Official Statistics

Income Dynamics: background information and methodology

Published 23 March 2023

Section 1: Purpose and context of the statistics

Income Dynamics (ID) contains analysis of income movements and the persistence of low income for various population groups.

Reporting on the percentage of children in the United Kingdom who live in households whose income has been less than 60% of median equivalised net household income, in at least three of the last four survey periods, is required under section 4 of the Welfare Reform and Work Act 2016, and, regarding Scotland, the Child Poverty (Scotland) Act 2017. Definitions for relevant key terms in these Acts are consistent with those given in the Glossary of this document. Data for reporting against the other three measures are available in the Households Below Average Income (HBAI) publication.

ID has been published annually alongside the HBAI series since 2017. While HBAI remains the best source of cross-sectional low income statistics, ID provides a longitudinal perspective. It fills a key gap that emerges if low income statistics are only considered on a cross-sectional basis.

By looking at low income longitudinally, we can look at durations of low income for different groups and what characteristics are associated with longer spells of low income. Longitudinal analysis also allows us to track how many individuals – and which individuals - enter and exit low income, and to calculate entry and exit rates. ID also explores the relationships between certain income, employment and demographic events, and low income entry and exit.

Potential users include:

• policy and analytical teams within the DWP
• the Devolved Administrations and other government departments
• local authorities
• Parliament
• Academics
• Journalists
• and the voluntary sector

The Department for Work and Pensions’ (DWP) responsibilities include understanding and dealing with the causes of poverty rather than its symptoms, encouraging people to work and making work pay, encouraging disabled people and those with ill health to work and be independent, and providing a decent income for people of pension age and promoting saving for retirement. The extent of progress towards these responsibilities will affect these statistics.

Changes for this publication

ID uses data from the Understanding Society survey. Understanding Society (referred to from now on as ‘USoc’) is a longitudinal survey, run by the University of Essex. Data collection is continual and takes place via survey ‘waves’, each of which is carried out across two calendar years. For its annual publication, ID uses the most recent USoc data release, which includes data from the most recent survey ‘wave’.

The following changes have been made to the ID 2010 to 2021 publication:

• A change to the sample used for analysis: USoc’s Immigrant and Ethnic Minority Boost (IEMB) sample has been included in ID for the first time; linked to this, we have made a change to the time period for part of our income mobility analysis
• The addition of two new ‘events’ to our analysis of events associated with low income entry and exit
• A new income sub-component has been included in the Private Benefit income component: this is due to a new category on income from Student and/or Tuition Fee Loans having been included in the USoc survey

These changes are discussed in more detail in this section and elsewhere in this report where relevant.

In addition, the coronavirus (COVID-19) pandemic continued to affect USoc fieldwork and data collection during the most recent survey period (Wave 12). Implications of the pandemic are also discussed in this section.

The inclusion of the Immigrant and Ethnic Minority Boost (IEMB) sample

As noted above, ID uses data from the Understanding Society (USoc) survey. USoc has been running since 2009 using a household panel design (see Section 2 for detail on the survey design). Over time, longitudinal surveys tend to experience attrition in sample size. This can affect quality because it can mean that they become less representative of the population they are aiming to describe. In addition, a survey which follows the same sample members over time cannot fully reflect the effect of immigration on that same population.

It is for these reasons that USoc launched an Immigrant and Ethnic Minority Boost sample (IEMB) in 2014 to 2015, which resulted in the addition of around 2,900 households to the survey. The target population groups for the IEMB were people born outside of the UK, and people who considered themselves or their parents or grandparents to be of Indian, Pakistani, Bangladeshi, Black Caribbean or Black African origin (the five largest established minority ethnic groups in the UK).

When ID was first published in 2017, the IEMB sample was not included, because persistent low income estimates require data from four consecutive years. This was not possible given the date of the IEMB launch. All subsequent ID statistical publications have, similarly, excluded the IEMB.

Because of the benefits to data quality, we initiated feasibility work in 2022 to explore the implications of integrating the IEMB. Early discussions with the University of Essex however, led us to understand that from the Wave 12 data release (the data on which the current publication is based), USoc would cease releasing the cross-sectional weights that we had previously used, and that we would need instead to use cross-sectional weights designed for use with the IEMB sample. We therefore focussed on integrating the IEMB sample for this 2010 to 2021 publication.

This change has improved the representativeness of ID statistics and enables us to be more confident regarding our estimates, particularly those based upon ethnic minority sub-groups, which have also benefitted from larger sample sizes and less suppression. The addition of the IEMB sample builds on steps already taken by ID (detailed here) to respond to recommendations made by the Office for Statistical Regulation (OSR) Review of income-based poverty statistics.

Users should note that we have revised our analysis of quintile movements (income mobility) to include the IEMB sample. Rather than using 2010 to 2011 as its starting point, quintiles analysis now starts from 2015 to 2016.

Further information on this change as well as on how we have incorporated the IEMB sample can be found in Section 2 below.

Events associated with entries into and exits from low income

We have introduced new analysis to Section 10 of the ID Summary report where we explore events associated with entry into and exit from low income. The following two new events allow us to further investigate the role that paid work plays in low income entry and exit:

• a change in the number of full-time workers in a household where the number of workers (as well as household size) remains the same. This aims to capture what happens when there is a transition from full-time to part-time work, and whether this is associated with low income entry; and when there is a transition from part-time to full-time work, and whether this is associated with low income exit
• a household changing from one in which there is paid work to one where there is no paid work or the other way around. A transition to a non-working household is explored in relation to low income entry, and a transition from a non-working to a working household is explored in relation to low income exit. Household size is kept constant for this event

The inclusion of Student and Tuition Fee Loans as an income sub-component

For the most recent survey period (Wave 12), USoc introduced a new sub-category to record income from Student and/or Tuition Fee Loans. This sub-component is classed as Private Benefit income. See Section 3 for further information.

The coronavirus (COVID-19) pandemic

The coronavirus pandemic and subsequent policy announcements resulted in several changes to USoc fieldwork, data collection and processing, affecting Wave 11 (2019 to 2020) and Wave 12 (2020 to 2021). Their impact on the USoc survey has been documented by the USoc team at the University of Essex, and relevant material is drawn on here to summarise the key points and what they mean for this ID publication (see also last year’s Background information and methodology report for discussion regarding the implications for Wave 11 specifically):

• the first national lockdown due to the coronavirus pandemic began in March 2020, which was month 3 of the 24-month Wave 12 survey period
• at this point in time, around two-thirds of USoc data collection was web-based, with around a third completed via face-to-face interviewing, and about 1% being conducted via telephone. Following lockdown, data collection for all respondents was swiftly moved online in the first instance, with telephone follow-up used as necessary. Although the remainder of 2020 saw varying degrees of restrictions regarding social mixing, which also continued into 2021, USoc maintained a web-first approach throughout the remainder of Wave 12
• fieldwork figures show that USoc response rates were largely protected over this period: the response rate for April to December 2020 was just 1.5 percentage points lower than that of the same period in 2019
• ID has therefore been able to release a full set of tables with no additional detriment to sub-samples beyond the attrition that is usual for this type of longitudinal panel survey
• while USoc response rates held up well, there were small changes in the sample profile in 2020 compared to 2019, as some groups – including those in higher age groups, those living alone, those not in work, and those with lower levels of education, were less likely to respond to the survey. USoc weights were adjusted to account for these changes
• in terms of the USoc questionnaire, a response option to capture the ‘furlough’ status of employees on the Coronavirus Job Retention Scheme was added to the economic status variable ‘jbstat’ from the end of July 2020. The variable ‘jbstat’ is only used by ID for analysis on the associations between various ‘events’ and low income entry and exit. This is because it offers a greater degree of detail than the economic status variable ‘employ’, which ID uses elsewhere (e.g. in persistent low income breakdowns). Our code was amended for last year’s publication to take furlough status into account, and this remains the case. Please see Section 3 for full details
• furlough payments made to employees during 2020 and 2021 were received via paycheques and are included as earnings rather than state benefits. This did not result in any revisions to our code. ID explores changes in earnings and other income components as part of its events analysis: the uptake of furlough is likely to have had little effect on these events. Please see Section 3 for full details
• a question asking about receipt of grants via the Self-Employment Income Support Scheme (SEISS) was introduced to the survey at the end of July 2020.
• Section 3 discusses the degree to which the coronavirus pandemic and changes discussed here may have affected ID measures. It is important to note that while we might observe some shifts in ID estimates, directly attributing observed changes to the pandemic is beyond their scope

For further information, please refer to:

• the USoc Main Survey Data and documentation
• USoc (2022) Understanding Society Main Study Changes due to the COVID-19 Pandemic (Wave 12 release); and
• Alvarez, Burton & Lynn (2021) COVID-19 and Mode Selection Effects in Understanding Society

Section 2: Source of the statistics

These statistics are derived from the Understanding Society (USoc) survey. USoc is an initiative funded by the Economic and Social Research Council and various Government Departments, with scientific leadership by the Institute for Social and Economic Research at the University of Essex, and survey delivery by NatCen Social Research and Kantar Public. The research data are distributed by the UK Data Service.

We have used the following dataset:

University of Essex, Institute for Social and Economic Research. (2022). Understanding Society: Waves 1-12, 2009-2021 and Harmonised BHPS: Waves 1-18, 1991-2009: Special Licence Access. [data collection]. 16th Edition. UK Data Service. SN: 6931, DOI: 10.5255/UKDA-SN-6931-15

We do not publish the dataset that we derive to produce ID statistics. Please contact us if you require further information about how we process and analyse USoc data.

We would like to thank Olena Kaminska from the University of Essex for her advice regarding the incorporation of the IEMB sample into our analysis. We would also like to thank Paul Fisher and Raj Patel at the University of Essex for their advice and assistance.

DWP bears sole responsibility for the analysis or interpretation presented here.

The USoc Sample

USoc is a UK household panel survey, which has been running since 2009. Individuals in households recruited at the first round of data collection are visited each year to collect information on changes to their household and individual circumstances. These individuals and their descendants form the core sample. Household members aged 16 or older are eligible for a full adult interview each year, allowing measurement of change. Household members not part of the core sample are also eligible provided they are living with at least one core sample member.

The USoc main sample is made up of four subsamples:

Subsample	Explanation
General Population Sample (GPS)	This is the largest part of the survey. The England, Scotland and Wales sample in this data release is based upon an initial sample of 47,520 addresses. In Northern Ireland, 2,395 addresses were selected in a single stage from the list of domestic addresses.
Ethnic Minority Boost Sample (EMBS)	The EMBS was designed to provide at least 1,000 adult interviews from each of five groups: Indian, Pakistani, Bangladeshi, Caribbean, and African based on sampling postal sectors with relatively high proportions of relevant ethnic minority groups, based upon 2001 Census data and more recent Annual Population Survey data. It screened in individuals living in households with at least one person who considered themselves, or a parent or grandparent to be an ethnic minority.
British Household Panel Survey (BHPS) cases	The sample issued for wave 2 (2010 to 2011) consisted of all members from the BHPS sample who were still active at Wave 18 of the BHPS and who had not refused consent to be included as part of the USoc sample. It should be noted that the BHPS sample contains different subsamples, including the original sample (first selected in 1991), boost samples in Scotland and Wales (first selected in 1999), and a Northern Ireland sample (selected in 2001).
Immigrant and Ethnic Minority Boost (IEMB)	This sample was introduced at Wave 6 and introduced around 2,900 new households to the survey. Data from the IEMB is included in ID for the first time in this 2010-2021 publication. The target population groups for the IEMB were either people born outside of the UK, or people who considered themselves, their parents or grandparents to be of Indian, Pakistani, Bangladeshi, Black Caribbean or Black African origin. Individuals with minority group origins of other than the five groups mentioned above were also screened in, if they lived in the sampled areas.

