Introduction
Access to food is influenced by the characteristics of the food environment, such as the quantity, type, and location of food outlets (Health Canadá, 2013). Studies show that the distribution of these retailers is uneven and varies according to the socioeconomic and demographic aspects of neighbourhoods, which can influence the eating patterns of residents and produce or aggravate social health inequalities (Black et al., Reference Black, Moon and Baird2014; Pessoa et al., Reference Pessoa, Mendes, Gomes, Martins and Velasquez-Melendez2015; Matozinhos et al., Reference Matozinhos, Gomes, Andrade, Mendes, Pessoa, Friche and Velasquez-Melendez2015; Llavero-Valero et al., Reference Llavero-Valero, Escalada-San Martín, Martínez-González, Basterra-Gortari, de la Fuente-Arrillaga and Bes-Rastrollo2021; Almeida et al., Reference Almeida, Garcez, Cunha, Schuch and Canuto2023). According to Espinoza’s model (Reference Espinoza, Egaña, Masferrer and Cerda2017), food environments can be considered an intermediate social determinant of health.
Specific metrics for food deserts and swamps have been proposed to better understand access to healthy food and its relationship with neighbourhood characteristics (Cummis and Mancintyre, Reference Cummins and Macintyre2002; Rose et al., Reference Rose, Bodor, Swalm, Rice, Farley and Hutchinson2009). Food deserts are socially vulnerable areas with limited or no access to healthy foods (Cummis and Mancintyre, Reference Cummins and Macintyre2002), while food swamps are areas where ultra-processed food sales predominate (Rose et al., Reference Rose, Bodor, Swalm, Rice, Farley and Hutchinson2009). In Brazil, food deserts are defined as areas with limited or no access to healthy foods independently of social vulnerability (CAISAN, 2018).
Food deserts have been more investigated in high-income countries, such as the USA, where they have been widely identified in poor and minority neighbourhoods through lower access to supermarkets, which are considered healthy food retailers (Walker et al., Reference Walker, Keane and Burke2010; Black et al., Reference Black, Moon and Baird2014). In the USA, purchasing healthy foods is also more difficult for populations living in food deserts due to higher prices (Powell et al., Reference Powell, Slater, Mirtcheva, Bao and Chaloupka2007). Food swamps are more common in Canada than in the USA and are considered better predictors of obesity and other chronic diseases than food deserts (Rose et al., Reference Rose, Bodor, Swalm, Rice, Farley and Hutchinson2009; Cooksey-Stowers et al., Reference Cooksey-Stowers, Schwartz and Brownell2017).
Although food deserts and swamps have been studied in some high-income countries, these food environments may not reflect the characteristics of low- and middle-income countries, such as Brazil (Turner et al., Reference Turner, Aggarwal, Paredes, Herforth, Nowski, Coates, Kalamatianou and Kadiyala2018). In Brazil, the Interministerial Chamber of Food and Nutrition Security (CAISAN) pioneered research into food deserts mapping the presence of food deserts in Brazilian state capitals. More recently, Honorório et al. conducted a study comparing the socioeconomic conditions of food deserts and swamps in Belo Horizonte, Brazil (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021).
Although this theme is advancing in Brazilian research, the northeast region of the country has been little investigated regarding social inequalities in food deserts and swamps. Therefore, the present study aimed to identify food deserts and swamps and their association with the socioeconomic and demographic conditions of census tracts of a metropolis in northeastern Brazil. We hypothesize that the presence of food deserts is associated with worse socioeconomic and demographic conditions in the area. Conversely, the presence of swamps is associated with better socioeconomic and demographic conditions.
Methods
Study design and unit of analysis
This ecological observational study was conducted in the city of Recife, the capital of the state of Pernambuco, Brazil, whose estimated 2021 population was 1,661,017 (IBGE, 2021), with a population density of 7,039.64 inhabitants per square kilometre (IBGE, 2010). According to social indicators for 2020 (IBGE, 2020), the city’s Gini coefficient is 0.612, making it the state capital with the greatest intra-urban social inequality in Brazil. Recife’s census tracts were considered the unit of analysis. Of its 1845 census tracts, 17 (0.9%) were excluded due to incomplete data. A census tract is a geographic region defined to take a census. Sometimes it coincides with the limits of cities, towns, or other administrative areas. The Brazilian Institute of Geography and Statistics uses the term census sector. As of the 2010 Census, there were approximately 314,000 sectors in Brazil (IBGE, 2010).
