Introduction

The public health burden of malnutrition (which encompasses obesity and undernutrition), in Latin American has been among the highest in the world (Batis et al., 2020). Both the increase and shift in available food have changed population diets, which have become one of the main drivers of malnutrition (Popkin et al., 2020). In recent years, available food has skewed towards energy-dense, nutrient-poor, and ultra-processed foods that are widely available, cheap, and heavily promoted by traditional and non-traditional media (Swinburn et al., 2013b). Hence, the food environment (broadly defined as any opportunity to obtain food at both micro and macro-levels) is a key mediator between the broader food system and people’s diets (Townshend & Lake, 2009; Turner et al., 2018). The International Network for Food and Obesity/non-communicable diseases Research, Monitoring and Action Support (INFORMAS) framework divides food environment into seven dimensions: food retail, provision, labelling, marketing/promotion, prices, composition, and trade and investment (Swinburn et al., 2013a).

In recent years, research has looked at the association between these different domains of food environments and the nutritional and health status of the population (Caspi et al., 2012; Engler-Stringer et al., 2014). Most of this research comes from high-income countries where food environments differ significantly from the food environments in Low- and Middle-Income countries (LMICs) (Turner et al., 2020). Therefore, research on the food environment in LMICs is needed to better understand the interaction between the food environment, people’s behaviours around food acquisition, and health outcomes.

Robust measurement of the food environment of a specific country is essential before its implications for nutrition and health outcomes can be deduced. However, the study and measurement of the food environment in LMICs is scarce because of the many obstacles to high-quality assessment: the lack of systematic data; the difficulty of processing data when available (big data mining); the lack of data processing from a public health point of view; the complexity of documenting and analysing the unregistered (mobile) food retail locations; and the lack of data localized to small-area level or even of geo-coded data (Pérez-Ferrer et al., 2019; Turner et al., 2020).

In Latin America, a region constituting mainly by LMICs, research on food environments remains scarce, although recent publications have reflected increasing interest (Pérez-Ferrer et al., 2019). The dimension most studied has been food retail. In this paper, we focus on Mexico City, the capital of the third largest country in Latin America and the second most populated city in the region, and where 35% of adults have obesity (Barquera et al., 2020; United Nations and Department of Economic and Social Affairs Population Division, 2019). To date, only three studies have addressed the food retail environment in Mexico City. Reyes-Puente et al. (2022) and González-Alejo et al. (2019) identified patterns of access to food retail, namely food deserts (areas with limited access to healthy food), food oases (areas with good access to healthy food) and food swamps (areas with high access to unhealthy food), and the association of these patterns with urban marginalization and socioeconomic status in Mexico City; however, these studies did not consider the wide variety of food outlets available, and the health classification of food outlets remained unclear, thereby hindering an accurate demarcation of food deserts, oases, and swamps. Rosales Chavez et al. (2021a, b) have been exploring street food availability and distribution across income levels, suggesting that street food vendors distribute differently on the basis of neighbourhood income levels; in spite of this, their work lacks spatial analysis techniques to explore spatial and socioeconomic patterns.

In view of the limitation of previous studies and considering the scarcity of research on the retail food environment in Mexico City, the objective of this study is threefold. First, to investigate further the spatial patterns of a broader selection of food outlets; second, to put forward a health-based classification of food outlets; and finally, to assess whether the spatial distribution is dependent on types of food outlet. This paper mines administrative data to produce systematic small-area level data on the spatial distribution of different types of food outlets -adjusted by population density- in Mexico City.

Methods

Data Source

This study uses data from the 2020 National Statistics Directory of Economic Units (DENUE for its Spanish acronym). DENUE is part of the economic census carried out in urban areas (localities with >2500 inhabitants) and collects data from permanent and semi-permanent businesses in Mexico where goods and services are sold. As such, it contains data on the localization of outlets selling food and beverages that are in a single physical location, placed permanently and delimited by building and fixed installations; this also includes dwellings where economic activities are carried out (Instituto Nacional de Estadística y Geografía, 2020). Semi-permanent establishments are also included in the DENUE. These are establishments that are settled in one place, although the structures are not fixed constructions but can be moved or dismantled (i.e. metal structures placed in parks or on pavements).

