124 (2022) 103608

Table 3
Multinomial and logistic regression models to assess the relationship of household characteristics and household urban food systems independent variables with the
dependent variable of which food retail location households spend the most money on food purchases. Relative risk ratios are presented with the significance values
and standard errors are shown in parentheses. Holding all other variables constant, risk ratios greater than one indicate that households are more likely to shop at a
given food retail location relative to the reference location (i.e., all other locations in the logistic regression model and open-air markets in the multinomial logit
model). Conversely, risk ratios less than one suggest that households are less likely to purchase food at a given food retail location. Higher or lower risk ratio co­
efficients from one indicate a stronger degree of risk or likelihood in either direction.
Dependent variable Kenya Zambia
Percent of sample spending the
Open-air Vendor Local Supermarket Other Open-air Vendor Local Supermarket Other
largest proportion of their budget at
market shop market shop
retail locations
28 30 16 22 3 78 5 12 4 2

Logistic Multinomial logit Logistic Multinomial logit

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Recipient of food from an urban 1.739*** 0.750 0.295*** 0.596** 0.415* 0.550*** 1.080 2.290*** 2.567*** 1.541
source in past year (1 = yes, 0 = (0.178) (0.209) (0.296) (0.239) (0.517) (0.120) (0.253) (0.157) (0.299) (0.329)
no)
Recipient of food from a rural source 0.883 1.423 0.635* 1.200 1.144 1.440** 1.119 0.678** 0.475* 0.638
in past year (1 = yes, 0 = no) (0.184) (0.217) (0.274) (0.237) (0.474) (0.148) (0.295) (0.190) (0.412) (0.431)
Engaged in urban agriculture (1 = 0.969 0.864 1.027 1.268 0.749 1.300** 1.273 0.638** 0.589* 0.921
yes, 0 = no) (0.187) (0.226) (0.263) (0.244) (0.543) (0.127) (0.251) (0.175) (0.294) (0.340)
Farm owner outside of city (1 = yes, 1.065 0.706 1.395 0.953 1.538 1.634*** 1.107 0.431*** 0.495* 1.215
0 = no) (0.184) (0.229) (0.267) (0.240) (0.464) (0.155) (0.309) (0.214) (0.392) (0.389)
Rural food sourcing in past year (no. 1.096*** 0.942 0.842*** 0.920* 0.876 0.876*** 0.905 1.258*** 0.888 1.123
of events) (0.035) (0.043) (0.064) (0.045) (0.106) (0.041) (0.128) (0.047) (0.154) (0.103)
Assets index (scale of 1–4) 1.259*** 0.709*** 0.709*** 1.134 0.689** 0.917** 0.725*** 1.200*** 1.475*** 0.879
(0.076) (0.086) (0.102) (0.109) (0.183) (0.042) (0.103) (0.054) (0.119) (0.119)
Log of income (USD) 0.925 0.868 0.994 1.516*** 1.207 0.729*** 0.719*** 1.423*** 2.975*** 1.275
(0.072) (0.088) (0.108) (0.102) (0.195) (0.057) (0.119) (0.076) (0.150) (0.161)
Other income recipient (1 = yes, 0 1.489** 0.722 0.642* 0.610** 0.781 0.784* 0.665 1.445** 2.209*** 0.934
= no) (0.178) (0.214) (0.268) (0.236) (0.469) (0.128) (0.302) (0.165) (0.297) (0.368)
City Cluster Type 1 (1 = yes, 0 = no) 0.930 1.114 2.264*** 0.636* 0.565 2.410*** 0.093** 0.854 0.430 0.217
(0.180) (0.210) (0.251) (0.251) (0.518) (0.326) (1.026) (0.433) (0.649) (1.032)
City Cluster Type 2 (1 = yes, 0 = no) 1.031 0.534* 2.438*** 0.069*** 0.596
(0.157) (0.339) (0.215) (0.639) (0.415)
City Cluster Type 3 (1 = yes, 0 = no) 1.147 0.789 2.110*** 0.047*** 0.346**
(0.138) (0.278) (0.194) (0.613) (0.426)
Constant 0.207*** 4.916*** 1.897 0.148*** 0.239* 10.999*** 0.240*** 0.018*** 0.001*** 0.024***
(0.305) (0.361) (0.435) (0.465) (0.806) (0.206) (0.369) (0.296) (0.661) (0.546)
Observations 810 1815
Log Likelihood − 463 − 912
Akaike Inf. Crit. 945 2207 2207 2207 2207 1847 2568 2568 2568 2568
*
p < 0.1.
**
p < 0.05.
***
p < 0.01.

households compare across the city clusters. Kenya. Instead, residents of these smaller cities were 2.64 times more
In both Kenya and Zambia, we found a positive association between likely to purchase food at local shops. Both results were statistically
household income and the likelihood of households spending most of significant at the 5% level. Relative to larger cities (City Cluster Type 4),
their money at supermarkets (Table 3). As the log of household income households in both mid-sized cities with lesser and greater connectivity
increases, households were significantly more likely to spend money at (City Cluster Types 2 and 3) were significantly more likely to purchase
supermarkets than any other food retail locations by 1.52 times in Kenya food at local shops.
and 2.98 times in Zambia. While households were significantly less
likely to purchase food at all other retail locations in Kenya based on
income, households with higher incomes in Zambia were significantly 4.4. Additional reasons underlying household food purchasing patterns
more likely to purchase food at local shops or supermarkets by 1.42
times, which further suggests that local shops and supermarkets in We posed additional questions in our survey to further elucidate why
Zambia are substitutes. These same food purchasing patterns held true households preferred to spend most of their food budget at a given retail
for household assets in Zambia, with a significant likelihood of food location. First, we asked a subset of questions aimed at understanding
purchases at local shops by 1.2 times and 1.48 times at supermarkets. the tradeoffs and price sensitivity of purchasing tomatoes across
Kenya differed in that with a one unit increase in the number of different food retail locations. We chose the example of tomatoes to
household assets, households were significantly more likely to purchase frame these questions because tomatoes are widely consumed in both
food at open-air markets by 1.26 times than any other location, which Zambia and Kenya, and given their perishability, they are also pur­
suggests that open-air markets were important across households chased frequently.
regardless of wealth in Kenya. We found that most households in our sample would choose to buy
Differences between the Kenyan and Zambian contexts arise their tomatoes at open-air markets and vendors, even if the prices were
regarding how households are embedded in their urban food systems by identical across these retailers, local shops, and supermarkets (Table 4).
city type. Households in the smallest cities (City Cluster Type 1) relative Freshness was cited as a main reason for continuing to purchase to­
to larger cities (City Cluster Type 4) were 2.4 times more likely to pur­ matoes at open-air markets in Kenya, while respondents in Zambia
chase food at open-air markets in Zambia, but this was not the case in preferred open-air markets for their negotiable prices and proximity.
Most respondents in Zambia associated the freshness of tomatoes with