How representative is USoc?

USoc was initially designed as a representative sample. Detailed ‘following rules’ (see the next section) set out how the sample is allowed to change over time, as sample members’ circumstances change. These rules aim to ensure that the survey remains representative of the UK population, although they do not enable it to capture the effect of immigration to the UK.

To help address this gap, the Immigrant and Ethnic Minority Boost (IEMB) sample was introduced at Wave 6 in 2014 to 2015. It resulted in an additional 2,900 households being added to the survey. The target population groups for the IEMB were either individuals born outside of the UK, or individuals who considered themselves or their parents or grandparents to be of Indian, Pakistani, Bangladeshi, Black Caribbean or Black African origin (the five largest established minority ethnic groups in the UK). In addition, those with minority group origins other than those of the five groups were also screened for eligibility if they lived in the sampled areas. The addition of the IEMB to ID this year therefore improves the representativeness of our statistics.

The initial USoc sample only included private households living in the UK. This meant that it excluded individuals living in institutions such as nursing or retirement homes, hostels for homeless people, or prison/young offender’s institutions. As a result, analysis for particular age bands, such as older people or younger people, may not be completely representative of all individuals in those age bands where the characteristics of people who live in private households (and who are included), differ in a fundamental way from those living in institutional settings (who are not included).

USoc does collect information on the reasons why household members may be currently absent, which includes being in a range of institutions. A person who moves into an institution (except prison) is still an eligible sample member and so if they do not complete the survey online, they are followed up by an interviewer by phone where feasible. USoc is collecting more information about people who go into care homes to facilitate follow-up, and now asks participants to nominate a proxy if they are unable to be interviewed because of ill-health or being away from home for a long period.

For further discussion on the representativeness of the USoc sample, including information on hard to reach groups and plans to improve sub-group coverage, please refer to:

• Benzeval, Bollinger, Burton, Crossley & Lynn (2020) The Representativeness of Understanding Society, and
• Borkowska (2019) Improving population and sub-group coverage: who is missing and what can be done about it?

Sample status and “following rules”

This section sets out the three different sample ‘statuses’ as well as the ‘following rules’ that are applied when there are specific changes to individual or household circumstances.

There are three possible sample statuses:

Original Sample Members (OSMs)

All members of USoc GPS households enumerated at Wave 1 - including absent household members and those living in institutions who would otherwise be resident - are Original Sample Members (OSMs). All ethnic minority members of an enumerated household eligible for inclusion in the EMBS are OSMs. Any child born to an OSM mother after Wave 1 and observed to be co-resident with the mother at the survey wave following the child’s birth is an OSM. In the former BHPS sample, OSMs are those who were enumerated at the first wave of the sample from which they come (Wave 1 for the original sample, Wave 9 for the Scotland and Wales boost samples, Wave 11 for Northern Ireland) or who were subsequently born to an OSM mother or father (or both). Following the incorporation into USoc from Wave 2 onward, for those in the former BHPS sample, as for all other USoc samples, only children born to an OSM mother will themselves become an OSM.

OSMs, of all ages, are followed for interview and remain eligible as long as they are resident within the UK. They remain potentially eligible sample members for the life of the survey. The case may arise where the only OSM in the household is a child. Provided they are co-resident with the child, other household members are then Temporary Sample Members (see below) and therefore eligible for interview, even if the OSM child is not yet old enough to be eligible for a youth or adult interview. If the OSM child moves house, they are followed to their new address and those living with the OSM child are eligible for interview. If the OSM child moves into an institution, where normally just the OSM/PSM would be interviewed and not co-residents, a split-off household is created containing only the OSM child and the household enumeration grid completed. The child OSM is an eligible sample member, even if they are not eligible for interview because of their age.

Temporary Sample Members (TSMs)

Any members of an enumerated household eligible for inclusion in the EMBS at Wave 1 who are not from a qualifying ethnic minority are Temporary Sample Members (TSMs) at Wave 1. This was the only category of TSM at Wave 1. In all other samples, any new person found to be co-resident with at least one OSM or PSM after Wave 1 is a TSM. This would include any child born to an OSM father after Wave 1 but not an OSM mother and observed to be co-resident with the father (or any other OSM) at the survey wave following the child’s birth. TSMs remain eligible for interview as long as co-resident in an OSM/PSM household. TSMs who are not co-resident with at least one OSM/PSM are not followed and become ineligible for interview. TSMs are identified as re-joiners if they are subsequently found to be living with at least one OSM/PSM, and then become eligible for interview.

Permanent Sample Members (PSMs)

Any TSM father of an OSM child born after Wave 1 and observed to be co-resident with the child at the survey wave following the child’s birth is a PSM. PSMs remain potentially eligible for interview for the life of the survey.

Only OSMs have positive longitudinal weights (see Weighting section, below), and they form the basis of the analysis included here. Account is also taken of the inevitable changes of address or location that some interviewees experience; as noted above the survey follows OSMs if they move to a new household. This is to ensure that the household or family history is not lost, and that there is no significant fall off in interview numbers.

For further information on the survey, see the USoc homepage, especially the Main Survey User Guide. The Frequently Asked Questions section of the USoc website also provides useful information on all aspects of the survey.

Mode of data collection

USoc interviews were initially typically carried out face-to-face in respondents’ homes. Mixed-mode data collection introduced in Wave 8 (2016 to 2017) resulted in increasing numbers of interviews being completed online. As noted above, from March 2020 following the implementation of nationwide ‘lockdown’ measures, USoc adopted a web-first approach, with a telephone interview available for online non-responders.

Survey period

USoc fieldwork is conducted over a two-calendar-year period, with each individual being interviewed on a yearly basis. Note that the periods of waves overlap, and that individual respondents are interviewed around the same time each year on an annual basis.

For the purposes of this publication, persistent low income analysis is analysed on the basis of ‘rolling’ four-survey period datasets. For example, the period 2010 to 2014 uses individuals present in all of the following four waves, 2010 to 2011, 2011 to 2012, 2012 to 2013 and 2013 to 2014, together with all children born to Original Sample Members before their 2010 to 2011 interview.

Usoc survey waves and samples

Survey Wave	Calendar Year	Sample Used
Wave 1	2009 to 2010	GPS and EMB
Wave 2	2010 to 2011	BHPS, GPS and EMB
Wave 3	2011 to 2012	BHPS, GPS and EMB
Wave 4	2012 to 2013	BHPS, GPS and EMB
Wave 5	2013 to 2014	BHPS, GPS and EMB
Wave 6	2014 to 2015	BHPS, GPS, EMB and IEMB
Wave 7	2015 to 2016	BHPS, GPS, EMB and IEMB
Wave 8	2016 to 2017	BHPS, GPS, EMB and IEMB
Wave 9	2017 to 2018	BHPS, GPS, EMB and IEMB
Wave 10	2018 to 2019	BHPS, GPS, EMB and IEMB
Wave 11	2019 to 2020	BHPS, GPS, EMB and IEMB
Wave 12	2020 to 2021	BHPS, GPS, EMB and IEMB

Wave 1 ran in 2009 and 2010 and used the GPS and EMB sample.

Wave 2 (2010 and 2011), Wave 3 (2011 and 2012), Wave 4 (2012 and 2013) and Wave 5 (2013 and 2014) all used the BHPS, GPS and EMB sample.

Wave 6 (2014 and 2015) used the BHPS, GPS, EMB samples and also the IEMB sample, which was introduced in that wave. Wave 7 (2015 and 2016), Wave 8 (2016 and 2017), Wave 9 (2017 and 2018), Wave 10 (2018 and 2019), Wave 11 (2019 to 2020) and Wave 12 (2020 to 2021) all used the same samples as Wave 6.

Samples used for ID analysis and integration of the IEMB

As noted in the introduction section, USoc’s Immigrant and Ethnic Minority Boost (IEMB) sample has been included in ID for the first time in this 2010 to 2021 publication.

The IEMB was introduced by USoc to improve the representativeness of the survey. Although it was launched in Wave 6 (2014 to 2015), the first publication of ID could not include IEMB sample members in persistent low income estimates because there was not four years’ worth of data from the IEMB sample at that point in time.

The inclusion of the IEMB for ID 2010 to 2021 follows a decision taken by the University of Essex to no longer release the cross-sectional weights which ID had previously used (which excluded the IEMB sample). Because ID relies upon the use of cross-sectional as well as longitudinal weights, work was done to integrate the IEMB for this publication.

As noted in Section 1 above, we had begun feasibility work to explore the integration of the IEMB during 2022. The decision taken by the University of Essex meant that we had to focus on implementation in time for this publication.

The revision of ID weights to include the IEMB

In order to integrate the IEMB, work was undertaken to revise the weights used across all ID outputs. Key steps in this process were as follows:

• developing understanding of the IEMB, including early discussion with Dr Olena Kaminska at the University of Essex
• scoping work to review the application of cross-sectional and longitudinal weights and to set out required changes across all ID code. ID uses several sets of code, including for producing publication tables as well as for quality assurance. Weights are used both for case selection, i.e., ensuring the correct samples from the USoc data files are used by the dataset we derive for ID analysis, as well as for producing weighted statistics (cross-sectional and longitudinal weights are used where applicable)
• documenting proposed changes to each section of relevant code, which were then sense-checked by another team member
• implementing the changes and then reviewing the code to ensure that the changes had been made correctly

Testing the implementation of IEMB weights

The revised code was tested by running it on the Wave 11 data (ahead of receiving the Wave 12 data) to derive the longitudinal dataset that we use for our analysis. We then carried out various quality assurance checks. These included:

• reviewing how the code change affected sample sizes for each wave, with particular attention paid to changes in the sample sizes for ethnic groups
• verifying variable and sample counts in our dataset with USoc variable documentation and relevant USoc publications (see below)
• checking unanticipated results with the University of Essex: integration of the IEMB sample resulted in increases in sample counts for waves 6 onwards by a number of individuals greater than the number of those from the IEMB sample. USoc confirmed that these increases were due to corrections of non-response across the sample (this is discussed in the next section)
• running our publication code to compare output from the newly derived dataset with outputs we had produced for publication in March 2022 (which were also based on the Wave 11 data)
• running our quality assurance code to ensure this reflected the changes observed on the publication outputs

The outcome of the testing established that:

• the single wave sample for 2014 to 2015 increased by around 12,700 individuals. Approximately 7,900 of these individuals were IEMB sample members, as expected
• the unanticipated additional increases in single wave sample sizes were raised with USoc colleagues at the University of Essex, who informed us that these were due to corrections to non-response included as part of the new weighting variable. The number of additional individuals that we identified in Wave 6 (2014 to 2015) beyond those in the IEMB was confirmed by USoc
• from 2014 to 2015 onwards, ID single wave income estimates, both before and after housing costs, were brought closer to those from HBAI: these estimates include mean and median values as well as measures at other points on the income distribution (see Table M8 in ID Methodology Tables)
• some single wave estimates of the percentages of individuals in relative low income (below 60% of median income) moved slightly closer to those published by HBAI, but these single wave statistics are generally closely aligned with HBAI estimates (see Table M9 in ID Methodology Tables)
• the four-wave sample sizes upon which persistent low income statistics are based increased by around 4,300 individuals in the four-wave period from 2014 and 2015 to 2017 and 2018. This reflected the point at which individuals from the IEMB sample who had been present in each of the four waves since entering the survey, were able to be included in these statistics
• we noted that just over half of the increase in the four-wave sample size represented individuals living in households with a head of household from an ethnic minority. The rest were living in households with a White head of household. This reflects corrections to non-response referred to above
• most of the headline (rounded) statistics on persistent low income rates for all individuals, children, working-age adults and pensioners (as presented in our Persistent low income trends table) remained unchanged when using the new weights on the Wave 11 data. Where there were changes to headline statistics, these were mainly by 1 percentage point (there was one instance of a 2 percentage point change, with none greater than that)
• most headline entry and exit rates remained unchanged

Revisions to income mobility analysis

In addition to short-term mobility analysis over the most recent two waves of the survey, ID previously included analysis comparing position in the income distribution at the start of data collection in 2010 to 2011 to the most recent wave.

Following the inclusion of the IEMB sample for the remainder of our analysis, including the short-term mobility analysis, we decided to take 2015 to 2016 as the starting point for the longer-term income mobility analysis. This allowed IEMB sample members to be included, resulting in a slightly larger sample size and more representative findings, as well as improving comparability to the rest of ID.

This change affected Tables 6.1 and 6.3 in Movements between quintiles, and Table 7.1 in Quintile movements over time.

As with other changes implemented as part of the inclusion of the IEMB sample, we reviewed the output once the starting points had been revised. This established that statistics on income movements between two points in time (Tables 6.1 and 6.3) identified less mobility than when this analysis was presented over a longer period of time. This is likely to reflect the fact that there is generally less income mobility over shorter periods, as illustrated by earlier ID publications.

When looking at analysis of the position in the income distribution in which individuals spend their time relative to their position at the starting point (Table 7.1), we found that revising the starting point resulted in changes to the statistics. This particular analysis is sensitive to the number of years upon which it is based, which means that comparing different outputs is not straightforward. While we did observe some changes in the statistics, we noted that there was little change in the overall pattern of this analysis, i.e. that it illustrates less income movement for those at the top and bottom of the income distribution.

For more information about IEMB sample, please refer to:

• McFall, Nandi & Platt (2020) Understanding Society: UK Household Longitudinal Study: User Guide to ethnicity and immigration research;
• TNS BMRB (2017) UK Household Longitudinal Study: Immigrant and Ethnic Minority Boost (IEMB) technical report; and
• Lynn, Nandi, Parutis and Platt (2017) Design and implementation of a high quality probability sample of immigrants and ethnic minorities: lessons learnt

Sample sizes, atrrition and quality

Table M.1b below shows how many individuals were available for longitudinal analysis in each wave. These are Original Sample Member cases with positive longitudinal weights, except where noted.

Table M.1b - Single-year cases available for analysis (longitudinal)

Wave	GPS+EMB cases	BHPS cases	All cases
2009-2010 (not used: see Section 4)	76,240	not included	76,240
2010-2011 (cross-sectional weights)	60,474	16,572	77,046
2011-2012	48,873	11,554	60,427
2012-2013	43,103	10,383	53,486
2013-2014	39,112	9,499	48,611
2014-2015	34,175	8,517	42,692
	GPS+EMB+IEMB [1]
2015-2016 [1]	38,954	8,352	47,306
2016-2017	34,409	7,698	42,107
2017-2018	30,532	7,166	37,698
2018-2019	27,833	6,667	34,500
2019-2020	25,119	6,178	31,297
2020-2021	22,477	5,616	28,093

Source: Understanding Society 2010 to 2021

[1] IEMB cases are included in the longitudinal sample count from 2015-2016 onwards, following the introduction of the boost in 2014-2015.

As with most longitudinal surveys, attrition reduces the USoc sample size over time. Attrition tends to affect some groups more than others. Recently published analysis has explored attrition in the GPS and IEMB samples of USoc up to and including Wave 11. It considers how various factors including age, employment status, health status, and income level, affect attrition. The findings reinforce those of an earlier study on the same topic, both finding that attrition is more likely to affect those at the lower end of the income distribution. It also reported that the IEMB experienced a greater rate of attrition than the GPS.

One of the roles of survey weights is to address the effects of attrition, taking into account how it affects some groups more than others. The new study assessed the efficacy of the USoc survey weights, finding that they effectively mitigate the effects of differential attrition.

Tables M.8 and M.9 in the 2010 to 2021 ID methodology tables compare ID income and income distribution measures for each wave of the USoc survey with the cross-sectional measures produced by HBAI. For a more detailed assessment of how USoc measures compare to HBAI statistics, please refer to Understanding Society and its income data.

Persistent low income analysis in ID is based on individuals with positive longitudinal weights in the final wave being considered, meaning they are OSMs, and include individuals who were previously BHPS sample members. Any individuals who are not in the survey for all four relevant waves are excluded.

Table M.2 - Four-wave cases available for analysis

Wave	All cases	Cases present for fewer than four waves [1]	Cases available for analysis
2010-2011 to 2013-2014	48,611	1,537	47,074
2011-2012 to 2014-2015	42,692	1,190	41,502
2012-2013 to 2015-2016	47,306	8,710	38,596
2013-2014 to 2016-2017	42,107	5,976	36,131
2014-2015 to 2017-2018	37,698	895	36,803
2015-2016 to 2018-2019	34,500	710	33,790
2016-2017 to 2019-2020	31,297	569	30,728
2017-2018 to 2020-2021	28,093	447	27,646

Source: Understanding Society 2010-2021

[1] This total includes new-borns who have not been counted in the longitudinal sample. In addition, this column includes IEMB sample members in the four-wave periods 2012-2013 to 2015-2016 and 2013-2014 to 2016-2017. This is because they had not yet been present in all four waves of the survey and were therefore not available for analysis of persistent low income.

Missing information

As well as attrition reducing the available sample, some of the variables used in analysis for ID have missing values. These may be due to item non-response or partial responses. We exclude individuals with missing data from relevant analysis.

Missing income information is imputed in the USoc dataset where possible. Table M.3 below shows the extent of missing data for the longitudinal analysis of incomes for individuals in each four-wave sample. The most common reason for not having income information is having individual responses but no household questionnaire. The percentage of individuals for whom this is the case has risen since the start of the survey, with a noticeable increase in the period 2014 to 2015 and 2017 to 2018. This may be linked to several factors, including a change in fieldwork procedures and the introduction of mixed-mode data collection (online reporting) from Wave 8 (2016 to 2017). The USoc team are exploring options for gathering missing household data. For further information, see the Understanding Society Main Survey User Guide (Waves 1-12), and ‘Understanding Society and its income data’.

Other reasons for missing income data include missing equivalisation factors and missing interview dates (which means we are unable to deflate incomes).

Table M.3 - Cases used for income analysis

Wave	Missing equivalised BHC income	Missing equivalised AHC income
2010-2011 to 2013-2014	29 cases (<1%) out of 47,074	29 cases (<1%) out of 47,074
2011-2012 to 2014-2015	779 cases (2%) out of 41,502	779 cases (2%) out of 41,502
2012-2013 to 2015-2016	1016 cases (3%) out of 38,596	1016 cases (3%) out of 38,596
2013-2014 to 2016-2017	1,551 cases (4%) out of 36,131	1,551 cases (4%) out of 36,131
2014-2015 to 2017-2018	2,943 cases (8%) out of 36,803	2,943 cases (8%) out of 36,803
2015-2016 to 2018-2019	2,778 cases (8%) out of 33,790	2,778 cases (8%) out of 33,790
2016-2017 to 2019-2020	2,751 cases (9%) out of 30,728	2,751 cases (9%) out of 30,728
2017-2018 to 2020-2021	2,644 cases (10%) out of 27,646	2,644 cases (10%) out of 27,646

Source: Understanding Society 2010 to 2021

Table M.4 in the 2010 to 2021 ID methodology tables shows the extent that information is missing for classificatory variables. In many cases less than 1% of cases have missing values. For highest qualification, only those completing an individual questionnaire are included, so non-responding individuals in partially responding households will have missing values for this variable. For highest qualification, there is also a known issue in feeding forward information collected on the BHPS (i.e. pre-populating the question with what the respondent said previously) or asking individuals turning 16 about their qualifications.

In the four-wave period (between 2014 to 2015 and 2017 to 2018) there was an unusually large number of missing values for the ID derived variable which identifies whether an individual was up-to-date with their bills. This was due to missing data on the USoc variable: ‘problems paying council tax’, in Wave 6 (2014 to 2015). This variable is one of the three variables used by ID to identify whether a household is behind with their bills. From Wave 7 (2015 to 2016), the amount of missing data for this variable decreased back to levels observed during previous four-wave periods.

Weighting

USoc data is designed to be used with weights. Weights are used to ensure that any analysis removes known bias and is as representative of the population as possible, based upon what is known about the survey design (particularly how this affects selection probability) as well as non-response. Longitudinal weights also take into account differential attrition probability, as mentioned above. USoc provides cross-sectional weights (for when only one wave of the survey is being analysed), and longitudinal weights (for any analysis using data - or measures based on data - from more than one wave, such as change in income).

ID uses weights which are produced by the University of Essex for analysing the combined GPS, EMBS, BHPS and, after Wave 6, the IEMB sample, and for producing population estimates by adjusting for unequal selection probabilities, differential non-response, and potential sampling error. Weighted analysis will adjust for the higher sampling fraction in Northern Ireland and for different probabilities of selection in the EMB and IEMB samples, as well as for response rate differences between subgroups of the sample. USoc produces weighting factors rather than grossing factors, so we do not produce estimates of numbers of individuals in low income. Instead we focus on percentages of the population.

USoc provides detailed information on weighting in specific weighting guidance and in its User Guide.

Section 3: Definitions and terminology within the statistics

ID includes analysis of low income as well as movements within the overall income distribution. This section sets out how income measures are calculated in ID.

