Web Appendix B: Detailed Methods

Size: px

Start display at page:

Download "Web Appendix B: Detailed Methods"

Gertrude Simmons
5 years ago
Views:

1 Web Appendix B: Detailed Methods Data Sources and Variable Description Trussell Trust Food Bank Data We obtained data directly from the Trussell Trust covering the fiscal years 2006/07 to 2013/14. All Trussell Trust food banks are required to collect data on the number of food parcels they distribute. The number of people in the household is recorded, which is used to calculate the number of people who receive food from a food parcel. A household can receive a food parcel more than once, and the number of unique households and individuals receiving food parcels is not tracked. Postcode information was used to link a food bank to a local authority using the Office of National Statistics postcode database. Data from individual food banks were merged to the level of local authority in the following ways: Food parcel distribution data for multiple food banks operating within a local authority were summed. The year the first food bank opened was taken as the initiation year for the local authority. The number of years with a food bank was calculated from the year the first food bank began distributing food parcels in the local authority. The number of food banks operating was the number of Trussell Trust food banks distributing food in a given year in the local authority. Local Authority Budgetary and Socio-Economic Data We collected data for the fiscal years from 2006/2007 to 2013/2014. Data for lower tier local authorities in England were collected from the Department for Communities and Local Government (DCLG) 1, and data for local authorities in Scotland and Wales were compiled from their respective Local Government Finance Statistics repositories 2 3. In England, some types of spending occur at the level of upper tier local authorities; this spending was divided to lower tier local authorities by population size. Spending at the local authority level covers central services, adult and child social care, leisure and cultural services, environmental and regulatory services, housing (excluding housing revenue account and capital spending on housing), planning and development services, and transportation services. We did not include spending on education, fire, or police. The latter two categories primarily do not occur at the level of local authorities, and spending on education has been changing in England with the introduction of Academies, whereby money for education no longer goes through local authorities in some areas. Spending on central welfare benefits going out to local authorities was obtained from the Department of Work and Pensions (DWP) 4. It covers spending on most income support benefits in the UK, with the exception of child benefit and tax credits. We did not include

2 spending on pensions. Further, data on council tax benefit spending was not available from the DWP for the year Spending on this benefit was therefore not included in total benefit spending for any of the years when 2013 data was included in the analysis. Data on the number of people claiming benefits administered by the Department of Work and Pensions were obtained from Nomis, a database of labour market statistics, as were unemployment rates for local authorities, which are based on estimates from the Annual Population Survey 5. Data on the total number of sanctions applied to Jobseeker s Allowance claimants were obtained from the DWP database, Stat X-plore 6. Adverse sanction decisions refer to the number of claimants that experienced their benefit payment being terminated for a period because they did not meet conditions for benefit receipt. 7 This variable was divided by the number of Jobseeker s Allowance claimants to obtain the rate that sanctions were applied per claimant over the year. Gross Value Added, a measure of the value of goods and services produced in small areas, measures total production of goods and services by residents and corporations for small groups of local authorities (Nomenclature of Territorial Units for Statistics Level 3) 8. All economic data were adjusted for inflation using the GDP deflator. The proportion of individuals identifying as Christian by local authority came from the 2011 Census; data was available from the Office of National Statistics and Scottish Government Analytic Samples The analytic samples were based on lower tier local authorities in England, Scotland, and Wales. We collected data from 375 of the 380 local authorities, excluding Eilean Sir, Shetland Islands, City of London, Isles of Scilly, and Orkney Islands due to their small population sizes. In these 375 local authorities, we tracked rates of Trussell Trust food bank initiation, highlighted in Figure 1. We highlight that in 2009/10, there were 29 local authorities with a food bank operating in 2009/10. By 2013/14, food banks had been initiated in an additional 222 local authorities, leaving 124 local authorities where Trussell Trust Foodbanks had not been initiated.

3 Table B1 Presence of Trussell Trust food banks in UK local authorities by year. Trussell Trust Foodbank in Local Authority Year No Yes 2006/ / / / / / / / Sample of Local Authorities for First Analysis: Where are food banks opening? In our first analysis, we restricted the sample to local authorities where there was no food bank present in 2009 (n=346). After a new food bank was initiated in a local authority, subsequent local authority-years were censored. There were a total of 222 food banks initiated and 124 local authorities where they were not initiated. Table B2 Analytic sample for cross-local logistic regression model for initiation of first food bank in local authorities. Total local authorities in analysis for year after censoring New food bank in Local authority and with data on unemployment Total local authorities New food bank in Local authority Year No Yes No Yes Total , ,071 The final analytic sample included the following number of local authority years, after excluding 49 observations missing data on unemployment in the previous year and censoring after a first food bank was initiated. We provide a table of descriptive statistics based on this sample of local authority-years (Table A2 of Web Appendix 1). Sample of Local Authorities for Second Analysis: Why are food banks distributing more food? For this analysis, we examine the rate of food parcel distribution in local authorities that had food banks operating. This included the 29 local authorities with a food bank operating in

