Estimating the risk of food poverty in England: using administrative data to address public health concerns

Transcription

1 Estimating the risk of food poverty in England: using administrative data to address public health concerns 16 September 2015 Dianna Smith Kirk Harland, Claire Thompson, Nicola Shelton, Storm Parker

2 Outline Lack of measurement of food poverty in the UK All-party parliamentary group: terms of reference Food Poverty Risk Index: two indicators of local population risk Classify areas using Census and benefits datasets, easily updated and accessible

3 Food poverty Defined: not clear in the UK Recent review finds a variety of measures Household Food Security Scale, subset USDA Food Sufficiency Indicator Radimer/Cornell scale Inconsistent measures in UK Parker 2013

4 Looking abroad Food security measured annually in the US Via USDA questionnaire Food banks run by a series of charities No referral required Also WIC, food stamps

5

6 What we don t know Who is at risk, or is currently in food poverty in England? No systematic data collected on individuals experiencing food poverty Food bank questionnaires not representative of the population who are food insecure (20-30%?)

7 Demand for emergency food Data from Trussell Trust : 61, : 128, : 346, million in food poverty in the UK? Low-income consumers buying less fruit, veg Poorest households spend 23% of income on food, wealthiest spend 4%

8 Guardian readers sent in known independent food banks to contribute to the Trussell Trust dataset Also FareShare, London Food Atlas Question: do people in need of emergency food have knowledge of, access to these databases/maps/ resources?

9 Profiling households at risk Qualitative data provides insight Educated assumptions are added to what is known Recent changes to referral forms hinder further data gathering Defining food poverty Below 60% of the median income? Skipping meals? Not able to access food in a socially acceptable way

10 Kneafsey et al 2012 Risk factors for food poverty

11 Considerations Household types at risk Low income Dependent children Claiming benefits Pensioners Possible colinearity between counting low income HH and those claiming benefits

12 Indicator 1: HH type risk Using data from 2011 Census Two broad groups: Working age and children (0-64) Low income households with dependent children Pensioners (65+) Single pensioners

13 Indicator 1 con t No cross-tabulation in the Census for lowincome households with dependent children Where low income is social grade D or E Combined data from KS tables, assumed independence

14 Calculating HH risk Estimated risk in households using 2011 Census KS105EW: Household composition QS611EW: Approximated Social Grade p (Ch,LI) = probability of a person in an MSOA to be in a household with dependent children and on a low income as defined by the occupation of the HRP

15 Indicator 1: Results Proportions of the total population at risk of food poverty by 2011 MSOA (n=6701) Mean 17.34% (2013 estimate) St Dev 6.9, 95% CI: Range % Grouped into quintiles of risk where 1 is lowest and 5 is highest proportion of the resident population at risk

16 Based on Indicator 1: household type, composition Quintiles of MSOA populations (%) deemed at risk based on 2011 Census data

17 Risk with locations of Trussell Trust food banks (2013 data)

18

19 Why not use the IMD? Strong correlation between household risk profile and 2010 IMD score (0.810, p<0.000) BUT This is expected given the IMD domains, and the timing of the data

20 Indicator 2: Benefits claimants Using data from Department for Work & Pensions ( ) Working age benefits claimants (16-59 for females, for males) Pensioners (60+ for women, 65+ for men) Rounded count data available by LSOA (2001 boundaries) Mid-year population estimates from ONS provide denominators*

21 Indicator 2 con t Adjusted DWP data for rounding Aggregated to 2011 MSOAs Calculated indirectly standardised crude rates for each year Change from each year to next, and from 2008 to each year using May data Compare magnitude of change between areas using 1-3 standard deviations as cut points

22 Indicator 2: Results Mixed patterns of changes to benefits claims by area Skew towards areas with high proportions of people claiming benefits Over time, areas moving towards high or low proportions rather than towards the middle

23 Indicator 2: Changes over time One-third of MSOAs have an increase in the proportion of pensioners claiming benefits from 2008 to 2013 Most MSOAs have no significant change in the proportion of working age population claiming benefits over the same time

24 Combining indicators Using the quintiles from Indicator 1, identify MSOAs in highest risk (Q5) which showed change > 2 SD (Very High) Results will identify areas at greatest estimated risk of food poverty in 2013 Also those areas in Q5 from Indicator 1 AND with an increase in benefits claimants from

25 All areas that are both high risk by area population profile AND >2 SD above average proportion of benefits claimants*

26 Areas with highest risk based on area household profile & change in benefits claims*

27 Limitations The benefits analysis does not capture information on benefit sanctions Need to consider local context: high cost of housing, lack of public transportation; this is a global model

28 Conclusions Some areas in the Northeast may be at higher risk of food poverty Similarly, areas of London have populations with higher risk measured this way Could be used as a local campaign tool for charities or local authority public health teams