Policy paper

Technical annex

Published 12 September 2023

1. Updates to the Index

In our 2022 State of the Nation report we first introduced our new Social Mobility Index and provided a description of its measurement framework in Technical Annex A. This measurement framework is still applicable to our updated index in this report. However, there have been some updates to the Index which we provide detail on below.

1.1 New mobility outcomes

In our measurement framework last year we showed that we wanted to include a range of mobility outcomes covering: occupational class, income, education, wealth and housing. However, we were only able to provide a picture of absolute and relative mobility on occupational class and income mobility. This year, we have conducted our own analysis to also estimate absolute mobility for education and housing mobility outcomes. We have also included relative mobility for education, housing and wealth.  Unfortunately, our data source for wealth mobility – the ONS Wealth and Assets Survey – does not allow us to estimate absolute levels of parental wealth.  We are therefore unable to report findings on absolute wealth mobility.

1.2 New intermediate outcomes

Last year we introduced a set of experimental indicators to cover career progression (Intermediate Outcome 4). These were experimental because we had not yet concluded on an appropriate data source and methodology for the selection of indicators. Last year our indicators were:

• 4.1: Acquisition of further training and qualifications
• 4.2: Occupational progression
• 4.3: Income progression
• 4.4: Class pay gap

See technical annex A from last year’s report for more details on these indicators. This year, we have developed indicators 4.1, 4.2 and 4.3 using the Labour Force Survey (LFS) instead of the UK Household Longitudinal Study (UKHLS), which we had used last year. This is because the LFS has a much larger sample size and therefore allows us to conduct intersectional analysis of these intermediate outcomes.

This year, we have removed indicator 4.4 on the class pay gap. The class pay gap compares the average pay of people from different class backgrounds who are currently in the same occupational class. Instead we decided to hone in on the returns to education and to compare how these vary between people coming from different socio-economic backgrounds.  Therefore, the following indicators have been added to intermediate outcome 3:

3.5.1 Income returns to education

The purpose of this indicator is to estimate how much one’s education contributes to their future earnings among young people from similar socio-economic backgrounds. This gives us a sense of the magnitude of the income returns to education – that is, the strength of the relationship between education and earnings. However, we cannot be sure that the relationship between educational attainment and earnings is a causal one – the relationship could be due to unmeasured factors such as cognitive skills.

3.5.2 Direct effect of social origins on hourly earnings

The purpose of this indicator is to estimate how much one’s socio-economic background contributes to their future earnings among young people with similar educational levels.  This gives us a sense of the magnitude of the direct effect of social origins on earnings – that is, the strength of the relationship between education and socio-economic background – after taking account of differences in educational level. As above, we cannot be sure that the direct relationship between socio-economic background and earnings is a causal one – the relationship could be due to unmeasured factors such as where people live.

1.3 New drivers

Driver 5: Environment favouring innovation and growth

As shown in chapter 4 of the main report, we have added an additional driver on innovation and growth to our Index. We included this driver because the extent to which the environment favours innovation and growth can potentially be important for social mobility. Much research and economic theory links innovation as a determinant of economic growth. A growing economy is important for social mobility because it will increase the amount of opportunities available for people to experience upward social mobility. Furthermore, this growth can take form in increased productivity, wages and employment opportunities.

We regard this as a largely ‘experimental’ section of the measurement framework, or Index, as we have yet to establish whether the selected indicators are causally linked to social mobility. The rationale for selecting each of these indicators is explained in the main report. In brief, a favourable technical, investment and educational infrastructure can be expected to promote local economic growth, stimulating innovation and expanding professional and business opportunities in the area. Conversely, areas with a less favourable technical infrastructure, less investment, and fewer human capital resources are more likely to miss out on economic growth.

The following 3 indicators – broadband speed, business R and D expenditure and the number of full-time research students – tap these technical, investment and human capital dimensions:

• Driver 5.1: Broadband speed
• Driver 5.2: Business spending on research and development
• Driver 5.3: Numbers of university research students

2. Data sources

In the State of the Nation 2023 report, we use a range of data sources to illustrate the components of the Index. We draw on the Labour Force Survey (LFS) study for the majority of indicators, including both drivers and outcomes. The LFS is a representative sample survey with a large sample size, covering the whole of the UK and enabling one to chart trends over time and across the regions of the UK from 1992 onwards. We supplement the LFS with a range of other large-scale and authoritative data sources.These include other government surveys such as the regular Wealth and Assets Survey, the Community Life Survey, the Annual Survey of Hours and Earnings (ASHE), and the Vacancy Survey. We also draw on data sources based on governmental administrative data such as the Department for Work and Pensions’ (DWP) Housing Below Average Incomes (HBAI) statistics and, for England only, the Department for Education’s (DfE) Education Statistics. We also draw on authoritative non-governmental studies such as the UK Household Longitudinal Study (UKHLS) and the European Social Survey.

All these data sources are conducted according to high methodological standards. For example, all the administrative data that we use are Designated National Statistics. However, we should emphasise that all data sources are subject to various limitations and potential biases. Thus ‘hard to reach’ groups (such as undocumented residents) will tend to be under-represented both in administrative datasets and in sample surveys of the population. In the case of sample surveys, all our data sources use weighting techniques in order to mitigate known biases (although weighting does not necessarily correct for all sources of potential bias). Weighting is not however used in the case of administrative data.

Sample surveys will also be subject to sampling error, and we therefore show confidence intervals wherever possible.

2.1 Office for National Statistics (ONS) Labour Force Survey

The Labour Force Survey (LFS) is a survey of households living at private addresses in the UK. Its purpose is to provide information on the UK labour market which can then be used to develop, manage, evaluate and report on labour market policies. The survey is managed by the Office for National Statistics (ONS) in Great Britain, and the Northern Ireland Statistics and Research Agency (NISRA) in Northern Ireland. The LFS excludes people, such as those resident in communal establishments or who are homeless or are not living in private households. While response rates to the LFS have fallen considerably over the years (down to around 50% just before the COVID-19 pandemic), comparisons between the results from the LFS and those obtained by the Census suggest that (after weighting) the extent of response bias is quite small.^{[footnote 1]}

The LFS is a nationally-representative government survey that covers England, Wales, Scotland and Northern Ireland. The survey has a rolling panel design over 5 quarterly waves, with one-fifth entering the survey and one-fifth leaving at each wave. The July to September wave has been used in each year for our analyses, as this wave has questions on respondents’ socio-economic background. It asks about the household composition, the main wage earner (including if no parent was earning) and the occupation of the main wage earner when the respondent was age 14 years. This has been included in each July to September wave since 2014, meaning that there is now 9 years’ worth of data. These data permit a granular measure of socio-economic background based on the official National Statistics Socio-economic Classification – NS-SEC).

A quarterly main LFS dataset typically contains around 75,000 individuals. However during the COVID-19 pandemic, from July to December 2020, the wave one LFS sample size was doubled to improve achieved sample sizes while response rates were impacted by a pause in face-to-face data collection.

Weights are used throughout the analysis. These weights ensure that estimates reflect the sample design so that cases with a lower probability of selection will receive a higher weight to compensate. They also compensate for differences in the non-response rate among different sub groups of the population. The LFS provides person weights, household weights and income weights. For most analyses we use person weights, but income weights are used when the outcome variable relates to income or earnings.^{[footnote 2]} We use these weights to help make our estimates derived from the LFS more representative of the actual population. We refer to these weights as probability weights in our Report and throughout this annex.

In some instances indicated in the report (notably geographical and intersectional analyses), data is pooled across multiple years in order to achieve sufficient sample sizes. As discussed in the Geographical Analysis section of this annex, the large sample size of the pooled LFS’s means that the data can be disaggregated into International Territorial Level 2 (ITL2) regions. However, it is not possible to disaggregate it further – such as into local authorities – because sample sizes become too small at this level of detail for reliable estimation.

Some important temporary changes were made to the methodology of the LFS as a result of the COVID-19 pandemic, with a move from face-to-face to telephone interviewing (in order to minimise social contact). This had implications both for response rates and for response bias.  Weighting methods were also changed in order to address the potential biases.^{[footnote 3]}

The version of the LFS that we have been given access to for this report includes some measures, such as where the respondent was living when aged 14, which are not available in the version that is available through the UK Data Archive for wider dissemination.

2.2 ONS Wealth and Assets Survey

This year we were able to estimate both housing and wealth mobility outcomes by utilising the Wealth and Assets Survey (WAS). The WAS is a regular survey of around 17,000 households conducted by the Office for National Statistics (ONS) every 2 years. The purpose of the survey is to measure the well-being of households and individuals in terms of their assets, savings, debt and planning for retirement.

The WAS was launched in 2006 and since 2016 has been conducted from April to March to coincide with the financial year. The first wave contained 30,000 individuals and this dropped to around 20,000 for the second to fifth waves. Although the 7th and latest wave was disrupted by the COVID-19 pandemic in 2020, ONS were still able to gather responses from approximately 17,500 households and note that the overall impact of the pandemic on the response rate was minimal.

The WAS has a design incorporating both a panel element and sampling of fresh respondents.  Thus respondents to an earlier round are reinterviewed as far as possible in subsequent rounds. At the same time, there is regular replenishment of the sample with fresh samples of new respondents being added in each round. Overall response rates have been consistently over 50%, although falling somewhat over time. The samples are weighted with design and non-response weights in order to mitigate bias.

ONS state that, as wealth is known to be unevenly distributed, they focus on sampling a disproportionately high number of wealthier households. This is to improve the efficiency of the survey, but results in a sample skew towards the relatively wealthy. Due to this skew, ONS do not generally report the mean values in their dataset releases, but instead focus on the median values. ONS also note that the self-evaluation of wealth by respondents tends to lead to higher estimates of worth than other property indicators may suggest. This implies there may be a tendency for absolute levels of wealth to be overestimated in the data.

2.3 ONS Annual Survey of Hours and Earnings (ASHE)

The Annual Survey of Hours and Earnings (ASHE) is managed by the Office for National Statistics.  It is based on a 1% sample of employee jobs taken from HM Revenue and Customs’ (HMRC’s) Pay As You Earn (PAYE) records.  It started in 2004, replacing the New Earnings Survey, and is carried out in April of each year with a sample of approximately 300,000.  Earnings of the self-employed are excluded.

ASHE is the most comprehensive source of information on the structure and distribution of earnings in the UK. ASHE provides information about the levels, distribution and make-up of earnings and paid hours worked for employees in all industries and occupations. The ASHE tables contain estimates of earnings for employees by sex, full-time or part-time status and other characteristics.

More information: Annual Survey of Hours and Earnings (ASHE) methodology and guidance.

2.4 ONS Vacancy Survey

The ONS Vacancy Survey produces monthly estimates of job vacancies across the whole economy. Questionnaires are sent to a sample of approximately 6,100 businesses every month, approached mainly via head offices. Responses are collected via an electronic questionnaire.  The survey covers all sectors of the economy and all industries in England, Scotland and Wales (Great Britain) with the exception of employment agencies (to avoid double-counting of vacancies) and agriculture, forestry and fishing. Northern Ireland businesses are not approached because of the risk of overlap with other surveys conducted by Northern Ireland departments. Overall, in 2019, the average response rate for the Vacancy Survey was 80.2%.

2.5 ONS General Household Survey (GHS)

The General Household Survey was an inter-departmental multi-purpose continuous survey carried out by the Office for National Statistics collecting information on a range of topics from people living in private households in Great Britain (thus excluding Northern Ireland).  The GHS was a nationally-representative sample survey that ran from 1971 to 2007. However, it did not include questions on parental occupation every year.

More information: General Household Survey.

2.6 Department for Business, Energy & Industrial Strategy (BEIS) /Nesta spatial data tool

BEIS and NestA have co-developed a research and development spatial data tool that allows users to compare indicators that show the scale of R and D systems at a sub-regional (ITL2) level.  The tool is an open data repository. As much as possible, the tool uses data from official sources such as ONS and Higher Education Statistics Agency (HESA) although a few proprietary sources are also used. Formally known as the National Endowment for Science, Technology and the Arts, Nesta is registered as a charity, which supports innovation. It was originally funded by a £250 million endowment from the UK National Lottery. The endowment is managed through a trust, and Nesta uses the interest from the trust to meet its charitable objectives and to fund and support its projects.

More information: BEIS/Nesta Research & Development spatial data tool: Methodology.

2.7 The Department for Digital, Culture, Media & Sport (DCMS) Community Life Survey

The Department for Digital, Culture, Media & Sport (DCMS) took on responsibility for publishing results from the Community Life survey (CLS) for 2016-17 onwards, after it was commissioned by the Cabinet Office in 2012. The survey is representative of adults in England aged 16+ and living in private residences. In 2020/21 the survey was conducted online and paper using ANOS methodology. There were 8787 online interviews and 2130 paper questionnaires returned. The household online response rate was 21.2%.

The Community Life Survey is a key evidence source for understanding more about community engagement, volunteering and social cohesion, sampling adults (aged 16+) throughout England. The CLS moved to a self-completion online and paper mixed method approach from 2016-17 onwards, with an end to the previous face-to-face method. Fieldwork for the Community Life Survey 2021/22 was conducted between October 2021 and September 2022, with samples issued on a quarterly basis. Overall, 10,126 interviews were achieved over the year, which represents a population-representative overall (online or paper) individual response rate of 23%.

More information: Community Life Survey: Technical report.

2.8 DCMS Taking Part Survey

The Taking Part survey is a continuous face to face household survey of adults aged 16 and over and children aged 5 to 15 years old in England. In 2018/9 the sample size was just over 8000 and the response rate was 50.5%. The survey is commissioned by the Department for Culture, Media and Sport (DCMS) and 3 partner organisations (Arts Council England, Historic England and Sport England). The survey has run since 2005 and is the main evidence source for DCMS and its sectors. Its main objective is to provide a central, reliable evidence source that can be used to analyse cultural, digital, and sporting engagement, providing a clear picture of why people do or do not engage.

More information: Taking Part Survey

2.9 Department for Education (DfE) Early Years Foundation Stage Profile results in England

The EYFSP is a teacher assessment of children’s development in England at the end of the early years foundation stage (the end of the academic year in which the child turns 5 years old – this is typically at the end of the reception year). All providers of state-funded early years education in England are within the scope of the EYFSP teacher assessments including academies, free schools and private, voluntary and independent (PVI) providers. However, the proportion of the age group that is covered by the EYFSP is not reported.

Data is collected by the DfE from local authorities covering state-funded schools and private, voluntary and independent (PVI) providers (including childminders) as part of the EYFS profile return. This data is then matched to other data sources, including the school and early years censuses, to obtain information on pupil characteristics such as ethnicity and eligibility for Free School Meals (the main measure of socio-economic background that is available in the dataset).  It is likely that there is some missing data on characteristics and eligibility for Free School Meals but this is not reported. The results are published by the DfE.

More information: Early years foundation stage profile results.

2.10 DfE National Curriculum Assessments at KS2 in England

Statutory testing and assessment for pupils in primary schools is the responsibility of the Standards and Testing Agency (STA), an executive agency of the Department for Education. KS2 tests must be administered by state-funded schools, the STA organises test marking and the return of results to schools. KS2 teacher assessments are also collected by STA and the information is collated and passed on within the department. Independent schools, non-maintained special schools and pupil referral units may take part in the KS2 assessments if they wish to do so.  The position of home-schooled children is not specified and the proportion of the age group taking part in the KS2 assessments is not reported.

The attainment data is combined with information on pupil characteristics such as ethnicity and eligibility for Free School Meals (the main source of socio-economic background available in the dataset) taken from the school census. Details of this data are provided in a separate quality and methodology document. It is likely that there is some missing data on ethnicity and eligibility for free school meals but this is not reported.  In addition eligibility for Free School Meals is only known for students at state-funded schools. Those who are in private education, or who are home schooled are therefore excluded from analyses which take account of disadvantage status. However, the DfE does not report what proportion of the age group are excluded in this way. Given that pupils in private education may be less likely to be disadvantaged than those in state-funded education, it is likely that measures such as the disadvantage gap underestimate the true extent of inequality among children and young people in English schools.

More information: Attainment in primary schools in England: Quality and methodology information (PDF).

2.11 DfE National Curriculum Assessments at KS4 in England.

The ‘total’ includes pupils for whom Free School Meal eligibility (FSM), Special Educational Needs status (SEN provision) or SEN primary need could not be determined. This figure also includes pupils at further education colleges: as FE colleges do not complete the school census, we do not have matched pupil characteristics data of pupils in FE colleges and therefore these pupils are not included in characteristics breakdowns. This means that there are some cases where the individual characteristics breakdowns will not add up to the all pupils figure. From 2014/15, disadvantaged pupils include pupils known to be eligible for free school meals (FSM) in any spring, autumn, summer, alternative provision or pupil referral unit census from year 6 to year 11 or are looked after children for at least one day or are adopted from care.

More information: Key stage 4 performance: methodology.

2.12 DfE Participation in education, training and employment age 16-18 statistics

These statistics from the Department for Education cover young people who reside in England and is based on their academic age, that is their age at the start of the academic year, 31st August. The data is at national level only and cannot be disaggregated to sub-national levels, or by characteristics other than gender.

The population at each age is based on Office for National Statistics (ONS) mid-year estimates, adjusted so that they relate to academic age and the end of the calendar year.  Participation data from administrative sources is then subtracted from this total.  Participation estimates are made by combining administrative data from schools, further education, work-based learning (apprenticeships) and higher education. Procedures are included to identify young people in more than one form of provision, to give a view of the cohort as a whole.

The labour market status (whether a young person is employed, unemployed or economically inactive) is then estimated from the Labour Force Survey (LFS) for each of the major groups:

• Full time education (FTE)
• Work based learning (WBL), comprises solely of apprenticeships from 2013
• Employer funded training (EFT)
• Other education or training (OET)
• Not in education or training (NET)

Those in the NET group whose labour market status is economically inactive or ILO^{[footnote 4]} unemployed are concluded to be NEET.

More information: Participation in education, training and employment age 16 to 18: Methodology.

2.13 Department for Work and Pensions (DWP) Households Below Average Income (HBAI) statistics

The Department for Work and Pensions’ (DWP) Households Below Average Income (HBAI) report presents information on living standards in the United Kingdom and is the foremost source for data and information about household income and inequality in the UK. It has provided annual estimates on the number and percentage of people living in low-income households since 1995.

The statistics are based on the Family Resources Survey (FRS). The FRS is a continuous household survey which collects information on a representative sample of private households in the United Kingdom with a sample size of around 10,000 households and a response rate of around 25%.

More information: Households below average income series: quality and methodology information report FYE 2022.

2.14 Higher Education Statistics Agency (HESA) UK performance indicators

The Higher Education Statistics Agency (HESA) produced UK performance Indicators from 2002/3 but these were discontinued after the 2020/21 statistics were published in 2022.

UK Performance Indicators (UKPIs) were statistics which compared universities and colleges against benchmarks for Widening participation, Non-continuation, and the Employment or further study of graduates. All the tables were based on undergraduate students who were residents of England, Scotland, Wales or Northern Ireland before starting their course. The statistics appear to have been based on returns from HE providers.^{[footnote 5]}

2.15 OECD online education database

This database includes raw data used for the computation of indicators published in Education at a Glance. The database is compiled on the basis of national administrative sources, reported by Ministries of Education or National Statistical offices according to international standards, definitions and classifications. The collected annual data cover the outputs of educational institutions, the policy levers that shape educational outputs, the human and financial resources invested in education, structural characteristics of education systems, and the economic and social outcomes of education.

2.16 UK Household Longitudinal Study

Some indicators that form the basis for the Index draw on Understanding Society, also known as the UK Household Longitudinal Study (UKHLS). The UKHLS is a longitudinal survey of the members of approximately 40,000 households in the UK with a household response rate of 57% at round 1.

The study is based at the Institute for Social and Economic Research at the University of Essex and is funded by the Economic and Social Research Council (ESRC) and the British Academy. The study covers the whole of the UK and also has booster samples for ethnic minority and immigrant groups. Information is collected on all members of each household, and each year recruited households are visited to collect information on changes to their individual and household circumstances.

The purpose of the UKHLS is to provide high-quality longitudinal data on subjects such as health, work, education, income, family, and social life. This helps to understand the long-term effects of social and economic change, as well as policy interventions designed to impact the general wellbeing of the UK population. To do this the study collects both objective and subjective indicators and offers opportunities for research within and across multiple disciplines including sociology, economics, geography, psychology and health sciences.

The UKHLS started with a representative sample of households in 2009/10.  It also incorporated respondents who had previously participated in the British Household Panel Study (see further below). Since 2009/10 there have been annual waves with repeat interviews of sample members. As with all panel studies of this kind, there is attrition over time, with some participants dropping out of the study. The sample is however replenished, through existing sample members who leave the original household and establish a new household of their own.

The attrition of the original sample members is not random but tends to be greater among some ethnic minority groups and among those from disadvantaged backgrounds. Weighting is therefore used in order to mitigate any resulting bias. However, weighting cannot guarantee that there will be no biases with respect to particular subgroups of the population.

The greatest strength of the UKHLS for social mobility research is that it follows up as many young people as possible from their original households where they lived as children into adulthood and the new household that they establish. This enables us to link data on parents with that on their adult children and thus permits analysis of intergenerational educational and income mobility.

More information: Panel attrition in the General Population Sample and the Immigrant and Ethnic Minority boost of Understanding Society (PDF).

2.17 British Household Panel survey (BHPS)

The British Household Panel Survey began in 1991 and was incorporated into the second wave of the UKHLS in 2010/11. The BHPS was managed by the Institute for Social and Economic Research (ISER) at the University of Essex and was funded by the ESRC.

The BHPS was household-based, interviewing every adult member of sampled households, and followed the same representative sample of individuals – the panel – over a period of years.  The wave 1 panel consisted of around 5,500 households and 10,300 individuals drawn from 250 areas of Great Britain, with a household response rate of 74%. Additional samples of 1,500 households in each of Scotland and Wales were added to the main sample in 1999, and in 2001 a sample of 2,000 households was added in Northern Ireland, making the panel suitable for UK-wide research. The waves 1-18 BHPS documentation, which includes the latest User Guide and Questionnaires, is available from the UK Data Archive.

More information: British Household Panel Survey

2.18 European Social Survey

The European Social Survey (ESS) is an academically driven cross-national survey that has been conducted every 2 years across Europe since 2001. The ESS was awarded European Research Infrastructure Consortium (ERIC) status in November 2013. It is directed by a Core Scientific Team led from City, University of London (UK) alongside 6 other partner institutions.  The UK has participated in every round since the inception of the ESS. The survey has been funded in the UK by the ESRC.

ESS samples are representative of all persons aged 15 and over (no upper age limit) resident within private households in the UK, regardless of their nationality, citizenship or language. Individuals are selected by strict random probability methods at every stage and a minimum effective achieved sample size of 1,500 is aimed for after discounting for design effects.

The ESS sample design in the UK is a clustered and stratified two-stage random probability design and excludes the following areas: Highlands and Islands, the Isle of Man and the Channel Islands. The sampling frame used is the Post Office Address File and is a sample of addresses. 5885 issued sample units in round 10 yielding 1249 valid interviews, a response rate of 21%.

3. Protected characteristics analysis

Under the Equality Act 2010 the protected characteristics in the UK are: age; disability; gender reassignment, marriage and civil partnership, pregnancy and maternity, race (including colour, nationality, and ethnic or national origins), religion or belief, sex and sexual orientation. In conducting our analysis we have focussed on intersectionality between socio-economic background and the protected characteristics sex, disability, and race / ethnicity (where suitable measures are available in the data sources). We also draw attention to age differences in the main report.  Unfortunately we have not been able to look at the intersectionality with other protected characteristics such as gender reassignment because of the absence of suitable measures and/or because of the imprecision of the estimates due to small sub-sample sizes for some protected characteristics.

3.1 Sex

Under the Equality Act 2010, sex (being a man or a woman) is a protected characteristic, as is gender reassignment. However, in the case of surveys such as the Labour Force Survey, sex is self-reported by the respondents themselves and is thus self-defined.  Respondents may differ in the precise meaning that they attach to the term.  In the case of education statistics, the DfE uses the term gender rather than sex and states that “The gender of the child is recorded as male or female in the school census or early years census. In exceptional circumstances a setting may be unsure as to which gender should be recorded for a particular child. The advice from the department is to record the gender according to the wishes of the pupil and/or parent.”

Our main data sources do not therefore distinguish between ‘sex registered at birth’ and ‘gender identity’ in the way that the 2021 Census did.  Empirically however the 2 concepts overlap very substantially. In the 2021 Census 94.0% of the population aged 16 years and over answered Yes to the question “Is the gender you identify with the same as your sex registered at birth?”.

3.2 Race/Ethnicity

Under the Equality Act 2020 race is taken to include colour, nationality, and ethnic or national origins. In practice however all our sources, both survey-based and administrative, are based on the Census approach to measuring ethnic group. The first Census to measure ethnic groups was the 1991 Census and there have been small modifications to the Census question subsequently.  In recent Censuses, respondents are asked ‘What is your ethnic group?’ and are asked to tick one of a number of boxes such as ‘White – British’, ‘Black or Black British – Caribbean’, ‘Asian or Asian British – Indian’. One should probably regard measures based on the Census questions as ones measuring ethnicity or ethnic identity rather than the Equality Act concept of race or ethnic origins.

In the case of the LFS, respondents are first asked to place themselves in one of 7 broad ‘top level’ categories (White, Mixed/multiple ethnic groups, Black/Black British, Asian/Asian British, Chinese, Arab and Other) and are then asked to place themselves within more specific groups within the selected broad category (for example, within the broad White category, respondents are asked to select one of the 4 categories (1) English/Welsh/Scottish/Northern Irish/British, (2) Irish, (3) Gypsy or Irish Traveller, (4) any other white background). See LFS User Guide 2022.

The number and content of the Census and survey categories have differed somewhat over time and also differ in Scotland and Northern Ireland from those used in England and Wales.  We have therefore attempted to construct harmonised categorizations that are reasonably compatible over time and between nations. When analysing the LFS and the UKHLS, we therefore use the following 10 category measure of ethnic group:

• White British
• White Other
• Mixed
• Asian or Asian British – Indian
• Asian or Asian British – Pakistani
• Asian or Asian British – Bangladeshi
• Black or Black British – Caribbean
• Black or Black British – African (including Black other)
• Chinese
• Other

In the case of Department for Education administrative statistics, ethnic group is shown either using a simplified ‘top level’ classification of 5 broad ethnic groups, namely:

• White
• Mixed
• Black
• Asian
• Other

Or alternatively, for some analyses of English administrative data, the DfE uses a more granular measure with the following 18 categories:

• White – British
• White – Irish
• White – Gypsy/Roma
• White – Traveller of Irish heritage
• White – any other white background
• Mixed – White and Black Caribbean
• Mixed – White and Black African
• Mixed – White and Asian
• Mixed –  Any other mixed background
• Asian – Indian
• Asian – Pakistani
• Asian – Bangladeshi
• Asian – Chinese
• Asian – any other Asian
• Black – Black Caribbean
• Black – Black African
• Black – Any other Black background
• Any other ethnic group

How granular a measure of ethnic group can be used will depend on the number of respondents in each category and on whether the measure is harmonised for use across the UK as a whole.  In general, we prefer to use the most granular classification that yields reliable estimates.

3.3 Disability

Under the Equality Act 2010, a person is disabled if they have a physical or mental impairment that has a ‘substantial’ and ‘long-term’ negative effect on their ability to do normal daily activities. ‘Substantial’ means that, for example, it takes much longer than it usually would to complete a daily task like getting dressed. ‘Long-term’ means 12 months or more.

To measure disability we use the LFS variable DISEA. This provides a measure of disability consistent with the Equality Act. It is derived from 2 questions, first, “Do you have any physical or mental health conditions or illnesses lasting or expecting to last 12 months or more?” and secondly “Does your condition or illness reduce your ability to carry out day-to-day activities?”.^{[footnote 6]} In the UKHLS, disability is measured somewhat differently. Respondents were asked: “Do you have any long-standing physical or mental impairment, illness or disability? ‘Long-standing’ means anything that has troubled you or is likely to trouble you over a period of at least 12 months.” This is a somewhat broader measure than that of the LFS variable DISEA since it does not explicitly refer to effects on the activities of daily living.

In the case of the Wealth and Assets Survey respondents were asked “Do you have any long-standing illness, disability, or infirmity? By long-standing I mean anything that has troubled you over a period of time or that is likely to affect you over a period of time.”

4. Methodology and analysis

4.1 Indicators methodology

Below we have included a more detailed description of the methodology used to derive each of our indicators. In some instances, the methodology used to create some of the charts for an indicator can be quite different, in those cases we have added an extra table to outline the specific methodology.

4.2 Mobility outcomes

Short-range absolute occupational mobility

• Definition: Percentages of men experiencing short-range occupational mobility (upward, downward, and total) by birth cohort.
• Unit of measurement: Percent
• Time period covered: Birth cohorts ranging from 1910-19 to 1990-99
• Methodology: Short-range occupational mobility means moving from one broad occupational category to an adjacent one. For example, moving from an intermediate origin to a working-class or professional occupation would be short-range mobility, as would moving from working class to intermediate, or professional to intermediate.
• Data source: The General Household Survey (1972 to 2005), British Household Panel Survey (1991 to 2009), Taking Part Survey (2005 to 2006), Understanding Society (2010 to 2019) and Labour Force Survey (2014 to 2022), respondents aged 25 to 65 years.
• Notes: The figures for total mobility are the sum of the percentages upwardly and downwardly mobile. This represents the percentage of the sample as a whole who were in a different social class position from the one in which they were brought up (based on 5 social classes: professional, intermediate, own account, skilled manual, unskilled manual. Classes differ from those used elsewhere in this report, due to data availability).
• Figures: 2.0 and 2.2

Long-range absolute occupational mobility

• Definition: Percentages of people of working-class origin in professional jobs, and people of professional origin in working-class jobs by birth cohort.
• Unit of measurement: Percent
• Time period covered: Birth cohorts ranging from 1910-19 to 1990-99
• Methodology: Long-range occupational mobility means moving either from a working-class origin to a professional occupation, or a professional origin to a working-class occupation.
• The figures for long-range upward mobility are percentages of those from a working-class background who went on to work in a professional occupation. The figures for long-range downward mobility are the opposite, that is, percentages of those from a professional background who went on to work in a working-class occupation.
• Data source: The General Household Survey (1972 to 2005), British Household Panel Survey (1991 to 2009), Taking Part Survey (2005 to 2006), Understanding Society (2010 to 2019) and Labour Force Survey (2014 to 2022), respondents aged 25 to 65 years.
• Notes: Analysis is based on 5 social classes: professional, intermediate, own account, skilled manual, unskilled manual. Classes differ from those used elsewhere in this report, due to data availability.
• Figure(s): 2.1 and 2.3

Absolute occupational mobility

• Definition: Percentage of people in each occupational class position by highest level of parental occupational class.
• Unit of measurement: Percent
• Time period covered: Figure 2.4: 2022, Figure 2.5: 2018 to 2022, Figure 2.6: 2022, Figure 2.7: 2014 to 2022, Figure 2.8: 2014 to 2022, Figure 2.9: 2018 to 2022
• Methodology: Percentage of people in each occupational class position by highest level of parental occupational class. Parental social class in the LFS is measured by asking respondents to recall the occupation of the main wage earner in their household when they were aged 14 years.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2022, respondents aged 25 to 64 years in the UK, data collected from July to September 2022. For ethnicity, disability and regional breakdown: Office for National Statistics, pooled Labour Force Survey 2018 to 2022, respondents aged 25 to 64 years in the UK, data collected from July to September each year.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 2.4, 2.5, 2.6, 2.7, 2.8 and 2.9

Relative occupational mobility

• Definition: Relative chances of achieving occupational class outcomes compared to people from other socio-economic backgrounds.
• Unit of measurement: Log odds
• Time period covered: Figure 2.10: 2014 to 2022, Figure 2.11: 2018 to 2022, Figure 2.12: 2018 to 2022, Figure 2.13: 2018 to 2022
• Methodology: The estimates are derived from the uniform difference (UNIDIFF) parameter. The UNIDIFF model assumes that all odds ratios differ by a common multiplier in comparison with a benchmark group (in this case, year 2014). This common percentage is expressed in log form, the log UNIDIFF parameter shown in the charts. When the logged value is near 0, there is no change in the odds ratios – relative mobility is constant across all groups. But when it is negative, the link between origins and destinations is weaker – the odds ratios are lower, and relative mobility is higher. At very large negative values, there would be almost no link at all between origins and destinations.
• Data source: Office for National Statistics, Labour Force Survey 2014 to 2022, respondents aged 25 to 64 years in the UK, data collected from July to September each year.
• Figure(s): 2.10, 2.11, 2.12 and 2.13

Absolute income mobility

• Definition: Percentage of people whose real income is higher than that of their parents when they were the same age by birth cohort.
• Unit of measurement: Percent
• Time period covered: Birth cohorts from 1960 to 1987
• Methodology: The upward absolute mobility rate is the percentage of children in each birth cohort whose pre-tax, post-transfer family real income (adjusted for inflation) at age 30 years was higher than their parents’ family income at age 30 years. This indicator is broken down by country and birth cohort from 1960 to 1987. Countries included are: US, Sweden, Denmark, Norway, Canada, Finland and the Netherlands.
• Data source: Manduca and others (2020). Trends in absolute income mobility in North America and Europe.
• Notes: For a full detail of the methodology see “Data and Methods” section in the original data source.
• Figure(s): 2.14

Relative income mobility

• Definition: Strength of intergenerational income mobility (intergenerational elasticity)
• Unit of measurement: Intergenerational Elasticity of income
• Time period covered: Years covered are various cohorts from 1975 to 2020, see data source section below for more detail.
• Methodology: Intergenerational elasticity of income is an estimate of the percentage change in one’s income associated with every 1% increase in one’s parent’s income.  There are major differences between the data and methodology of the various studies used to make these estimates, and there is also considerable imprecision in the estimates themselves. The 1975 to 1978 study measured fathers’ and sons’ weekly earnings. The age of the sons was not specified. The 1991 study measured parental household income and their children’s earnings at age 33 years. The 2004 study measured parental household income and their children’s earnings at age 34 years. The 2009 to 2016 study measured parental household gross income and adult children’s gross household income at age 25 years or older. The 2020 study measured parental household gross income and sons’ (and daughters’) gross personal income at age 25 years or older (average age 30 years). The 2020 results are estimated with a linear regression model. Regression model of the form Y^Son = α + βY^parent + u, where Y^Son represents the son’s income (logged) and Y^parent represents the parents’ income (logged). α is the intercept, β is the regression coefficient representing the strength of association between parents’ and adult children’s income, and u is an error term. The regression coefficient (also known as intergenerational elasticity) has a natural interpretation. For example, a coefficient of 0.3 means that if 2 families have (log) incomes that differ by 10%, their sons’ (log) income will differ by about 3%. For detailed methodology on intergenerational elasticity for 2020 figures see the “Regression (Intergenerational Elasticity (IGE) and rank-rank co-efficient)” section below
• Data source: 1975 to 1978: Atkinson and others (1981) based on follow up of the 1950 social survey of York; 1991 and 2004: Blanden and Machin (2008) based on National Child Development Study and Birth Cohort Study 1970; 2009 to 2016: Rohenkohl (2019) based on linked British Household Panel Study and UK Household Longitudinal Study (UKHLS) data; 2020: own calculations based on the UKHLS 2009 and 2020.
• Figure(s): Table 2.15

Absolute educational mobility

• Definition: Percentage of people by their highest level of qualification and the highest level of qualification of their parents.
• Unit of measurement: Percent
• Time period covered: 2020 for all figures for this indicator
• Methodology: Parental education is measured by using whichever parent has the higher level of qualification. If there is data on only one parent, then only this data is used. The available measure of parental education in the UKHLS distinguishes those with university degrees, those with some post-school qualification, those with a school qualification, and those without any qualification. The respondents’ own highest level of qualification has been re-coded into the same 4 categories. For the ethnicity and regional breakdown the sample has been restricted to people whose parents did not have a degree. The percentages in the figure can be interpreted as the proportion of those from a non-graduate family who are upwardly mobile educationally.
• Data source: The UK Household Longitudinal Survey (UKHLS), 2020 calendar year, respondents aged 25 to 64 years in the UK.
• Notes: The data used is weighted using the UKHLS probability weights.
• Figure(s): 2.16, 2.17, 2.18, 2.19 and 2.20

Relative educational mobility

• Definition: Parent:adult children odds ratios relating to university degrees
• Unit of measurement: Odds ratio
• Time period covered: Figure 2.21: 1991 to 2020, Figure 2.22: 2020, Figure 2.23: 2020, Figure 2.24: 2020, Figure 2.25: 2020
• Methodology: Education is a binary measure of attainment of an undergraduate degree qualification. A higher odds ratio indicates greater intergenerational persistence while a lower odds ratio indicates greater relative mobility. For detailed description on the construction of odds ratios please see the Odds ratio section.
• Data source: UK Household Longitudinal Survey (UKHLS), 1991 to 2020. Respondents aged 28 to 37 years in the UK.
• Notes: The data used is weighted using the UKHLS population weights.
• Figure(s): 2.21, 2.22, 2.23, 2.24 and 2.25

Absolute housing mobility

• Definition: Respondents’ home ownership status by their parental home ownership status.
• Unit of measurement: Percent
• Time period covered: For all figures for this indicator: 2016 to 2020
• Methodology: This indicator shows the current tenure by parental tenure. For example, 71% of those who own a house in adulthood had parents owning a house when a teenager. The error bars show 95% confidence intervals. The data used is weighted using the WAS individual weights.
• Data source: Wealth and Assets Survey (WAS) waves 6 (from 2016 to 2017) and 7 (from 2018 to 2020). Respondents aged 25 to 64 years in the UK.
• Figure(s): 2.26, 2.27, 2.28 and 2.29

Relative housing mobility

• Definition: Odds ratios of the relationship between parental and respondent home ownership among younger respondents.
• Unit of measurement: Odds ratio
• Time period covered: Figure 2.30: 1991 to 2020, Figure 2.31: 2016 to 2020, Figure 2.32: 2016 to 2020, Figure 2.33: 2016 to 2020
• Methodology: Ratio of the odds (of owning a house or not) among those whose parents owned a house to the odds among those whose parents had not. For detailed description on the construction of odds ratios please see the “Odds ratio” section. The data used is weighted using the WAS individual weights.
• Data source: Wealth and Assets Survey (WAS) waves 6 and 7 (respondents aged 30 to 34 years) and Bell and others (2022, table 5, respondents aged 28 to 37 years) in the UK.
• Figure(s): 2.30, 2.31, 2.32 and 2.33

Levels of wealth

• Definition: Financial wealth, pension wealth, physical wealth, property wealth and total wealth by age group.
• Unit of measurement: Pounds (£)
• Time period covered: Figure 2.34: 2016 to 2020
• Methodology: Level of financial wealth, pension wealth, physical wealth, property wealth and total wealth by age group. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index. For further details see the Office for National Statistics (2022) household total wealth in Great Britain: April 2018 to March 2020. The data used is weighted using the WAS individual weights.
• Data source: Wealth and Assets Survey (WAS) waves 6 (from 2016 to 2018) and 7 (from 2018 to 2020), respondents aged 25 to 64 years.
• Figure(s): 2.34

Relative wealth mobility:

• Definition: Strength of intergenerational wealth mobility (intergenerational elasticity)
• Unit of measurement: Intergenerational elasticity of wealth
• Time period covered: 2016 to 2020, see data source section below for more detail.
• Methodology: Intergenerational elasticity of wealth is an estimate of the percentage change in one’s wealth associated with every 1% increase in one’s parent’s wealth. Logarithms of both parents’ and adult children’s wealth (for those whose wealth was not 0). The degree of intergenerational persistence can then be estimated with a linear regression model as with wealth mobility. Regression model of the form Y^Son = α + βY^parent + u, where Y^Son represents the son’s wealth(logged) and Y^parent represents the parents’ wealth(logged). α is the intercept, β is the regression coefficient representing the strength of association between parents’ and adult children’s wealth, and u is an error term. The regression coefficient (also known as intergenerational elasticity) has a natural interpretation. For example, a coefficient of 0.3 means that if 2 families have (log) wealth that differ by 10%, their sons’ (log) wealth will differ by about 3%. Total wealth estimates for respondents are derived by adding up the value of different types of assets owned by households and subtracting any liabilities. Estimates of parental wealth are imputed using a 2-stage least squares method. For detail see “Two-stage imputation” section. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index. More detailed on the 2020 results or “For detailed methodology on intergenerational elasticity for 2020 was calculated see “Regressions” section below.
• Data source: Wealth and Assets Survey (WAS) waves 6 (from 2016 to 2018) and 7 (from 2018 to 2020), respondents aged 25 to 64 years.
• Figure(s): No figure for this indicator

4.3 Intermediate outcomes

Intermediate Outcome 1.1: Level of development at age 5 years

• Definition: Percentage of students achieving a ‘good level of development’ at age 5 years by eligibility for FSM.
• Unit of measurement: Percent
• Time period covered: Figure 3.4: Academic year 2012/13 to 2021/22, Figures 3.5, 3.6, 3.7 and 3.8: Academic year 2021/22
• Methodology: This indicator captures the percentage of students achieving a good level of development at age 5 years in England. A child achieving at least the expected level in the early learning goals within the 3 prime areas of learning and within literacy and numeracy is classed as having ‘a good level of development’. The EYFSP is a teacher assessment of children’s development at the end of the early years foundation stage (the end of the academic year in which the child turns 5 years old – this is typically at the end of the reception year). All providers of state-funded early years education in England are within the scope of the EYFSP teacher assessments including: academies, free schools and  private, voluntary and independent (PVI) providers. To capture socio-economic background, results are  split by eligibility for free school meals (FSM). FSM eligibility is defined as collected in the school census which states whether a child’s family have claimed eligibility. Parents are able to claim FSM if they receive certain benefits.
• Data source: Department for Education. Early years foundation stage (EYFS) profile results from the 2021 to 2022 academic year, 2022.
• Notes: The EYFS was significantly revised in September 2021 which means we cannot directly compare the outcomes for 2021 to 2022 with earlier years. Due to the COVID-19 pandemic, the early years foundation stage profile (EYFSP) results in England publication was cancelled for 2019 to 2020.
• Figure(s): 3.4, 3.5, 3.6, 3.7 and 3.8

Intermediate Outcome 1.2: Attainment at age 11 years

• Definition: Percentage of students reaching the expected standard in reading, writing and maths at KS2 by disadvantage status.
• Unit of measurement: Percent
• Time period covered: Figure 3.9: Academic year 2015/16 to 2021/22, Figures 3.11, 3.12 and 3.13: Academic year 2021/22
• Methodology: Proportion of pupils who meet the expected standard in all 3 of reading, writing and maths (combined) at key stage 2 (age 11 years). Disadvantaged pupils are defined as those who were registered as eligible for free school meals at any point in the last 6 years, and children looked after by a local authority (LA) or who left LA care in England and Wales through adoption, a special guardianship order, a residence order or a child arrangements order. For more details see Methodology of KS2 attainment.
• Data source: Department for Education. National curriculum assessments at KS2 in England, 2022.
• Notes: Attainment in all of reading, writing and maths is not directly comparable to some earlier years (2016 and 2017) because of changes to teacher assessment frameworks in 2018. No data was collected for the 2 academic years starting in 2019 and 2020 due to the COVID-19 pandemic.
• Figure(s): 3.9, 3.11, 3.12 and 3.13

Intermediate Outcome 1.2: Disadvantage attainment gap index at key stage 2 (KS2)

• Definition: Disadvantage attainment gap index at key stage 2 (KS2)
• Unit of measurement: Disadvantage gap index units
• Time period covered: Academic year 2010/12 to 2021/22
• Methodology: To create the index comparisons are made by ordering pupil scores in reading and maths assessments at the end of key stage 2 and assessing the difference in the average position of disadvantaged pupils and others. The mean rank of pupils in the disadvantaged and other pupil groups are subtracted from one another and multiplied up by a factor of 20 to give a value between -10 and +10 (where 0 indicates an equal distribution of scores). Disadvantaged pupils are defined as: those who were registered as eligible for FSM at any point in the last 6 years, children looked after by a local authority or have left local authority care in England and Wales through adoption, a special guardianship order, a residence order or a child arrangements order.
• Data source: Department for Education. National curriculum assessments at KS2 in England, 2022.
• Notes: Between the academic years 2018 to 2019 and 2021 to 2022, there was a break in assessments due to the pandemic.
• Figure(s): 3.10

Intermediate Outcome 1.3: Attainment at age 16 years

• Definition: Percentage of students achieving a pass (grade 5 or above) in both GCSE English and maths by disadvantage status.
• Unit of measurement: Percent
• Time period covered: Figure 3.14: 2018 to 2022, Figures 3.16, 3.17 and 3.18: Academic year 2021/22
• Methodology: This covers the attainments in Maths and English GCSEs of students at state-funded schools using a positional measure of attainment (students achieving a pass at grade 5 or above in both subjects). Pupils are defined as disadvantaged if they are known to have been eligible for FSM at any point in the past 6 years (from year 6 to year 11), if they are recorded as having been looked after for at least one day or if they are recorded as having been adopted from care. For more details see Methodology of KS4 performance.
• Data source: Department for Education (DfE). National curriculum assessments at key stage 4 in England, 2022.
• Notes: The 2021 to 2022 year assessment returned to the summer exam series, after they had been cancelled in 2020 and 2021 due to the impact of the COVID-19 pandemic. During this time alternative processes were set up to award grades (centre assessment grades, and teacher assessed grades).
• Figure(s): 3.14, 3.16, 3.17 and 3.18

Intermediate Outcome 1.3: Disadvantage attainment gap index at key stage 4 (KS4)

• Definition: Disadvantage attainment gap index at key stage 4 (KS4)
• Unit of measurement: Disadvantage gap index units
• Time period covered: Academic years 2010/11 to 2021/22
• Methodology: The disadvantage gap index is intended to provide a more resilient measure of changes over time in attainment that may have been affected by, for example, the GCSE reforms introduced in 2017 and associated changes to headline measures (for example, moving away from 5 or more GCSEs to average attainment 8 scores). The disadvantage gap index summarises the relative attainment gap (based on the average grades achieved in English and Maths GCSEs) between disadvantaged pupils and all other pupils. The index ranks all pupils in state-funded schools in England and asks whether disadvantaged pupils typically rank lower than non-disadvantaged pupils. A disadvantage gap of zero would indicate that pupils from disadvantaged backgrounds perform as well as pupils from non-disadvantaged backgrounds. We measure whether the disadvantage gap is getting larger or smaller over time. While the absolute differences (in English and Maths GCSE grades) may differ between years the gap index measures results in terms of how disadvantaged pupils are ranked in comparison to non-disadvantaged pupils therefore it offers greater comparability between years. Pupils are defined as disadvantaged if they are known to have been eligible for FSM at any point in the past 6 years (from year 6 to year 11), if they are recorded as having been looked after for at least one day or if they are recorded as having been adopted from care. For more details see ‘Statistical working paper measuring disadvantaged pupils’ attainment gaps over time (updated)’.
• Data source: Department for Education (DfE). National curriculum assessments at key stage 4 in England, 2022.
• Notes: The 2021 to 2022 year assessment returned to the summer exam series, after they had been cancelled in 2020 and 2021 due to the impact of the COVID-19 pandemic. During this time alternative processes were set up to award grades (centre assessment grades, and teacher assessed grades).
• Figure(s): 3.15

Intermediate Outcome 2.1: Destinations following the end of compulsory full-time education

• Definition: Proportion of young people aged 16 to 24 years who are in education and training, employment, or NEET by socio-economic background (SEB)
• Unit of measurement: Percent
• Time period covered: Figure 3.19: 2022, Figures 3.20, 3.21, 3.22 and 3.23: 2014 to 2022
• Methodology: Education and training captures young people who satisfy the following definition: people aged 16 to 24 years who are in full-time education or training of any type. Those in training were included with those in education due to small sample sizes.
  
  Employment is defined as those aged 16 and over, who are in employment if they did at least one hour of work in the reference week (as an employee, as self-employed, as unpaid workers in a family business, or as participants in government-supported training schemes) and those who had a job that they were temporarily away from (for example, if they are on holiday).
  
  NEET is defined as ‘not in employment, education or training’ in the week before the survey.
  
  SEB refers to the main wage earner’s occupation when the respondent was aged 14 years. Where there was no earner in the family, SEB is included in the lower working class.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2022, respondents aged 16 to 24 years in the UK, data collected from July to September 2022. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey 2014 to 2022, respondents aged 16 to 24 years in the UK, data collected from July to September each year.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.19, 3.20, 3.21, 3.22 and 3.23

Intermediate Outcome 2.2: Entry of young people into higher education

• Definition: Percentage of young people aged 18 to 20 years in the UK enrolled in higher education by socio-economic background (SEB)
• Unit of measurement: Percent
• Time period covered: 2021
• Methodology: The proportion of young people aged 18 to 20 years who began studying in higher education by socio-economic background in 2021. The data refers to participation rates of young people aged 18 to 20 years.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 18 to 20 years in the UK.
• Notes: This indicator only covered age 19 last year. The age range has been extended to 18-20 to increase the sample size and improve the precision of the estimates. The data used is weighted using the LFS probability weights.
• Figure(s): 3.24

Intermediate Outcome 2.3: Highest qualification of young people

• Definition: Percent of highest level of qualification achieved by young people aged 25 to 29 years by socio-economic background
• Unit of measurement: Percent
• Time period covered: Figure 3.25: 2021, Figures 3.26, 3.27 and 3.38: 2014 to 2021
• Methodology: This indicator includes breakdowns of the percentage of the highest qualifications achieved by socio-economic background.
  
  Parental social class is measured by the main wage earner’s occupation when the respondent was aged 14 years.
  
  For the ethnicity breakdown the outcome measure is simplified to whether the respondent has a university degree or not due to the small sample sizes of some ethnic groups. The estimated percentages and confidence intervals are derived from a logistic regression model on the likelihood of attaining a degree by ethnic group and SEB, controlling for sex. The model assumes that class effects are the same within each ethnic group. Further tests indicate that this assumption cannot be rejected. The percentages shown are those for men. Percentages are shown only for those with lower working-class and higher professional-class backgrounds for illustrative purposes. For more detail see the “Intersectional analysis and modelling section”.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 25 to 29 years in the UK. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.25, 3.26, 3.27 and 3.28

Intermediate Outcome 3.1: Economic activity of young people

• Definition: Percentage of young people aged 25 to 29 years who were economically active by socio-economic background.
• Unit of measurement: Percent
• Time period covered: Figure 3.29: 2021, Figures 3.30, 3.31 and 3.32: 2014 to 2021
• Methodology: Proportions of people aged 25 to 29 years who were economically active in 2021 by socio-economic background. Economically active is defined as either being in work, or available for and actively looking for work. Parental social class is measured by the main wage earner’s occupation when the respondent was aged 14 years. For ethnicity, the estimated percentages and confidence intervals are derived from a logistic regression model on the likelihood of being economically active by ethnic group and SEB, controlling for sex. The model assumes that class effects are the same within each ethnic group. We will test this assumption in further work. The estimated percentages are those for men. Percentages are shown only for those with lower working-class and higher professional-class backgrounds for illustrative purposes. For more detail see the “Intersectional analysis and modelling section”.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 25 to 29 in the UK. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.29, 3.30, 3.31 and 3.32

Intermediate Outcome 3.2: Unemployment among young people

• Definition: Percentage of young people aged 25 to 29 years who were unemployed by socio-economic background
• Unit of measurement: Percent
• Time period covered: Figure 3.33: 2021, Figures 3.34, 3.35 and 3.36: 2014 to 2021
• Methodology: Proportions of people aged 25 to 29 years who were unemployed in 2021 by socio-economic background. Unemployment refers to those without a job, who have actively sought work in the last 4 weeks and are available to start work in the next 2 weeks. Parental social class is measured by the main wage earner’s occupation when the respondent was aged 14 years. For ethnicity, the estimated percentages and confidence intervals result from a logistic regression model on the likelihood of being unemployed by ethnic group and SEB, controlling for sex. The model assumes that class effects are the same within each ethnic group. A formal test confirms this assumption. The estimated percentages are those for men. Percentages are shown only for those with lower working-class and higher professional-class backgrounds for illustrative purposes.  For more detail see the “Intersectional analysis and modelling section”.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 25 to 29 in the UK. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.33, 3.34, 3.35 and 3.36

Intermediate Outcome 3.3: Occupational level of young people

• Definition: Percentage of young people aged 25 to 29 years in different social class positions by socio-economic background.
• Unit of measurement: Percent
• Time period covered: Figure: 3.37: 2021, Figures 3.38, 3.39 and 3.40: 2014 to 2021
• Methodology: Proportions of people aged 25 to 29 years  in different social class positions in 2021 by socio-economic background. Parental social class is measured by the main wage earner’s occupation when the respondent was aged 14 years. For ethnicity, because of small sample sizes the outcome measure is whether the respondent has a professional occupation (either higher or lower professional). The estimated percentages and confidence intervals are derived from a logistic regression model on the likelihood of being in a professional occupation by SEB and ethnic group, controlling for sex. The model assumes that class effects are the same within each ethnic group. A formal test shows that this assumption does not hold for the Chinese group. For more detail see the “Intersectional analysis and modelling section”.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 25 to 29 in the UK. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.37, 3.38, 3.39 and 3.40

Intermediate Outcome 3.4: Earnings of young people

• Definition: Mean hourly earnings of young people aged 25 to 29 years by socio-economic background (SEB)
• Unit of measurement: Pound (£)
• Time period covered: Figure 3.41: 2021, Figures 3.42, 3.43 and 3.44: 2014 to 2021
• Methodology: The frequency of earnings was chosen to be hourly earnings to avoid a possible skewing of earnings data resulting from people of one socio-economic background being more likely to work part-time than people of other backgrounds. Self-employed respondents and those without earnings are excluded.
  
  Parental social class is measured by the main wage earner’s occupation when the respondent was aged 14 years.
  
  For ethnicity, the estimated means and confidence intervals are derived from a linear regression model of log hourly earnings by SEB and ethnic group, controlling for sex. The model assumes that class effects are the same within each ethnic group. However, the assumption does not hold for the White Other group. The means shown are those for men. Means are shown only for those with lower working-class and higher professional-class backgrounds but other SEBs are included in the analysis.
  
  Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index.
• Data source: Office for National Statistics, Labour Force Survey (LFS) 2021, respondents aged 25 to 29 years in the UK. For sex, ethnicity and disability breakdowns: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.41, 3.42, 3.43 and 3.44

Intermediate Outcome 3.5: Returns in earnings to education for young people

• Definition: Percentage differences in hourly earnings of young people with different levels of highest qualification (aged 25 to 29 years in the UK) relative to those with lower level (below GCSE grade 1 or equivalent), controlling for socio-economic background, sex and age.
• Unit of measurement: Percent
• Time period covered: 2019 to 2021
• Methodology: The percentage differences estimates are derived using a log linear regression model. This model consists of log hourly pay as the dependent variable. The explanatory variables included in the model are: educational attainment, socio-economic background, gender and age. The returns by educational attainment and socio-economic background are derived from their respective coefficients. The reference group to derive the estimates is men who were from a lower working-class background and had lower level (below CSE grade 1 or equivalent) qualifications. The data is pooled for the years 2019-2021 to obtain more accurate estimates. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index.
  
  For ethnicity, hourly earnings were estimated from a linear regression model of log hourly pay by ethnic group and SEB (2 categories only, namely professional and non-professional), controlling for educational level and age. Since interaction terms between ethnicity and SEB were of marginal significance, they are not included in the model. Estimates are shown for those with the lowest levels of education and aged 27 years.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS) from 2019 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.45

Intermediate Outcome 3.5: Returns in earnings to education for young people over time

• Definition: Hourly earnings in pounds (£) of young people aged 25 to 29 years by highest qualification controlling for socio-economic background, sex and age.
• Unit of measurement: Pounds (£)
• Time period covered: 2014 to 2021
• Methodology: Hourly earnings were estimated from a linear regression model of log hourly pay, controlling for educational level, SEB, gender and age. The estimates shown refer to the hourly earnings of  men who were from a lower working-class background. For sex, ethnicity and disability estimates are shown for people aged 27 years from lower working-class backgrounds. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index.
• Data source: Office for National Statistics, Labour Force Survey (LFS) from 2014 to 2016 and from 2019 to 2021, respondents aged 25 to 29 years in the UK. For sex, ethnicity and disability: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.46, 3.47, 3.48 and 3.49

Intermediate Outcome 3.5: Direct effect of social origins on hourly earnings

• Definition: Percentage differences in hourly earnings of young people from different social class origins (aged 25 to 29 years in the UK) relative to those from lower working-class backgrounds, controlling for highest educational level, sex and age.
• Unit of measurement: Percent
• Time period covered: 2019 to 2021
• Methodology: Percentage differences were estimated from a linear regression model of log hourly pay, controlling for educational level, socio-economic background, sex and age. The reference group is men who were from a lower working-class background and had lower-level qualifications (below CSE grade 1 or equivalent). We pool the data for years 2019 to 2021 in order to obtain more accurate estimates. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS) from 2019 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.50

Intermediate Outcome 3.5: Direct effect of social origins on hourly earnings (pounds)

• Definition: Estimated mean hourly earnings of young people aged 25 to 29 years by socio-economic background, controlling for educational level and age.
• Unit of measurement: Pounds (£)
• Time period covered: 2014 to 2021
• Methodology: Hourly earnings were estimated from a linear regression model of log hourly pay, controlling for educational level, SEB, age, and the specific protected characteristic. Estimates are shown for people with the lowest levels of education and aged 27 years. Interactions between sex and SEB, and ethnicity and SEB were not significant and have therefore not been included. Earnings have been adjusted for inflation.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 29 years in the UK.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.51, 3.52 and 3.53

Intermediate Outcome 4.1: Further training and qualifications

• Definition: Percentages of young people born in 1989 who had obtained degrees at age 25 (in 2014) and at age 32 (in 2021) in the UK, by socio-economic background
• Unit of measurement: Percent
• Time period covered: 2014 and 2021
• Methodology: Take the percent of young people in 1989 of young people born in 1989 who had obtained university degrees at age 25 (in 2014) and at age 32 (in 2021) respectively. The LFS waves used are independent of each other meaning that the comparison is made between 2 independent surveys rather than following the same individuals from age 25 to 32.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS)
• Notes: The data used is weighted using the LFS probability weights
• Figure(s): 3.54

Intermediate Outcome 4.2: Occupational progression

• Definition: Probability of access to the professional classes for men by parental class controlling for age and survey year in the UK by socio-economic background
• Unit of measurement: Probability (ranging from 0 to 1)
• Time period covered: 2014 to 2021
• Methodology: Take the average marginal effects derived from a logistic regression model of access to the professional classes controlling for age, age squared, survey year and social class background. For more detail see “Regressions” section.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021, respondents aged 25 to 44 years in the UK in work at the time of the survey.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.55 and 3.56

Intermediate Outcome 4.3: Income progression

• Definition: Income progression by socio-economic background and age
• Unit of measurement: Pounds (£)
• Time period covered: 2014 to 2021
• Methodology: Estimates are derived from a linear regression of annual income controlling for age, age squared (to account for the changing importance of age as people get older), survey year, number of dependent children and socio-economic background. For more detail see “Regressions” section. Earnings are adjusted for inflation by using 2021 as the base year and the Consumer Price Index.
• Data source: Office for National Statistics, pooled Labour Force Survey (LFS) from 2014 to 2021 (pooled), respondents aged 25 to 44 years in the UK in paid employment.
• Notes: The data used is weighted using the LFS probability weights.
• Figure(s): 3.57 and 3.58

4.4 Drivers

Driver 1.1: Distribution of earnings

• Definition: The gap in hourly earnings for full-time employees
• Unit of measurement: Ratio of 90th percentile relative to 10th percentile
• Time period covered: 1997 to 2021
• Methodology: Values to calculate the 90:10 ratio are taken from ‘Earnings and hours worked, place of work by local authority: ASHE table 7, Gross hourly pay for full-time employees from 1997 to 2021.
• Data source: Annual Survey of Hours and Earnings (ASHE)
• Notes: For 2022, ASHE Table 7 is yet to be updated, however, ASHE Table 6 provides a provisional estimate of the 90th and 10th percentiles for full-time employees. We use these figures to calculate the ratio for 2022 and note this is a provisional figure only.
• Figure(s): 4.4

Driver 1.2: Childhood poverty

• Definition: Percentage of children in relative poverty after housing costs in the UK and in England, Northern Ireland, Wales and Scotland
• Unit of measurement: Percent
• Time period covered: 1994 to 2022
• Methodology: Data are calculated using 3-year averages (including the current year and 2 preceding years). For example, the figure for 2021 represents the average of the financial years (FY) starting in 2019, 2020 and 2021. FY are reported by the year in which they start. For example, 2021 represents the financial year ending in 2022 (FY 2021 to 2022).A household is said to be in relative poverty if their equivalised income is below 60% of the median income. ‘Equivalised’ means adjusted for the number and ages of the people living in the household.
• Data source: Annual Survey of Hours and Earnings (ASHE)
• Figure(s): 4.5

Driver 1.3: Distribution of parental education

• Definition: Percentages of adults in families with dependent children having different levels of education in the UK
• Unit of measurement: Percent
• Time period covered: 2014 to 2021
• Methodology: The sample (N=160,238) was established by selecting those respondents with dependent children (dependent children defined as those aged 0 to 15 years, and those aged 16 to 18 years who are in full-time education) in their family. Respondents who are aged less than 21 years are excluded and the median age of the included respondents is 40 years. The great majority of the selected respondents are likely to be the parents/carers of the dependent children. However, the dataset could include some adults who are living at home with parents who have dependent children. The data used is weighted using the LFS probability weights.
• Data source: Labour Force Survey (LFS)
• Figure(s): 4.6

Driver 1.4: Distribution of parental occupation

• Definition: Percentages of adults in families with dependent children in different levels of occupation in the UK by socio-economic background
• Unit of measurement: Percent
• Time period covered: 2014 to 2021
• Methodology: The sample (N=160238) was established by selecting those respondents who had dependent children in their family (those aged 0 to 15 years and those aged 16 to 18 years who are in full-time education). Respondents who are aged less than or equal to 20 years are excluded and the median age of the included respondents is 40 years. The great majority of the selected respondents are likely to be the parents or carers of the dependent children, but it could include some adults who are living at home with parents of dependent children.^{[footnote 7]} The data used is weighted using the LFS probability weights.
• Data source: Labour Force Survey (LFS)
• Figure(s): 4.7

Driver 2.1: Further education and training opportunities

• Definition: Percentage of young people aged 16 to 18 years participating in education, training and employment in England
• Unit of measurement: Percent
• Time period covered: 2011 to 2021
• Methodology: Work-based learning (WBL); Not in education, employment or training (NEET); NEET includes anybody who is not in any forms of education or training and who is not in employment. This means that a person identified as NEET is either unemployed or economically inactive. Historically, there have been very small overlaps of students studying in further education and higher education and WBL at the same time. The total number of young people in training is calculated by omitting these overlaps. Of note, 16 to 17 year olds are required to remain in (at least part-time) education and training in England (but not in Wales, Scotland or Northern Ireland) following raising the participation age legislation in 2013.
• Data source: Department for Education, participation in education, training and employment, 2022.
• Notes: Participation estimates for the 2020 and 2021 cohorts impacted by COVID-19 may not fully reflect engagement and attendance.
• Figure(s): 4.8

Driver 2.2: Availability of high-quality school education

• Definition: Average pupil attainment scores (out of 1000) on PISA reading, maths, and science assessments
• Unit of measurement: Attainment scores ranging from 0 to 1000 used by PISA.
• Time period covered: 2006 to 2018
• Methodology: Proxy to measure opportunities for high-quality school education. Average scores for young people aged 15 years on PISA’s overall reading, mathematics and science. The reading, mathematics and science scale ranges from 0 to 1000.
• Data source: Organisation for Economic Co-operation and Development (OECD)
• Notes: Participation estimates for the 2020 and 2021 cohorts impacted by COVID-19 may not fully reflect engagement and attendance.
• Figure(s): 4.9

Driver 2.3: Access to higher education

• Definition: Percentage of 19 year olds enrolled in secondary or tertiary education, UK and international average
• Unit of measurement: Percent
• Time period covered: 2010 to 2020
• Methodology: Proxy measure of the participation rate relative to the number of young people aged 19 years in the population. Enrolment rates in secondary and tertiary education are expressed as net rates. These are calculated by dividing the number of students aged 19 years enrolled in these levels of education by the size of the population of 19 year olds. Generally, figures are based on headcounts and do not distinguish between full-time and part-time study. In some OECD countries, part-time education is only partially covered in the reported data.
• Data source: Organisation for Economic Co-operation and Development (OECD)
• Figure(s): 4.10

Driver 2.4: Availability of high-quality higher education

• Definition: Non-continuation (dropout) rates of full-time entrants during their first year at a higher education provider
• Unit of measurement: Percent
• Time period covered: Academic years 2014
• Methodology: Years represent the academic year of entry. Percentage of UK-domiciled full-time entrants who did not leave within 50 days of commencement and did not continue in higher education after their first year, academic years of entry 2014 to 2015 and 2019 to 2020.
• Data source: Higher Education Statistics Agency (HESA)
• Figure(s):  4.11

Driver 3.1: Vacancy rate

• Definition: Number of vacancies per unemployed person in the UK (seasonally adjusted)
• Unit of measurement: Number of vacancies per unemployed person
• Time period covered: 2001 to 2022
• Methodology: A proxy for job opportunities is calculated by ONS as the ratio of the number of unemployed (as estimated from the LFS) relative to the number of vacancies (as estimated in the Vacancy Survey) and published here as the reciprocal. Ratios were calculated using quarter 4 (October to December) from 2001 to 2021. A higher value indicates greater opportunities for job seekers.
• Data source: Office for National Statistics (ONS)
• Figure(s): 4.12

Driver 3.2: Youth unemployment

• Definition: Percentage of young people aged 16 to 24 years in the UK, from 2014 to 2021, who were unemployed
• Unit of measurement: Percent
• Time period covered: 2014 to 2021
• Methodology: The LFS follows the internationally-agreed definition for unemployment recommended by the International Labour Organisation (ILO) – a UN agency. Unemployed people are those without a job, who have actively sought work in the last 4 weeks and are available to start work in the next 2 weeks; or are out of work, have found a job and are waiting to start it in the next 2 weeks. Those who are economically inactive are excluded from the calculations (for example in full-time education, looking after the home, or permanently sick and disabled). The data used is weighted using the LFS probability weights.
• Data source: Labour Force Survey (LFS)
• Figure(s): 4.13

Driver 3.3: Type of employment opportunities for young people

• Definition: Breakdown of occupational class of young people aged 22 to 29 years in the UK
• Unit of measurement: Percent
• Time period covered: 2014 to 2021
• Methodology: The 5 social classes distinguished here represent a shortened version of the ONS NS-SEC classification, which has 8 classes. We have grouped the ONS NS-SEC classes as shown in chapter 1 of the report. The data used is weighted using the LFS probabilityLFS population weights. A formal statistical test shows that compared with 2014, access to the higher-professional class has become significantly different since 2018.
• Data source: Labour Force Survey (LFS)
• Figure(s): 4.14

Driver 3.4: Labour market earnings of young people

• Definition: Median real hourly pay for people aged 22 to 29 years in the UK
• Unit of measurement: Pounds (£)
• Time period covered: 1997 to 2021
• Methodology: Values taken from earnings and hours worked by employees, place of work by local authority: ASHE table 6.5a. Hourly pay: gross from 1997 to 2022. Earnings are inflation-adjusted using the Consumer Price Index (base year = 2021) (see the technical annex for a complete explanation and sources). ASHE covers employee jobs in the UK. It does not include self-employed people or employees not paid during the reference period.
• Data source: Annual Survey of Hours and Earnings (ASHE)
• Figure(s): 4.15

Driver 4.1: Civic engagement

• Definition: Percentage of adults who have engaged in democratic processes within the last 12 months in England
• Unit of measurement: Percent
• Time period covered: 2014 to 2021
• Methodology: The plot shows the percentages of adults who were civically engaged. This means engagement in democratic processes, both in person and online, including signing a petition or attending a public rally within the last 12 months. This does not include voting. Data is taken for the 8 financial years to March 2021. 95% confidence intervals available for 2018 to 2019 and 2019 to 2020 only.
• Data source: Community Life Survey
• Figure(s): 4.16

Driver 4.2: Level of trust, fairness and helpfulness

• Definition: Mean levels of trust, perceived fairness and helpfulness, 0 to 10 point scales, in the UK
• Unit of measurement: 0 to 10 point scale
• Time period covered: 2002 to 2018
• Methodology: Fairness was measured on a scale running from 0 (indicating “most people try to take advantage of me”) to 10 (indicating “most people try to be fair”). Helpfulness was measured on a scale running from 0 (indicating “people mostly look out for themselves”) to 10 (indicating “people mostly try to be helpful”). Trust was measured on a scale running from 0 (indicating “you can’t be too careful”) to 10 (indicating “most people can be trusted”).
• Data source: European Social Survey, data for the UK, rounds 1 to round 9 (from 2002 to 2018)
• Figure(s): 4.17

Driver 5.1: Broadband speed

• Definition: Ratio (relative to first available year) of the median broadband speed in the UK
• Unit of measurement: Ratio compared to initial year (2014)
• Time period covered: 2014 to 2019
• Methodology: Nesta provides scores at the ITL2 regional level, but not a national average figure. So, we show here the figure for the median UK area to track changes over time.
• Data source: Department for Business, Energy and Industrial Strategy and Nesta Research & Development spatial data tool
• Figure(s): 4.18

Driver 5.2: Business expenditure on research and development

• Definition: Ratio (relative to first available year) of the median business (R and D) expenditure in the UK
• Unit of measurement: Ratio compared to initial year (2007)
• Time period covered: 2007 to 2018
• Methodology: Nesta provides scores at the ITL2 regional level, but not a national average figure. So, we show here the figure for the median UK area in order to track changes over time.
• Data source: Department for Business, Energy and Industrial Strategy and Nesta Research & Development spatial data tool.
• Figure(s): 4.19

Driver 5.3: University research students

• Definition: Ratio (relative to first available year) of the median number of full-time equivalent research students enrolled in universities in the UK
• Unit of measurement: Ratio compared to initial year (2015)
• Time period covered: 2015 to 2018
• Methodology: Nesta provides scores at ITL2 region level, but not a national average figure. We therefore show here the figure for the median UK area to track changes over time.
• Data source: Department for Business, Energy and Industrial Strategy and Nesta Research & Development spatial data tool
• Figure(s): 4.20

4.5 Confidence intervals and test of significance

In the case of indicators derived from survey data, we calculate tests of statistical significance in order to check whether differences between the estimates for groups or over time are likely to be due to sampling error rather than represent real differences in the population. We show the 95% confidence intervals when making comparisons between percentages. A confidence interval is a range around a value that conveys how precise the measurement of the value is. A 95 percent confidence interval shows that, if a sample survey of the given size were conducted 100 times, the true population value would lie within the confidence interval in 95 of the 100 samples whereas in only 5 samples would the population value lie outside the confidence interval. In general, the smaller the size of the sample, the larger will be the confidence interval. Confidence intervals for estimates concerning small subgroups of a sample, such as some ethnic minorities or local areas, will tend to be large. In other words, there will be less precision in the estimates.

For most analyses of categorical data, such as the percentages experiencing absolute upwards mobility as in figure 2.7, we show the 95% confidence intervals. The standard formula for the 95% confidence interval is:

CI = +/- 1.96√{(p(1-p)/N)} (1)

Where p represents the proportion of the group obtaining a given outcome, and N is the number of respondents in the group under consideration.

We frequently use confidence intervals in order to check whether rates of, for example, upward mobility differ between groups. A simple rule of thumb is that, if the confidence intervals do not overlap, then there is a statistically significant difference between the 2 groups. In other words, it is unlikely that a difference of that magnitude between the 2 groups could have occurred simply as a result of sampling error. Note however that a small overlap can also be significant at the 5% level and we therefore use a formal difference of proportions test in these cases.

We also use a range of other formal tests of significance where different statistical techniques are employed.

4.6 Odds ratios

When estimating levels of relative mobility using categorical data, we use odds ratios. The odds ratio is a statistic which can be used for comparing the mobility chances of people from different socio-economic backgrounds. It can be thought of as a way of assessing the outcome of a competition between people from 2 different backgrounds to achieve an advantaged outcome and to avoid a disadvantaged outcome.

Consider the example of housing mobility. In figure 2.26, we see that 71% of people whose parents were homeowners became owners themselves, compared with only 46% of those whose parents were renters. One natural way to measure the inequality is to measure the ratio. In this example, those whose parents were homeowners had 71/46 = 1.5 times as good a chance of owning a home compared with those whose parents were renters.

However, this approach can produce paradoxical results when looking at changes over time, or when comparing groups. For example, if we compare the inequality in becoming a homeowner among people with and without a disability (as in figure 2.28), we find that the inequality in chances of becoming an owner are greater among the disabled (a ratio of 1.79) than among those without a disability (a ratio of 1.40). However if we look at the inequality in chances of becoming a renter, the story is reversed, and the inequality is smaller among those with a disability (a ratio of 1.69) than among those without a disability (a ratio of 1.81). We thus obtain different answers to our question depending on which perspective we take.

The solution is to look both at the chances of becoming an owner and at the chances of becoming a renter. And this is what the odds ratio does. In this case the odds for the children of an owner becoming an owner or a renter themselves is 71:29 (= 2.5:1) and the odds for the children of a renter to become an owner or a renter themselves is 46:54 (= 0.9:1). The overall odds ratio is therefore 2.5/0.9 = 2.8:1. More generally, if the data is set up in a 2 x 2 table (as in this case) with a, b, c and d representing the frequencies in each cell of the table, then the odds ratio is (a/b) / (c/d) = ad/bc. An odds ratio of 1 indicates that the odds of the 2 groups of reaching one outcome and avoiding the other are identical. In other words, there is no inequality in the relative chances of the 2 groups.

We can also estimate the confidence interval for the odds ratio. One has to first take the natural logarithm of the odds ratio. The formula for the 95% confidence interval of the log odds ratio is:

+/- 1.96√( 1/a + 1/b + 1/c + 1/d) (2)

We then exponentiate these values in order to obtain the 95% confidence interval of the (unlogged) odds ratio.

4.7 UNIDIFF

In the case of housing mobility, we had a binary measure of housing status, namely whether one was an owner or a renter. In the case of occupational mobility, however, we have 5 categories of occupation. In the case of the binary housing status measure, there is only one odds ratio that can be estimated. However, in a 5 by 5 cross-categorization of social class, such as in figure 2.4 there will be a large number of possible odds ratios. To investigate differences in relative mobility when making comparisons across tables (for example over time or across groups), we therefore use the UNIDIFF model (uniform difference model), which provides a single number to show whether the level of relative mobility differs across the tables being compared.

The UNIDIFF model assumes that the log odds ratios have the same pattern in the 2 tables that are being compared, but that in the second table they are scaled up or scaled down by a common factor. This common factor is known as the UNIDIFF parameter estimate. When it is one, there is no difference between the 2 tables in the magnitude of the log odds ratios – relative mobility is therefore the same in the 2 tables. As the log odds ratio decreases, the link between origins and destinations has weakened and relative mobility has improved.

Note that the UNIDIFF parameter always takes one of the 2 tables as the reference table (or reference layer). The parameter is then the estimate of the magnitude of the multiplier that best represents the extent to which the log odds ratios in the second table are greater or smaller than those in the first (baseline) table. A positive parameter estimate indicates that the log odds ratios are higher in the second table, while a negative parameter estimate indicates that the log odds ratios are smaller. The equation for the relationship between the odds ratios in a set of table  is:

logθ_ijk = φ_k ∙ logθ_ij (3)

Where θ_ijk represents a particular odds ratio in table k, and φ_k is the table-specific parameter that scales the log odds ratios relative to the log odds ratio in the baseline table, logθ_ij.^{[footnote 8]}

4.8 Regression (Intergenerational Elasticity (IGE) and rank-rank coefficient)

In principle, one could use statistics such as odds ratios to analyse income mobility in the same way as sociologists analyse occupational mobility. We could, for example, divide the income distribution into 5 ordered quintiles, each containing 20 percent of the sample. However, there is no particular need to use odds ratios since income is not an inherently categorical variable in the way that social class is. Whereas there are clear substantive distinctions between the occupational classes of the NS-SEC, income bands such as quintiles differ quantitatively not qualitatively. Economists therefore prefer to use statistical techniques that are designed for analysing continuous rather than categorical data. However, because of the skewed distribution of income, economists take the natural logarithm of both parents’ and adult children’s incomes. This is because taking the logarithm of income instead of just income can help with the statistical challenges provided by the rightward skew often found in income data. When analysing income mobility, economists thus typically fit a regression model of the form:

Y_i^Child = α + βY_i^parent + e_i (4)

where Y_i^Child represents the adult child’s income (logged) and  Y_i^parent represents the parents’ income (logged). α is the intercept, β is the regression coefficient representing the effect (slope) of parents’ income on their adult children’s income, and ei is an error term (capturing the distribution of the residuals which are assumed to be normally distributed around the line of best fit with an expected value of 0). The intercept is sometimes taken to be a measure of absolute income mobility while the coefficient β, often termed the intergenerational elasticity or IGE, is regarded as a measure of relative mobility. The coefficient β also has a natural interpretation. For example, a coefficient of 0.3 means that if 2 families have (log) incomes that differ by 10%, their adult children’s (log) income will differ by about 3%. In other words, for every 10% increase in one’s parents’ income, on average the child will be expected to earn 3% more.

The IGE is not, however, the ideal measure of relative income mobility, as it can be affected by changes in the degree of income inequality between the parental and child generations.  A simple solution to this problem is to use parents’ and children’s percentile position in the income distribution rather than their (logged) income. This is referred to as the rank-rank coefficient. This corresponds closely with our conception of relative mobility as focusing on people’s position within the queue. It also has some other methodological advantages – for example, it can cope with incomes of zero – and has generally been found to be a more robust measure. Exactly the same regression techniques can be used as before, but Y_i^Son and Y_i^parent in equation 4 would instead capture sons’ and parents’ ranks in their respective income distributions. For further details see Gregg et al (2017).^{[footnote 9]}

4.9 Two-stage imputation of parental wealth

Given that parental wealth is not available in the WAS data, we use a two-stage imputation called the Two Stage Two Sample Least Squares method to estimate the intergenerational elasticity coefficient. This method was used to investigate intergenerational income mobility by Björklund and Jäntti (1997), and recently replicated in the wealth mobility context by Gregg and Kanabar (2022).^{[footnote 10][footnote 11]}

As parental wealth is not available in the WAS, we estimate wealth mobility in 2 stages:

1. Since we do have the educational attainment and housing tenure of the parents, we first estimate the relationship between wealth and these 2 variables. We use wave 3 of the WAS as this is the first wave in which one’s wealth is captured, this covers the years 2010 to 2012. We use respondents aged 64 to make our estimated relationship more representative of the one the parents of the respondents would have.

2. Using this estimated relationship, we predict the wealth of the parents by using their known status of educational attainment and housing tenure. This is sometimes referred to as a ‘synthetic parent’ as we have predicted the outcome of interest rather than used data in which it is observed. Using our predicted values of parental wealth, we then estimate the intergenerational elasticity of wealth, in the same regression specification used to estimate the intergenerational elasticity of income. We use wave 7 as this is the most recent data and include wave 6 in our sample to boost its size. Together these cover the years 2016 to 2020.

4.10 Intersectional analysis of protected characteristics

Intersectionality has been defined as “a metaphor for understanding the ways that multiple forms of inequality or disadvantage sometimes compound themselves and create obstacles that often are not understood among conventional ways of thinking.”^{[footnote 12]} The concept is often used in the context of protected characteristics and their inter-relationship with socio-economic background. When considering intersectionality between protected characteristics and socio-economic background we focus on 2 conceptually distinct questions. The first question is whether, among people who come from similar socio-economic backgrounds, there is an additional disadvantage arising from a protected characteristic such as sex, disability or race.  For example, we ask whether young black people are disadvantaged with respect to employment compared with their majority-group peers from the same socio-economic background. In other words our intersectional analysis focuses on whether there are additional disadvantages arising from protected characteristics such as ethnicity on top of those arising from socio-economic background. This can be expressed in the form of the following equation:

Y_i^Child = α + βY_i^parent +  γX_i^pc + e_i (5)

Where X_i^pc represents a protected characteristic, Y_i^Child is the outcome of interest for the child, Y_i^parent represents the measure of interest for the parent, α is the constant and e_i captures the residual. This is termed an additive model since it estimates a relationship between the protected characteristic and the adult child’s outcome in addition to the relationship with the parental characteristic.

The second question is whether particular combinations of socio-economic background and a protected characteristic ‘compound themselves and create obstacles that are not understood among conventional ways of thinking’.^{[footnote 13]} An example of compounding is the finding that the disadvantages experienced by disabled young people with respect to employment are magnified when they come from lower working-class backgrounds. In other words, not only are there additional disadvantages arising from disability but also these disadvantages are further magnified when the individual concerned has both a disability and a lower working-class background. In other words, the issue is whether disadvantages are compounded in the case of specific combinations of SEB and protected characteristics. This can be expressed in the form of the following equation:

Y_i^Child = α + βY_i^parent + γX_i^pc + δY_i^parent*Yi^pc + e_i (6)

where δY_i^parent*Yi^pc represents an interaction between the parental characteristic and the protected characteristic.

For simplicity, when reporting intersectionality between socio-economic background and a protected characteristic such as sex or disability, we simply show charts comparing the mobility patterns for the 2 groups under consideration and use standard tests of significance as described above. However, we always conduct supplementary regression analyses in order to check whether the estimated coefficients γ and δ for the protected characteristic and for the interaction between the protected characteristic and parental characteristic are statistically significant.

In the case of the protected characteristic of ethnicity, the sample sizes for some ethnic groups become quite small and the estimates therefore become imprecise and may be potentially misleading. We use 2 methods for improving precision in such cases. First, we can combine categories, for example in the case of educational mobility comparing those with and without a university degree instead of the full 6-category measure of highest qualification. Second, we can estimate regression models of the same general form as equations (5) and (6). If the coefficient for the interaction term in equation (6) proves not to be statistically significant, we simply show the results (the fitted estimates) for equation (5). This model assumes that socio-economic background has the same effect size on the outcome variable among all ethnic groups. If the coefficient for one or more interaction terms is statistically significant, however, we then show the fitted estimates for the best-fitting model including any significant interaction terms.

Note that in the intersectional analyses with a binary dependent variable (such as capturing whether one attains a professional class occupation or not) we fit logistic regression models rather than the linear additive models shown in equations (5) and (6). The equation for the logistic regression model is:

Ln(p/1-p)^Child = α + βY_i^parent (7)

Where p/1-p represents the odds of the adult child obtaining one outcome and avoiding the other.

4.11 Geographical analysis

In our SoN 2023 report, we show maps and plots for some of the drivers and intermediate outcomes that use the Labour Force Surveys as the data source. In order to obtain more precise estimates of regional variation, we pool multiple years of LFS data together.

For our geographical analysis wherever possible we use the ITL2 classification (International Territorial Levels 2), which in our version has 41 categories. The ITL classification replaces the previous EU NUTS classification. While we could analyse the LFS data using upper-tier local authorities (ITL3), the precision of the estimates would be much lower and the amount of material to digest would be much greater.

Since we are in general interested in whether the area in which individuals grew up shapes their mobility chances, we use the LFS measure of the area where the respondent was resident at the age of 14, not their current area of residence.^{[footnote 14]}

Below is a list of the 41 ITL2 areas included in the report (the ITL1 regions each of these ITL2 lies in are provided in brackets).

1. Inner London – West (London)

2. Inner London – East (London)

3. Outer London – South (London)

4. Outer London – East and North East (London)

5. Outer London – West and North West (London)

6. Bedfordshire and Hertfordshire (East of England)

7. Berkshire, Buckinghamshire and Oxford (South East England)

8. Cheshire (North West, England)

9. Cornwall and Isles of Scilly, (South West England)

10. Cumbria (North West England)

11. Derbyshire and Nottinghamshire (East Midlands, England)

12. Devon (South West England)

13. Dorset and Somerset (South- West England)

14. East Anglia (East of England)

15. East Yorkshire and Northern Lincolnshire (Yorkshire and the Humber, England)

16. Essex (East of England)

17. Gloucestershire, Wiltshire and Bristol and Bath area (South West England)

18. Greater Manchester (North West England)

19. Hampshire and Isle of Wight (South East England)

20. Herefordshire, Worcestershire and Warwickshire (West Midlands, England)

21. Kent (South East England)

22. Lancashire (North West England)

23. Leicestershire, Rutland and Northamptonshire (East Midlands, England)

24. Lincolnshire (East Midlands, England)

25. Merseyside (North West, England)

26. North Yorkshire (Yorkshire and the Humber, England)

27. Northern Ireland (Northern Ireland)

28. Northumberland and Tyne and Wear (North East England)

29. Shropshire and Staffordshire (West Midlands, England)

30. South Yorkshire (Yorkshire and the Humber, England)

31. Surrey, East and West Sussex (South East England)

32. Tees Valley and Durham (North East England)

33. West Midlands (England)

34. West Yorkshire (Yorkshire and the Humber)

35. West Wales and The Valleys (Wales)

36. East Wales (Wales)

37. Highlands and Islands (Scotland)

38. Eastern Scotland (Scotland)

39. West Central Scotland (Scotland)

40. Southern Scotland (Scotland)

41. North Eastern Scotland (Scotland)

Even when pooling the LFS’s in order to maximise sample size, the sample sizes for particular ITL2 areas can become quite small. This is especially the case when we look at specific age groups as we do for the intermediate outcomes. Moreover, when comparing a large number of categories, there is the likelihood that, if we are using the 95% level of significance, we will by chance alone obtain 5 apparently significant results in every 100. We can think of these as being ‘false positives’.

In order to reduce the risk of obtaining false positives we use multi-level models for comparing estimates across areas. These multilevel models take account of the clustering of individuals within each area and estimate both the variance of individuals’ outcomes within areas and the variance in average outcomes between areas. We thus have 2 levels – the individual level and the area level – and there will be error terms (residuals) at both the individual and the area level (denoted by the subscripts i and j respectively). In contrast to equations 3, 4, 5 and 6 which take account only of error at the individual level, the multi-level model for a continuous outcome shown in equation (8) includes both level 1 and level 2 error terms.

Yi^Child = α₀ + βX_i^parent + e_i + u_j (8)

where α₀ represents the intercept, β represents the effect of parental socio-economic position on the adult child’s outcome, e_i is the level 1 error term and u_j is the level 2 error term. This particular model assumes that β is the same across all areas while u_j can be thought of as the residuals for the different areas around α₀. This is therefore sometimes described as a random intercepts model. Other models, such as a random slopes model, and models with level 2 explanatory variables, can also be fitted to the data.^{[footnote 15]} In effect the multilevel model, when estimating differences between areas, takes account both of the number of respondents in each area and of the level 2 variance between areas. In other words, the model does not treat each area in isolation from the other areas but also takes account of the likelihood of an area having an extreme value. By taking account of the level 2 variance between areas in this way, the model shrinks extreme values towards the mean value. When we report values for different areas (as for example in caterpillar plots), we report the value of the (shrunk) residual for each area. We should note that these residuals are derived from a model which also fits a relationship between parental socio-economic position and the adult child’s outcome. They therefore tell us whether children from the same socio-economic background have better or worse outcomes in the specific area than their peers do in other areas. We can think of these as measures of absolute mobility chances in the different areas.

We use these models only for geographical differences in intermediate outcomes (controlling for SEB) when using the LFS data. We are not able to use them when analysing administrative data, since we do not have access to the individual-level data. Nor do we use multilevel models when estimating area differences in drivers.

4.12 Composite indices methodology

There is considerable imprecision in the results for individual ITL2 areas with respect to those intermediate outcomes and drivers in the measurement framework that are based on sample surveys. This is because the sample surveys, such as the LFS on which we place most reliance, contain relatively small samples within each area. The problem is exacerbated when we focus on results for restricted age groups (as we do with the intermediate outcomes).  As we can see from the caterpillar plots: the confidence intervals for the area estimates overlap substantially – we therefore cannot reject the null hypothesis that most areas are the same with respect to net intermediate outcomes.

This imprecision means that we cannot confidently draw conclusions about the way in which intermediate outcomes vary between different geographical areas of the UK and whether some areas are associated with more favourable outcomes than others. To mitigate this problem, we construct 5 composite indices (3 for drivers and 2 for intermediate outcomes) by combining results from several different indicators.

To construct the composite indices we follow the procedure used for constructing the UN’s Human Development Index (HDI).^{[footnote 16]} To put all indicators on a common metric, they must be rescaled. We do this using the following formula:

I^*_n = I_n - I_min / I_max - I_min (9)

Where I^*_n is the rescaled score and I_n is the original score. This will have the effect of rescaling the best performing area’s score on the indicator to 1 and the least well performing area’s score to 0.  Different indicators can then be aggregated into an index.

Each of our composite indices aggregates 3 indicators. We aggregate the indicators into an index by taking the geometric mean of the (normalised) component scores:

Index = 3√I^*₁ ∙ I^*₂ ∙ I^*₃ (10)

In deciding which indicators to combine, one main criterion is whether they ‘go together’ in the sense that they are highly correlated at the area level. Factor analyses (at the area level) reveal 3 factors for the drivers and 2 for the intermediate outcomes. Note that the sample size for these analyses is only 41, and so factor loadings may be unreliable.

Our second criterion is whether the selection of indicators makes conceptual sense, in that theory and research suggest causal processes that could generate the correlations between the observed indicators. In a few cases we ignore on conceptual grounds an indicator which has a relatively high loading but where the correlation is likely to be spurious rather than the result of causal processes.

Table 1 shows the indicators used for the 2 composite indices for intermediate outcomes, and their loadings on the relevant factor, while table 2 shows the indicators used for the composite indices for drivers.

Table 1. Summary of composite indices for the intermediate outcomes.

Promising prospects:

Indicator	Labour Force Survey (LFS) data used	Factor loadings
IN2.3 Highest qualification (university degree)	Net levels of a university degree among young people in each area after controlling for socio-economic background (SEB)	0.356
IN3.3a Occupational level (professional occupation)	Net proportions of young people in higher professional-class jobs in each area after controlling for SEB	0.750
IN3.4 Hourly earnings	Mean hourly earnings among young people in each area after controlling for SEB	0.829

Precarious situations:

Indicator	Labour Force Survey (LFS) data used	Factor loadings
IN3.1 Economic inactivity	Net levels of inactivity among young people in each area after controlling for SEB	0.508
IN3.2 Unemployment	Net levels of unemployment among young people in each area after controlling for SEB	0.231
IN3.3b Occupational level (lower working-class occupation)	Net proportions of young people in lower working-class jobs in each area after controlling for SEB	1.028

Table 2: Summary of composite indices for the drivers of social mobility

Sociocultural advantage:

Indicator	Measurement	Factor loadings
Driver (DR) 1.3 Parental education (university degree)	Percentages of parents in the area with a university degree	1.000
DR1.4a Parental occupation (higher professional)	Percentages of parents in the area with a higher professional occupation	0.915
DR3.3a Young people’s occupation (higher professional)	Percentages of young people in the area with a higher professional occupation	0.528

Childhood poverty and disadvantage:

Indicator	Measurement	Factor loadings
DR1.2 Childhood poverty	Data from the Department for Work and Pensions on Households Below Average Income	0.873
DR1.4b Parental occupation (lower working)	Percentages of parents in the area with a lower working-class occupation	0.851
DR3.2 Youth unemployment	Percentages of young people in the area who are unemployed	0.778

Research and development (R and D) environment:

Indicator	Measurement	Factor loadings
DR5.1 Broadband speed	The median broadband speed in the UK	0.300
DR5.2 Business expenditure on R and D (logged)	The median business R and D expenditure in the UK	1.014
DR5.3 University research students	Based on the median number of research students enrolled in the UK	0.361

5. Glossary

5.1 Absolute social mobility

This is the idea that some people have different life outcomes from their parents.

Absolute mobility rates look at the proportion of the population who are in different positions (often occupational class or income) from their parents, and are usually given as a simple percentage.

For example, a person experiences upward absolute income mobility if their income is greater than their parents’ income. They experience upward absolute occupational mobility if their occupation class is higher than their parents’. Mobility can also be downward.

5.2 Apprenticeships

A work-based training system, where apprentices earn a qualification after completing a blended mix of study and work.

Apprentices must complete 20% of their training off the job, be paid the apprenticeship minimum wage (£4.81/hour for those aged 19 years and over) and pass an end point assessment.

5.3 Caterpillar plots

A chart which shows both an estimate of a data point and a confidence interval (see definition in this glossary), arranged according to the magnitude of the estimates. The confidence interval usually is represented by a line going through the point which itself represents the estimated value in the chart.

5.4 Composite index

An index consisting of multiple indicators. In our report they all consist of 3 different indicators. For example, our composite index on Promising Prospects consists of indicators on the highest level of qualification achieved, occupational class achieved and hourly earnings.

5.5 Confidence intervals

A range of values for which there is a defined probability (usually 95%) that an estimate lies within this range. For example if we have a confidence interval of 5 to 10 at the 95% level, we are 95% confident that our estimate lies within the range of values from 5 to 10.

5.6 Class pay gap

The difference in average pay between people who occupy the same type of job but come from different class backgrounds.

5.7 Drivers

These capture the background conditions that make social mobility easier. For example, the availability of good education is a driver, because it helps people to be upwardly mobile. So our measures of drivers tell us about these nationwide background conditions. They do not tell us what the UK’s rates of mobility have been, and they are not broken down by socio-economic background.

5.8 Early years foundation stage profile (EYFSP)

The early years foundation stage profile (EYFSP) sets standards in England for the learning, development and care of a child from birth to 5 years old. All schools and Ofsted-registered early years providers must follow the EYFSP, including childminders, pre-schools, nurseries and school reception classes. There are different early years standards in Scotland, Wales and Northern Ireland.

5.9 Economically inactive

Individuals that are out of work and not looking for a job. Reasons for this include sickness, looking after family, and being a student, amongst other reasons.

5.10 Free school meals (FSM)

In England, a free school meal (FSM) is a statutory benefit available to school-aged children from families who receive other qualifying benefits and who have been through the relevant registration process. FSM eligibility is often used as a proxy measure for disadvantage in school-aged children.

Free school meal eligibility is the only available measure of pupils’ social background in England, Wales and Northern Ireland. In Scotland, a completely different area-based measure of social background is used, the Scottish Index of Multiple Deprivation (SIMD).

FSM only includes those who have both applied for and been deemed eligible by the relevant local authority. It excludes an unknown number of people who might have been deemed eligible had they applied, and may also include some relatively affluent families.

5.11 Further education (FE)

Typically refers to classroom-based learning at further education (FE) colleges or providers. Students can start at age 14 or 16 years, depending on the college.

5.12 General Certificate of Secondary Education (GCSE)

The GCSE is an academic qualification taken in England, Wales and Northern Ireland, normally at age 16. State schools in Scotland use the Scottish Qualifications Certificate instead.

5.13 Higher education (HE)

Typically refers to post-secondary education, or tertiary education, leading to award of an academic degree. Higher education is an optional final stage of formal learning that occurs after completion of secondary education.

5.14 Income mobility

See social mobility, absolute social mobility and relative social mobility.

5.15 Intermediate (occupations)

See NS-SEC.

5.16 Intermediate outcomes

These capture the progress that people make from their starting point to an intermediate point, such as employment in their 20s, or educational attainment at 16. We break outcome measures down by people’s socio-economic background, so that we can see how different starting points affect progress to eventual end points.

5.17 International Territorial Level (ITL)

This is the internationally comparable regional geography for the UK. The regions that we use are part of a system developed by the Office for National Statistics (ONS). The level of the system we use, ITL2, divides the UK into 41 regions. Each region has between 800,000 and 3,000,000 inhabitants and contains about 4 upper-tier local authorities.

5.18 Key stage 2 (KS2)

The 4 years of schooling in England and Wales, described as year 3, year 4, year 5 and year 6, when pupils are aged between 7 and 11 years. Key stage 2 SATs (National Curriculum tests) are taken by pupils aged 11 years at the end of year 6, which is, usually, the final year in primary school.

5.19 Key stage 4 (KS4)

The 2 years of schooling known as year 10 and year 11, when pupils are aged between 14 and 16 years. Most pupils take their final general certificate of secondary education (GCSE) exams at the end of year 11.

5.20 Labour Force Survey (LFS)

The LFS is a large-scale nationally-representative government survey that covers England, Wales, Scotland and Northern Ireland. The survey is managed by the Office for National Statistics (ONS) in Great Britain, and the Northern Ireland Statistics and Research Agency (NISRA) in Northern Ireland.

Its purpose is to provide information on the UK labour market which can then be used to develop, manage, evaluate and report on labour market policies.

5.21 Multilevel Model

This is a statistical model which takes account of the clustering of individuals within higher-level units such as regions or schools.

5.22 Mobility outcomes

These capture the progress people have made compared to their parents at a later point in life (often in one’s 50s).

5.23 National minimum wage

The minimum wage that an employer must pay its workforce. For those aged over 23 years, it is currently set at £9.50, an amount known as the ‘national living wage’. There are lower national minimum wages amounts for younger people. In this report, this is referred to as the minimum wage.

5.24 Not in education, employment or training (NEET)

The annual publication of national participation figures of young people aged 16 to 18 years includes a measure of those who are NEET, or not in education, employment or training.

5.25 The National Statistics Socio-economic Classification (NS-SEC)

This is the best national measure to monitor occupational social mobility. We define an individual’s socio-economic background according to the occupation of their higher- earning parent. This year we use a 5 part grouping:

1. Higher professional: NS-SEC 1. Examples of which include CEOs, doctors and engineers.

2. Lower professional: NS-SEC 2. Examples of which include teachers, nurses and journalists.

3. Intermediate: NS-SEC 3 and 4. Examples of which include shopkeepers, taxi drivers and roofers.

4. Skilled working class: NS-SEC 5 and 6. Examples of which include mechanics, electricians and housekeepers.

5. Routine working class: NS-SEC 7 and 8. Examples of which include cleaners, porters and waiters.

5.26 Occupational mobility

See social mobility, absolute social mobility and relative social mobility

5.27 Odd ratio

The odds ratio is a statistic which can be used for comparing the mobility chances of people from different socio-economic backgrounds.  It can be thought of as a way of assessing the outcome of a competition between people from 2 different backgrounds to achieve an advantaged outcome and to avoid a disadvantaged outcome.

5.28 The Organisation for Economic Co-operation and Development (OECD)

The OECD is an international organisation of 38 countries committed to democracy and the market economy. We use the 37 other member states as international comparators to the UK.

5.29 Professional, professional/ managerial (occupations)

See NS-SEC.

5.30 Program for International Student Assessment (PISA)

PISA is the OECD’s Programme for International Student Assessment. PISA measures 15-year-olds’ ability to use their reading, mathematics and science knowledge and skills to meet real-life challenges.

5.31 Pupil Premium

A sum of money given by the UK government to schools in England to improve the attainment of disadvantaged children.

5.32 Quintile

20% of a population which is usually ordered in either ascending or descending order of some metric. After being ranked, the population is split into 5 equal sized groups, each of these groups is one quintile.

5.33 Random effects model

A type of statistical model in which the model parameters are random variables.

5.34 Relative poverty

This is one measure of poverty. A household is in relative poverty if its income is below 60% of the average (median) net household income in the same year. In other words, the pound amount of the poverty line changes each year based on current median income in the country.

We can look at relative poverty in 2 ways: before and after housing costs. We look at poverty after housing costs to see how much households have in disposable income.

5.35 Relative social mobility

Relative mobility involves a comparison of the positions of people from different social backgrounds. For example, in the case of income, relative mobility tells us how strongly children’s ranking within their income distribution is associated with their parents’ ranking. Relative mobility is low if almost everyone ends up with a similar rank as their parents. For example, if parents in the bottom decile of earnings have children that mostly end up in the bottom decile of earnings.

While absolute and relative social mobility often go together, they are not the same concept. For example, if a society creates more professional jobs, absolute occupational mobility should improve. But if most of these professional jobs go to people from professional backgrounds, relative social mobility may remain static.

5.36 Social mobility

Social mobility refers to movement within a given stratification system. Social mobility can be either inter-generational (when children move away from their parents’ position) or intra-generational (when people move away from their own initial position).  There are different dimensions of stratification;   we focus on the key mobility dimension of occupational class, and add further dimensions like income, wealth, education and housing, as the data allows. See also the distinction between relative social mobility and absolute social mobility.

5.37 Social Mobility Index (SMI)

The Social Mobility Index is a long-term measurement framework for social mobility in the UK. It replaces the Commission’s original Social Mobility Index, launched in 2016. This new Index provides a critical starting point to improve the evidence base and goes well beyond solely reporting on the drivers of mobility.

We report on social mobility outcomes which show where people end up in comparison with where they started. This is across a range of outcomes of interest, including occupational class, income, education, and either at an earlier stage in their lives in their 20s and 30s (intermediate outcomes), or a later stage in their 40s and 50s (mobility outcomes).

5.38 Socio-economic classification, background

See NS-SEC.

5.39 Universities and Colleges Admissions Service (UCAS)

The Universities and Colleges Admissions Service – a nonprofit organisation which conducts the application process for UK universities.

5.40 Understanding Society, The UK Household Longitudinal Survey (UKHLS)

The UKHLS is a longitudinal survey of the members of approximately 40,000 households in the UK. The study is based at the Institute for Social and Economic Research at the University of Essex.

The purpose of UKHLS is to provide high-quality longitudinal data on subjects such as health, work, education, income, family, and social life. This helps to understand the long-term effects of social and economic change, as well as policy interventions designed to impact the general wellbeing of the UK population.

5.41 Unidiff parameters

The Unidiff parameter provides a single number to show whether the level of relative mobility (as measured by odds ratios) differs across the tables being compared.

5.42 Wealth and Assets Survey (WAS)

A longitudinal survey of around 17,000 households conducted by ONS every 2 years. The purpose of the survey is to ‘measure the well-being of households and individuals in terms of their assets, savings, debt and planning for retirement’.

5.43 Working-class occupations

See NS-SEC.

1. See the following ONS paper: ‘Non-response Weights for the UK Labour Force Survey? Results from the Census Non-response Link Study for comparisons based on the 2011 Census’ ↩

2. For further details see the LFS USER Guide, volume 1, section 10. ↩

3. For further details of the changes, see the ONS Labour Force Survey Performance and quality monitoring report: April to June 2021. ↩

4. Based on the methodology used by the International Labour Organisation. ↩

5. Some information about performance indicators and the reasons for discontinuation can be found on the HESA website. ↩

6. For more detail, see the Labour Force Survey, ‘User guides, volumes 3 and 4’, 2023. Published on ONS.GOV.UK. ↩

7. We have checked these results using the Labour Force Survey (LFS) measure of parental background (based on the occupation of the main earner when the respondent was aged 14 years). While the LFS measure of parental background shows slightly higher proportions in the professional classes, which is to be expected given its methodology, the trends over time show the same pattern as those from the measure used above. ↩

8. For further details see: Breen, Richard, ‘Methodological preliminaries’ in Education and Intergenerational Social Mobility in Europe and the United States, edited by R Breen and Walter Mueller, Stanford University Press. ↩

9.   Gregg, P., Macmillan, L. & Vittori, C. (2017). Moving Towards Estimating Sons’ Lifetime Intergenerational Economic Mobility. Oxford Bulletin of Economics and Statistics, 79(1), pp. 79-100. ↩

10. Björklund, A. and M. Jäntti. (1997).  “Intergenerational income mobility in Sweden compared to the United States,” American Economic Review, 87(5), 1009–18. ↩

11. Gregg, P. and Kanabar, R. (2022), Intergenerational wealth transmission in Great Britain. Review of Income and Wealth. ↩

12. Crenshaw, Kimberle´ Williams (1989) “Demarginalizing the Intersection of Race and Sex: A Black Feminist Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Politics.” University of Chicago Legal Forum 1989:139–67, p. 149 ↩

13. Crenshaw, Kimberle´ Williams (1989) “Demarginalizing the Intersection of Race and Sex: A Black Feminist Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Politics.” University of Chicago Legal Forum 1989:139–67, p. 149 ↩

14.   For further details see the ONS international geographies methodology. ↩

15.   For further details see Snijders, Tom and Roel Bosker (1999) Multilevel Analysis: an introduction to basic and advanced multilevel modelling. London: Sage Publications. ↩

16. For further details see Anand, Sudhir and Amartya K Sen (1994) Human Development Index: Methodology and Measurement.  UNDP, Human Development Report Office, Occasional Papers 12. https://hdr.undp.org/system/files/documents/oc12pdf.pdf ↩