Food retail information
The following information was collected about food retail to survey the food environment in Recife: name, address, and code from the National Classification of Economic Activities (CNAE), a system for standardizing economic activity that is used by Brazilian tax agencies to identify economic activities (IBGE, 2016). A previous study carried out in Recife described the types of foods most frequently sold by each food retailer (Menezes et al., Reference Menezes, Oliveira, Almendra, Lira, Costa, Leal and Santana2022; Oliveira et al., Reference Oliveira, Menezes, Almendra, Lira, Aquino, Souza and Santana2022). We utilized this information to categorize food retailers as follows: (I) food retailers that predominantly sell natural or minimally processed foods (50% or more); (II) mixed food retailers (those selling natural or minimally processed and ultra-processed foods); (III) food retailers selling predominantly ultra-processed foods (50% or more); and (IV) supermarkets and hypermarkets (Supplementary Table 1). Information from all 20.199 registered food retailers in Recife was analysed.
Following previous Brazilian studies (Leite et al., Reference Leite, de Assis, do Carmo, Nogueira, Netto and Mendes2021; Menezes et al., Reference Menezes, Oliveira, Almendra, Lira, Costa, Leal and Santana2022; Oliveira et al., Reference Oliveira, Menezes, Almendra, Lira, Aquino, Souza and Santana2022; Borges et al., Reference Borges, Vargas, Honório, Mendes and Canuto2023), supermarkets and hypermarkets were categorized separately due to the wide variety of foods they sell and the lack of consensus in the literature about their real influence on the purchasing profile of the population (Larson et al. Reference Larson, Story and Nelson2009; Machado et al., Reference Machado, Claro, Canella, Sarti and Levy2017).
Food deserts and swamps
Food deserts were identified using the methodology of the Technical Mapping Study of Food Deserts in Brazil (CAISAN, 2018), in which census tracts below the 25th percentile in the density of healthy food retail (i.e., those selling predominantly fresh or minimally processed foods, mixed businesses, supermarkets and hypermarkets) per 10,000 inhabitants are considered food deserts. Following Honório et al. (Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021), food swamps were classified according to sales of a list of ultra-processed foods proposed by CAISAN (2018). This method considers food swamps to be census tracts above the 25th percentile in the density of unhealthy food retail (i.e., those selling mainly ultra-processed foods) per 10,000 inhabitants (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021).
The following data were collected on food retail: address, National Classification of Economic Activities registration, and geographic coordinates (latitude and longitude) (Find Latitude and Longitude, 2022). The businesses were then georeferenced through the Geographic Information System and QGIS 2.14.9 software.
Socioeconomic and demographic variables
The following variables were extracted from the 2010 Census database for each tract: total and per capita family income, total number of households, and total population. Data for each household included the literacy of the head of household, the residents’ race, and household characteristics (IBGE, 2010). The per capita family income of the census tract was calculated using the ratio between the total income and number of residents in the tract, categorized into quartiles: 1st quartile: BRL 78.20–264.88; 2nd quartile: BRL 264.89–417.91; 3rd quartile: BRL 417.92–939.96; 4th quartile: > BRL 939.96, based on 2010 exchange rates.
Literacy was categorized as literate (those who can read and write a simple note) or illiterate (those who can only write their own name) (IBGE, 2010); race was categorized as White, Black, Indigenous, or mixed race according to self-report (IBGE, 2010); and households were evaluated regarding basic services: water supply, garbage collection, and sewage system (IBGE, 2010).