10
C. Hannah et al. Cities 124 (2022) 103608

Table 4 Table 5
Percent of household respondents indicated their preferred location and reason Price elasticity of demand for purchasing food at local shops and supermarkets
for purchasing four medium tomatoes among open-air markets, vendors, local relative to open-air markets in Kenya and Zambia.
shops, and supermarkets, given the assumption that the price of four medium Kenya Zambia
tomatoes is the same across all food retail locations. Respondents could select
more than one reason, and customer service was included in the Kenya survey Median price at open-air markets 20 Kenyan 4 Zambian
Shillings (Ksh) Kwacha (ZK)
based on the consumer context of Central Kenya.
Baseline price increase from median price 10 Ksh 1 Ksh
Kenya Open-air market Vendor Local shop Supermarket Price increase from median price 15 25 2 ZK 3 ZK
(n = 1010) (n = 610) (n = 294) (n = 112) (n = 29) Ksh Ksh
% Change in price from baseline price 75% 125% 50% 75%
Preferred location 60 29 8 3
Baseline demand increase for supermarkets 168 households 94 households
Clean 2 2 0 59
(17% of sample) (5% of sample)
Closest 8 74 79 14
Demand increase for supermarkets 172 174 75 51
Freshness 63 4 10 28
% Change in demand from baseline demand 2% 1% − 20% − 32%
Negotiable prices 26 10 8 0
Baseline demand increase for local shops 58 households 38 households
Customer Service 11 7 16 3
(6% of sample) (2% of sample)
Other reason 16 12 5 24
Demand increase for supermarkets 56 54 15 7
% Change in demand from baseline demand − 3% − 4% − 61% − 53%
Zambia Open-air market Vendor Local shop Supermarket Price elasticity of demand for supermarkets 0.03 0.01 − 0.40 − 0.43
(n = 2007) (n = 1159) (n = 478) (n = 112) (n = 258) over open-air markets
Price elasticity of demand for local shops − 0.05 − 0.03 − 1.21 − 0.71
Preferred location 58 24 6 13 over open-air markets
Clean 5 3 20 81
Closest 44 82 78 1
Freshness 18 6 5 67
This result may be driven by the fact that supermarkets are more
Negotiable prices 64 23 4 1
Other reason 22 3 1 4 prevalent in Kenya and these households might be less price sensitive
overall. In Zambia, on the other hand, respondents indicated a greater
sensitivity to price differences. Only a very small fraction of households
supermarkets, but there were no concrete reasons underlying these would be willing to pay more for tomatoes at a supermarket or local
country-level differences found in households' perceptions of tomato shop, and this amount dropped substantially as the hypothetical price
freshness. We expect that a combination of the local contexts of our difference increased. Thus, the willingness to pay higher prices to shop
cities and collective consumer behaviors drive these aggregated con­ at a supermarket or a local shop is limited in both countries, but prices
sumer preferences. Households who would choose to purchase identi­ appear to play a larger role in Zambia, where supermarkets are less
cally priced tomatoes at vendors and local shops rather than at open-air common. Given the intrinsic qualities of tomatoes (e.g., freshness,
markets or supermarkets explained that they would do so due to the negotiable prices, supporting local economies) that were associated with
close location of these retailers, while the main reason for choosing tomatoes from open-air markets relative to tomatoes from other food
supermarkets was an association with cleanliness. retail locations (Table 5), this result is not surprising.
Additional reasons for purchasing food at open-air markets and Finally, we introduced a set of two scenarios to explore how house­
vendors included cultural norms or having a family history of buying at holds would hypothetically spend money on food if they experienced a
a certain location, a location's convenience, and having a wide variety of change in their weekly food expenditure budget. Given the purported
food options. Respondents in Kenya and Zambia also valued instances growth of supermarkets and westernized diets in some locations in SSA
when vendors gave extra tomatoes, the ability to buy larger quantities, (Reardon et al., 2021), we investigated what types of food households in
the enjoyment or “adventure” of buying from vendors, and supporting our sample of secondary cities would prefer to buy with a higher food
the livelihoods of local vendors and farmers. A final key benefit under­ expenditure budget. In the first scenario, we asked respondents what
lying open-air markets and vendors were opportunities to purchase food food products they would buy more of if they had 400 Ksh or 50 ZK more
with credit based on good relationships and trust with vendors. to use towards food purchases each week. In the second scenario, we
Next, we conducted a simple experiment to assess the respondent's asked what food products they would buy less of if they had 400 Ksh or
willingness to pay for four medium tomatoes at a supermarket or a local 50 ZK less to use towards food purchases each week.
shop. We asked respondents whether, given the actual prices of to­ Counter to evidence that African diets are shifting towards processed
matoes at open-air markets and vendors, they would be willing to buy foods sold at supermarkets (Reardon et al., 2021), households in our
four medium tomatoes at a local shop or a supermarket if prices were study showed a predisposition towards non-processed, whole foods.
higher at these retail locations by 75 and 125% in Kenya, and 50 and Given a larger food spending budget, households tended to prioritize
75% in Zambia. To measure the responsiveness of demand for pur­ healthier, non-processed, and whole foods associated with commodities
chasing food at local shops and supermarkets over open-air markets, we sold at open-air markets over processed foods that are associated with
calculated the price elasticity of demand for purchasing tomatoes at a more western diets. With a hypothetical increase in food expenditure
supermarket or local shop. Our base prices were obtained from the amount, respondents in both Kenya and Zambia stated they would spend
household survey, where we asked respondents what price they paid for more money on maize, meat, oil, and beans than processed foods and
four medium tomatoes the last time they purchased tomatoes. We drinks (Fig. 5). Conversely, a decreased food expenditure budget per
calculated the price elasticities using the median price of four medium week led to respondents indicating they would spend less money on
tomatoes in each country, which was 20 Ksh (0.20 USD) in Kenya and 4 processed foods and drinks compared to all other food commodities. If
ZK (0.31 USD) in Zambia during the 2019 field work survey. given more money, a greater proportion of households in Zambia (i.e.,
We did not observe a systematic difference in the prices of four 67%) would spend more of their food expenditure on meat relative to
medium tomatoes purchased at open-air markets compared to local any other food commodity, while a smaller proportion of households in
shops and supermarkets in our study sites. Overall, the demand for su­ Kenya (i.e., 40%) would spend more on meat. These differences suggest
permarkets and local shops was limited in both countries, where only a that meat may have a higher income elasticity of demand, where meat
small portion of each country's sample size would purchase tomatoes at purchases are more income-constrained, in Zambia compared to Kenya.
supermarkets and retail shops over open-air markets (Table 5). In Kenya, In Kenya, households cited that with more money, they would also
the price demand is relatively inelastic, suggesting that those who are purchase more vegetables, fruits, ground nuts, onions, potatoes, and
willing to buy tomatoes at supermarkets or stores are not price sensitive. cereals, such as rice and flour. Only six respondents (less than 1% of our

11
C. Hannah et al. Cities 124 (2022) 103608

Fig. 5. Types of food products that households would buy more or less of with a hypothetical increase or decrease in food expenditure amounts per week.