How is income measured

ID bases its income measure on a monthly net household income variable provided in the USoc dataset. The income measure used in ID is total weekly net (disposable) equivalised household income. This comprises total income from all sources after tax, national insurance and other deductions in the latest period before the interview. It comprises income from all household members, including dependants.

The components of the income measure used in ID are:

Term	Definition
Labour income	Includes usual pay and self-employment earnings. This also includes income from second jobs.
State support	Include tax credits and all state benefits, including State Pension.
Pension income	Includes occupational pensions income.
Investment income	Includes private pensions/annuities, rents received, income from savings and investments.
Private benefit income	Includes trade union/friendly society payments, maintenance or alimony and sickness or accident insurance. For the first time, the Wave 12 survey gathered data on income from Student and/or Tuition Fee Loans. This is allocated to the Private benefit income category.
Miscellaneous income	Includes educational grants, payments from family members and any other regular payment.

Income is net of the following items:

• income tax payments
• National Insurance contributions
• Pension contributions
• council tax (note domestic rates in Northern Ireland are not deducted)

Changes to income sources due to the coronavirus (COVID-19) pandemic

As noted above, ID derives its income measure from a monthly net household income variable produced by USoc. Policies implemented in response to the pandemic resulted in some changes to income sources and, subsequently, to the way in which information on or linked to these sources was gathered by USoc for both Wave 11 and Wave 12. The key changes were as follows:

Support received by employees via the Coronavirus Job Retention Scheme (CJRS): This scheme was announced by the government in March 2020. Employers who were unable to maintain their workforce because of the coronavirus pandemic could put their employees on ‘furlough’ and apply for a grant. Government and employer contributions varied during the scheme to ensure that an employee received at least 80% of their monthly wage up to £2,500 a month, including National Insurance and pension contributions. The scheme was in place until the end of September 2021, thereby covering around 19 months of the 24-month Wave 12 period.

Information on income received via the CJRS was gathered by the USoc questionnaire as part of ascertaining employee pay (amount received via employee paycheque). No additional response options or questions were included to establish whether employee pay was linked to being on furlough, and any income received via the scheme was therefore included as labour earnings.

USoc did introduce a new response option to the economic status variable ‘jbstat’, from the end of July 2020, which enabled respondents to identify if they were on furlough. This variable is used in ID’s analysis of events associated with low income entry and exit.

Support received by self-employed individuals via the Self-Employment Income Support Scheme (SEISS): The government introduced the Self-Employment Income Support Scheme (SEISS) to assist self-employed people whose income was affected by the coronavirus pandemic. SEISS grants were available over the period from March 2020 until April 2021.

A new question was introduced into the USoc survey at the end of July 2020 to capture receipt of SEISS grants. It is however understood that SEISS grants were generally reported via existing self-employment questions in the survey, which are used to calculate income from self-employment based on the reporting of profits over an earlier timeframe, usually the previous financial year.

Universal Credit £20 uplift: From April 2020, the government temporarily increased the standard allowance in Universal Credit (UC) by £1,040.04 per year and the basic element in Working Tax Credit by £1,045 per year. From April 2020, both new and existing UC claimants and existing Working Tax Credit claimants received an additional £20 per week on top of annual uprating. From April 2021, the temporary £20 per week UC uplift continued until 6 October 2021. For working households receiving Tax Credits, a one-off payment of £500 was paid in April 2021, replacing weekly uplift payments. These amounts should be captured by the USoc questionnaire as it asks, for each source of unearned income and state benefit, what amounts are received and what period of time they cover.

The extent to which ID estimates have been affected by changes linked to the coronavirus pandemic is discussed later in this section.

Housing costs

Income After Housing Costs (AHC) is derived by deducting housing costs from the above income measure.

Housing costs in ID include the following:

• rent (gross of housing benefit)
• water rates, community water charges and council water charges
• mortgage interest payments (net of tax relief)
• ground rent and service charges

In the case of renters, these housing costs will include service and water charges because this is how the information is requested on the USoc questionnaire. For owner-occupiers, these amounts will not be included.

The main difference between HBAI and ID in terms of housing costs data is that, for owner-occupiers, HBAI includes structural insurance payments, whereas ID does not: no information is collected on structural insurance payments in USoc.

For Northern Ireland households, water provision is funded from taxation and there are no direct water charges. Therefore, it is already taken into account in the Before Housing Costs measure.

Adjustments for inflation

Incomes are adjusted for inflation so they are in real terms corresponding to the middle January of the latest USoc wave (for Wave 12, i.e. calendar years 2020 and 2021, January 2021 is used). Like HBAI, ID uses a bespoke variant of the Consumer Price Index (CPI) to adjust for inflation to look at how incomes are changing over time in real terms. The most recent release of this index included minor corrections to previously published data for the period January to June 2021. These are detailed on the webpage for the release. In line with HBAI, corrections have not been applied to deflators which have previously been used for published ID statistics.

Equivalisation

An adjustment called ‘equivalisation’ is made to income to make it comparable across households of different size and composition. ID uses net disposable weekly household income, after adjusting for the household size and composition, as an assessment for material living standards - the level of consumption of goods and services that people could attain given the net income of the household in which they live. In order to allow comparisons of the living standards of different types of households, income is adjusted to take into account variations in the size and composition of the households in a process known as equivalisation. ID assumes that all individuals in the household benefit equally from the combined income of the household. Thus, all members of any one household will appear at the same point in the income distribution.

How household income is equivalised

Equivalence scales conventionally take an adult couple without children as the reference point, with an equivalence value of one. The process then increases relatively the income of single person households (since their incomes are divided by a value of less than one) and reduces relatively the incomes of households with three or more persons, which have an equivalence value of greater than one.

The main equivalence scales used in ID are the modified OECD scales, which take the values shown in Table M.5 below. These are in line with those used in HBAI. The equivalent values used by the McClements equivalence scales are also shown for comparison alongside modified OECD values. The McClements scales were used for the main estimates in the predecessor Low Income Dynamics publication with results based on the modified OECD equivalence scales published in an Appendix. In the modified OECD and McClements versions two separate scales are used, one for income BHC and one for income AHC.

Table M.5 – Comparison of modified OECD and McClements equivalence scales

Equivalence scales
	Modified OECD to equivalise BHC results rescaled to couple without children=1[1]	OECD ‘companion’ Scale to equivalise AHC results	McClements BHC	McClements AHC
First Adult	0.67	0.58	0.61	0.55
Spouse	0.33	0.42	0.39	0.45
Other Second Adult[2]	0.33	0.42	0.46	0.45
Third Adult	0.33	0.42	0.42	0.45
Subsequent Adults	0.33	0.42	0.36	0.40
Children aged under 14yrs[3]	0.20	0.20	0.20	0.20
Children aged 14yrs and over[3]	0.33	0.42	0.32	0.34

Notes:

[1] Presented here to two decimal places

[2] For the McClements scale, the weight for ‘Other second adult’ is used in place of the weight for ‘Spouse’ when two adults living in a household are sharing accommodation, but are not living as a couple. ‘Third adult’ and ‘Subsequent adult’ weights are used for the remaining adults in the household as appropriate. In contrast to the McClements scales, apart from for the first adult, the OECD scales do not differentiate for subsequent adults

[3] The McClements scale varies by age within these groups; appropriate average values are shown in the table

The construction of household equivalence values from these scales is quite straightforward. Consider a single person, a couple with no children, and a couple with two children aged twelve and ten, all having unadjusted weekly household incomes of £300 (BHC). The process of equivalisation, as conducted in ID, gives an equivalised income of £448 to the single person, £300 to the couple with no children, but only £214 to the couple with children.

How the low income threshold is calculated

Relative low income sets the threshold as a proportion of the median income. The amount therefore changes each year as median income moves. It is used to measure the number and proportion of individuals who have incomes at a certain level below the median.

Changes in the rates of relative low income can be driven by changes in single year relative low income estimates or by individuals spending a longer or shorter time in relative low income.

The percentage of individuals in relative low income will increase if:

• the average income stays the same, or rises, and individuals with the lowest incomes see their income fall, or rise less, than average income; or
• the average income falls and individuals with the lowest incomes see their income fall more than the average income

The percentage of individuals in relative low income will decrease if:

• the average income stays the same, or rises, and individuals with the lowest incomes see their income rise more than average income; or
• the average income falls and individuals with the lowest incomes see their income rise, or fall less, than average income, or see no change in their income

Measuring Income Before and After Housing Costs

Before Housing Costs (BHC) measures allow an assessment of the relative standard of living of those individuals who were actually benefiting from a better quality of housing by paying more for better accommodation. On this basis, income growth over time incorporates improvements in living standards where higher costs reflected improvements in the quality of housing.

After Housing Costs (AHC) measures allow an assessment of living standards of individuals whose housing costs are high relative to the quality of their accommodation. Income growth over time may also overstate improvements in living standards for low income groups, as a rise in Housing Benefit to offset higher rents (for a given quality of accommodation) would be counted as an income rise.

For completeness, ID presents analyses on both a BHC and AHC basis. This is principally to take into account variations in housing costs that themselves do not correspond to comparable variations in the quality of housing.

Households and families

ID presents information on an individual’s household income by various household and family (sometimes referred to as a ‘benefit unit’) characteristics. There are important differences between households and families.

A household is one person living alone or a group of people (not necessarily related) who either share living accommodation OR share one meal a day and who have the address as their only or main residence. A household can consist of one or more families or ‘benefit units’, which are single adults or a married or cohabiting couple and any dependent children. For example, a group of students with a shared living room would be counted as a single household even if they did not eat together, but a group of bed-sits at the same address would not be counted as a single household because they do not share living space or eat together.

A husband and wife living with their young children and an elderly parent would be one household but two families or ‘benefit units’. The husband, wife and children would constitute one benefit unit and the elderly parent would constitute another. It should be noted that the term ‘benefit unit’ is used as a description of groups of individuals regardless of whether they are in receipt of any benefits or tax credits. A household will consist of one or more benefit units, which in turn will consist of one or more individuals (adults and children).

Income measurements used in ID

Relative low income

Low income can be defined in various ways. ID uses a measure of relative low income. Relative low income measurements identify which individuals have income which falls below a certain threshold based upon the range of incomes in the population. In ID, we concentrate on individuals with household incomes below 60% of median income. Information on those with household incomes below 70% of the median income is available in the detailed tables published alongside this release.

Persistent low income

An individual is classified as being in persistent low income if they live in a household in relative low income for at least three of their last four consecutive interviews. It is therefore possible that some of those classed as being in persistent low income were not in relative low income at the time of their most recent interview. However, they will have experienced low income in each of the previous three years and, as a result, their long-term living standards are not anticipated to be different to other individuals in persistent low income who were in low income in the most recent interview.

This issue is relevant because the income distribution is particularly dense around the 60% and 70% of median income thresholds. In addition, some short periods of recorded high income may be due to measurement error and not reflect any real improvement in living standards.

Low income entries and exits

The methodology used here is the same as in the predecessor Low Income Dynamics publication. This analysis looks at individuals moving into and out of relative low income across a two-wave period, and uses a measure of equivalised household income consistent with that used elsewhere in ID.

