Dr. Angelo Moretti, Manchester Metropolitan University, email@example.com
Dr. Adam Whitworth, University of Sheffield, firstname.lastname@example.org
Dr. Megan Blake, University of Sheffield, email@example.com
In the UK many are not food secure. Food security is the ability to consistently afford, access and utilise the food needed to maintain good health and wellbeing. When we think of food insecurity, we tend to think of it in relation to low-income countries. More recently, however, as a nation, we are beginning to recognise that a significant proportion of the UK population is not food secure. Much of this awareness-raising has been as a result of efforts by food charities and campaigns such as that spearheaded by Marcus Rashford.
While this awareness-raising has been welcome, the focus has been on those who are at the sharpest end of food insecurity; those who are skipping meals for a whole day not out of choice. Food banks have been set up in communities where people have recognised this problem of hunger with the intention of meeting immediate food needs. Hunger is understood as having been hungry at least once in the previous month but were unable to get food. This is our first measure.
We identify two further measures.
Those who struggle include people who have cut back on food or skipped meals. In addition, they have received support from their community with food essentials, or they indicated they could not get to the shops, could not get a delivery, or were too ill to get food. Those who experience these additional indicators of food insecurity are not typically included in the statistics, which tend to focus on financial reasons for food insecurity.
The last measure are those who worry about being able to adequately supply the food they need for themselves and their families. This latter group are typically considered marginally food secure because they have enough food. However, they may have traded down on the nutritional quality of the food they purchase (Drewnowski 2012). We have included this category because there is firstly mental stress associated with food worry. Secondly, these are people who are at risk of having low or very low food security, for example, through an unexpected expense, illness or relationship breakdown. We have seen many people over the period of the pandemic who have fallen from this group into the other types of food insecurity.
The burden of these forms of food insecurity includes immediate threats to health and wellbeing (Blake 2019). This burden includes the stress of trying to manage a budget that may not extend sufficiently, the worry about providing adequate nutrition, and the mental load associated with trying to navigate limitations imposed by transportation, inadequate equipment, cost, physical ability and household food preferences. Trading down on food quality and nutrition extracts a price to physical health in terms of diet-related illness, but it also results in narrower diets and the loss of understanding about what certain foods are and how to cook them. Finally, research demonstrates that those who struggle to access food are also isolated, which has an impact on quality of life and wellbeing (Blake 2019).
Food insecurity is concentrated into places. What this means is that cumulatively the effects of food insecurity include reductions in the ability of a community to be resilient in the face of crisis because community-based social networks have been lost. Local foodscapes have become food deserts because the demand for and the supply of healthy food is not present in the place where people live at a price they can afford. The burden of food insecurity also means that people in these communities struggle to see how they can contribute to achieving wider social and environmental goals that shape life in the UK today and help define us as a society.
Many local governments have spent considerable time and resources enabling access to food by supporting food banks whilst also moving those using food banks onward. Many are also looking at ways to support those who are moderately food insecure and those who are worried about being able to purchase food that contributes to their health, social and local economic wellbeing. Until now, however, there has not been an estimate of these three levels of food insecurity at the local authority scale across the whole of the UK.
The UK Local Government food security estimates were developed to help provide local level benchmarking. The purpose is to inform the types of services and support that are needed to achieve food security for everyone and to move beyond a focus on food banks.
The data sources and methodology used to create the estimations are briefly outlined in this document. Further background, context, results and implications will be shortly published in the academic literature, and this document will be updated with references to those publications. If, in the meantime, you have further questions, please do contact the research team.
Data & methods
Despite increasing academic and policy interest in food insecurity and its geographical variation, there has been a lack of large-scale and UK-wide quantitative detail regarding its local prevalence. This is due to the lack of adequate data and/or methodological innovation. Our local estimates use a methodological technique known as small area estimation. In particular, we use a multilevel regression-based small area estimation approach to small area estimation that moves through four key steps.