4 2009 and the additional 222 local authorities that had a food bank initiated from 2010 forward. Data on food parcel distribution was not available for some local authorities with food banks for some years, which reduced the number of observations in some years, as shown below (Table B3). Table B3 Analytic sample for cross-local linear regression model for food parcel distribution in local authorities with food banks. Data on Food Parcel Distribution Food Parcels Distributed (per 100 persons) Trussell Trust Foodbank in Local Authority Year No Yes n Mean SD 2010/ / / / Total An additional 17 local authorities were excluded from the analysis because they were missing data on unemployment for a given local authority-year, giving a total 575 observations included in the analysis. We provide table of descriptive statistics for this sample of local authority-years (Table A3 of Web Appendix 1). Statistical Analysis We used logistic regression to examine the odds of a first Trussell Trust food bank being initiated in a given local authority in a given year. First, we examined whether there was evidence that food banks have opened in areas that had greater economic hardship, austerity measures, and/or sanctioning rates, as indicated by unemployment, Gross Value Added, the yearly reduction in spending on central welfare benefits, the yearly reduction in spending on local authorities services, and the average rate of sanctions applied to Jobseeker s Allowance claimants. In addition, we controlled for the percent of the population identifying as Christian because Trussell Trust food banks are initiated by Christian church groups, and therefore, this area-level characteristic could also be related to the likelihood that a Trussell Trust food bank opened in a given area. We modelled these relationships using a lagged variable approach, hypothesizing that it would take time for a individuals in a given local authority to mobilize sufficiently to initiate a food bank in response to recently implemented funding cuts and economic conditions. Because the ability to run a food bank in a given year could also be determined by the local economic conditions given the reliance on donated labour and food, we also included a contemporaneous measure of Gross Value Added in the models. We tested the fit of these models by examining joint F-tests when variables were included for only one year prior and when included for both one year and two years prior. We selected the most parsimonious model.

5 We used the predictive margins command in Stata to provide us with estimates for the predicted probability of a food bank opening based on our model when there was no cut in local authority spending in the prior year or two years prior, and then report the predicted probability of a food bank opening when there was a 3% cut in the one year prior and two years prior. To examine determinants of food parcel distribution in a given area, we used a linear regression model to examine what predicted the level of food parcel distribution in a given local authority for a given year where a food bank was operating. In addition to the variables outlined above, we also included indicators of supply-side factors given the argument that it is increasing number of food banks and aggressive food bank outreach that has driven the rise in the number of people receiving food parcels. Here, we selected the number of years a food bank has been operating in a local authority as an indicator of operational capacity, as the longer it has been open, the more time there is to establish connections with local businesses and care professionals, the former being related to the supply of donated food available to give away, and the latter influencing the number of vouchers given out. Years of operation could also influence food bank use because it gives more time for a food bank to become known in a community. The number of food banks was used as an indicator of the supply of food banks, which allowed us to the test the assertion that where there are more food banks, more food parcels will be given out. This variable was included on a per capita basis. Standard errors were estimated using Huber-White standard errors to adjust for the nonindependence of local authorities in the sample. In sensitivity analyses, we included other measures of economic hardship for an area in our models, namely, the rate of Jobseeker s Allowance claimants per working age adult and child poverty rates. We selected the most parsimonious models for presentation in the main text, and adjusting for these variables did not change our findings.

6 References for Web Appendix B 1. Department of Communities and Local Government. Local authority revenue expenditure and financing. London: gov.uk, The Scottish Government. Scottish Local Government Financial Statistics, Welsh Government. Local authority revenue outturn expenditure, Department for Work and Pensions. Benefit expenditure by local authority from 2000/01 to 2013/14. London: gov.uk, Office for National Statistics. Nomis: Official Labour Market Statistics. Durham: Durham University, Department of Work & Pensions. Jobseeker's Allowance and Employment and Support Allowance sanctions. London: GOV.UK, Watts EE, Fitzpatrick S, Bramley G, et al. Welfare sanctions and conditionality in the UK. York: Joseph Rowntree Foundation,, Office for National Statistics. Regional GVA NUTS3, London: Office for National Statistics, Office for National Statistics Census, Key Statistics for Local Authorities in England and Wales. London, National Records of Scotland. Religion by council area, Scotland, Edinburgh, 2013.

Evidence of rising food insecurity in UK and EU: potential drivers and the role of social protection

Evidence of rising food insecurity in UK and EU: potential drivers and the role of social protection Rachel Loopstra Division of Diabetes and Nutritional Science, King s College London Department of Sociology,