When analysing transitions into and out of low income, the threshold used is the standard 60% of median income. Analysis of transitions between one wave and the next only includes what are defined as ‘clear’ transitions. For example, for an exit or entry to occur, household incomes must change such that they cross the threshold and are at least 10% higher or lower than 60% median income in the following wave. This requirement is put in place because all survey estimates - including household incomes and measures based on them such as the low income threshold - are subject to sampling and measurement error.

The exit or entry rate for individuals is calculated as, using exit rates as an example, the number of individuals in low income in one wave who exited low income in the following wave, expressed as a percentage of all those who were in low income. Roughly equal numbers of individuals move into as move out of low income across each two-wave period. Rates of entry are, however, much lower than rates of exit, because entry rates are expressed as a percentage of those who were not in low income in the first of the two waves under consideration. Exit rates are expressed as a percentage of those who were in low income in the first of the two waves under consideration (a much smaller number than those not in low income).

The unit of analysis is the individual. However, as individuals live in households and we assume that all members of the household benefit equally from the household’s income, they will be affected by changes at the household level. This could come about either through changes in income levels, or by changes in the household composition which affects incomes through the equivalisation process.

For the publication tables containing detailed breakdowns we have presented the rates as an average over three, two-wave periods as this will, to some extent, remove volatility that can be associated with smaller sample sizes, and help to create a more robust and stable series.

The addition of income from Student and/or Tuition Fee loans to USoc Wave 12 data collection is likely to have contributed to the increase in low income exits seen across the most recent two-wave period (2019 and 2020 to 2020 and 2021). Please see discussion regarding this in the Events section, below.

Events associated with low income entry and exit

This analysis is based on the same individuals included in low income entry and exit analysis (above). It aims to help us understand how certain events are linked to these low income entries and exits by considering changes to income sources (including earned and non-earned income), changes in employment, and demographic changes. For AHC analysis only, we also look at changes in tenure and housing costs.

The method used is based on an approach used on the British Household Panel Survey (BHPS) previously (for example, Jenkins S.P and Rigg J.A with the assistance of Devicienti, F. (2001) ‘The dynamics of poverty in Britain’, DWP Research Report No. 157); and also in Low Income Dynamics, the predecessor of ID, based on the BHPS.

ID includes findings on the following events.

Event	Definition
Earnings	A change in monthly household earnings of at least 20% and a minimum of £10. No change in the number of workers in the household.
Benefit Income	A change in monthly household benefit income of at least 20% and a minimum of £10. No change in household size.
Investment Income	A change in monthly household income from investment of at least 20% and a minimum of £10. No change in household size.
Occupational pension income	A change in monthly household occupational pension income of at least 20% and a minimum of £10. No change in household size.
Other Income	A change in monthly household ‘other’ income of at least 20% and a minimum of £10. No change in household size. This event was affected by the addition of income from Student and/or Tuition Fee Loans in Wave 12 (see below for discussion).
Number of Workers	A change in the number of workers in a household. This is considered both for where there is an accompanying household size change and where there is no change in household size.
Number of Full-Time Workers	A change in the number of full-time workers in a household. This is considered both for where there is an accompanying household size change and where there is no change in household size.
Changing the number of hours worked	New for ID 2010 to 2021, this event considers where there is a change from full-time to part-time work (low income entries) or the other way around (low income exits).
Working or Workless households	New for ID 2010 to 2021, this event looks at where a household changes from containing no adult in work (workless) to having at least one adult in work (working), or the other way around.
Household Type	A change in household type.
Lone Parent Household	A change in lone parent status – either becoming a lone parent (low income entries) or changing from being a lone parent (low income exits).
Couple to Single Person Household	A change from living in a couple household to single person status – no children in either case.
Single Person to Couple Household	A change from single person status to living in a couple household – no children in either case.
Number of Children	A change in the number of children – an increase for low income entry or a decrease for low income exit.
Tenure	AHC only, a change in tenure across the four tenure types (outright ownership; buying with a mortgage; social renting or private renting).
Housing Costs	AHC only, an increase (low income entries) or decrease (low income exits) in monthly housing costs of at least 20% and a minimum of £10.

For a change in monthly income or housing costs to be considered an event, it must have changed by at least 20% and a minimum of £10. This requirement aims to ensure that only meaningful changes in income components are included. To be included in analysis of low income exits, income components must have increased, and to be included in analysis of low income entries, income components must have decreased.

Certain employment and demographic changes are controlled for when examining income and housing costs events. This is to try and remove the effect of major changes that are likely to have a bearing on the event being considered. For example, changes in household earnings are only included if the number of workers stays the same. For all other income events, the number of people in the household must not change.

For each event, we present three statistics in our ID publication tables across all two-wave periods since 2010 to 2011. To help explain the three measures, statistics from the most recent two-wave period presented in Table 9.1n, are used as an example:

	Prevalence (%)	Entry rate (%)	Share of entries (%)
Fall in earnings	11	27	37

A prevalence statistic – this tells us how common an event is among the population at risk of either entering or exiting low income. When considering the relationship between a fall in earnings and low income entry, for example, the prevalence statistic tells us what percentage of those who were not in low income in the first wave experienced a fall in earnings between the two waves. Table 9.1n tells us that over the most recent two-wave period, this was 11%.

A statistic expressing the rate of entry or exit if an individual experienced this event – the extract above shows us that 27% of those who were not in low income in 2019 to 2020 and who experienced a decrease in earnings, entered low income in 2020 to 2021.

A share statistic, indicating how common an event was among all low income entries or exits – Table 9.1n shows that 37% of all those who entered low income experienced a fall in earnings.

Because those in low income are a smaller group than those who are not, and are, by definition, closer to the low income threshold, they have a greater chance of exiting low income if they experience a particular event than those not in low income have of entering low income if they experience the same event. This explains why exit rates tend to be higher than entry rates.

While we have aimed to rule out major confounding events, this analysis does not rule out all factors which may have a bearing upon low income entry or exit, and should not be interpreted as implying causality.

New events introduced for ID 2010 to 2021

We have developed two new events which allow us to further investigate the role that paid work plays in low income movements.

The first of these aims to capture how a change in the number of hours worked is linked to low income entry or exit. As with other events, each is defined differently according to whether looking at a low income entry or a low income exit. In order to capture change at the household level, for low income entries, we defined this particular event as a fall in the number of full-time workers in the household where the number of workers in the household stays the same as well as the number of household members. By defining the event in this way, this excludes the possibility of the fall in the number of full-time workers being due to a loss of employment altogether or to a change in household composition. For low income exits, we require this event to involve an increase in the number of full-time workers where the number of workers stays the same as well as the number of household members.

The second event aims to capture how a change in household-level working status (i.e. whether working or workless) is linked to low income entry or exit. For low income entries, this event requires the household to change from one in which at least one adult is working to one where no adult is working. For low income exits, the change must be the other way around. As with the previous event, household size is kept constant to avoid including scenarios where the change in working status is linked to household composition changes, such as relationship breakdown or formation.

Income from Student and/or Tuition Fee loans

As noted elsewhere in this report, information on income from Student and/or Tuition Fee loans was included as part of Private Benefit income for the first time in the Wave 12 USoc survey. For the purposes of Events analysis, Private Benefit income is classed as ‘Other’ income. It is likely that the new addition of this income source has affected the statistics presented on Table 9.5x, which show that an increase in ‘Other’ income was experienced by 14% of all individuals in low income (both BHC and AHC) in the most recent two-wave period compared to a prevalence rate of 5% BHC and 6% AHC in the two-wave period immediately preceding it.

Table 9.5x also shows that of all individuals who exited from low income over the most recent two-wave period, 21% of them (both BHC and AHC) had an increase in an ‘Other’ income source. This compared to 9% BHC and 8% AHC in the preceding two-wave period. Note that it is not possible to fully attribute these individuals’ exits from low income to the new inclusion of income from a Student or Tuition Fee Loan, because these individuals may have experienced other events that affected their income.

How have changes related to the coronavirus pandemic affected ID measures?

Earlier sections of this note have set out how USoc data collection responded to policies implemented due to the coronavirus pandemic. This section considers how different ID measures may have been affected.

Policy changes and questionnaire content: in response to the coronavirus pandemic, policies were implemented relating to three income sources:

• the CJRS or ‘furlough’ scheme for employees: the USoc questionnaire was adapted to gather information on employees being supported by the CJRS via a revision to the economic status variable ‘jbstat’ but otherwise, employees reported their income in the usual way
• the SEISS for those who were self-employed: a new question on SEISS was added at the end of July 2020
• the £20 uplift to UC; and the £20 uplift and subsequent £500 lump sum Working Tax Credit payment: information on these payments will have been gathered by questions which ask about the social benefits that are received

Implications for ID income measures: except for analysis on events associated with low income entry and exit, ID does not publish analysis of individual income components. This means that changes specifically made to employee or self-employed earnings cannot be identified via the statistics, or that a change in benefit income cannot be specifically allocated to, say, a change in the level of UC income.

While many employees experienced immediate changes to their earnings following the start of lockdown, it is reasonable to assume that data collection will not have captured all of them. Employee earnings are recorded via a question which asks about usual pay. We have not carried out any analysis on pre- and post- lockdown earnings, but it is plausible that some individuals in receipt of reduced earnings linked to the furlough scheme will have reported pre-pandemic earnings. Attributing any changes in earnings to the CJRS is not fully possible, since furlough status was not added as a response option to the economic status variable ‘jbstat’ until the end of July 2020, meaning that at least some of those individuals in receipt of this support cannot be identified. Furthermore, other employees not in receipt of support via the CJRS will have experienced changes to their income over this period, some of which will have been unrelated to the pandemic.

Although a new question on SEISS grants was introduced to the survey, it is understood that these grants were generally included in profit amounts reported for self-employed survey participants.

Any changes resulting from additional payments to UC and Working Tax Credit claimants were gathered via questions based on the last payment received, and should therefore have been captured effectively via the USoc survey.

Beyond issues associated with data collection and measurement, the way in which ID statistics are produced is also relevant.

Persistent low income: the most recent four-wave period between 2017 to 2018 and 2020 to 2021 included two waves that were affected by the coronavirus pandemic (Waves 11 and 12), and two waves that were not (Waves 9 and 10). Headline rates of persistent low income have not shown any unusual changes compared to previous four-wave periods. It is worth noting that because persistent low income statistics are based on a relative income threshold that is linked to median income, they are likely to be better at reflecting distributional rather than absolute changes in income levels.

Low income entry and exit rates: these are considered over a two-wave period. Given this, it might be more reasonable to expect to have seen changes in these rates over the period between 2019 to 2020 and 2020 to 2021, reflecting changes to income levels associated with the coronavirus pandemic. ID statistics on low income entries and exits have shown some small changes over the two most recent two-wave periods. Rates of low income exit, for example, have shown increases for working-age adults and children, both BHC and AHC. There were also small increases in the low income entry rate for children and working-age adults in the period 2018 and 2019 to 2019 and 2020.

