Understanding the post-2010 increase in food bank use in England: new quasi-experimental analysis of the role of welfare policy

Until around 2009–10 food banks (charitable organisations distributing free food parcels) used to be a very marginal phenomenon in the UK. However, the number of food banks increased very rapidly in the 2010s. The Trussell Trust, which support the largest network food banks in the UK, had around 35 food bank centres in 2010/11, 650 in 2013/14 and nearly 1300 in 2019/20 [1, 2]. (By food bank centres we mean individual venues. A food bank may have more than one such venue). The dynamics of the volume of food aid broadly followed this trend, with 61,000 food parcels distributed in 2010/11, 0.9 m in 2013/14 and 1.9 m in 2019/20 [1, 2]. Less is known about food banks operating independently of the Trussell Trust network (estimated to represent around 40–45% of all UK food banks, many of which are members of the Independent Food Aid Network) but one study suggests that they underwent a similar growth [3].

Reliance on food banks is a symptom of household food insecurity, particularly its severe type: around 80% of people referred to food banks in the Trussell Trust network are severely food insecure [1]. ‘Household food insecurity’ is an established, internationally used concept and is defined here as ‘a household-level economic and social condition of limited or uncertain access to adequate food’ [4]. In the UK, researchers commonly classify households into four categories: food secure, marginally food insecure, moderately food insecure and severely food insecure. Broadly speaking, ‘marginal’ food insecurity signifies uncertain access to adequate food (anxiety about adequate food supplies in the near future, for financial reasons), ‘moderate’ food insecurity indicates inability to afford a balanced diet, and ‘severe’ food insecurity indicates inability to afford enough food. (In this paper, as often in relevant literature, the term ‘food insecurity’ collectively signifies the moderate and severe levels). Food insecurity is typically measured via household surveys, with respondents asked standardized questions.

The most recent survey data prior to the Covid-19 pandemic shows that 4% of UK households were severely food insecure in 2019/20, while a further 4% were moderately food insecure [5]. While these figures regard respondents’ experiences in ‘the last 30 days’, research in the United States (US) suggests that the reported levels of food insecurity would have been much higher if the questions referred to ‘last 12 months’ [6]. As it is estimated that around 2.5% of all UK households used a food bank at any point in 2019/20 [2], it follows that a substantial proportion of severely food insecure people in the UK do not use food banks (consistent with Canadian evidence [7]). Barriers to accessing food aid may take the form of a geographical distance to a food bank, its opening hours, the necessity to obtain a referral or, in the case of The Trussell Trust, the fact that it is a Christian organisation operating largely from Christian churches, which may discourage non-Christians [8]. Furthermore, some people prefer to go without food than to experience the feelings of shame and stigma associated with relying on charitable food aid [9, 10]. However, even those who use food banks should not be perceived as automatically sheltered from negative effects of food insecurity, mainly because food aid does not address the underlying causes of household food insecurity [11], meaning that it amounts to a short-term ‘sticking plaster’. Additionally, access to food aid may not be available ‘as and when’ needed, resulting in periods of lacking food. A food parcel would also often lack fresh vegetables and fruit [12] while providing disproportionately high sugar and carbohydrate [13, 14], meaning that a balanced diet still cannot be achieved.

Since food insecurity implies a nutritionally inadequate diet, it is a phenomenon of significant importance for public health. Poor dietary intake has been linked to a number of diseases and chronic conditions, including cardiovascular disease, Type 2 diabetes, some types of cancer, and osteoporosis [15, 16]. In addition, inadequate dietary intake during pregnancy and early childhood can increase the risk for birth defects, anaemia, low birth weight, preterm birth, and developmental risk [15‐18].

There is also some evidence that diets with a high glycaemic index and load (e.g., diets containing high amounts of refined carbohydrates and sugars) may have a detrimental effect on psychological wellbeing [19]. Poor dietary intake also contributes to obesity, which is associated with many serious physiological, psychological, and social consequences for children and adults, including high blood pressure [20, 21], heart disease [22], diabetes [20, 21], pregnancy-related complications [23], decreased life expectancy [24], asthma [25, 26], depression [27, 28], and stigmatization [29, 30].

Since absence of food insecurity is a desired outcome from a public health perspective, the major increase in food bank use is of concern; some researchers have warned of a ‘public health emergency’ (e.g., [31]). It would be in the public interest if this situation attracted governmental and wider societal efforts at reversing the trend, by addressing the root causes of food insecurity (as opposed to handing out more food, which does not resolve food insecurity [11]). A prerequisite for an effective response is to understand the issue, and in particular to understand what the drivers are, the focus of this paper.

