View Report Details
Do Food Deserts exist? A multilevel geographical analysis of the relationship between retail food access, socio-economic position, and dietary intake
Project Code: N09010
University of Newcastle
White, D ; Bunting, D; Williams, D; Raybould, D; Adamson, A; Mathers, P
Diet is strongly patterned socio-economically and spatially, but it remains unclear whether geographical variations in retail provision contribute significantly to these associations. It has been suggested that ‘Food Deserts’ exist in areas where there is poor availability of the foods that make up a ‘healthy’ diet at an affordable price.
To determine the relationship between dietary intake and socio-economic factors at individual, household and neighbourhood levels and retail access to a ‘healthy’ and
affordable diet, and thus determine whether ‘food deserts’ exist and, if so, to describe their characteristics.
A cross-sectional, multilevel study was undertaken, involving simultaneous surveys of diet, social factors, health and food shopping behaviour in a representative sample of households and individuals in Newcastle upon Tyne and all retail outlets selling food, as well as data on access to retail outlets by private and public transport and socio-economic data on local areas. Surveys were piloted prior to main data collection. 5044 individuals (out of 6162 – 83% response rate) in 3153 households (out of 17801 contacted – 18% response rate) completed the main surveys. 560 retail outlets selling food (out of 658 approached – 85% response rate) were surveyed to collect information on the type, size and opening hours of the store and the range, cost and quality of 33 commonly eaten food items. Data on foods within this overall ‘basket’ were weighted in accordance with typical levels of consumption in England. Data from household individual and retail surveys were linked and analysed statistically and spatially using simple and multivariable techniques in order to explore the relationship between food retailing, socio-economic factors and dietary patterns.
Geographically referenced data on the road network and public transport in Newcastle, as well as socio-economic data from the 1991 Census were integrated with these data sets to enable spatial analyses.
The main survey produced a sample of households that was slightly biased towards higher socio-economic status, though geographically representative of all residential areas of Newcastle; the sample of individuals was slightly biased in favour of older age groups and women.
77% of households did their main food shopping at a multiple supermarket, 14% in a discount supermarket, 3% in a department store and 2% in other local shops. Households in the top socio-economic quintile were twice as likely to use a multiple supermarket compared with those in the bottom quintile. 33% said the main reason for using their usual food shop was because it was ‘near to home’. Disadvantaged individuals tended to cite cost and convenience whilst more advantaged individuals tended to cite range of foods available and quality as their main reason for using their main food shop. In regression analyses, shopping at a multiple supermarket was independently associated with having a university education, having greater dietary knowledge, living in a more affluent area, being in employment and travelling to their main food store by car. Shopping at a discount supermarket was independently associated with living in a more deprived area, having a lower income, having poorer dietary knowledge, travelling to shops on foot or bicycle, and living closer than 1km to their main food shop.
64% of householders used a private car to get home from their main food shop, 16% travelled by foot, 14% used public transport and 8% used a taxi. Those in the lowest socioeconomic quintile were 10 times as likely to travel by taxi or foot compared to those in the highest quintile. People travelled all over the city, and beyond, to shop and did not appear to be limited by the distance of shops from their homes (except those travelling on foot). People shopping at a significant distance from their home did so usually because they linked food shopping with other journeys (‘trip-chaining’). Individuals in the lowest socio-economic quintile had a median travel distance to their main food shop of 1.3 km and those in the highest a distance of 2.4 km. Those using multiple supermarkets travelled about twice as far as individuals using discount supermarkets. These differences reflect socio-economic patterns in shop use and ownership and use of a car versus walking to shops.
Most people (65%) reported no problems with food shopping. The main problem identified with food shopping was carrying food home, identified by 18% of householders. Those in the lowest socio-economic quintile were 10 times more likely to have problems carrying shopping home than those in the top quintile.
Average amount spent on food was about £26.00 per adult equivalent per week. Those in the lowest socio-economic quintile spent a median of £22.00 per adult equivalent per week on food, whilst those in the highest quintile spent on average £31.50. Overall, an estimated 17.8% of household income per adult equivalent was spent on food by our sample of households. More affluent households spent a significantly lower proportion of their income on food, even though they spent more per adult equivalent on food.
In univariable analyses, consumption of fresh fruit and vegetables was patterned by food expenditure and neighbourhood deprivation, and by age, sex and self-rated health.
Individuals in the lowest Townsend Deprivation Score quintile and those who spent less on food ate more fat as a percentage of total energy intake and drank less alcohol than individuals in the top quintiles. Healthy eating, measured by all dietary indices, was most strongly and consistently predicted by higher dietary knowledge. Regression models, at best, explained <10% of variance in dietary indices. The strongest independent predictors of dietary patterns at household and individual levels were demographic, socio-economic and behavioural factors, with dietary knowledge the most important overall.