sample) in Kenya said they would buy more sugar. In Zambia, re­ cities and two large cities in India would continue to shop at traditional
spondents noted the same preference to spend more money on vegeta­ retail locations in the short run, but with the caveat that traditional
bles, fruits, ground nuts, onions, potatoes, and cereals, but also retailers (i.e., small local shops) would need to update their facilities to
referenced tomatoes, sweet potatoes, Irish potatoes, cabbage, okra, avoid being supplanted by cleaner and more organized modern retailers.
chicken, fish, and additional meat products, such as goat, pork, sausage, Yet, the scale at which supermarkets are introduced to SSA's secondary
and kapenta (dried fish). In terms of processed foods, less than 1% of cities would need to be significantly pronounced to overturn the
respondents in Zambia would spend more money on bread and sugar. important role that open-air markets and vendors play in the food pur­
Several households in both Kenya and Zambia also noted that they chasing behaviors of low- to middle-income households in our study.
would rather save the money than spend it on food purchases. The outcomes of our study deviate from earlier research on urban
food systems in developing countries that have associated dietary shifts
5. Discussion with the increasing dominance of modern food retailers in large cities
(Battersby, 2019; Battersby & Peyton, 2016; Demmler et al., 2017;
5.1. Persistence of open-air markets Otterbach et al., 2021; Reardon et al., 2021). We found that neither
household food purchases at supermarkets, nor the preference for pro­
Our findings across 18 cities in Kenya and Zambia provide insights cessed foods, were very widespread, even in our larger secondary cities
into how households in SSA's secondary cities navigate urban food (i.e., Karatina, Kenya and Choma, Zambia). This result leads us to
systems. Namely, households in our sample spent most of their food question the dominant storyline that modern retail presence is associ­
budget at open-air markets and vendors, which reflects food purchasing ated with increased consumption of western food products. Informing
trends elsewhere in SSA (White et al., 2018; Crush & Frayne, 2010; Smit, this narrative, dietary research in Africa has narrowly focused on the
2018). We expect that low- to middle-income households in the sec­ relationship between supermarkets and dietary behaviors, even though
ondary cities of our study sites will continue to shop at open-air markets the informal retail sector plays the most important role in Africa's urban
in the foreseeable future, because our results show that households food retail environment (Holdsworth & Landais, 2019). For example,
intrinsically value these markets in terms of the relational experiences Demmler et al. (2017)'s study in three cities in central Kenya demon­
that they provide. strates that food purchases at supermarkets contribute to higher body
Open-air markets in our study sites also meet the dietary preferences mass index with an increased probability of being overweight or obese.
of low- to middle-income urban households and offer product quality In Lusaka, Zambia, Khonje et al. (2020) found that the proportion of
and availability that is not always matched at supermarkets and local food expenditure shares spent at modern retailers such as supermarkets
shops, especially for fruits, vegetables, and whole foods. These food was associated with being overweight among adults.
commodities appear to be a higher priority for our sampled households Yet, our study moves beyond linking food purchases to biometric
than processed foods, such as bread and sugar, which suggests an indicators. We assess the consumer demand for shopping for vegetables
adherence to a more traditional African diet compared to a westernized at supermarkets versus open-air markets and the underlying preferences
diet. Similarly, the convenience, accessibility, and competitive pricing of of where consumers prefer to shop across multiple retail options. At the
open-air markets and vendors suggests that households are also likely to same time, however, our respondents may have indicated that they
continue shopping at both open-air markets and vendors in the sec­ prefer healthy foods since that perspective may be perceived to be more
ondary cities in our study. Our findings are reflected in Goswami and socially desirable. While we cannot confirm from our data whether our
Mishra (2009)'s study, which suggested that consumers in two small study's respondents do indeed purchase the healthier wholefoods that

12
C. Hannah et al. Cities 124 (2022) 103608

they have indicated a preference for, or if they had a higher food activities and processed food purchases begin to rise; and in stage three,
expenditure budget, our findings do suggest that the impact of incoming supply chains extend further to include international markets, where
supermarkets on urban diets is complex. For instance, a qualitative study supermarkets emerge and processed foods become common (Reardon
of food consumption in Malawi similarly demonstrated that dietary et al., 2019). Changes in consumer behavior and dietary preferences
transitions are complex and nuanced and are more largely based on the occur over these transformation stages, which broadly moves from
socio-cultural factors that shape urban food systems (Riley & Dodson, consumers purchasing less processed, unpackaged foods to more pro­
2016). Moreover, in a framework that identifies dietary behaviors cessed, packaged foods and from purchasing food at smaller retail stores
among adults and adolescents in a typical African urban food system, the to purchasing from supermarkets (Reardon et al., 2021). These processes
top individual-level factors that contribute to diet include food habits, of transformation have been observed in Asia and Latin America
preferences, and socioeconomic status (Osei-Kwasi et al., 2021). Thus, (Reardon et al., 2003; Reardon & Timmer, 2014), but to what extent
we may be observing a socio-cultural preference for healthier whole­ have these urban food systems transformations occurred in SSA's sec­
foods and a consumer habit for purchasing food at open-air markets that ondary cities?
is more suitable to the low- to middle-income households in our study. Insights from our study show that Zambia's secondary cities may be
While a rise in supermarkets has suggested a change in dietary be­ at an earlier stage in the urban food system transformation process
haviors and nutrition (Battersby, 2019; Battersby & Peyton, 2016; compared to Kenya. Households in Kenya purchased food across a more
Demmler et al., 2017), this pattern may hold true for some households in diverse range of food retail sources compared to Zambia, which reflects
some cities, but not in the secondary cities in our study, as our findings the greater diversity of retail locations, including supermarkets, in
show. Mackay (2019) similarly found that residents in two Ugandan Kenya. In contrast, Zambia has experienced less of a food system
secondary cities tended to purchase food from open-air markets and transformation and open-air markets continue to serve a prominent role
local shops, regardless of gender, age, and class. Thus, the storyline that for household food purchases. Contextualizing the urban food systems of
modern retail presence is associated with increased consumption of these smaller cities along the trajectory of urban food systems trans­
western food products does not necessarily hold true for all smaller formation can help urban planners to better understand where different
secondary cities in Africa. types of retailers are more likely to persist, and to implement urban
governance strategies in a way to support urban populations, including
5.2. Trajectories of urban food systems transformation in secondary cities low- to middle-income households.