Analysis of the events associated with low income entry and exit is the only element of ID which considers changes in income components and employment status. It looks at these changes over a two-wave period. Relevant events are considered here over the two most recent two-wave periods, i.e. from 2018 and 2019 to 2019 and 2020, and from 2019 and 2020 to 2020 and 2021:

• association between low income entry and a fall in earnings: although there was a slight increase in the percentage of individuals who experienced a fall in household earnings over the first of these two-wave periods, from 11% to 12%, this fell back to 11% in the most recent two-wave period (BHC and AHC figures were the same). To be counted as having experienced a fall in earnings for this analysis requires the decrease to be at least 20%. It is unlikely therefore to have captured all individuals in receipt of support via the CJRS, since workers on furlough received at least 80% of their pre-pandemic earnings. Furthermore, the ONS have estimated that around half of employees who were on furlough during 2020 and 2021 received top-up payments from their employers. See Table 9.1n in the Events tables file.

• association between low income exit and an increase in benefit income: looking at low income exits, we might anticipate being able to observe effects linked to increased UC and Working Tax Credit payments. Over the first of the two periods considered here, low income exits analysis showed no increase in the prevalence of an increase in benefit income, but an increase was observed in the most recent two-wave period (from 31% to 34% BHC and from 30% to 32% AHC). This event was also associated with slightly higher chances of exiting from low income. It will be important to view these changes as part of the longer-term trend. See Table 9.2x in the Events tables file.

• association between low income entry and a reduction in the number of workers in a household: Over the two most recent two-wave periods, there were some small increases in events statistics which considered associations between a fall in the number of workers and low income entry. These increases were, however, generally seen in the first of these two-wave periods, and were followed by decreases in the most recent two-wave period. They cannot be directly attributed to the coronavirus pandemic but may reflect some employment loss associated with it during 2020. See Tables 9.6n and 9.10n in the Events tables file.

• association between low income entry and a reduction in the number of full-time workers in a household: Events statistics on changes in the number of full-time workers showed some changes across the two most recent two-wave periods. As with events considering workers overall, there was some evidence of small increases in these statistics across the first two-wave period, some of which were followed by small decreases in the second two-wave period. When looking at reductions in the number of full-time workers where the household size stays the same and also where the number of workers stays the same (new event for this year), there is evidence of a longer-term upward trend in all three of the events statistics, but particularly the risk and share statistics. For this reason, it is difficult to assess what changes might be linked to the coronavirus pandemic from these figures. See Tables 9.7n and 9.8n in the Events tables file.

Variables used for sub-group analysis

When longitudinal analysis is carried out on particular sub-groups, we allocate individuals to sub-groups according to their status in the first wave being considered, and use weights in the final wave being considered. This means that their status in any intervening waves may change, and have a bearing upon what is being measured in the final wave, but will not be taken into account.

For example, if looking at persistent low income over a four-wave period, consider a single working-age adult living alone who was working in the first interview of the period being considered and not in relative low income, but then was not working for the next three interviews and was in relative low income for each of those waves.

This individual would be identified as in persistent low income, but would also be classified as in a family where all adults are working despite not having worked for the past three waves.

Variables used to classify age and pensioner status

All age-related derived variables in ID use the USoc age variable ‘age_dv’. USoc check the variable ‘age_dv’ for consistency, and make imputations when errors are found in order to improve its accuracy.

We use the USoc variable ‘pensioner_dv’ to identify if an individual is of State Pension age (SPa). ‘Pensioner_dv’ is derived using a respondent’s full date of birth (via ‘age_dv’). This allows precision in determining SPa.

State support

The Government pays money to individuals to support them financially under various circumstances. An individual is in receipt of state support if they receive one or more benefits, or are being paid Tax Credits. Most of these benefits are administered by DWP, with the exception of Housing Benefit and Council Tax Reduction which are administered by local authorities, and Child Benefit and Tax Credits which are administered by HM Revenue and Customs. National roll-out of UC for all new relevant claims completed in December 2018, and it is now the primary working-age benefit. It replaces income-based Jobseekers’ Allowance, income-related Employment and Support Allowance, Income Support, Working Tax Credit, Child Tax Credit and Housing Benefits. However, existing claimants with legacy benefits who have not have a change in circumstances have remained on their legacy benefits. For more information about the way in which information on state support is gathered and processed, please see the USoc Wave 12 questionnaire and ‘Understanding Society and its income data’.

Ethnicity

USoc includes detailed ethnicity classifications which are combined into the following publication splits:

White:

• British/English/Scottish/Welsh/Northern Irish
• Irish
• Gypsy or Irish Traveller resident in England, Scotland or Wales
• Any other White background

Mixed:

• White and Black Caribbean
• White and Black African
• White and Asian
• Any other mixed background

Asian:

• Asian/ Asian British: Indian
• Asian/ Asian British: Pakistani
• Asian/ Asian British: Bangladeshi
• Asian/ Asian British: Chinese
• Asian/ Asian British: Any other Asian background

Black:

• Black/ Black British: Caribbean
• Black/ Black British: African
• Black/ Black British: Any other Black Background

Other:

• Other Ethnic Group: Arab
• Other Ethnic Group: Gypsy or Irish Traveller resident in Northern Ireland
• Other Ethnic Group: Any other ethnic group.

Source: Understanding Society questionnaire documentation.

To ensure ID statistics are harmonised with GSS standards Gypsy or Irish Travellers resident in Northern Ireland are included in the Other ethnic group rather than the White ethnic group. Gypsy or Irish Travellers resident in England, Scotland or Wales are still included in the White ethnic group.

Long-standing illness or disability

The way in which information on long-standing illness or disability was gathered by the USoc survey changed over Waves 7 to 9. These changes affect how this information is reported on in ID.

Up to and including Wave 7 (2015 to 2016)

USoc included two questions to determine the presence of a limiting or non-limiting long-standing illness or disability:

“Do you have any long-standing physical or mental impairment, illness or disability? By ‘long-standing’ I mean anything that has troubled you over a period of at least 12 months or that is likely to trouble you over a period of at least 12 months.”

If respondents reported having a long-standing illness or disability, a follow-up question asked:

“Does this / do these health problem(s) or disability(ies) mean that you have substantial difficulties with any of the following areas of your life?”

Twelve areas are listed:

• mobility (moving around at home and walking)
• lifting, carrying or moving objects
• manual dexterity (using your hands to carry out everyday tasks)
• continence (bladder and bowel control)
• hearing (apart from using a standard hearing aid)
• sight (apart from wearing standard glasses)
• communication or speech problems
• memory or ability to concentrate, learn or understand
• recognising when you are in physical danger
• your physical co-ordination (e.g. balance)
• difficulties with own personal care (e.g. getting dressed, taking a bath or shower)
• other health problem or disability

If a respondent answered that they have substantial difficulties with any of these areas, they were said to have a limiting long-standing illness or disability. If they identified none of these, then the long-standing illness or disability was said to be non-limiting. Prior to Wave 7, although the presence of a long-standing impairment, illness or disability was asked of all adult respondents, whether it was limiting or not was not asked of proxy respondents (i.e. where someone else answers interview questions on behalf of another individual). In Wave 7 (2015 to 2016), this information was also gathered via proxy interviews, resulting in a decrease in the amount of data coded as missing for this variable in 2015 to 2016.

Changes in Waves 8 and 9

In Wave 8 (2016 and 2017), USoc changed the way in which this information was captured, but only to the main questionnaire (not the proxy questionnaire). This change removed the link between the presence of a long-standing impairment, illness or disability and whether any such illness or disability was limiting or not: in Wave 8, all adult respondents were asked whether they had a limiting health problem or disability regardless of whether they reported a long-standing illness or disability.

Those who reported a long-standing impairment, illness or disability were asked:

“Do you have any health problems or disabilities that mean you have substantial difficulties with any of the following areas of your life?”

Those who did not report a long-standing impairment, illness or disability were asked:

“Even though you don’t have any long-standing health problems, do you have any health problems or disabilities that mean you have substantial difficulties with any of the following areas of your life?”

This question was followed by a list of the same twelve areas noted above.

While ID only derives its statistics on limiting and non-limiting illness or disability for those who report having a long-standing illness or disability, it should be noted therefore, that from 2016 to 2017, limiting conditions may not necessarily be linked to the long-standing illness or disability reported. Section 7 of the Summary report has been drafted to reflect these changes. See also USoc survey documentation for further details.

In Wave 9 (2017 and 2018), the proxy questionnaire was brought in line with the main questionnaire.

Section 4: What to be aware of when interpreting ID statistics

Accurate interpretation of statistics produced using survey data relies upon understanding the way in which the data is collected. This section sets out some of the key factors that are relevant to ID.

Wave 1 (2009 to 2010) income information

There are known issues with the income information in the first USoc wave covering 2009 to 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 USoc are not comparable with subsequent waves and are likely to be of lower quality. We have therefore excluded the first wave from any analysis presented in this publication.

Survey Data

The figures in ID come from USoc, a longitudinal survey. In 2020 to 2021, the survey gathered information on around 28,000 individuals. In addition to capturing detailed information on incomes, USoc gathers rich contextual information on household and individual circumstances, such as employment, education level and disability.

Surveys gather information from a sample rather than from the whole population. The sample is designed carefully to allow for this, and to be as accurate as possible given practical limitations, such as time and cost constraints. Results from sample surveys are always estimates, not precise figures. This means that they are subject to a margin of error which can affect how changes in the numbers should be interpreted, especially in the short-term. Year-on-year movements should be treated with caution.

In addition to sampling errors, consideration should also be given to non-sampling errors. Non-sampling errors arise from the introduction of some systematic errors in the sample compared to the population it is supposed to represent. As well as response bias, errors include inappropriate definition of the population, misleading questions, data input errors or data handling problems – in fact any factor that might lead to the survey results systematically misrepresenting the population. There is no simple control or measurement for such non-sampling errors, although the risk can be minimised through careful application of the appropriate survey techniques from the questionnaire and sample design stages through to analysis of results.