The data are publicly available and produced by the Mexican Institute of Statistics and Geography (INEGI). The DENUE data were collected by INEGI as follows: 1) street audits were carried out by field workers to identify economic units; 2) field validation of the economic units identified in step one verified the localization of the units, and 3) a questionnaire collected data on owners and employees (e.g. number of employees, demographics), and outlets´ characteristics such as name, street address and type of street. Use of mobile computing devices for data capturing, processing, and transmission allowed digitalized cartography to assign a geographic reference code to each establishment in real time (Instituto Nacional de Estadística y Geografía, 2020).

Additionally, we mapped an approximation of the socioeconomic distribution of Mexico City in terms of urban marginalization levels based on the marginalization index that comprises 11 indicators from the following domains: education, health access, housing and basic needs (CONAPO2020). However, it has been suggested that the index is not useful for Mexico City (Cortés & Vargas, 2011), as it is a measure for rural areas and uses standards from the past century, thereby reducing variability at the top of the distribution, i.e. the very rich and the middle classes fall into the same category “low marginality”. Therefore, we used the index in this paper solely as a reference, but caution is recommended when using it as independent variable.

Classification of Food Outlets

The DENUE classifies food outlets based on the North American Industry Classification System (NAICS), which is used in Canada, the USA and Mexico to monitor industrial and economic activities across the North American region (Instituto Nacional de Estadístíca y Geografía, 2018a, b). Although the use of the NAICS provides a common ground for researchers in those countries, it was created with an economic purpose in mind, and less is known about the usefulness of this classification for research into health and nutrition.

The classification of food outlets is still subject to ongoing debate in food environment research (Wilkins et al., 2017). There is no universal classification for food outlets and, such classification may not be feasible because of differences in the food environment around the world. Even literature from a single country can vary in its classification of food outlets, e.g. in Mexico, where some authors have developed their own classificattion system (Hernandez-Barrera et al., 2016; Lopez-Barron et al., 2015; Rosales Chavez et al., 2021b) while others have retained the NAICS (Molina et al., 2017; Navarro & Vélez, 2019; Reyes-Puente et al., 2022) or have adapted it (González-Alejo et al., 2019; Hernández-F et al., 2021; Pérez-Ferrer et al., 2020; Pineda et al., 2021). This study proposes a reclassification of the NAICS that considers a broader variety of food outlets. The aim is to present a more adequate and detailed picture of the food offered in Mexico City (Table 1). In order to provide a health characterization, food outlets were classified into three categories: healthy, mixed and unhealthy. This health categorization was based on work by Hernandez-Barrera et al. (2016) and Pérez-Ferrer et al. (2020) and was discussed among local dietitians:

  1. 1)

    Healthy: food outlets that sell primarily fresh and unprocessed food such as fruits, vegetables, lean protein food sources such as poultry and fish, maize-based tortillas, wholemeal bread, salads, and unsweetened beverages.

  2. 2)

    Unhealthy: food outlets that primarily sell processed and ultra-processed food products, or foods high in added sugars (sweets, ice cream, sugar-sweetened beverages), or high in saturated fats (fried food, fast food).

  3. 3)

    Mixed: vendors who offer products from both healthy and unhealthy food categories.

Table 1 Description and classification of food outlets in Mexico City
Full size table

Because in-store assessment was not conducted, both the food outlets and the healthfulness classifications were based primarily on the descriptions provided by INEGI; these descriptions include the most common food items sold by each type of food outlet. When items sold in a specific outlet were not clear, we were aided by the establishment’s name, Internet searches, Google Street view and local knowledge. This involved an important amount of data cleaning as many food outlets were misclassified in the raw data.