The role of projected incoming supermarkets to secondary cities 5.3. Planning for resilient and equitable urban food systems
poses questions about the prospective trajectories of urban food system
transformations in these cities. Urban consumers in the secondary cities The continual and rapid growth of cities in SSA means that urban
in our study have indicated a preference for local foods and have even planners have a critical hand in shaping urban food systems that are
indicated that they also prefer to shop at open-air markets to support more resilient and equitable for populations living in secondary cities.
local farmers and businesses. Given these findings, we ask how much Achieving this will mean grappling with the relationship between the
influence large supermarket chains will have on household food pur­ spatial and political dimensions of retail expansion, which is often uti­
chases in the food systems of secondary cities in the coming decades, lized by municipal governments as a tool for local economic develop­
relative to open-air markets and vendors. Further research is needed to ment (Frimpong Boamah et al., 2020; Skinner, 2016). Yet, a desire
understand these urban food systems to develop urban planning policies among African policymakers to westernize cities in SSA has resulted in
most suitable to secondary cities. urban policies that favor modern retailers (i.e., supermarkets), but at the
Indeed, some secondary cities currently have one supermarket, cost of marginalizing traditional retail sectors (i.e., open-air markets,
which is likely adequate given the population sizes of these cities. Su­ street vendors, and local shops) (Giroux et al., 2020; Rousseau, Boyet, &
permarket retailers will also logically consider population densities and Harroud, 2020). For example, imposing national-level bans on street
incomes before investing in the development of new stores and would vending may adversely impact people living in smaller and less con­
therefore be unlikely to develop in smaller cities, especially those with nected urban areas, where supermarkets are currently not as prevalent
populations as low as 5000. Our results suggest that even with potential and where the informal sector is essential for sustaining food security
rapid supermarket expansion in these smaller SSA cities, we expect that and local livelihoods (Crush & Frayne, 2011; Kazembe et al., 2019).
low- to middle-income households are less likely to purchase food at Decisions made without appropriate food systems planning and
these supermarkets in the foreseeable future. engagement - such as market relocation or upgrading, and preferential
Of course, supermarkets and western food diets may eventually zoning for supermarkets - risk overlooking the urban food security and
infiltrate further into the urban food systems of some secondary cities. socio-economic justice repercussions of these decisions (Battersby,
Yet, open-air markets and vendors will hold their significance in food 2017a).
systems of secondary cities while supermarkets will play a more sub­ Our findings have highlighted the persistence of open-air markets in
stantial role in primate cities. However, the pace, scale and location of secondary cities, which emphasizes the need for urban municipal gov­
supermarket adoption requires further research to improve our under­ ernments to support these markets. One way in which open-air markets
standing of how secondary cities will address food insecurity in the should be integrated into city planning procedures is by ensuring that
context of their unique urban food systems in the future. At present, the they are allocated sufficient land near existing and planned low- and
necessary data available to investigate these processes remains scarce. middle-income areas. Urban planners also need to consider how rural
However, our study offers an initial empirical understanding of urban and urban farmers, vendors, and customers access markets to supply,
food system profiles in secondary cities based on a standardized survey sell, and purchase food. Ensuring that open-air markets are centrally
that could allow for future research on how these urban food systems located, within reasonable proximity to transportation networks, and
could evolve in the future. embedded in the ways in which urban populations are socially and
Theories of urban food systems transformation present a starting geographically organized is important for supporting the large consumer
point to explain how city types shape urban food retail environments in base that use these open-air markets in our secondary cities. Indeed, the
SSA and the differences in urban food system characteristics across city development and large donor investment of new modern markets can
types. In the first stage of transformation, food supply chains are short fail when they are not positioned to support these social and spatial
and food systems are comprised of small enterprises with low food processes in the cities. Investments in transportation and market infra­
processing; in the second stage, supply chains stretch from rural to urban structure are therefore equally essential for the sustainability of markets.
areas with small to medium-sized enterprises driving most food systems The diversity of food retail options that comprise the food systems of

13
C. Hannah et al. Cities 124 (2022) 103608

secondary cities in SSA suggests that the type of governance support from within designated markets (Giroux et al., 2020). We also found in
afforded to open-air markets should also extend to street vending and our field observations that some street vendors in the city of Choma
local shops. Our findings show that the range of retail sources, from avoid harassment by selling their goods very early in the morning, or
open-air markets to supermarkets, is similarly essential to food access in after 5 pm when the authorities have finished work for the day.
urban areas. Urban policies that are neglectful of, or hostile towards, the In this regard, policymakers should consider introducing regulations
informal food sector risk undermining its contribution to urban food or protocols that allow vendors to continue their activities in ways that
security. Rather than sanctioning unregulated traders, which has been do not violate health and safety standards, create congestion, or nega­
the case throughout Zambia and in parts of Kenya (e.g. Kisumu City) in tively impact the operations of other food retailers such as open-air
recent years, policymakers should recognize the important links that markets and supermarkets. These regulations can include allocating
vendors create between suppliers and consumers across informal and areas within the city or times of day that vendors can trade and investing
formal sectors (Giroux et al., 2020). For instance, based on our field in public facilities like toilets and water points. To ensure that these
observations, some vendors buy items in bulk from supermarkets (e.g., a regulations fit local conditions, legislation needs to be clear and
25 kg bag of maize meal or a dozen eggs), and then repackage them in consistent at national, county or district, and city-levels of enforcement
smaller quantities to resell to customers who can only afford to buy regarding the use of space for vending.
small amounts on a day-to-day basis. Thus, while supermarket expan­
sion can economically undermine and spatially marginalize unregulated 5.4. Policy considerations for urban food systems governance in
vendors, supermarket growth can also create opportunities for other secondary cities
retail actors in the food system (Battersby & Watson, 2018).
Although restrictions on vendors have predominantly been imple­ We posed additional questions in -air markets, vendors, local shops,
mented in primary cities, more attention will need to be paid to how and supermarkets will likely meet the needs and expectations of a
vendors might be integrated into the food systems of secondary cities as greater diversity of actors within the urban food systems of secondary
these cities continue to grow. One of the challenges for local govern­ cities. Implementing this “cohabitation” approach (Nickanor et al.,
ments in smaller cities will be the degree to which municipal decision 2019) will require policymakers to embrace the hybrid nature of urban
makers have the autonomy and capacity to manage and enforce re­ food systems in SSA, and to work closely with city-level planners, do­
strictions on vendors. National or even county- or district-level decisions nors, and private sector actors to ensure that city visions around food
on vending regulations may not fit the contexts of smaller secondary retail are shaped collaboratively, and with the urban poor in mind. Key
cities. Moreover, the state's reach is largely peripheral and city officials steps could include establishing city-level institutions, such as a
may be more inclined to make their own authoritative decisions on municipal department of food or food policy councils, which has been
vending regulations. done recently in Lusaka, and integrating food system concerns into the
For instance, the four Kenyan cities in our study each had their own work of existing urban planning departments (Pothukuchi & Kaufman,
unique rules according to local conditions that regulated vending busi­ 1999).
nesses. Vendors were restricted to what times of the day and in what For smaller secondary cities, coordinating with city officials will be
locations they could conduct business so as not to compete with super­ key, especially for cities that are not yet large enough to justify the
markets, local shops, and open-air markets. City officials also required establishment of a municipal department of food or food policy councils.
vendors to pay a small fee to occupy a public space, where vending For instance, at the time of fieldwork, key urban food systems gover­
outside of these parameters came with repercussions in the form of a nance decisions and oversight in our Kenyan cities were made at the
warning, fine, or arrest. These secondary city-level governance in­ county-level. At the same time, each city also had their own county-level
stitutions present an example of vending regulation that suits local representatives, including a market master that oversees the imple­
municipal contexts without requiring national oversight. mentation of county-level policies, suggesting some limitations to policy
However, in larger cities, such as in Kisumu, Kenya, the harmony autonomy that secondary cities wield in the governance of urban food
between street vendors and other retailers are challenged with contested systems.
uses of space, hostility, and conflict (Racaud, 2018). Similarly, in At present, however, policy decisions related to food systems
Lusaka, Zambia, government authorities have pressured street vendors governance globally are largely made at higher levels, despite decen­
to relocate to designated council-run markets, or else face eviction. tralization policies that purportedly afford power to local authorities for
Despite a strong resistance from street vendors, a cholera outbreak in city management and service delivery. Yet cities are mobilizing across
2017 resulted in the implementation of even harsher regulations, the globe to take greater responsibility to provide access to urban food
including fines of up to K5,000 (US$500) for both vendors and their governance decision making to citizens (Moragues-Faus, 2021). In this
customers (Mwango et al., 2019). These government crackdowns on process, urban governance and planning processes should engage
street vending have also been highly politicized, with vendors being directly with the dynamics of each city's unique urban food system,
used as voting fodder in election years (Resnick, 2019). Thus, in these which is inclusive of the social, spatial and economic dimensions of
larger cities, national legislation and oversight may be necessary and access to food from different sources (Moragues-Faus & Battersby,
more appropriate for regulating the roles of vendors in SSA's urban food 2021).
systems.
Yet, the balance of adhering to national laws while also promoting 6. Conclusion
municipal authority presents a host of complex, yet understudied nu­
ances in the implementation of urban food systems governance. For Our study directly addresses the challenges associated with the lack
example, we observed in the field that while the laws banning street of data availability and methodologies in articulating urban food sys­
vending in Zambia technically apply nationally, food system governance tems planning opportunities in secondary cities across the globe. We
in secondary cities is often limited by financial and personnel capacity, present the first multi-city analysis of SSA urban food systems that uses a
and the enforcement of regulations has therefore been less consistent in standardized survey with a large-N sample of households in diverse
certain contexts. For example, Giroux et al. (2020) found that in the secondary cities in Kenya and Zambia. We also address a knowledge gap
secondary city of Mumbwa, Zambia, street vendors are often warned regarding how household-level food acquisition varies across cities by
multiple times about noncompliance with the law before they face re­ examining the nature of urban food systems from the perspective of
percussions such as fines or the confiscation of goods. Officials are also household food acquisition across diverse types of secondary cities in
likely to confiscate only perishable goods that pose a food safety risk, Kenya and Zambia.
and vendors are regularly encouraged (rather than forced) to operate Our study further illustrates some underlying reasons for why low- to