The Health Vulnerability Index (HVI) was also used to assess the degree of vulnerability in census tracts. The HVI is used as a proxy for socioeconomic vulnerability (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021; Leite et al., Reference Leite, de Assis, do Carmo, Nogueira, Netto and Mendes2021). This is a composite indicator that combines socioeconomic dimension (residents per household, percentage of illiterate people, percentage of private households with a per capita income of up to half a minimum wage, average nominal income of those responsible, percentage of people of brown, black and indigenous) and environmental dimension (sanitary sewage, water supply, and disposal of solid waste). All indicators were standardized to values between zero and one to facilitate comparison and aggregation of indicators measured on different scales.
The final composite indicator assigns distinct weights to each variable and dimension, determined through discussions involving 16 experts in the field. For the socioeconomic dimension, the weights are as follows: residents per household (0.073), percentage of illiterate people (0.289), percentage of private households with a per capita income below half a minimum wage (0.288), average nominal income of household heads (0.173), and percentage of people identifying as brown, black, and indigenous (0.185). Similarly, for the environmental dimension, the weights are sanitary sewage (0.424), water supply (0.375), and disposal of solid waste (0.201). The composite indicator also incorporates a weighted sum of both dimensions, with a weight of 0.604 assigned to socioeconomic variables and 0.396 to environmental variables.
The resulting HVI ranges from 0 to 1, where values closer to 1 indicate high social vulnerability and values closer to zero indicate low or negligible social vulnerability. Census tracts are subsequently categorized based on their HVI scores into medium risk (mean ± 0.5 SD), low risk (<−0.5 SD), high risk (+0.5 to +1.5 SD), and very high risk (>+1.5 SD). Notably, the high-risk and very high-risk categories were combined due to the limited number of high-risk tracts.
Data analysis
The absolute and relative frequency of food deserts and swamps were described according to census tract HVI, and the association was tested using Pearson’s chi-square test. Data normality was determined using the Kolmogorov-Smirnov test. Using Student’s t-test, the association between tracts considered food deserts/food swamps and their socioeconomic conditions was determined. The significance level was set at p < 0.05. All analyses were performed in QGIS 2.14.9 and SPSS version 20.
Results
Of the total of 20,199 identified food retail, most predominantly sold ultra-processed foods (77.3%, n = 15,607), followed by mixed items (14.6%, n = 2952), natural or minimally processed foods (7.4%, n = 1488), and supermarkets and hypermarkets (0.75%, n = 152). Regarding the classification of census sectors according to the HVI, 34% (n = 623) were classified as low health vulnerability, 35.3% (n = 645) as medium, and 30.6% (n = 560) as high or very high. In total, 28.5% (n = 521) of Recife’s census tracts were classified as food deserts and 73.47% (n = 1343) were classified as food swamps. The frequency of food deserts varied according to vulnerability, being more prevalent in areas with high HVI (12.47%). In contrast, food swamps were more frequent in tracts with a low or medium HVI (Table 1). There was a dose-response association in both associations.
Food deserts and swamps according to census tract (n = 1828) Health Vulnerability Index. Recife, 2019
* Pearson’s chi-square test.
As shown in Table 2, census tracts classified as food deserts had a higher mean number of households whose per capita income was ≤0.5 the federal minimum salary (35% [SD, 21%] vs. 27% [SD, 19%]; p < 0.001), as well as a lower mean monthly nominal income (BRL 1730.66 [SD, 1941.88] vs. BRL 1812.39 [SD, 1921.9]; p = 0.004), than in tracts considered non-food deserts. Food deserts also had more residents per household, more illiterate people, and a higher percentage of Black, Indigenous, and mixed-race residents than non-food deserts. Food deserts also had a higher mean number of households with inadequate water supply, sewage, and garbage collection than non-food deserts.
Socioeconomic conditions of census tracts in considered food deserts and swamps in Recife, Brazil, 2019
* Student’s t-test.
As also shown in Table 2, census tracts classified as food swamps had a lower mean number of households whose per capita income was ≤0.5 the federal minimum salary (28% [SD, 19%] vs. 35% [SD, 21%]; p < 0.001) and a higher mean nominal income than tracts classified as non-food swamps (BRL 1829.64 [SD, 1963.9] vs BRL 1456.67 [SD, 1854.10]; p < 0.001). In addition, food swamps had fewer illiterate people, fewer non-Whites, and a lower mean number of residents per household than non-swamps. The mean number of households with adequate essential services (water supply, sewage, and garbage collection) was higher in food swamps than in non-food swamps.