Food Outlets Density

Food environment data can be aggregated in different ways depending on the purpose of the study. In the present study, the main purpose was to produce data at the highest possible disaggregation level that can still have policy relevance and can be useful for future spatial or hierarchical modelling. Thus, we use census tracts or the Basic Geostatistical Area recognized in Mexico (AGEB for its Spanish Acronym) as defined by the national statistical office. A limitation, however, of using census tracts is that their arbitrary size and scale can lead to problems such as the Modifiable Areal Unit Problem (MAUP), a source of bias when point data are aggregated into spatial units and the resulting sum is influenced by original underlying patterns distorted by the choice of district boundaries (Nelson & Brewer, 2017; Wilkins et al., 2017). A recommendation that might help to counteract the MAUP (but not solve it) is to use small geographical scale variables (Nelson & Brewer, 2017).

To summarize food outlets in census tracts, intensity metrics were used, specifically standardized metrics known as densities (Wilkins et al., 2017). Adensity variable for each type of food outlet was computed by diving the total count of outlets by the population density of each census tract. The use of this density measure aimed to account for the area (km2) but also for the population living in that area. Since the density measure was highly skewed toward zero, Rosner’s test was used to explore density outliers. Some of those outliers corresponded to those census tracts with very low population density but high number of outlets. It was decided not to exclude those values, because it was considered that they could be an informative part of the analysis. Due to the skewness of the density distribution a log transformation was performed, this to facilitate further analyses.

Statistical Analysis

The analysis was based on two main statistics: Global Moran’s Index and univariate Local Moran’s I. Global Moran’s I was used to formally assess the existence of spatial patterns for each food outlet type (Anselin, 1996). It is the most widely used statistic to estimate global spatial autocorrelation and is a succinct measure of geographical clustering; the underlying null hypothesis is that the target variable is randomly distributed across space, so when there is a pattern, the null hypothesis is rejected. The present study relied on a spatial weight matrix calculated using a second-order queen’s contiguity. Monte Carlo based computation of Moran’s Index was used as is a robust approach to estimate this statistic and test the hypothesis (Gimond, 2021). A limitation of the global Moran’s Index is that do not provide the location of such clustering, so we also measured local spatial autocorrelation by using Local Indicator of Spatial Association (LISA), specifically univariate Local Moran’s Index. This method identifies where the clustering is located and whether this placement varied according to the type of food outlet. We used second-order queen’s contiguity spatial weights as this were the most sensible approach considering the size of the census tracts. Local Moran provided the location as follows: a) hot spots or High-High (HH), high values with neighbours with high values; b) cold spots or Low-Low (LL), low values with neighbours with low values; c) High-low (HL), high values surrounded by neighbours with low values; and d) Low-High (LH) low values surrounded by neighbours with high values. These last two categories represent spatial outliers which are index regions having higher or lower-than-average rates, and the neighbours having lower or higher-than-average rates (Kramer, 2022).

The statistical software R, version 4.1.2 (R Core Team, 2021) was used to prepare the spatial data, create maps and calculate Moran’s I. The package “EnvStats” was used to identify outliers (Millard, 2013); the package “tmap” was used to create maps (Tennekes, 2018); and the package “spdep” was used to calculate Moran’s I (Bivand & Wong, 2018). To calculate LISA, we used the software GeoDA version 1.20.0.8 (Anselin, et al., 2006).

Results

According to the 2020 geostatistical framework, Mexico City is divided into 16 municipalities and 2431 census tracts (Fig. 1). The average geometric area of census tracts in Mexico City is 0.32 km2 (0.008 – 7.39 km2), and the average population is 3759 inhabitants (1 – 21,198).

Fig. 1
figure 1

Municipalities and spatial distribution of the Urban marginalization levels in  2020 within Mexico City

Full size image

A characteristic of Mexico City is the heterogeneity of land uses within its borders. For instance, the southern municipalities combine rural and urban land uses, having large territories with low population density due to the presence of conservation areas (e.g. Milpa Alta), with surrounding mountains and volcanoes (Magdalena Contreras and Tlalpan) or with water canals (Xochimilco). Because of this, population density is higher in the northern and eastern zone of the city (León Torres et al., 2016). Socioeconomic status also varies (Fig. 1).