Step 1: Identifying the predictors of food insecurity
First, estimate the key relationships between predictor variables and the food insecurity outcome variables of interest in a suitable survey dataset. The survey data that we use come from a nationally representative survey of 4,231 adults across the UK conducted in late Jan/early Feb 2021. Local authority codes are included in the survey to enable the small area estimation. The survey was commissioned by the Ford Foundation and conducted by YouGov. For greater detail about the survey and its methodology, see this Food Foundation report.
We provide local estimates of three different indicators of food insecurity in order to offer a sense of the depth and variation of experiences geographically:
Hunger: Have you/anyone else in your household ever been hungry but not eaten because you couldn't afford or get access to food over the past month?
Struggle: A positive response to any of the following three indicators:
● Sought Help: Since Christmas 2020, did you seek help from any of these places: local authority; food bank; advice agency (e.g. CAB); government website; local charity; a volunteer; a religious institution; somewhere else?
● Skipped or shrank meal: have you/anyone in your household had smaller meals than usual or skipped meals in the past month?
● Reasons: Reasons why they did not have enough food. Positive answers to any of the three possible reasons asked around: money, access or supply.
Food worry: How worried, if at all, are you currently about getting the food you need? (very worried and fairly worried are taken as positive responses)
How these categories compare to other Food Foundation reports from this data:
The Food Foundation reported that food insecurity had been about 7.4% of the adult population in January 2021. This measure focuses on three questions from their survey and is a combined food security measure (see their report) --
1. Did you/anyone else in your household have smaller meals than usual or skip meals because you could not afford or get access to food?
2. Have you/anyone else in your household ever been hungry because you could not afford or get access to food?
3. Have you/anyone else in your household not eaten for a whole day because you could not afford or get access to food?
The hungry measure includes those who indicated a positive response to the second question.
The struggle category includes the first question as well as responses to some additional questions. Included also are those who have sought help accessing food, or who indicated elsewhere in the survey a reason for why they did not have enough food. We have included these in the analysis because food insecurity includes the ability to consistently access, afford and utilise food. As such, this definition engages all the dimensions of food insecurity that are included in the FAO definitions.
The report from the Food Foundation also does not include those who indicated that they are worried about having enough food, which we include as our third category.
Overlap between the measures
All of the percentages below are survey weighted. As intended, survey respondents are in one category only and all survey respondents are included in exactly one category.
Hunger only: 0%
Struggle only: 4.6%
Worry only: 6.1%
Hunger and Struggle: 1.3%
Struggle and Worry: 2.4%
Hunger and Worry: 0.03%
Hunger, Struggle and Worry: 2.9%
Because of the overlap, it is not possible to add the three measures together to arrive at a total. Within measures you can aggregate the data to regional and national percentages.
A multilevel logistic regression model (with survey individuals nested in local authority districts) is estimated on the survey data in order to obtain model predictions of each of these three outcome measures, including the random effects. These give information on the between-local authorities variations. In order to fit our models, key socio-demographic variables that are able to explain the spatial variability of our indicators are chosen from other data sources. These are: age, whether the person suffers from long-term health problem or disability, approximated social grade, ethnicity, index of multiple deprivation, and a variable indicating whether the person lives in a one-person household or not. These explanatory data need to be good predictors of food insecurity, also available for local authorities (see Step 2), and not highly correlated with other explanatory factors in the model.
Step 2: Making the local estimates
Second, the local estimates are made for each of the outcomes and for each local authority in the UK. The estimated coefficients from the models are applied to the same set of covariate/explanatory variables at the local level to provide synthetic estimates of food insecurity. These local authority covariate data come from a range of Census and administrative data sources.
These are then combined with direct survey-based estimates, which are obtained using the available survey weights, taking into account the sampling design. The rationale for using such a combination of estimators is to optimise the final estimates in terms of the minimisation of their bias and variance when these are compared with either the survey-based or synthetic estimates separately. Whilst the direct survey-based estimates are unbiased, they are highly unreliable for small sample sizes (large variance). They also cannot be produced when local authorities have zero area sample sizes. In contrast, the synthetic estimates have lower variance but are biased. In order to optimise their combination, when the direct and synthetic estimates are combined, more weight is given to the direct survey estimates when the sample size is large (and, hence, their variance is lower) and, on the contrary, more weight is attached to the synthetic estimate when the area sample size is small (and, hence, its variance is larger).