While demographic characteristics of people experiencing food insecurity (such as younger age, non-white ethnicity, low education, having a disability, being unemployed) have been studied [32], the body of research into drivers of food insecurity in the UK is very limited [2]. In contrast, several studies of drivers of food bank use have been conducted. They have tended to point at issues with the welfare system, particularly those resulting from changes introduced after 2010, such as tightened benefit eligibility, increased benefit conditionality and reduced generosity of benefit levels (e.g. [33]). Generally labelled ‘Welfare Reform’ (WR) and primarily justified by the need to reduce state deficit [34], these changes were targeted at working age population, which probably explains why people of pension age have been so under-represented among food bank users [1]. Key elements of WR include:

Factors not directly related to the benefit system, such as job loss, ill health, homelessness, domestic violence or bereavement have also been identified as drivers of food bank use, albeit (in the assessment of some researchers) of overall smaller weight than benefit issues [1, 33, 35].

The common weakness of these studies is that the evidence they provided is long way from the ‘gold standard’ of causal research, a randomized experiment. The evidence has been either qualitative in nature – and as such dismissed by policy makers as ‘not robust’ [36] – or quantitative, but limited to description of the population of food bank users, without contrasting them with non-users of food banks (and dismissed in turn as ‘self-selecting data’ that ‘can’t prove anything’ [37]). Attempts at quasi-experimental analyses have been few and usually limited to the weakest quasi-experimental design, simple balancing on covariates by means of multiple regression conducted on cross-sectional data (e.g. [38, 39]). While these analyses found statistically significant associations between certain elements of WR and food bank use, the findings cannot be described as robust due to the possibility of the effect being biased by omitted variables. Only two studies attempted to use a stronger quasi-experimental design, where ‘fixed effects’ regression modelling has been employed to evidence the causal link between benefit sanctioning and food bank use [40, 41]. Thanks to this design these two studies neutralised bias from time-invariant omitted variables (see below). However, benefit sanctions are just one factor on a long list of potential drivers of food bank use; there is a need to investigate those other factors using as strong quasi-experimental design as possible.

Table 2 presents results of the main FD model. The outcome variable was the number of food parcels distributed by the food banks in the Trussell Trust network per 1000 working age (WA) population in a financial year.

The first predictor in Table 2, the number of the food banks in the Trussell Trust network per 1000 working age population, controls for the size of the potential ‘supply’ of food parcels. The coefficient means that on average, an increase in the number of food bank centres by one in the local authority is associated with 358 additional food parcels per year, an 8% year-on-year increase in a typical local authority (relative to the 2019/20 level). As an increase in the number of food bank centres by one means, on average, an increase of 28% in the number of food bank centres,¹ the effect of this additional ‘supply’ is much less than proportional.

The next predictor, unemployment rate, shows that a one percentage point higher unemployment would have led to 0.85 more food parcels per 1000 working age population, equivalent to an extra 107 in a typical local authority, a 2% increase in the number of food parcels distributed. This suggests that food bank need was not very strongly driven by unemployment in this period.

The following predictor is the standard allowance of the main income replacement benefit (currently UC, previously Jobseeker’s Allowance), adjusted for inflation. The coefficient is negative, meaning that an increase in the real value of this allowance is associated with a decrease in the uptake of food parcels: the more income benefit claimants have, the less they use food banks. A £1 per week increase in UC standard allowance (in 2011 pound value) was associated with an annual decrease of 2.6%, or 118 food parcels in a typical local authority.² This provides evidence that the real terms reduction in the basic working age benefit allowances has been an important factor in increasing food bank need.

The next predictor refers to the proportion of working age population who are in receipt of out-of-work benefits, that is UC (out-of-work parts only), Jobseeker’s Allowance, Employment and Support Allowance, or Income Support. One percentage point more of the working age population on such benefits would have reduced food parcels by 2.7 per 1000 working age population, 291 per year in a typical local authority, or about 6.5% of the 2019/20 level. While the coefficient’s negative sign may seem somewhat unintuitive, one needs to consider that some people are excluded from the welfare safety net because of their immigration status, while some others struggle to apply for benefits or to sustain the benefit claim [1]. Having stable benefit income usually protects from destitution and food bank use.

The following predictor is an interaction of the preceding two. Although it is not statistically significant, it is included in the model for a substantive reason: the value of the main out-of-work benefit will have stronger impact in local authorities with a larger proportion of people on out-of-work benefits. The interaction has the expected negative sign.

Next, an increase of 10 percentage points in those receiving UC as proportion of all claimants of working age benefits, was associated with an annual increase of 3.6 per 1000 in the number of food parcels, 454 in a typical local authority, an increase of 8.4% on the 2019/20 level. This indicates that the rollout of UC has been a major driver of increased food bank need.

An increase of 100 in the number of benefit sanctions was associated with an annual increase of 24 in the number of parcels.