Mexico City had 186,458 food retail outlets registered in November 2020. Of these, 92% were permanent (171,541) and 8% were semi-permanent (14,916). The most prevalent were small grocery stores, followed by restaurants and Mexican fast food (Table 2).

Table 2 Total number and proportion of food outlets by category for Mexico City, 2020
Full size table

Choropleth maps for each type of outlets are presented in supplementary material.

Global Moran’s I (Table 3) showed that spatial clustering was present for all types of food outlets, suggesting that location of food outlets do not follow a random spatial pattern.

Table 3 Global Moran’s I for each type of food outlet in Mexico City, 2020
Full size table

Healthy Food Outlets

The two types of food outlet classified as healthy were those that sell fruit and vegetables and those that sell maize-based tortillas, a staple and culturally important food item in Mexico. Fruit and vegetable outlets (Fig. 2A) showed high-density clusters at the northern extremity, in the eastern municipality of Iztapalapa and in some parts towards the southern border. Low densities were mainly in the western parts of the city. Some low-density clusters were surrounded by high-density census tracts; this could mean that people live in census tracts with low availability of fruits and vegetables, but that they have a high-density census tract in relatively close proximity, which might facilitate access.

Fig. 2
figure 2

Densities of healthy food outlets in census tracts of Mexico City; LISA cluster map

Full size image

Tortilla shops (Fig. 2B) had high-density clusters towards the eastern, southern and northern borders. Beyond a lower number of HL census tracts lay a larger low-density cluster in the west and mid-west. A low-density cluster in the south covered two municipalities, Tlalpan and Xochimilco.

Mixed Food Outlets

Food outlets categorized as mixed included mainly shops and restaurants that sell a variety of food, both healthy and unhealthy; among these were food stores where it was possible to buy food for immediate consumption or to supply household consumption. Small grocery stores (Fig. 3A) had high-density clusters throughout the city but primarily on the northern border, in the north-east (around the international airport), in the east around the Central de Abastos (the largest public wholesale market in Mexico), and on the southern border with larger clusters in the south-east. Low-density clusters occurred in the north-west and mid-west, with the latter coinciding with the location of the longest main road in the city (Avenida Insurgentes), which crosses the city from north to south. Finally, another cluster occurred in the south in the municipality of Tlalpan.

Fig. 3
figure 3

Densities of mixed food outlets in census tracts of Mexico City; LISA cluster maps

Full size image

Mini supermarkets (Fig. 3B) were concentrated primarily in the mid-north, which is consistent with the location of the historic centre of the city. Another cluster appeared in the south and some smaller clusters in the east, near the Central de Abastos. Low-density clusters were concentrated primarily in the west.

Supermarkets (Fig. 3C) were few, being absent from 90% of census tracts. Nonetheless, there were high densities in a few census tracts scattered in the west and mid-south. Most notably, the largest cluster was on the western border and coincided with Santa Fe, one of the major business districts of Mexico City.

Retail of animal products (Fig. 3D) had high-density clusters in the east, three in the south, and the largest in the south-west. Smaller high-density clusters were scattered in the north. The largest low-density cluster was on the western border. A north-to-south line of low-density tracts in the west appeared to be surrounded by HL clusters.

Further  categories of mixed-food outlets were those that sell and serve food for consumption on the premises, including restaurants, bakeries and cafés. For à la carte restaurants (Fig. 4A), high-density clusters were identified specifically in the west and mid-west, north and mid-south. Hot spots were also located in the north-east, around the international airport and the Central de Abastos. Low-density census tracts were scattered throughout the city, with a cluster in the west.

Fig. 4
figure 4

Densities of further categories of mixed food outlets in census tracts of Mexico City; LISA cluster maps

Full size image

Mexican fast-food outlets sell tacos (generally maize, with varied fillings), tamales (steamed maize dumplings), tortas (sandwiches) and other traditional “on the go” foods. Their spatial distribution (Fig. 4B) resembled that of à la carte restaurants, with high-density clusters in almost the same spots. However, high-density clusters were also conspicuous in the south-west and north-east.