14
C. Hannah et al. Cities 124 (2022) 103608

middle-income households across 18 cities in Kenya and Zambia pre­ References

dominantly purchase food at open air markets. We conducted quanti­
tative analyses that identified household- and city-level factors related Abrahams, C. (2009). Transforming the region: Supermarkets and the local food
economy. African Affairs, 109(434), 115–134.
to household food purchasing patterns. We also considered the quali­ Acharya, G., Cassou, E., Jaffee, S., & Ludher, E. K. (2021). RICH food, smart city: How
tative perceptions of household respondents that further explain building reliable, inclusive, competitive, and healthy food systems is smart policy for
household food purchasing behaviors. Based on these findings, we urban Asia. World Bank.
Andersson Djurfeldt, A. (2015). Multi-local livelihoods and food security in rural Africa:
expect that low- to middle-income households will continue to mainly Multi-local livelihoods and food security in rural Africa. Journal of International
purchase food from open-air markets, regardless of the purported rise in Development, 27(4), 528–545.
supermarkets across Africa. Implications of these findings for future Andreasen, M. H., Agergaard, J., Kiunsi, R. B., & Namangaya, A. H. (2017). Urban
transformations, migration and residential mobility patterns in African secondary
research point to the need for further systematic study of the dynamics cities. Geografisk Tidsskrift-Danish Journal of Geography, 117(2), 93–104.
of urban food systems planning, development, and governance in Asante, L. A. (2020). Urban governance in Ghana: The participation of traders in the
smaller secondary cities across the globe. A key policy and research redevelopment of Kotokuraba market in Cape Coast. African Geographical Review, 39
(4), 361–378.
question for the future remains. How can urban food systems planning
Asante, L. A., & Helbrecht, I. (2019). Urban governance and its implications for the
across diverse urban retailers proceed in a way that best supports the micro-geographies of market trading in Ghana: A case of the Kotokuraba market
diversity of cities and demographic groups within cities to access and project in Cape Coast. GeoJournal, (May).
afford sufficient, safe, and nutritious food to meet basic dietary needs Balsevich, F., Berdegué, J. A., Flores, L., Mainville, D., & Reardon, T. (2003).
Supermarkets and produce quality and safety standards in Latin America. American
and preferences? Journal of Agricultural Economics, 85(5), 1147–1154.
Moving forward, national policy approaches to the persistent role of Battersby, J. (2017a). Food system transformation in the absence of food system
open-air markets in SSA's secondary cities should account for the nuance planning: The case of supermarket and shopping mall retail expansion in Cape Town,
South Africa. Built Environment, 43(3), 417–430.
and heterogeneity associated with different city types. In doing so, local Battersby, J. (2017b). MDGs to SDGs – New goals, same gaps: The continued absence of
city officials and departments should be trusted and supported to urban food security in the post-2015 global development agenda. African
implement plans that are most appropriate and beneficial to their own Geographical Review, 36(1), 115–129.
Battersby, J. (2019). The food desert as a concept and policy tool in African cities: An
urban food system contexts and socioeconomic populations. Municipal opportunity and a risk. Sustainability: Science Practice and Policy, 11(2), 458.
governments have a salient role to play in policy development and Battersby, J., & Peyton, S. (2016). The spatial logic of supermarket expansion and food
change, as they operate at the local level and therefore have a better access. In J. Crush, & J. Battersby (Eds.), Rapid urbanisation, urban food deserts and
food security in Africa (pp. 33–46). Cham: Springer International Publishing.
understanding of key grassroots issues, as well as greater opportunities Battersby, J., & Watson, V. (2018). Addressing food security in African cities. Nature
to engage directly with local stakeholders (Gore, 2018). Thus, proac­ Sustainability, 1(4), 153–155.
tively pursuing an integrated food system planning agenda that can Battersby, J., & Marshak, M. (2017). Mapping the invisible: The informal food economy of
Cape Town, South Africa. Southern African Migration Programme.
appropriately meet the food systems needs of these secondary cities will
Battersby, J., & Muwowo, F. (2018). Planning and governance of food systems in Kitwe,
require policymakers and planners to work together across governance Zambia: A case study of food retail space. In Urban food systems governance and
scales, from the local to the regional and national levels. poverty in African cities (pp. 128–140). Routledge.
Battersby, J., & Watson, V. (2019). The planned ‘city-region’in the new urban agenda: An
appropriate framing for urban food security? The Town Planning Review.
CRediT authorship contribution statement Béné, C. (2020). Resilience of local food systems and links to food security - A review of
some important concepts in the context of COVID-19 and other shocks. Food Security,
12(4), 1–18.
Corrie Hannah: Conceptualization, Methodology, Formal Analysis, Béné, C., Fanzo, J., Prager, S. D., Achicanoy, H. A., Mapes, B. R., Toro, P. A., &
Investigation, Data Curation, Writing – Original Draft, Review & Edit­ Cedrez, C. B. (2020). Global drivers of food system (un)sustainability: A multi-
ing, Visualization, Project Administration, Supervision. Julia Davies: country correlation analysis. PLoS One, 15(4), Article e0231071.
Berger, M., & van Helvoirt, B. (2018). Ensuring food secure cities – Retail modernization
Conceptualization, Methodology, Investigation, Writing – Original
and policy implications in Nairobi, Kenya. Food Policy, 79(August), 12–22.
Draft, Review & Editing. Rachel Green: Conceptualization, Methodol­ Bigsten, A., Kebede, B., Shimeles, A., & Taddesse, M. (2003). Growth and poverty
ogy, Investigation, Writing – Original Draft, Review & Editing. Andrew reduction in Ethiopia: Evidence from household panel surveys. World Development,
Zimmer: Visualization (map), Methodology, Investigation, Writing – 31(1), 87–106.
Blekking, J., Tuholske, C., & Evans, T. (2017). Adaptive governance and market
Original Draft, Review & Editing. Patrese Anderson: Validation, heterogeneity: An institutional analysis of an urban food system in sub-Saharan
Formal Analysis, Writing – Original Draft, Review & Editing. Jane Africa. Sustainability: Science Practice and Policy, 9(12), 2191.
Battersby: Writing – Original Draft, Review & Editing. Kathy Baylis: Blekking, J., Waldman, K., Tuholske, C., & Evans, T. (2020). Formal/informal
employment and urban food security in sub-Saharan Africa. Applied Geography, 114
Validation, Methodology, Funding Acquisition, Writing – Review & (January), Article 102131.
Editing. Nupur Joshi: Writing – Original Draft, Review & Editing. Tom Boyer, D., Sarkar, J., & Ramaswami, A. (2019). Diets, food miles, and environmental
Evans: Conceptualization, Methodology, Investigation, Writing – Re­ sustainability of urban food systems: Analysis of nine Indian cities. Earth’s Future, 7
(8), 911–922.
view & Editing, Funding Acquisition, Project Administration, Buettner, T. (2015). Urban estimates and projections at the United Nations: The
Supervision. strengths, weaknesses, and underpinnings of the world urbanization prospects.
Spatial Demography, 3(2), 91–108.
Caswell, J. A., & Yaktine, A. L. Committee on Examination of the Adequacy of Food
Declaration of competing interest Resources and SNAP Allotments, Food and Nutrition Board, Committee on National
Statistics, Institute of Medicine, National Research Council. (2013). Individual,
household, and environmental factors affecting food choices and access. US: National
None. Academies Press.
Cattaneo, A., Nelson, A., & McMenomy, T. (2021). Global mapping of urban–rural
catchment areas reveals unequal access to services. Proceedings of the National
Acknowledgements Academy of Sciences of the United States of America, 118(2).
Chai, B., & Seto, K. C. (2019). Conceptualizing and characterizing micro-urbanization: A
This work was supported by the United States National Science new perspective applied to Africa. Landscape and Urban Planning, 190(October),
Article 103595.
Foundation [grant number DEB-1924309, SES-1832393 and SES- CIESIN, & ITOS. (2013). Global roads open access data set, version 1 (gROADSv1).
1360463]. We thank the editor and anonymous reviewers for their Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC).
comments and suggestions for improving the manuscript. Claasen, N., & Lemke, S. (2019). Strong ties, weak actors? Social networks and food
security among farm workers in South Africa. Food Security, 11(2), 417–430.
Crush, J., & Caesar, M. (2017). Food remittances: Rural-urban linkages and food security
Appendix A. Supplementary data in Africa. [Link]
Crush, J., & Caesar, M. (2018). Food remittances and food security: A review. Migration
and Development, 7(2), 180–200.
Supplementary data to this article can be found online at [Link] Crush, J., & Ceasar, M. (2020). Food remittances and food security. In Handbook on urban
org/10.1016/[Link].2022.103608. food security in the global south. Edward Elgar Publishing.