In order to evaluate the goodness of our small area estimates, a series of model diagnostics validation is carried out regarding model assumptions, and the comparison of the synthetic estimates against the direct survey estimates locally. All checks perform satisfactorily.
Step 3: Constraining the local estimates to known regional totals
Third, to maximise the robustness of the local estimates, they are next constrained to known regional totals. These regional totals of the three food insecurity measures are weighted direct survey estimates taken from the survey. Since they are at the regional level, they have large survey sample sizes and, therefore low variance (i.e. relatively narrow confidence intervals) in addition to being unbiased. These regional estimates are therefore treated as 'true' for the purposes of the constraining step. The constraining step acts to ensure that the aggregated total of the local authority estimates sum to its regional total. It does this by firstly calculating the ratio between the direct survey estimate of food insecurity in each region with the sum of our small area estimated local authority food insecurity across local authorities in the same region. It then inflates/deflates our local authority small area estimates accordingly such that their regional sum comes to equal the direct survey estimate for the region. For example, after the constraining step, the direct survey estimate of the number of food-insecure households in Scotland equals the sum of our small area estimated number of food-insecure households in each local authority district in Scotland.
Step 4: Creating the confidence intervals
It is important that users of small area estimates are aware that they are estimates and, as such, that they come with confidence intervals that reflect the degree of expected uncertainty around their central 'most likely' point estimate. In small area estimation research, it is common practice to use the mean squared error to produce the measures of uncertainty, and we follow this approach. We adopt a bootstrap approach to do so.
The UK LGFI Outputs
There are three final estimates that result from the estimation: one for each measure of food insecurity across every local government area in the UK. These indexes are derived by ranking local government areas according to the percentage estimate and then to be grouped into quintiles. Because the estimations use a harmonised index of multiple deprivation, it is possible to compare the estimated percentages of food insecurity across the four nations of the UK. These estimations are available as a download.
For those interested in seeing the overall patterns of food insecurity across the UK an interactive map is available. Users are able to turn on and off the individual layers to investigate patterns of food insecurity across the three measures or leave all the levels visible to get an overall visual impression. We highlight that there is evidence of food insecurity in every locality across the UK, however, there are some areas that are more food secure compared to others. Across the whole of the UK, approximately one-third of local authorities fall into the two quintiles with the least food insecurity in all three measures. These are largely concentrated in the east of England. Conversely, more than half of the localities have at least one measure of food insecurity that ranks among the highest two quintiles, with high concentrations of these being in Wales, Northern Ireland, the North of England and the Southwest of England.
For those interested in a particular local authority, the map is zoomable, and you can click on the area to determine the estimated percentages across all three measures.
**Please note, the measures are not additive, and we point you to the Food Foundation reports to understand national and regional levels.
Although we plan to apply the estimations to ongoing work, we recognise that it could be of considerable benefit to other researchers, policymakers, and analysts and have thus made the estimations freely available via the Food Foundation. We encourage researchers and policymakers to:
Consider ways to provide support that extends beyond addressing the immediate food needs of the severely food insecure and use these data as a benchmark to demonstrate improvement over time.
Use the data to inform local and national policy debates that have implications for either exacerbating food insecurity or increasing food security across the UK.
The research team would be delighted to hear how the UK LAFI estimations are being used and can be contacted via our emails above. Specifically Angelo Moretti and Adam Whitworth can offer guidance regarding the estimation and Megan Blake can provide insight into food security and insecurity in the UK, its causes and effects as well as potential solutions at the local authority scale.
If you would like further information about the Food Foundation data, please contact firstname.lastname@example.org.
An excel file of the estimates is available upon request.
The University of Sheffield copyright and database rights reserved 2021