Clustering of outlets for non-alcoholic beverages (Fig.4C) did not differ substantially from those of Mexican fast-food outlets and à la carte restaurants, although there were additional low-density census tracts in larger clusters on the southern border. Bakeries (Fig. 4D) had fewer high-density census tracts, with small clusters in the centre, mid-south, south-west, east and south-east. Low-density clusters were found in the west, mid-south and east.

Unhealthy Food Outlets

Finally, the category of unhealthy food outlets constitutes international-style fast-food outlets, convenience stores, and retailers of sweets, desserts and savoury snacks. International-style fast-food outlets (Fig. 5A), whether transnational fast-food chains or local restaurants, sell mainly burgers, pizza, fried chicken, and hot dogs. High-density census tracts were clustered predominantly in the centre and mid-south. Other high-density census tracts were scattered throughout the city and in some cases formed small clusters in the south-west and north-east. Only 85 low-density clusters were identified, with the largest cluster being in the west.

Fig. 5
figure 5

Densities of outlets for unhealthy foods in census tracts of Mexico City; LISA cluster maps

Full size image

Chain convenience stores sell mainly ultra-processed food and a very small selection of staple foods and household items. They were absent from 52% of the census tracts (Fig. 5B), and the majority of high-density clusters were towards the west side of the city with the largest cluster expanding from north to south. Low density clusters were identified in the east and on the northern border.

Retail of sweets and desserts (Fig. 5C) shows high-density clusters in the east, north and mid-north, with a few other smaller clusters scattered across the south. Two main clusters of low-density tracts were identified: one in the city centre and a large one in the west.

Savoury snacks outlets, specializing in fried snacks such as potato chips and traditional fried wheat snacks, showed few significant census tracts in LISA analysis (Fig. 5D). Nonetheless, the few small clusters of high density were scattered mainly to the east of the city, with two additional clusters in the south-west and the north. Low-density tracts were scattered throughout the city.

Discussion

The pattern of spatial distribution of food outlets in Mexico City differed according to the type of establishment. All types showed clustering of high density (hot spots) and low density (cold spots), but with the geographic location of clusters changed depending on type of establishment.

Unhealthy and Mixed Food Outlets Have the Highest Density in Mexico City

An important finding from this study is that outlets with higher numbers of high-density clusters are those categorized as mixed (small grocery stores, Mexican fast foods, non-alcoholic beverages and restaurants) and unhealthy (chain convenience stores and fast foods). A higher density of unhealthy food near places where people carry out their daily and recreational activities can have several implications for the diet and health of inhabitants of Mexico City. Evidence from Canada, the USA and the UK suggests that higher availability and accessibility to fast-food restaurants increases the likelihood of obesity in both adults and children (Fraser & Edwards, 2010; Hollands et al., 2013; Mylona et al., 2020; Otterbach et al., 2021). Nonetheless, the sole study in Mexico on this aspect (Pineda et al., 2021) concluded that although the statistical model suggests that fast food outlets showed higher odds of being associated with a higher body mass index (BMI) in adults, the coefficients were not statistically significant. Therefore, more research with different methodology including ecological and hierarchical modelling, as well as among other age groups, should explore this association further.

In the case of convenience stores, Pineda et al. (2021) suggested that in Mexico 10% increase on the density of these would be associated with a 0.34 kg increase in weight in adults (Pineda et al., 2021). Nonetheless, a study in Mexico City (Rodríguez-Guerra et al., 2022) found no significant association of a medium and high density of convenience stores with a lower score in the overall diet quality. Additionally, availability of these food outlets can influence purchase behaviours, since a study in Mexico (Hernández-F et al., 2021) suggested that a higher density of convenience stores and small grocery stores was associated with a higher likelihood of the purchase of ultra-processed foods and sugar-sweetened beverages.

In the present study, an important finding was that hot spots of convenience stores were mainly in the west, an area with a low level of marginalization, whereas hot spots of small grocery stores were in the east, an area with a high level of marginalization. In line with this, a study by Rosales Chavez et al. (2021a) concluded that there was a higher density of street food vendors in middle-income census tracts than in low- or high-income census tracts. This might indicate that the incomes in census tracts and neighbourhoods might influence the location of specific types of food outlets.