Regarding geographical distribution, food deserts predominated in highly vulnerable regions, especially the northern outskirts of the city, with scattered points in areas of low and medium HVI, as shown in Fig. 1. Food swamps were widely distributed among the 3 HVI categories, except for tracts of low vulnerability in the extreme north of the city, as shown in Fig. 2.
Census tracts (n = 1828) mapped according to the Health Vulnerability Index (HVI) (a) and food deserts (b). Recife. Brazil, 2019.
Census tracts (n = 1828) mapped according to the Health Vulnerability Index (HVI) (a) and food swamps (b), Recife, Brazil, 2019.
Discussion
Notably, there were few food deserts, but they predominated in highly vulnerable areas on the outskirts of the city. The socioeconomic conditions and essential services were worse than in areas not classified as deserts. In contrast, food swamps were many and widely distributed, predominating in areas of medium and low vulnerability, mainly downtown and in upscale neighbourhoods. Such areas also have a better income profile and access to essential services.
Fewer food deserts were found in Recife than in other large cities in southern (48.3%) and southeastern (38%) Brazil (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021; Borges et al., Reference Borges, Vargas, Honório, Mendes and Canuto2023). However, the characteristics of food deserts (i.e., income, demographic profile, and access to basic services) were similar in all 3 cities. The CAISAN study (2018) identified food deserts in several Brazilian capitals but did not find a fixed distribution pattern in income. These differences may be linked to urban characteristics. For example, Recife’s low-income neighbourhoods are contiguous with high-income neighbourhoods, where food swamps are highly prevalent. In contrast, high-income and low-income neighbourhoods are separated in the cities investigated in southern and southeastern Brazil (Belo Horizonte, 2021; Borges et al., Reference Borges, Vargas, Honório, Mendes and Canuto2023).
North American studies have also found more food deserts in low-income regions with a higher prevalence of Black and Latino residents (Walker et al., Reference Walker, Keane and Burke2010; Black et al., Reference Black, Moon and Baird2014). The lower income of minority populations and racial segregation are considered important factors in the occurrence of food deserts (Walker et al., Reference Walker, Keane and Burke2010; Black et al., Reference Black, Moon and Baird2014). However, due to the different methodologies of Brazilian studies, caution should be exercised when comparing the findings in the food deserts and swamps analyses. Indeed, the present study evaluated a wide variety of food retail, as in previous Brazilian studies (CAISAN, 2018; Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021; Borges et al., Reference Borges, Vargas, Honório, Mendes and Canuto2023), while studies in other countries have focused on proximity to supermarkets (Walker et al., Reference Walker, Keane and Burke2010; Black et al., Reference Black, Moon and Baird2014); in these countries, the presence of supermarkets is a marker of access of foods considered to be healthy (Walker et al., Reference Walker, Keane and Burke2010; Black et al., Reference Black, Moon and Baird2014). North American studies also used different measures of socioeconomic position limiting comparison.
A North American study adopted the measure of income ≤200% of the poverty line and distance greater than 1.6 km in an urban area or greater than 16 km in a rural area to classify a neighbourhood as a food desert (Cooksey-Stowers et al., Reference Cooksey-Stowers, Schwartz and Brownell2017). Food swamps are also assessed by different measures. The Center for Disease Control and Prevention (CDC), the USA, proposed a widely adopted measure: the modified retail food environment index (mRFEI) (CDC, 2011). This index evaluates the proportion of healthy food retail stores in relation to the total available in each census sector and classifies neighbourhoods that have a small value, but greater than zero, as a food swamp (CDC, 2011). Despite the different measurement methods, the Brazilian methodology is the most appropriate to study the topic in Brazil as it considers the most frequent food retailers present in Brazil (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021).