ID is based on data from a longitudinal household survey. It is subject to the following nuances of using survey data, with additional issues which specifically affect longitudinal surveys:

Term	Definition
Attrition	Some respondents will inevitably drop out between interviews. To minimise attrition, the USoc team maintains a database of information on respondents so they can send communications to respondents. The database builds on contact information collected during the survey interviews, and is updated throughout the year. A between-wave-mailing is also used to help maintain contact with participants and update addresses. The mailing has a report of research findings, an address confirmation slip and materials to encourage registration with the participant website.
Sampling error	Results from surveys are estimates and not precise figures. In general terms the smaller the sample size, the larger the uncertainty.
Non-response error	As with any survey, analysis based on USoc is at risk from a systematic bias due to non-response. This is when households that had been selected for interview do not respond to the survey. Individuals within households may also be non-responders even if the rest of the household does respond. In an attempt to correct for these biases, the results are weighted to adjust for non-response, taking account of previous responses.
Item non-response	Item non-response occurs where a respondent has given a full interview, but has refused or given a ‘don’t know’ answer to a particular question, which consequently leads to a missing value for that item. The USoc team, based at The Institute for Social and Economic Research (ISER) have used imputation for some USoc variables to correct for item non-response, whereby a valid value is imputed in to replace the missing value, with the aim of reducing potential bias caused by the missing values. Imputation is used to derive overall incomes where there is missing data.
Survey coverage	The initial USoc survey only sampled private households in the United Kingdom. This meant that it excluded individuals living in institutions such as nursing or retirement homes, hostels for homeless people, or prison/young offender’s institutions. As a result, analysis for particular age bands, such as older people or younger people, may not be completely representative of all individuals in those age bands where the characteristics of people who live in private households (and who were included), differ in a fundamental way from those living in institutional settings (who were not included).
Sample size	Although USoc has a large sample size for a household survey, small sample sizes for some more detailed analyses may mean results are more volatile.
Measurement error	Other social surveys underestimate incomes from certain sources when compared with administrative data. Like these surveys, it is likely that USoc also does not fully capture all income streams. However, the longitudinal nature of the survey means that this may improve over time with the use of dependent interviewing (where respondents can be reminded of previous responses) and panel conditioning (where familiarity with the questionnaire means that respondents respond more accurately to later waves of the survey). See Dr Paul Fisher’s paper Does repeated measurement improve income data quality? (ISER Working Paper Series, 2016-11) for further details.

Reporting Uncertainty

As noted above, survey results are always estimates, not precise figures and so subject to a level of uncertainty. Two different random samples from one population, for example the UK, are unlikely to give exactly the same survey results, which are likely to differ again from the results that would be obtained if the whole population was surveyed. We are unable to calculate sampling uncertainties for these statistics.

Rounding and suppression

Figures are rounded to the nearest percentage point independently and as a result, differences may not sum exactly due to rounding. Any proportions based on a sample population of 100 or less are suppressed.

Survey-based income data

The following issues need to be considered when using any survey-based income information:

Term	Definition
Lowest incomes	Comparisons of household income and expenditure suggest that those households reporting the lowest incomes may not have the lowest living standards.
Benefit receipt	Relative to administrative records, surveys tend to under-report benefit receipt.
Self-employed	Almost all analyses in the ID publication include the self-employed (the exception is the events associated with low income entry and exit which rely upon identifying the number of full-time workers in a household). A proportion of this group are believed to report incomes that do not reflect their living standards and there are also recognised difficulties in obtaining timely and accurate income information from this group. This may lead to an understatement of total income for some groups for whom this is a major income component, although this is likely to be more important for those at the top of the income distribution.
High incomes	Unlike in the HBAI series, no adjustment is made to correct for a likely undercount of ‘very rich’ households in survey data. However, this should not affect low income statistics based on median incomes.
Gender analysis	In any analysis of gender, it must be remembered that ID attempts to measure the living standards of an individual as determined by household income. This assumes that both partners in a couple benefit equally from the household’s income, and will therefore appear at the same position in the income distribution. Any difference in figures can only be driven by gender differences for single adults, which will themselves be diluted by the figures for couples. The lower-level gender disaggregation in the family type classification is therefore likely to be more informative. Research has suggested that this assumption with regard to income sharing is not always valid, as men sometimes benefit from shared household income at the expense of women. This means that it is possible that ID results broken down by gender could understate differences between the two groups. A discussion of the evidence on the distribution of money within couple households, particularly those on low incomes, is included in this research briefing published by the Women’s Budget Group.
Students	All analyses in ID includes students. Information for students should be treated with some caution because they are often dependent on irregular flows of income.
Older people	The effect of the exclusion of some of the older people who live in residential homes is likely to be small overall except for results specific to those aged 80 and above.
Ethnicity analysis	Smaller ethnic minority groups exhibit year-on-year variation which limits comparisons over time. USoc includes an ethnic minority boost which means it has larger sample sizes for non-white ethnic groups than if such a boost was not present.
Disability analysis	No adjustment is made to disposable household income to take into account any additional costs that may be incurred due to the illness or disability in question. This means that using income as a proxy for living standards for these groups may be somewhat upwardly biased.
Regional analysis	Although the USoc sample is large enough to allow some analysis to be performed at a regional level, it should be noted that no adjustment has been made for regional cost of living differences. It is therefore assumed that there is no difference in the cost of living between regions, although the After Housing Costs measure will partly take into account differences in housing costs.

Section 5: ID and other statistics

Other measures of persistent low income

ID is a successor series to DWP’s Low Income Dynamics (LID) publication, the final issue of which was published in September 2010. LID was based on the British Household Panel Survey (BHPS). There are methodological differences between the two approaches and data sources which mean that direct comparisons are not possible.

Statistics on persistent low income have also been published by the Office for National Statistics (ONS): see Persistent Poverty in the UK and EU. These statistics were last published in 2019, and are based on the European Union Statistics on Income and Living Conditions (EU-SILC). As well as drawing on a different data source, these statistics use slightly different definitions, and count persistent low income differently i.e. individuals who are in low income households for at least three of the last four years, including the latest year. This difference means that UK estimates published using this method tend to be a little lower than those presented in ID. For details of SILC calculations see EU statistics on income and living conditions (EU-SILC) methodology - monetary poverty, while “The relationship between EU indicators of persistent and current poverty” by Stephen P. Jenkins and Philippe Van Kerm contains a commentary on the EU-SILC persistent low income methodology. Persistent low income data sourced from EU-SILC for other European countries is available via the eurostat website.

Other publications which use ID statistics

Poverty and Income Inequality in Scotland includes figures on persistent poverty for children and other population groups in Scotland. This year’s release will, for the first time, present findings on low income entry and exit for Scotland. These statistics are produced for the Scottish Government by the DWP.

Persistent poverty in Wales includes headline figures on persistent poverty for children, working-age adults, and pensioners in Wales. This release is published by the Welsh Government.

Ethnicity Facts and Figures: Persistent low income published by the Race Disparity Unit at the Cabinet Office presents ID data on persistent low income and low income entry and exit, by ethnic group.

Other statistics on income, wealth and economic wellbeing

ID is released alongside other government statistical publications focused on income and low income. The publications listed below can be considered alongside ID to give a more complete picture of income and low income.

This is not intended to be an exhaustive list. More information on income data as well as sources of data on earnings can be found in the ONS Income and earnings statistics guide. ONS have also developed a new income and earnings interactive tool, which can be used to quickly identify sources of statistics on income and earnings, as well as information about their key features. See also ‘Explaining income and earnings: important questions answered’, also published by the ONS.

The Households Below Average Income (HBAI) series presents information on living standards in the UK based on household income measures for the financial year ending 2022. HBAI uses data from the FRS. Estimates are provided for average incomes, and for the number and percentage of people living in low income households. Tables M.8 and M.9 in the ID publication methodology tables compare single year income statistics derived from USoc with HBAI figures, and show a good level of coherence. HBAI remains the best source of cross-sectional low income statistics.

The Family Resources Survey (FRS) is a continuous household survey. It publishes a range of annual statistics on household circumstances, including income, disability, tenure and pension participation.

The Pensioner’s Incomes Series provides a more detailed analysis of pensioners’ incomes, based on data collected by the FRS.

European comparisons: statistics on levels of low income in other European countries is available via the eurostat website.

International Comparisons: the OECD provides international comparisons on trends and levels in Gini coefficients before and after taxes and transfers, average household disposable incomes, relative poverty rates and poverty gaps, before and after taxes and transfers.

The effect of taxes and benefits on household income: these ONS statistics provide a detailed breakdown of household income, including estimates of both direct and indirect taxes, cash benefits and in-kind benefits provided by the state by decile and quintile groups, ranked by equivalised disposable income.

Personal Incomes statistics published by HMRC include data on UK taxpayers, their incomes and the income tax they are liable for, based on the annual Survey of Personal Incomes.

Information on state benefits including claims and caseload numbers on benefits administered by DWP can be found via the DWP benefits statistics collection.

Commentary on average household income is available via a separate bulletin, as is Household income inequality.

Wealth in Great Britain:

The Wealth & Assets Survey (WAS) is a key source of information on how households in Great Britain are managing economically. WAS is a large-scale longitudinal survey with seven rounds currently published. The most recent publication presents headline results from the seventh round (2018 to 2020), covering household debt, financial wealth, pension wealth, property wealth and physical wealth.

Personal and economic well-being in Great Britain:

Based on the ONS’s Opinions and Lifestyle Survey (OPN) which ran weekly during the course of the coronavirus pandemic, as well as the Survey on Living Conditions (SLC), this publication reported on how the pandemic has affected people’s employment, income, savings and debt, as well as their well-being.

Improving Lives: Helping Workless Families indicators and evidence base:

Improving Lives is a compendium of nine indicators which tracks progress in tackling the disadvantages that affect families’ and children’s outcomes. It includes measures of worklessness and problem debt.

Estimates of income and low income for small areas

We do not publish data below the level of region, due to sample sizes. However, there are some related data sources that present information at smaller geographies:

Children in Low Income Families: Local Area Statistics:

These statistics were first published in March 2020. They provide estimates of the number and proportion of children living in relative and absolute (before housing costs) low income by local area across the United Kingdom. They replaced Official Statistics previously published separately by DWP (Children in out-of-work benefit households) and HMRC (Children in low-income families local measure).

Income estimates for small areas, England and Wales: Financial Year Ending 2018:

ONS produce model-based estimates of income at Middle layer Super Output Area (MSOA) level. The publication is available here.

Other sources of information on low income: a number of non-government bodies also produce evidence on low income, including analysis of quantitative data as well as qualitative research on the experiences of people affected by low income. Such organisations include charities, campaign groups, academic research institutes, and think tanks.

Section 6: Revision to the statistics

As noted above, ID uses data produced by the longitudinal survey, Understanding Society (USoc), run by the University of Essex.

Each annual USoc data release provides a revised set of datasets for each wave of the survey. Revisions to previously released data are made where this improves data quality, for example where new information is gathered which was previously missing or incorrect. Statistics derived for a certain time period in this ID publication may therefore be different to those derived for the same time period in a previous ID publication, and may also be subject to future revisions. For this reason it is best to always refer to the most recent ID publication. Please refer to the USoc user guide and specific information on revisions for more information.

Section 7: Status of the statistics

Official statistics

ID is Official Statistics. The Code of Practice for Statistics sets out the standards to which producers of Official Statistics should commit.

Quality Statement

We have worked closely with the University of Essex to review their income derivations. We also compare single year income distribution and low income statistics with HBAI. While we would expect differences between these sources because of different survey vehicles, timings and definitions, as can be seen in the relevant tables, there is a good level of consistency between different data sources. See Table M.8 and M.9 in the ID methodology tables for comparisons between HBAI and ID income distribution and relative low income statistics.

We welcome feedback

We would welcome any feedback on these statistics and would also be interested in knowing how you make use of these data to inform your work and any changes or additions you would like us to consider to improve its usefulness.

Feedback is always welcome. Please contact us via email:

teamincome.dynamics@dwp.gov.uk

Useful links

See our collections page for all material on ID including the detailed tables.

Section 8: Glossary

Adult

All individuals who are aged 16 and over, unless defined as a dependent child (see Child). All individuals aged 16 or over in the household are interviewed as part of USoc.

Benefit units or Family

A single adult or a married or cohabiting couple and any dependent children.

Child

An individual is defined as a (dependent) child for the purposes of this analysis if they are aged under 16 or aged 16 to 18 and in school or non-advanced further education, not married, and living with a parent. If an individual, aged 16 to 18 and in full-time education, did not receive an interview (to determine their educational status), they were assumed to be a dependent child. It does not include any children who have a spouse, partner or child living in the household.

Contemporary median income

The average income for the period covered by the survey. Household incomes are adjusted for inflation so they are in real terms corresponding to the middle January of the latest USoc wave.

Long-standing illness or disability

To establish whether someone has a long-standing illness or disability, USoc asks the following question:

“Do you have any long-standing physical or mental impairment, illness or disability? By ‘long-standing’ I mean anything that has troubled you over a period of at least 12 months or that is likely to trouble you over a period of at least 12 months.”

A follow-up question is used to establish whether respondents have any limiting health problems or disabilities. Up to and including Wave 7, this was only asked of those respondents who identified that they had a long-standing illness or disability:

“Does this / do these health problem(s) or disability(ies) mean that you have substantial difficulties with any of the following areas of your life?”

Twelve areas are listed (full details are in Section 3). If an individual identifies none of these, then the long-standing illness or disability is said to be non-limiting.

Up to and including Wave 7, the above question used to establish the presence of limiting health problems or disabilities was only asked of those respondents who identified that they had a long-standing illness or disability. From Wave 8 onwards, the follow-up question was changed so that it was asked of all adult respondents, regardless of whether they had reported a long-standing illness or disability. The wording of the question was changed accordingly, as set out in full in Section 3, although the 12 listed areas remain the same.

Economic status of the family

Families are allocated to the first applicable category:

• all adults working – Benefit units where all adults regard themselves as working
• some adults working – Benefit units with two adults where only one is working
• workless - Benefit units where no one is working

Individuals who have retired are counted as workless. Therefore, analysis of economic status based on ‘all individuals’ will include retired people in the ‘workless’ total.

Entry and Exit rates

The standard threshold, 60% of median income Before Housing Costs (BHC) is used when analysing transitions into and out of low income.

For an entry into low income to occur, the individual must be in a household whose income is at least 10% below the low income threshold, while in the previous wave they were in a household whose income was above the relative low income threshold.

For an exit from low income to occur, the individual must be in a household whose income is at least 10% above the low income threshold in a wave, while in the previous wave they were in a household whose income was below the relative low income threshold.

Equivalisation

Income measures used in ID take into account variations in the size and composition of the households in which people live. This process is called equivalisation. Equivalisation reflects the fact that a family of several people needs a higher income than a single individual in order for them to enjoy a comparable standard of living.

Ethnicity

Ethnicity information is combined into five groups for the purposes of analysis:

• White – made up of White: British/ English/ Scottish/ Welsh/ Northern Irish; White: Irish; White: Gypsy or Irish Traveller who are resident in England, Scotland or Wales; and White: any other White background
• Mixed – made up of Mixed: White and Black Caribbean; Mixed: White and Black African; Mixed: White and Asian; and Mixed: Any other mixed background
• Asian – made up of Asian/ Asian British: Indian; Asian/ Asian British: Pakistani; Asian/ Asian British: Bangladeshi; Asian/ Asian British: Chinese; and Asian/ Asian British: Any other Asian background
• Black – made up of Black/ Black British: Caribbean; Black/ Black British: African; and Black/ Black British: Any other Black background
• Other – made up of Other Ethnic Group: Arab; Other Ethnic Group: Gypsy or Irish Traveller resident in Northern Ireland; and Other Ethnic Group: Any other ethnic group

Families/family unit

The terms ‘families’ and ‘family units’ are used interchangeably with benefit units. See Benefit unit definition.

Family type

For some analyses, individuals are classified into family type. Individuals are classified according to the status of the benefit unit in which they live. All individuals in a benefit unit (adults and children) will therefore be given the same classification. The classifications are defined below and individuals are allocated to the first applicable category.

• Pensioner couple - a couple where one or more of the adults are State Pension age or over
• Single male pensioner - single male adult of State Pension age or over
• Single female pensioner - single female adult of State Pension age or over
• Mixed-age couple – one of the couple are above State Pension age and one below
• Couple with children - a non-pensioner couple with dependent children
• Single with children - a non-pensioner single adult with dependent children
• Couple without children - a non-pensioner couple with no dependent children
• Single male without children - a non-pensioner single adult male with no dependent children
• Single female without children - a non-pensioner single adult female with no dependent children

Head of household

The head of household, also known as the household reference person, is defined as the owner or renter of the accommodation in which the household lives. If there are multiple owners or renters, the default is the eldest of them is the Household reference person. The Household reference person may change as the household composition changes.

Household

One person living alone or a group of people (not necessarily related) who either share living accommodation or share one meal a day and who have the address as their only or main residence is defined as a household. A child is classed as living in a household if their household is defined as above. A household will consist of one or more benefit units.

Housing costs

Housing costs are rent or mortgage interest payments, plus service and water charges. Consistent with HBAI, for mortgage-holders only the mortgage interest payments are included as housing costs i.e. capital repayment amounts are excluded.

Income

The Before Housing Costs (BHC) income measure used in ID is weekly net (disposable) equivalised household income. This comprises total income from all sources of all household members including dependants.

Income is adjusted for household size and composition by means of equivalence scales. Incomes are adjusted for inflation so they are in real terms corresponding to the middle January of the latest USoc wave (January 2020 in this case).

Income on ID includes:

• Labour income – usual pay and self-employment earnings. Includes income from second jobs
• State support – tax credits and all state benefits including State Pension and Universal Credit.
• Pension income – occupational pensions income
• Investment income – private pensions/annuities, rents received, income from savings and investments
• Private benefit income – includes trade union/friendly society payments, maintenance or alimony and sickness or accident insurance, and income from Student Loans and/ or Tuition Fee Loans.
• Miscellaneous income – educational grants, payments from family members and any other regular payment

Income is net of the following items:

• income tax payments
• National Insurance contributions
• council tax

Income After Housing Costs (AHC) is derived by deducting housing costs from the above BHC income measure.

Income distribution

The spread of incomes across the population.

Low income

Low income is defined using thresholds derived from percentages of median income for the whole population. Individuals are said to be in relative low income if they live in a household with an equivalised income below a percentage of median income BHC or AHC. Relative low income statistics fall if income growth at the lower end of the income distribution is greater than overall income growth.

Mean

Mean equivalised household income of individuals is found by adding up equivalised household incomes for each individual in a population and dividing the result by the number of people.

Median

Median household income divides the population, when ranked by equivalised household income, into two equal-sized groups. Contemporary median income refers to the median income in the survey period being considered.

Pensioner

Pensioners are defined as all those adults above State Pension age (SPa).

For women born on or before 5th April 1950, SPa is 60. For men born on or before this date, SPa is 65. From 6th April 2010, the SPa for women born on or after 6th April 1950 increased gradually between April 2010 and November 2018 to be the same as the SPa for men. The SPa for men remained at 65.

From December 2018, the SPa for both men and women increased, to reach 66 in October 2020. Further increases to bring the SPa to 67 are proposed to be phased in between 2026 and 2028. These changes are detailed here.

Pensioner couples and Mixed-age couples

In the ID pensioner tables, the “pensioner couple” category includes individuals above SPa in couples where the other partner is below SPa. Similarly, the mixed-age couple category in the Working-age adult tables covers adults below SPa in couples where the other partner is above SPa.

Persistent low income

Individuals are described as being in persistent low income if they live in a household in relative low income (below 60% or 70% of median income) in at least three out of four consecutive annual interviews. One possible measure of persistence of low incomes would be to consider only those individuals whose household income was low in each year of the period in question. However, this would exclude people who experienced slightly higher income for very short periods, but whose long-term living standards were not actually very different from those captured in low income in each year.

This issue is relevant because the income distribution is particularly dense around the 60% and 70% of median income thresholds. In addition, some short periods of recorded high income may be due to measurement error and not reflect any real improvement in living standards.

Qualifications

Highest qualification is derived as follows:

• Degree level and above – University Higher Degree (e.g. MSc, PhD); First degree level qualification including foundation degrees, graduate membership of a professional institute, PGCE
• Higher education diploma – Diploma in higher education; Teaching qualification (excluding PGCE); Nursing or other medical qualification not yet mentioned
• A-level etc – A Level; Welsh Baccalaureate; International Baccalaureate; AS Level; Higher Grade/Advanced Higher (Scotland); Certificate of sixth year studies
• GCSE etc – GCSE/O Level; Standard/Ordinary (O) Grade / Lower (Scotland)
• Other qualifications – CSE; Other school (inc. school leaving exam certificate or matriculation)
• No qualifications

Quintile values

Income values which divide the whole population, when ranked by household income, into five equal-sized groups. This helps to compare different groups of the population.

Quintile groups

Five equal-sized groups - the lowest quintile describes individuals with incomes in the bottom 20% of the income distribution. Quintile is often used as a standard shorthand term for quintile group.

Region and country

Regional classifications are based on the standard statistical geography of the former Government Office Regions: nine in England, and a single region for each of Scotland, Wales and Northern Ireland. These regions are built up of complete counties or unitary authorities. Tables also include statistics for England as a whole. For the definitive Office for National Statistics geography classifications and codes see the ONS Open Geography portal.

Although the USoc sample is large enough to allow some analysis to be performed at a regional level, it should be noted that no adjustment has been made for regional cost of living differences, as the necessary data are not available. In the analysis here it is therefore assumed that there is no difference in the cost of living between regions, although the AHC measure will partly take into account differences in housing costs.

Sampling error

The uncertainty in the estimates which arises from taking a random sample of the household population.

State support

The Government pays money to individuals to support them financially under various circumstances. An individual is in receipt of state support if they receive one or more benefits, or are being paid Tax Credits. Most of these benefits are administered by DWP, with the exception of Housing Benefit and Council Tax Reduction which are administered by local authorities, and Child Benefit and Tax Credits which are administered by HM Revenue and Customs. For more information regarding the way in which information on state support is gathered via USoc, please see the Wave 12 questionnaire.

Tax Credits

Working Tax Credits and Child Tax Credits are paid by HM Revenue & Customs. Tax Credits are being phased out, as they are being replaced by Universal Credit.

Tenure

A four-way split is presented:

• Owned outright
• Owned with mortgage
• Social rented – local authority rent or housing association rent
• Private rented – rented from employer, rented privately or other renting

Homes which are part-owned and part-rented are classified as being rented. A small proportion of cases have a tenure classification of ‘other’. These are included in sample sizes but not shown separately in tables.

Working-age adult

Individuals below State Pension age (SPa) who are over 16, not in further education and not classed as a dependent child.