Considering that small grocery stores are the predominant type of food outlet in Mexico City, further research is needed to characterize better these outlets. For the purpose of the present study, small grocery stores were considered “mixed” because they sell both healthy and unhealthy food (see Pérez-Ferrer et al., 2020); however, no empirical study to date has conducted in-store audits of these types of stores, so the ratio of healthy and unhealthy food is still not known. Further exploration could help to understand better the food items sold in these outlets.

Where Are the Clusters of Healthy Food Outlets?

Another finding of note was that among the three health categories of food outlets used in the present study, the healthy category was the one whose constituents had the most similar spatial distribution: both fruit and vegetable and tortilla shops had high-density clusters along the southern border and in the north and east, urban areas with medium to very high marginalization (see Fig. 1). This contributes to the debate as to whether those in Mexico with lower socioeconomic status have healthier diets. For instance, a study by Perez-Tapayo et al. (2020) concluded that Mexican adults with lower socioeconomic status and living in rural areas had better diets than did urban dwellers with higher socioeconomic status. This could suggest that a healthier food environment could contribute to better diets, although further research is still needed, especially to assess the role of food availability on healthy diets. Although food outlets availability might not necessarily imply consumption, it is important to consider other contributory factors, such as economic and physical access, since they could affect or mediate the relationship between food availability and food consumption.

What Is Happening in the West Part of the City?

In the present study, low-density clusters in the western part of the city occurred, albeit with small variations in size and location, for all types of establishments except restaurants, supermarkets and convenience stores. We did not consider this area a food desert, mainly because this low density applied to all three types of food outlet: healthy, mixed and unhealthy. The western region of Mexico City is characterized by very low or low marginalization levels, which indicates that people with high socioeconomic status reside there. On the basis of the spatial distribution of urban marginalization levels, it is feasible to propose that socioeconomic status could influence distribution of food outlet types. Further mixed-methods research could include ground truthing of the area, as well exploring food acquisition patterns, mobility patterns and the existence of other food acquisition channels, such as mobile markets or mobile vendors of fruit and vegetables. This could provide more insights on whether food availability and food access might be an issue in the western region of Mexico City and how people behave around it.

The Case of Supermarkets

Supermarkets are considered in the USA and some other countries to be a healthy food source, as they have been associated with a lower prevalence of obesity and overweight and higher intake of fruit and vegetables (Morland & Diez Roux, 2006; Zhou et al., 2021). However, findings from developing countries are mixed or inconclusive; for instance, in Kenya, buying in supermarkets was associated with a higher BMI (Kimenju et al., 2015). Evidence from Mexico is still inconclusive as to the direction of the association between density of supermarkets and BMI; findings so far point toward a negative association, but this has not been statistically significant (Molina et al., 2017; Pineda et al., 2021). In this study, we classified supermarkets as a mixed type of food outlet because although they present healthy options, they also offer a wide range of unhealthy and ultra-processed foods. The low numbers of supermarkets for a city this large and populous were surprising. This could indicate two things: that supermarkets do not represent the main source of food for residents of Mexico City, owing to the availability of other food acquisition channels such as small grocery stores, convenience stores, public markets and tianguis (open-air mobile markets that occur in neighbourhoods on certain days of the week); or (although unlikely) that the DENUE might have underreported the number of supermarkets in the city. A cross-validation exercise with other sources could resolve this.

The Role of the Central de Abastos

Another consistent finding was the relevance that the wholesale market Central de Abastos has in the food environment of the city. It is located in the eastern municipality of Iztapalapa and is considered to be the largest public market in Mexico or even the world, occupying 328 hectares and selling >30,000 tons of food a day (Hume, 2019). Its distribution network supplies food throughout Mexico and to diverse food retailers such as restaurants, public markets, chain supermarkets and other smaller retailers. Although the business is largely wholesale, it is also possible to buy food for household consumption. This study has suggested that the presence of such a large food supplier affects the surrounding census tracts by increasing the presence of food outlets such as small grocery stores, mini-supermarkets and à la carte restaurants. Research on the influence of Central de Abastos on individual diets and health outcomes might shed light on the health impact of large urban public markets.