Regarding the income and racial profile of food deserts, the lower access to healthy foods among Black, Indigenous, and mixed-race residents reflects historical racial inequalities in Brazilian society. According to a 2021 survey by the Brazilian Institute of Geography and Statistics (IBGE), the average income of White workers was almost double that of Black and mixed-race workers in Brazil. Accordingly, in these populations, the poverty rate is twice that of Whites. Unemployment also disproportionately affects Blacks (16.5%) and people of mixed race (16.25%) compared to Whites (11.3%) (IBGE, 2022).
This study is the first to associate food environment characteristics with basic service provision in a city in northeastern Brazil. The lower socioeconomic condition of residents and inadequate access to essential services prevalent in Recife’s food deserts shows that vulnerable groups may face additional barriers to maintaining a healthy diet. This could be due to the ability of unfavourable environmental conditions to increase individual socioeconomic disadvantage (Macintyre, Reference Macintyre2007; Ford and Dzewaltowski, Reference Ford and Dzewaltowski2008). Thus, individuals from a historically lower socioeconomic position (non-White race, lower education and income, and rural vs urban residence) are less likely to have a nutritionally balanced diet (Canuto et al., Reference Canuto, Fanton and Lira2019), and a healthy diet is even more inaccessible for residents of food deserts (Macintyre, Reference Macintyre2007; Ford and Dzewaltowski, Reference Ford and Dzewaltowski2008).
The city’s numerous food swamps, characterized by an excessive supply of ultra-processed foods (Rose et al., Reference Rose, Bodor, Swalm, Rice, Farley and Hutchinson2009), especially in less vulnerable areas, reflect the findings of other studies in southeastern Brazil (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021; Grilo et al., Reference Grilo, Menezes and Duran2022). The higher percentage of White residents, higher income, and better access to basic services in the food swamps of Recife were similar to food swamps in Belo Horizonte, where 58.8% of the census tracts were classified as food swamps (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021).
The predominance of food swamps in the most affluent areas of Recife follows a trend toward a greater density of food retail, including those that sell ultra-processed foods, as a neighbourhood’s socioeconomic level increases, resulting in high exposure to unhealthy foods in Brazilian urban areas (Almeida et al. al., Reference Almeida, Novaes, Pessoa, do Carmo, Mendes and Ribeiro2021; Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021). However, Canadian (Luan et al., Reference Luan, Law and Quick2015) and North American studies (Hager et al., Reference Hager, Cockerham, O’Reilly, Harrington, Harding, Hurley and Black2017) have found that food swamps exist independently of the neighbourhood’s socioeconomic status.
One explanation for the predominance of food swamps in these census tracts in Recife could be the purchasing power of the population and the characteristics of the foods. The high presence of food swamps may also be a reflection of the non-random distribution of food retail. These businesses generally appear in wealthier areas of the city, where there are more potential consumers. The precarious infrastructure and higher crime rates in the most vulnerable areas can discourage the development of food retail (Walker et al., Reference Walker, Keane and Burke2010).
The abundant supply of ultra-processed foods in food swamps can discourage healthy eating and negatively affect the health of the population. Evidence suggests that individuals living in areas with more healthy food retail eat more fruits and vegetables (Jaime et al., Reference Jaime, Duran, Sarti and Locket2011; Pessoa et al., Reference Pessoa, Mendes, Gomes, Martins and Velasquez-Melendez2015) and are less likely to become obese (Matozinhos et al., Reference Matozinhos, Gomes, Andrade, Mendes, Pessoa, Friche and Velasquez-Melendez2015; Oliveira et al., Reference Oliveira, Menezes, Almendra, Lira, Aquino, Souza and Santana2022). However, living in areas with more fast food restaurants and less fresh and whole foods is associated with a higher prevalence of overweight and chronic disease (Mendonça et al., 2010; Chen et al., Reference Chen, Creger, Howard, Judd, Harrington and Fointaine2019; Gómez-Donoso et al., Reference Gómez-Donoso, Sánchez-Villegas, Martínez-González, Gea, Mendonça, Lahortiga-Ramos and Bes-Rastrollo2020; Llavero-Valero et al., Reference Llavero-Valero, Escalada-San Martín, Martínez-González, Basterra-Gortari, de la Fuente-Arrillaga and Bes-Rastrollo2021).