Lastly, an increase in the number of households subject to ‘bedroom tax’ of 100 was associated with an annual increase in the number of food parcels of 46.

Overall, these predictors explain approximately 31% of the variance of annual increase in food bank use. While this is a respectable figure by social research standards, clearly the model does not provide a complete picture of drivers of food bank use, although we tested a wide range of variables.

Model diagnostic procedures have shown that the model fit is satisfactory (see Additional file 4 for details).

In the situation where conducting a randomised control experiment into the role of welfare policies in driving food bank use is impossible, findings presented in the previous section constitute, in our view, robust quasi-experimental evidence that four aspects of the welfare system partly drove the uptake of food parcels in the UK between 2011/12–2019/20. These elements were: the reducing value of the standard allowance received by claimants of out-of-work benefits; benefit sanctions; the roll-out of UC; and ‘bedroom tax’ in social housing. The finding about benefit sanctions independently verifies the result of two earlier studies [40, 41], while the finding about the other three aspects is an original contribution of this study.³

The implications of this evidence are three-fold. On an evidential level, it strongly implies that the current welfare system in the UK is partly responsible for generating food bank use. Since 2010, the structure of this system has been deliberately changed to provide overall less support to working age claimants [52]. While this has been primarily justified by the need to reduce state deficit, a more ideological agenda focused on the need to reduce apparent welfare dependency and incentivise paid employment has also arguably been central to the aims of the reforms [34].

Secondly, on the level of social and political action directly relevant to the nutritional condition of the population this evidence means that in order to reduce food bank use, the UK would need to reform or amend welfare policies that have been evidenced to drive it. The evidence presented here can be exploited to prioritise reforms, by identifying effects of different potential reforms on the uptake of food parcels. We have carried out additional policy forecasting [53] - based on the main model and assuming the Covid-19 pandemic has not happened - which indicated that retaining the £20/week uplift to UC (introduced in April 2020 and ended in October 2021) would result in the largest reduction of the uptake, followed by suspending the ‘bedroom tax’, suspending benefit sanctions and finally halting the intake of new claimants onto UC. The intake of new claimants onto UC obviously cannot be halted, but the most destitution-generating aspects of UC can be reformed (the initial five-week wait in particular, but also debt recovery arrangements and problems with applying online; see [48, 49]).

It needs to be remembered, however, that the list of social policies creating demand for food parcels might not be limited to the four measures evidenced here. A large volume of qualitative and descriptive quantitative evidence exists pointing at the role of PIP assessments, Work Capability Assessments, low LHA rates, Benefit Cap, and the two-child limit (among other policies) in driving food bank use (and destitution in general; see [49]). Some of these factors have been explored in the modelling but dropped from our final model due to lack of statistical significance. Lack of statistical significance results either from insufficient volume of data (while the effect is genuine) or due to there not being any effect. The above-mentioned existing evidence suggests to us that in this case the lack of significance could be due to insufficient amount of data.

Thirdly, the evidence in this paper provides support for ‘Health in All Policies’ approach to policymaking. Health in All Policies is an approach to all public policies (whether health related or not) that systematically takes into account the health implications of decisions, seeks synergies, and avoids harmful health impacts in order to improve population health and health equity [54]. The rationale for this approach lies in recognition of the fact that ‘health, wellbeing and health inequalities [ …] are largely determined by living conditions and wider social, economic, environmental, cultural and political factors. These in turn are controlled by policies and actions outside the health sector’ [55]. Therefore, policymaking that is not explicitly health-focused but has potential negative health implications needs to be aware of these implications and take them into account in the cost-benefit analysis of the prospective policy. The fact that the Westminster Government’s own impact assessment of the ‘bedroom tax’ states that the policy would not have impact on health and well-being [56] is a case in point.

Food bank use is a seriously concerning phenomenon from the public health perspective. The vast majority of food bank users are severely food insecure, which signifies inadequate diet. Inadequate diet has been linked to a range of negative physical and mental health outcomes. Hence, the increase in in people using food banks over the 2010s calls for an action to alleviate the level of severe food insecurity in the population, as manifested by food bank use. This however requires sound, evidence-based understanding of factors that have been driving food bank use in the UK in the last decade. This paper provided quasi-experimental evidence that four aspects of the welfare benefit system have been driving food bank use: falling value of welfare benefits, the structure of UC, the ‘bedroom tax’ in social housing, and benefit sanctions. These findings mean that one way to improve food security would be to reform the welfare benefit system through a combination of increasing the minimum value of benefits, reshaping UC (particularly shortening the initial waiting period), suspending or lessening the ‘bedroom tax’ and suspending or lessening benefit conditionality. Going beyond the immediate focus of this research, the findings also suggest that health inequalities in England could be reduced, and health outcomes improved, if ‘Health and Health Equity in All Policies’ approach was adopted by policymakers.