Policy Implications

Results from the present study could guide policymakers towards an improvement of the food environment, particularly to direct efforts towards specific areas of the city. For instance, to increase the availability of fruits and vegetables in low-density clusters located in the west and centre of the city in the municipalities of Miguel Hidalgo and Coyocán. Another potential policy initiative derived from the large presence of small grocery stores in the city could be the promotion of healthier small grocery stores by regulating the ratio of unhealthy food or ultra-processed food sold in this type of outlets.

In term of urban planning policies, and following what has been implemented in other countries, findings from this paper could contribute to target specific areas where urban planning could modify the food environment. For instance, in England there are restrictions on opening new fast-food outlets near schools or in areas where existing outlets has surpassed a certain threshold of all retail outlets and where childhood obesity rates are above a certain threshold (Brown et al., 2022). Although the context might difer, these results could help to focalize similar actions by regulating the new location of unhealthy food outlets.

Strengths and Limitations

This paper is not without limitations. The DENUE, despite being a good resource for food environment research, only registers static food outlets, ignoring mobile food vendors. In view of the cultural importance of street food in Mexico City (Long-Solís, 2007), this provides an incomplete picture of the food environment accessible to the residents. Assessment of the street food environments necessitates fieldwork, as has been used by Rosales-Chavez and colleagues (Rosales Chavez et al., 2021a, b), but this can be costly and time consuming.

Use of population density to adjust the data assumes that people buy their food in the vicinity of where they live. However, this might not be appropriate in a large city where people may live in one area but carry out their daily activities in another. An important factor to be considered is the daily mobility of inhabitant of Mexico City and its metropolitan area. On average, Mexico City residents spend 88 minutes a day in transit, either by public transport (51%), walking (66%) and private transport (22%) (Hayden, 2021; Instituto Nacional de Estadístíca y Geografía, 2018a, b). If we add to this the time people spend at work or in education, we can assume people spend a significant amount of time out of home. Thus, as suggested by Hayden (2021), only considering the food environment around people’s homes can be a limitation as we assume that people are stationary. Therefore, it is important to consider that inhabitants are indeed exposed to a wider food offer while in transit, especially from street food vendors that tend to locate in busy circulatory networks like transport hubs. Additionally, there are socioeconomic and demographic factors such as social class and occupation that can play a role in the interaction between food environment and daily mobility. It has been suggested that those with higher income might be able to afford eating in restaurants or in workplaces, while low-income people rely more on street food vendors which tend to be much cheaper and widely available (Hayden, 2021). Accurate assessment of the food environment to which people are exposed will require a survey to collect, geolocation of their homes, workplace or school, as well as their mode of transport and transportation routes.

Another restriction of this paper is that only carries univariate analysis limiting the extrapolation of results to health and nutrition outcomes. Even so, this study provides the basis for further analyses that could strengthen the existing evidence of the association between the retail food environment and obesity and diet-related chronic diseases.

Among the strengths of this study is that it is the first paper to characterize in more detail the food environment of Mexico City by using 14 types of food outlet. Moreover, it is the first study to present spatial analysis techniques to assess spatial patterns in the retail food environment of Mexico. Additionally, identification of cold and hot spots can provide the basis for future research; for instance, identification of the clustering of unhealthy food outlets can help to explore the health implications. The spatial patterns should be investigated further to assess whether they can be explained by other factors.

Conclusion

This article provides systematized information, at the small-area level, regarding retail food environments in Mexico City; this entailed data mining and the proposal of a classification of these data based on public health. The information was grouped into 14 types of food outlet and was projected in choropleth maps as a population-based density measure at census tract level. Characterization of the retail food environment of Mexico City could help to identify the types of food outlet that might be deleterious to nutrition and health. This could guide researchers and policymakers towards an improvement, particularly to direct effort towards specific areas of the city.