15
C. Hannah et al. Cities 124 (2022) 103608

Crush, J., Frayne, B., & Hayson, G. (2020). Introduction to urban food security in the Mackay, H. (2019). Food sources and access strategies in Ugandan secondary cities: An
global south. In Handbook on urban food security in the global south. Edward Elgar intersectional analysis. Environment and Urbanization, 31(2), 375–396.
Publishing. Maxwell, D. (1999). The political economy of urban food security in sub-Saharan Africa.
Crush, J., Frayne, B., & Pendleton, W. (2012). The crisis of food insecurity in African World Development, 27(11), 1939–1953.
cities. Journal of Hunger & Environmental Nutrition, 7(2–3), 271–292. Mbataru, P. (2017). Agribusiness in the construction, deconstruction and reconstruction
Crush, J., & Frayne, B. (2010). The invisible crisis: Urban food security in southern of a mountain urban centre: The case of Karatina Town, Kenya. Rural-Urban
Africa. [Link] Dynamics in the East African Mountains, 275.
Crush, J., & Frayne, B. (2011). Supermarket expansion and the informal food economy in Melesse, M. B., van den Berg, M., Béné, C., de Brauw, A., & Brouwer, I. D. (2020). Metrics
southern African cities: Implications for urban food security. Journal of Southern to analyze and improve diets through food systems in low and middle income
African Studies, 37(4), 781–807. countries. Food Security, 12(5), 1085–1105.
Crush, J., & Riley, L. (2018). Rural bias and urban food security. In , 42. Urban food Milan Urban Food Policy Pact. (2015). Milan urban food policy pact. [Link]
systems governance and poverty in African cities. [Link]/the-milan-pact/.
das Nair, R. (2020). The ‘supermarket revolution’ in the south. In Handbook on urban food Minten, B., & Reardon, T. (2008). Food prices, quality, and quality’s pricing in
security in the global south. Edward Elgar Publishing. supermarkets versus traditional markets in developing countries. Review of
Davies, J., Hannah, C., Guido, Z., Zimmer, A., McCann, L., Battersby, J., & Evans, T. Agricultural Economics, 30(3), 480–490.
(2020). Barriers to urban agriculture in sub-Saharan Africa. Food Policy, (November), Minten, B., Reardon, T., & Sutradhar, R. (2010). Food prices and modern retail: The case
101999. of Delhi. World Development, 38(12), 1775–1787.
De Zeeuw, H., Van Veenhuizen, R., & Dubbeling, M. (2011). The role of urban agriculture Moragues-Faus, A. (2021). The emergence of city food networks: Rescaling the impact of
in building resilient cities in developing countries. The Journal of Agricultural Science, urban food policies. Food Policy, (June), 102107.
149(S1), 153. Moragues-Faus, A., & Battersby, J. (2021). Urban food policies for a sustainable and just
Demeke, A. B., Keil, A., & Zeller, M. (2011). Using panel data to estimate the effect of future: Concepts and tools for a renewed agenda. Food Policy, (July), 102124.
rainfall shocks on smallholders food security and vulnerability in rural Ethiopia. Moseley, W. G., & Battersby, J. (2020). The vulnerability and resilience of African food
Climatic Change, 108(1–2), 185–206. systems, food security, and nutrition in the context of the COVID-19 pandemic.
Demmler, K. M., Klasen, S., Nzuma, J. M., & Qaim, M. (2017). Supermarket purchase African Studies Review, 63(3), 449–461.
contributes to nutrition-related non-communicable diseases in urban Kenya. PLoS Mwango, M., Kaliba, M., Chirwa, M., & Guarín, A. (2019). Informal food markets in
One, 12(9), Article e0185148. Zambia: Perspectives from vendors, consumers and policymakers in Lusaka and Kitwe.
Devereux, S., Béné, C., & Hoddinott, J. (2020). Conceptualising COVID-19’s impacts on Nickanor, N., Crush, J., & Kazembe, L. (2019). The informal food sector and cohabitation
household food security. Food Security, (July), 1–4. with supermarkets in Windhoek, Namibia. Urban Forum, 30(4), 425–442.
Ericksen, P. J. (2008). Conceptualizing food systems for global environmental change Nickanor, N., & Kazembe, L. (2017). The supermarket revolution and food security in
research. Global Environmental Change: Human and Policy Dimensions, 18(1), Namibia. Southern African Migration Programme.
234–245. Nkrumah, B. (2018). Edible backyards: Climate change and urban food (in)security in
FAO. (2009). Declaration of the world food summit on food security. Rome: Food and Africa. Agriculture & Food Security, 7(1), 45.
Agriculture Organisation. Opiyo, P. O., & Agong, S. G. (2020). Nexus between urban food system and other urban
Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed systems: Exploring opportunities for improving food security in Kisumu, Kenya.
methods designs-principles and practices. Health Services Research, 48(6 Pt 2), Economic Geography, 5(1), 20–28.
2134–2156. Osei-Kwasi, H. A., Laar, A., Zotor, F., Pradeilles, R., Aryeetey, R., Green, M., Griffiths, P.,
FEWSNET. (2011). Kenya livelihood zone descriptions. In Famine Early Warning Systems et al. (2021). The African urban food environment framework for creating healthy
Network. [Link] nutrition policy and interventions in urban Africa. PloS One, 16(4), Article
FEWSNET. (2014). “Zambia Livelihood Zone Descriptions.”. In Famine Early Warning e0249621.
Systems Network. [Link] Otterbach, S., Oskorouchi, H. R., Rogan, M., & Qaim, M. (2021). Using Google data to
july-2014. measure the role of big food and fast food in South Africa’s obesity epidemic. World
Frayne, B. (2010). Pathways of food: Mobility and food transfers in southern African Development, 140(105368), Article 105368.
cities. International Development Planning Review, 32(3–4), 291–310. Owusu-Sekyere, E., Amoah, S. T., & Teng-Zeng, F. (2016). Tug of war: Street trading and
