Annex 2: Methodology
This section provides key information on the methodology used to produce persistent poverty statistics. A more detailed methodological paper is available from the DWP website: https://www.gov.uk/government/statistics/income-dynamics-experimental
The figures in this publication are derived from the Understanding Society survey. Understanding Society is a large scale longitudinal survey that captures information about people's social and economic circumstances, attitudes, behaviours and health. Being longitudinal, the same individuals are interviewed each year allowing identification of those who have been in poverty over a number of years rather than just at a single point in time.
Comparison with other sources
Poverty estimates presented in the National Statistics Poverty and Income Inequality in Scotland publication come from a different source - the Department for Work and Pensions' ( DWP) Households Below Average Income dataset which is produced from the Family Resource Survey. This is the best source of household income data available in the UK. However it does not track individuals or households over time and so cannot be used to calculate persistent poverty rates.
The FRS and Understanding Society use different income definitions and cover different time periods and so figures which come from the two surveys are not comparable with each other. It should also be noted that an individual can be in persistent poverty without being in relative poverty in the most recent year (if they were in relative poverty in the 3 previous years) and so those in persistent poverty are not simply a sub-group of those in relative poverty.
This publication presents analyses on two bases: Before Housing Costs ( BHC) and After Housing Costs ( AHC). This is principally to take into account variations in housing costs that themselves do not correspond to comparable variations in the quality of housing.
This publication presents persistent poverty rates for 2 overlapping periods: 2010-2011 to 2013-2014 (referred to as 2010 - 2014) and 2011-2012 to 2014-2015 (referred to as 2011 - 2015).
Persistent poverty statistics are based on tracking an individual over a 4 year period. Each set of results are therefore based on 4 waves of the Understanding Society survey. This publication presents persistent poverty statistics based on waves 2-5 and waves 3-6. Each wave of interviews is conducted over a 2 year period as shown in the table below.
An individual is in persistent poverty if they are in relative poverty for at least 3 years in any 4 year period. This means that the same individual can be in persistent poverty in both, either or none of the time periods covered in this publication.
|Wave||Start Year||End Year||In 2010-2011 to 2013-2014 statistics?||In 2011-2012 to 2014-2015 statistics?|
There are known issues with the income information in the first Understanding Society wave covering 2009-2010. See Dr Paul Fisher's paper Does repeated measurement improve income data quality? ( ISER Working Paper Series, 2016-11) for details of why income data on the first wave of Understanding Society are not comparable with subsequent waves and are likely to be of lower quality. The first wave has therefore been excluded from any analysis presented in this publication.
Understanding Society is a survey of private households (although it does collect information from households about their children if a child has moved into an institution). This means that people who were in residential institutions, such as nursing homes, barracks, prisons or university halls of residence at the start of the survey are excluded from the scope of the analysis presented here.
Reliability of estimates
The figures are estimates based on sample surveys and are therefore subject to sampling variation. Caution should be exercised in the interpretation of small year-on-year fluctuations.
As with most longitudinal surveys, attrition reduces the Understanding Society sample size over time. As well as attrition reducing the sample size, we have missing data for many of the variables we are using in the analysis. We exclude individuals with missing data from relevant analysis, but include individuals whenever we can. Weights have been applied which adjust for unequal selection probabilities, differential nonresponse, and potential sampling error.
Whether an individual is counted as a child, working age adult or pensioner is determined by their age during the first survey period. So, for example, an individual aged 15 in 2011 and aged 19 in 2015 will be counted as a child for the 2011-2015 period.
Email: Andrew White