Respondents, on average, ate out at a restaurant, pub, café, chip shop or other take-away 2-3 times per week. Eating out at a chip shop was associated with lower socio-economic position, lower dietary knowledge and worse dietary indicators, whilst eating out at a restaurant, pub, café or other take-away was associated with greater affluence, higher levels of knowledge and better dietary indicators.
Individuals in the lowest socio-economic quintile were less aware of current healthy eating messages than those in the top quintile. Poor knowledge was consistently associated with less healthy eating, measured by a variety of indicators, as well as with other measures of a less healthy lifestyle (e.g. low levels of physical activity and smoking).
Obesity and physical activity
BMI was found to be strongly socio-economically patterned, with less affluent respondents reporting significantly higher BMI overall. Being above ideal weight was also more common among men, older age groups and white Europeans. Being underweight was particularly associated with the youngest age group, being single, being a student and having inadequate cooking facilities. Physical activity was also strongly patterned socioeconomically and demographically. Although less affluent groups were generally more active at work, higher socio-economic groups engaged in more leisure time activity. Higher levels of leisure time activity were strongly associated with eating a healthier diet and with higher levels of dietary knowledge.
Shops in Newcastle
The most common category of shop (216, 39%) was convenience stores, but a small number of discount and multiple supermarkets (38, 7%) act as the main food shop for more than 90% of the population. Shops were distributed fairly evenly across the city, although some wards were better provided for than others. There is a concentration of food stores in the city centre (West City ward), which has the highest total number of stores and grocery stores per thousand households, and checkouts per thousand population. The socioeconomic characteristics of the area in which a shop was located correlated positively with the Townsend Deprivation Score of the household that shopped there, indicating that people from poorer areas tend to shop in stores in poorer areas, and vice versa. In general, shops were located in less socio-economically deprived areas (TDS of shops 2.6 (SD 3.5) vs. an overall TDS of Newcastle EDs of 5.09). General stores, multiples and delicatessens tended to be in more affluent areas (TDS -0.54, 1.03 and 0.97, respectively). Local discount stores, ethnic food stores, fishmongers, freezer centres and greengrocers tend to be in more deprived areas (TDS 4.0, 4.17, 3.27, 3.19 and 3.11, respectively).
Availability of food in shops
Few shops (40, 7%) sold the full range of foods surveyed (33 items). Those that did were primarily multiple supermarkets, discount supermarkets and departments stores. The full range of ‘unhealthy’ items was more available (in 50% of shops) than the full range of ‘healthy’ items (available in 25% of shops). The availability of the full range of foods, the ‘healthy’ and the ‘unhealthy’ baskets was not socio-economically patterned by area.
Convenience stores were more likely to sell the full range of ‘unhealthy’ items than ‘healthy’ items. Carbonated drinks, crisps, Kit Kats, whole milk and white bread were the only ‘less healthy’ items available in more than 50% of food shops (available in 84%, 79%, 61%, 61% and 53% of stores respectively). The only widely available ‘healthy’ foods were fruit juice (available in 52% of food shops) and semi-skimmed milk (available in 61% of food shops). Availability was associated with size of store, as measured by number of checkouts (the larger the store the greater likelihood of selling more of the 33 items). Larger stores were also more likely to sell a full range of fresh fruit and vegetables and the 10 less healthy items.
Cost of food
Our basket of 33 items cost £19.06 on average (IQR: £17.03-20.40), though the cheapest was £14.44 and the most expensive £23.57. TDS was significantly associated with cost of fruit and vegetables, which were more expensive in more affluent areas. However, in regression analysis, only one variable independently predicted cost: number of checkouts – prices were generally cheaper in the larger stores. For ‘baskets’ of items, discount and multiple supermarkets were cheapest overall (for 33 items). Discount and multiple supermarkets were significantly cheaper for the pre-packed 11 ‘healthier’ items and 10 ‘less healthy’ items. The basket of 10 fresh fruit and vegetables was cheapest in the market stalls and greengrocers. Overall, convenience stores and department stores were the most expensive shops, although only 3 out of 216 convenience stores sold all 33 items. There was also significant price variation within chains (‘price-flexing’) - for example, the full range of items cost a median of £20.50 at Tesco but ranged from £15 to £21.26 between Tesco stores.
Quality of fresh produce
Overall, 91% of individual fresh fruit and vegetables were of good quality. Quality was positively associated with size of store, as measured by number of checkouts, and number of fresh fruit and vegetables sold, but not number of ‘less healthy’ items or total number of items sold. Good quality fresh fruit and vegetables were consistently available only at multiple supermarkets and department stores, but quality of produce did not vary by ward.