Frayne, B., McCordic, C., & Shilomboleni, H. (2014). Growing out of poverty: Does urban city governance in Kumasi, Ghana. Development in Practice, 26(7), 906–919.
agriculture contribute to household food security in southern African cities? Urban Peyton, S., Moseley, W., & Battersby, J. (2015). Implications of supermarket expansion
Forum, 25(2), 177–189. on urban food security in Cape Town, South Africa. African Geographical Review, 34
Frayne, B., & Crush, J. (2017). Food supply and urban-rural links in southern African (1), 36–54.
cities. In Food and nutrition security in southern African cities (pp. 34–47). Pingali, P. (2007). Westernization of Asian diets and the transformation of food systems:
Frayne, B., Pendleton, W., Crush, J., Acquah, B., Battersby-Lennard, J., Bras, E., Implications for research and policy. Food Policy, 32(3), 281–298.
Chiweza, A., et al. (2010). The state of urban food insecurity in southern Africa. Pothukuchi, K., & Kaufman, J. L. (1999). Placing the food system on the urban agenda:
Frimpong Boamah, E., Amoako, C., & Asenso, B. K. (2020). Spaces of market politics: The role of municipal institutions in food systems planning. Agriculture and Human
Retailscapes and modernist planning imaginaries in African cities. Applied Values, 16(2), 213–224.
Geography, 123(October), Article 102265. Prain, G., & Lee-Smith, D. (2010). Urban agriculture in Africa: What has been learned? In
Giroux, S., Blekking, J., Waldman, K., Resnick, D., & Fobi, D. (2020). Informal vendors G. Prain, D. Lee-Smith, & N. Karanja (Eds.), African Urban Harvest: Agriculture in the
and food systems planning in an emerging African city. Food Policy, (November), Cities of Cameroon, Kenya and Uganda (pp. 13–35). New York, NY: Springer New
101997. York.
Gore, C. D. (2018). How African cities lead: Urban policy innovation and agriculture in Racaud, S. (2018). In , Feburary. Ambiguous resource: ‘Informal’ street trading in Kisumu,
Kampala and Nairobi. World Development, 108(August), 169–180. Kenya (pp. 17–18). Articulo.
Goswami, P., & Mishra, M. S. (2009). Would Indian consumers move from Kirana stores Reardon, T., Echeverria, R., Berdegué, J., Minten, B., Liverpool-Tasie, S., Tschirley, D., &
to organized retailers when shopping for groceries? Asia Pacific Journal of Marketing Zilberman, D. (2019). Rapid transformation of food systems in developing regions:
and Logistics, 21(1), 127–143. Highlighting the role of agricultural research & innovations. Agricultural Systems,
Haklay, M., & Weber, P. (2008). OpenStreetMap: User-generated street maps. IEEE 172(June), 47–59.
Pervasive Computing/IEEE Computer Society [and] IEEE Communications Society, 7(4), Reardon, T., & Timmer, C. P. (2014). Five inter-linked transformations in the Asian
12–18. agrifood economy: Food security implications. Global Food Security, 3(2), 108–117.
Haysom, G., & Tawodzera, G. (2018). ‘Measurement drives diagnosis and response’: Gaps Reardon, T., Timmer, C. P., Barrett, C. B., & Berdegué, J. (2003). The rise of
in transferring food security assessment to the urban scale. Food Policy, 74(January), supermarkets in Africa, Asia, and Latin America. American Journal of Agricultural
117–125. Economics, 85(5), 1140–1146.
Haysom, G., & Fuseini, I. (2019). Governing food systems in secondary cities in Africa: An Reardon, T., Timmer, C. P., & Berdegué, J. A. (2005). Supermarket expansion in Latin
enquiry into Africa’s urban transition in secondary cities – Through a Food Lens. America and Asia. New Directions in Global Food Markets, 47–61.
[Link]. Reardon, T., Tschirley, D., Lenis, S. O., Liverpool-Tasie, T. A., Fanzo, J., Minten, B.,
Holdsworth, M., & Landais, E. (2019). Urban food environments in Africa: Implications Vos, R., et al. (2021). The processed food revolution in african food systems and the
for policy and research. The Proceedings of the Nutrition Society, 78(4), 513–525. double burden of malnutrition. Global Food Security, 28(March), Article 100466.
Kazembe, L. N., Nickanor, N., & Crush, J. (2019). Informalized containment: Food Resnick, D. (2019). The politics of crackdowns on Africa’s informal vendors. Comparative
markets and the governance of the informal food sector in Windhoek, Namibia. Politics, 52(1), 21–51.
Environment and Urbanization, 31(2), 461–480. Resnick, D., Sivasubramanian, B., Idiong, I. C., Ojo, M. A., & Tanko, L. (2019). The
Khonje, M. G., Ecker, O., & Qaim, M. (2020). Effects of modern food retailers on adult enabling environment for informal food traders in Nigeria’s secondary cities. Urban
and child diets and nutrition. Nutrients, 12(6). Forum, 30(4), 385–405.
Kraemer, K., Cordaro, J. B., Fanzo, J., & Gibney, M. (2016). In Chapter 1.1 Ten forces Riley, L., & Dodson, B. (2016). Intersectional identities: Food, space and gender in urban
shaping the global food system (pp. 19–30). [Link]. Malawi. Agenda, 30(4), 53–61.
Lee-Smith, D. (2010). Cities feeding people: An update on urban agriculture in equatorial Rousseau, M., Boyet, A., & Harroud, T. (2020). Politicizing African urban food systems:
Africa. Environment and Urbanization, 22(2), 483–499. The contradiction of food governance in Rabat and Casablanca, Morocco. Cities, 97,
Legwegoh, A. F., & Riley, L. (2014). Food, place, and culture in urban Africa: 102528.
Comparative consumption in Gaborone and Blantyre. Journal of Hunger & Satterthwaite, D. (2017). The impact of urban development on risk in Sub-Saharan
Environmental Nutrition., 9(2), 256–279. Africa’s cities with a focus on small and intermediate urban centres. International
Mabrouk, F., & Mekni, M. M. (2018). Remittances and food security in African countries. Journal of Disaster Risk Reduction., 26, 16–23.
African Development Review, 30(3), 252–263.

16
C. Hannah et al. Cities 124 (2022) 103608

Satterthwaite, D., McGranahan, G., & Tacoli, C. (2010). Urbanization and its implications Toriro, P. (2018). 11 urban food production in Harare, Zimbabwe. In , 154. Urban food
for food and farming. Philosophical Transactions of the Royal Society of London. Series systems governance and poverty in African cities.
B, Biological Sciences, 365(1554), 2809–2820. Tuholske, C., Andam, K., Blekking, J., Evans, T., & Caylor, K. (2020). Comparing
Schiavina, M., Freire, S., & MacManus, K. (2019). GHS population grid multitemporal measures of urban food security in Accra, Ghana. Food Security, 12(2), 417–431.
(1975, 1990, 2000, 2015) R2019A. Eur. Comm. JRC. Tuholske, C., Caylor, K., Evans, T., & Avery, R. (2019). Variability in urban population
Seto, K. C., & Ramankutty, N. (2016). Hidden linkages between urbanization and food distributions across Africa. Environmental Research Letters, 14(8), 085009.
systems. Science, 352(6288), 943–945. UN-DESA. (2014). World urbanization prospects: The 2014 revision, highlights.
Skinner, C. (2016). Informal food retail in Africa: A review of evidence. In Consuming urban Department of Economic and Social Affairs. In , 32. Population newsletter/issued by the
poverty project working paper, no. 2. Population Division of the Department of Economic and Social Affairs, United Nations.
Smart, J., Nel, E., & Binns, T. (2015). Economic crisis and food security in Africa: United Nations, Department of Economic and Population Division Social Affairs. (2018).
Exploring the significance of urban agriculture in Zambia’s Copperbelt Province. World urbanization prospects: The 2018 revision (Online ed.). New York, NY:
Geoforum; Journal of Physical, Human, and Regional Geosciences, 65(October), 37–45. Department of Economic and Social Affairs PD.
Smit, W. (2016). Urban governance and urban food systems in Africa: Examining the Vernon Henderson, J., & Kriticos, S. (2018). The development of the African system of
linkages. Cities, 58(October), 80–86. cities. Annual Review of Economics, (August).
Smit, W. (2018). Current Urban Food Governance and Planning in Africa. Urban Food Weatherspoon, D. D., & Reardon, T. (2003). The rise of supermarkets in Africa:
Systems Governance and Poverty in African Cities (p. 94). Routledge Taylor & Francis Implications for agrifood systems and the rural poor: The rise of supermarkets in
Group. Africa. Development Policy Review: The Journal of the Overseas Development Institute, 21
Sonnino, R., Tegoni, C. L. S., & De Cunto, A. (2019). The challenge of systemic food (3), 333–355.
change: Insights from cities. Cities, 85(February), 110–116. White, S., Kampanje-Phiri, J., Phiri, C., Hamm, M., & Richter, K. (2018). Regional supply
Tacoli, C. (2019). The urbanization of food insecurity and malnutrition. Environment and chains and the food economy of Malawi. Expanding livelihood opportunities and
Urbanization, 31(2), 371–374. enhancing food security. GCFSI Research Brief.
Tawodzera, G., & Crush, J. (2016). Migration and food security: Zimbabwean migrants in World Bank. (2020). World bank open data. [Link]
urban South Africa. [Link] Zezza, A., & Tasciotti, L. (2010). Urban agriculture, poverty, and food security: Empirical
[Link]. evidence from a sample of developing countries. Food Policy, 35(4), 265–273.
Tefft, J., Jonasova, M., Adjao, R., & Morgan, A. (2017). Food systems for an urbanizing Zhong, T., Si, Z., Crush, J., Scott, S., & Huang, X. (2019). Achieving urban food security
world. World Bank, Washington, DC and Food and Agriculture Organization of the through a hybrid public-private food provisioning system: The case of Nanjing,
United Nations. China. Food Security, 11(5), 1071–1086.
Thornbush, M. (2015). Urban agriculture in the transition to low carbon cities through Zimmer, A., Guido, Z., Tuholske, C., Pakalniskis, A., Lopus, S., Caylor, K., & Evans, T.
urban greening. AIMS Environmental Science, 2(3), 852–867. (2020). Dynamics of population growth in secondary cities across southern Africa.
Thornton, A. (2020). Urban food democracy and governance in north and south. Cham: Landscape Ecology, 1–16.
Palgrave Macmillan.

17