The concepts of food swamp and food desert complement each other as they demonstrate the effects of competition in the food environment between a high supply of ultra-processed foods and a scarcity of natural or minimally processed foods. Evidence suggests that food swamps are stronger predictors of obesity and other chronic non-communicable diseases than food deserts (Rose et al., Reference Rose, Bodor, Swalm, Rice, Farley and Hutchinson2009; Cooksey-Stowers et al., Reference Cooksey-Stowers, Schwartz and Brownell2017). A study carried out in the USA identified a higher prevalence of obesity in adults living in swamps than those living in food deserts, and this situation was worsened by socioeconomic factors, such as the lack of public transport in the neighbourhood (Cooksey-Stowers et al., Reference Cooksey-Stowers, Schwartz and Brownell2017). Another study demonstrated that living in a food swamp was associated with increased hospitalization rates for diabetes complications in the USA (Phillips and Rodriguez, Reference Phillips and Rodriguez2020). Moreover, in Canada, proximity to fast-food restaurants has been associated with an elevated likelihood of developing diabetes mellitus among young adults (Polsky et al., Reference Polsky, Moineddin, Dunn, Vidraceiro and Boot2016).
Limitations and strengths
This study has some limitations. The first involves the age of the data, since the sociodemographic data are from the 2010 Census, while the food retail data are from 2019. Due to COVID-19 and a political crisis, Brazil did not have a Census in 2020. The second limitation is the use of a HVI not adapted for the context of Recife. However, similar Brazilian studies have used the same HVI (Honório et al., Reference Honório, Pessoa, Gratão, Rocha, Castro, Canella, Horta and Mendes2021; Leite et al., Reference Leite, de Assis, do Carmo, Nogueira, Netto and Mendes2021; Borges et al., Reference Borges, Vargas, Honório, Mendes and Canuto2023). The third limitation refers to the non-inclusion of outdoor farmer’s markets in the study, which may have led to an underestimation of the number of retailers that predominantly sell natural or minimally processed foods. Finally, the utilization of secondary data for food establishments can be a limitation, as it does not encompass informal market establishments.
However, this study’s strengths include its pioneering investigation into the association between food deserts/swamps and social inequality (especially regarding essential services) in a city in northeastern Brazil. The study’s second strong point is its inclusion of a wide variety of food retailers, including street vendors, rather than simply supermarket retailers.
Conclusions
In Recife, a metropolis in the northeast region of Brazil, the food environment includes few food deserts, which are predominantly in vulnerable areas with worse socioeconomic conditions and access to essential services and mainly affect non-White and illiterate residents. On the other hand, food swamps predominate in more affluent areas with better basic services. The characteristics of Recife’s food deserts and swamps point to socioeconomic and racial inequalities in the food environment that can limit access to healthy foods and facilitate the consumption of ultra-processed foods. Thus, this investigation identified priority regions for increasing exposure and access to healthy foods and discouraging the sale of ultra-processed foods, especially in more vulnerable areas. For example, public facilities could be created to support healthy eating in food deserts and swamps, taxes could be added to ultra-processed foods and tax subsidies could be provided to natural or minimally processed food sellers. Furthermore, the social inequalities in food deserts and swamps, especially regarding access to basic services, indicate that public managers must consider new structural policies to mitigate social inequality.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0021932024000087
Acknowledgements
This study was supported by the Brazilian Ministry of Education (CAPES: Code 001).
Funding statement
Larissa Loures Mendes – The National Council for Scientific and Technological Development (CNPq) n. 4/2021 – Research Productivity Scholarship – PQ – Process n. 312979/2021-5. Raquel Canuto – The National Council for Scientific and Technological Development (CNPq) n. 9/2022 – Research Productivity Scholarship – PQ – Process n. 306656/2022-1.
Competing interests
The authors declare no conflicts of interest.
Ethics standard
Since this was an ecological study, no human beings were recruited, and institutional ethics committee approval was not required.