Geographical proximity to shops
A shop selling any five food items was within 250m of the majority of streets in the city. Only a few rural streets were more than 1000m from a shop selling any five food items. Some parts of the city were further than 1000m from a shop selling 10 ‘less healthy’ food items. Some parts of the city were further than 1000m from a shop selling all 10 fresh fruit and vegetables. The proportion of areas served by shops selling all 10 fruit and vegetables of good quality, or of good quality and less than half of the median basket price, was relatively small, although the best coverage was in the less affluent areas along the riverside. Despite this, approximately 80% of households shop at stores selling 100% of the 33 food items we surveyed. More affluent households generally live further from the nearest shop selling each of our sample food baskets, including both ‘healthy’ and ‘less healthy’ baskets. People who shop at local and convenience stores, and to a lesser extent discount
supermarkets and department stores, appear to be disadvantaged with respect to availability of food items. However, perhaps surprisingly, there was no difference in availability for car users and people shopping by other modes of transport.
In regression analyses, retail factors were found to play little independent role in predicting healthiness of diets and none in the models for fruit and vegetables, NSP or Fat. Higher fat intake was associated with: lower dietary knowledge, risky or hazardous alcohol consumption, being male, higher cost of 33 food items (with maximum cost substitution) in households’ usual main food store, lower cost of weekly household food shopping, lower physical activity and shorter distance to nearest food store.
Overall retail provision for Newcastle residents was found to be good, with (as a measure of volume) around 8 checkouts per 1000 population. However, despite the large number of shops selling food, only 22 stores sold all 33 basic food items included in our survey. Of these, 14 were multiple supermarkets, four were discount supermarkets, one was a department store and three were convenience stores. Nevertheless, in 24 out of 26 wards, there was at least one shop selling 27 or more of the 33 food items on our list, and most of the ‘healthier’ and ‘less healthy’ pre-packed items and all of the fresh fruit and vegetables.
We have demonstrated marked differences in diets both socio-economically and geographically and, although these are not always in the anticipated direction, overall less
healthy diets were associated with social disadvantage and having a poorer dietary knowledge. However, using regression analysis, we were unable to demonstrate an
independent relationship between most indicators of healthier eating and factors relating to the local retail environment. The main exception to this was a paradoxical one: eating a diet higher in fat was associated with higher cost of 33 food items (with maximum cost substitution) in households’ usual main food store, but lower cost of weekly household food shopping. This may be because those consuming higher fat diets were more likely to shop at smaller local stores with higher prices and poorer availability.
Our analysis suggests that the strongest predictors of ‘less healthy’ eating are, in the case of fruit and vegetables, NSP and overall dietary index, predominantly demographic, socioeconomic and behavioural – factors such as age, sex, ethnicity, the presence of children within the household and levels of physical activity and alcohol consumption. Similar factors predicted fat intake, but higher cost of 33 food items (with maximum cost substitution) in households’ usual main food store, lower cost of weekly household food shopping and shorter distance to nearest food store were also independent predictors of fat intake.
Overall, it seems unlikely that the explanation for the wide variations in dietary intake seen across the study population lie within the retail domain. The vast majority of households do their food shopping at a large, multiple supermarket and travel there by car – an average of 1.9 km (about 1 mile) from their home. This is also the case for the majority of those who used public transport or taxis to get to and from their main food stores (average distance travelled 1.5-2.4 km). Together, these households
account for 84.5% of the sample. It can, therefore, be concluded that most people travel outside their immediate locality to perform their food shopping. This leaves those who do their shopping on foot – 620 individuals in 490 households in our sample or about 12% of the total households and individuals. It is only for this group that local retail provision would seem to be a possible, although not necessarily a necessary, determinant of diet. If we look at the dietary indices of this group of individuals who do their shopping on foot, we can see that they did have relatively low fruit and vegetable and NSP consumption and relatively low overall dietary index.
So, ‘do food deserts exist?’ The answer must be a qualified: ‘only for some’. And that ‘some’ is a minority of people who, for a variety of reasons, do not or cannot shop outside their immediate locality, and for whom, in addition, this locality suffers from poor retail provision of foods that make up a ‘healthy’ diet. Our findings suggest that the key predictors of healthy eating are primarily dietary knowledge, relative affluence and a ‘healthy’ lifestyle, so we must question whether those people whose diet is ‘less healthy’ than desirable would eat more healthily if supplied with improved retail provision. Our study does not provide evidence to support retail provision as a primary cause of consuming an ‘unhealthy’ diet, although poor retail provision may be an important contributing factor in some, well defined, circumstances (e.g. when individuals are dependent on local retail provision and that provision is less than ideal). Approaches to tackling the problem of food poverty need to address poor knowledge and skills related to the acquisition and preparation of a ‘healthy’ diet, as well as the question of retail access. Further research is needed to explore in greater detail the relationship between diet and retailing, in particular among those without access to a car.
Some of the files on this site may be in a format that your computer can't read. However, you can download Readers and Viewers for the following document types below: