Policy paper

Technical annex A: methodology

Updated 21 July 2022

Introduction

The new Social Mobility Index proposes a new long-term measurement framework for social mobility in the United Kingdom (UK). It establishes a set of empirically-based indicators for assessing progress in a rigorous way, allowing us to monitor the mobility outcomes, intermediate outcomes, and known drivers of social mobility. Its development has also exposed the areas where the data needed to effectively monitor social mobility is inadequate. Therefore, the new index is a dynamic work in progress.

This annex, alongside the State of the Nation 2022 report, details the new measures and delves into the literature and expert consultations conducted to select them. This work is about our methodology and rationale – it does not provide systematic new analysis of the currently available data.

The new index has been built using an evidence-based approach. It draws on research about the factors that are likely to be causally related to increasing rates of upward mobility for groups with historically lower rates, and for reducing the influence of parental circumstances on children’s social mobility chances.

In addition to a review of the existing research literature, we have consulted with a wide range of experts from different disciplines, such as economics, geography, psychology and sociology, and with stakeholders from government, business, schools and universities and third sector organisations throughout the development of the index. We have carried out a thorough literature review, developed a theory of change, and carried out analyses of past trends in social mobility.

The scope of the index

The index consists of a set of concepts and empirical indicators that measure changes in social mobility both across an individual’s life, and across the different parts of the UK, as well as the social and economic conditions that help drive change. It includes 10-yearly, 5-yearly, and annual measures, providing timely information for the Commission to be able to signal where action might be needed. It is designed to provide longer-term measures of a range of outcomes and early-warning signs of emerging problems.

Reflecting the Commission’s monitoring responsibilities for the whole of the UK, the new framework covers the UK as a whole, not just England, and will allow for regional comparisons. It includes a range of social mobility outcomes, and also aims to report the interplay between socio-economic background and characteristics such as sex, ethnicity, disability and geography, where data allows.

Limitations of the index

Data on its own cannot tell us why a problem has emerged or what interventions should be attempted in order to mitigate it. Further work needs to be carried out both to double-check the finding and to understand the causal mechanisms that generate it. Effective policymaking requires understanding of the causal processes involved. A major function of the index, then, is to help identify where deeper analysis is required in order to inform policy action.

This framework is also not designed to be exhaustive. In selecting drivers and indicators for inclusion, the main criterion has been whether there is sufficiently convincing evidence that the concept in question has a causal (direct or indirect) influence on aggregate rates of social mobility. As research in this area develops, and the evidence base improves, we expect that new concepts will be added, and that existing ones could be dropped.

Improving data

Our selection of intermediate outcomes has been limited by the availability of reliable data. As we note in this document and the main report, there are many surprising gaps in the current availability of data, which limits the Commission’s ability to report and advise the government accurately.

For example, to the best of our knowledge, there is no UK-wide, regularly-updated database relating young people’s educational achievements to their social backgrounds. The closest we can get is free school meals (FSM) eligibility, which captures roughly the poorest 15% of students in England. This forces us to divide pupils into 2 groups that broadly correlate with higher and lower income and occupational-class background.

However, subtle differences are lost in these groupings. Within-group variation is likely to be significant, and, even more seriously, the amount of within-group variation may itself vary across the UK. This makes FSM eligibility problematic for geographical analysis. It also does not capture family income over time. This means we cannot, for example, distinguish between children from families with permanently low incomes from ones with short periods of low income. Moreover, the criteria for FSM eligibility have expanded recently, making comparison over time more difficult.

We will look carefully at what can be done to improve this. We are committed to pursuing independent research looking at how the government can improve data on social mobility that will enhance not only this index but many other areas.

The components of the framework

The index aims to use evidence-based measures that track progress towards improved social mobility over the next generation’s life. As well as measures of eventual mobility outcomes, the index includes ‘drivers’, or enablers, of social mobility, along with intermediate-outcome measures for assessing progress. While the Commission has a duty to report every year, it only makes sense to report on the drivers of change and the intermediate outcomes annually, as only these are likely to change significantly each year.

10-yearly measures: mobility outcomes

The measurement framework starts with 10-yearly measures of the main sorts of social mobility within the adult population – the long-term outcomes. This is the ‘rear-view mirror’ look – an assessment of how social mobility has changed. In order to achieve a holistic view of social mobility, the index covers occupational, income, educational, housing and wealth mobility.

Historically, social mobility research focused primarily on occupational mobility.^{[footnote 1]} Occupations are associated with a wide range of important outcomes: people’s incomes, employment conditions and security, risks of unemployment, and health and wellbeing are all correlated with their type of occupation. Parental occupation can also readily be measured in survey research (though with some degree of measurement error), facilitating the study of intergenerational social mobility. It should be noted that sociologists and economists have also been researching the role of grandparents, but due to the scarcity of data, the measurement framework has not been extended in this way.

More recently, economists and the Commission itself have studied income mobility.^{[footnote 2]} Incomes provide a convenient metric for measuring a person’s starting point and their outcomes in life, although obtaining measures of parental income is much more challenging than in the case of parental occupation.

Educational mobility has also become of increasing interest and is widely used in comparative research by the Organisation for Economic Co-operation and Development (OECD) and World Bank.^{[footnote 3]} While correlated with occupational class and income, education captures a distinct dimension of stratification and is particularly important for social relationships and for health. It is also rather more straightforward to measure parental education than it is to measure either parental occupation or income.

Wealth is another distinct dimension of stratification and inequality, and parental wealth can have major implications for children’s life chances over and above parental income. Wealth shapes life chances, protecting against income shocks, providing income in retirement and enabling parents to invest in their children’s future. Wealth inequalities can hinder long-term social mobility.^{[footnote 4]} The asset-rich have better outcomes in employment, earnings, health and psychological wellbeing.^{[footnote 5]} Wealth is much more unequally distributed than income.^{[footnote 6]} The wealth share of the top 10% increased and that of the bottom 40% and 60% decreased after the financial crisis.^{[footnote 7]} For most people, housing property is the single most important component of their wealth, and access to home ownership has become of increasing public concern. Housing conditions also have important consequences for children too.^{[footnote 8]}

Annual indicators: the intermediate mobility outcomes

For the annual measures of social mobility, the framework focuses on the experiences of young people as they move through education and into the labour market. The overall rate of social mobility in a society changes very slowly, since many adults remain in the same class or occupational level for several years. It will scarcely show much change from year to year and will not be informative in an annual index. The measures for the annual index, therefore, focus on the experiences of each new cohort of young people leaving school and entering the labour market, which can be expected to change from year to year. These experiences will in turn shape their future prospects of social mobility. We term these ‘intermediate outcomes’, since they reflect both prior changes in the drivers, as well as causally affecting future chances in the labour market. This annual index gives us the ability to predict how future social mobility rates will be affected and to steer today’s government responses accordingly.

5-yearly indicators: intermediate outcomes, by protected characteristics and geography

The new index also aims to look at the impact of ‘protected characteristics’ such as sex, ethnicity and disability, as well as geography, on social mobility chances. (Additional ‘protected characteristics’ should be added to this shortlist, as data becomes available). As with the old index, we envision the new one will also assess how social mobility outcomes compare in local geographic places. These kinds of analyses need large sample sizes, to see how social mobility processes play out among different subgroups or subareas of the population. To achieve this, the framework pools 5-years’ worth of annual data, and will report every 5 years on these measures.

Drivers: the enablers of social mobility

Including the drivers of change provides an early-warning system to identify areas for potential policy action. By ‘drivers’, we mean the social and economic conditions that have a causal effect on aggregate rates of social mobility.

The extent of completed social mobility experienced by the adult population will have been shaped by a range of causal processes going back many years or even decades. For this reason, measuring final social mobility outcomes has been described as ‘looking in the rear-view mirror’.^{[footnote 9]} But if we measure what is currently happening to the drivers of change, we can look forward and anticipate what future social mobility trends might be. By measuring drivers we can assess whether rates of social mobility in the future are likely to improve or deteriorate, and whether the government is putting in place the foundations for promoting social mobility.

No drivers of drivers

Since the index is essentially descriptive, it does not include ultimate causes, or ‘drivers of drivers’. For example, the conditions of childhood are included as drivers, but the causes of these conditions are not. Similarly, educational opportunity is included as a driver, but the drivers of educational opportunity are not (for example, what makes a good school?). While these issues are of critical importance, detailed analysis and policy recommendations around the ultimate causes of drivers go beyond the descriptive scope of the index. We expect to publish separate detailed reports on these issues in the future.

Aggregate rates versus individual chances

Like the previous index, the focus is getting an accurate count of the number of people that have experienced mobility, with a view to the actions that government and other bodies can undertake in order to promote these rates. We do not address the distinct, although related, issue of an individual’s chances of mobility and what parents can do to help their children’s chances. We plan to address these in future work. We focus on measuring the extent to which rates of social mobility are improving over time in the UK and across the different parts of the UK, and which specific groups in society have greater or lesser chances of social mobility.

In summary, the index covers:

• 10-yearly measures of social mobility outcomes in the adult population, covering occupational, income, educational, housing and wealth mobility

• 5-yearly measures of the intermediate outcomes, based on the pooled annual data and exploring overlap between socio-economic background and other characteristics (such as sex, ethnicity, disability) as well as subnational geographic analysis

• Annual measures of how socio-economic background is related to the drivers and intermediate outcomes such as educational attainment and post-school transitions into the labour market

• The drivers of change, setting out the social and economic conditions that affect social mobility

In addition, there are some important measures that are not available annually, but should be included when they can be. For example, the results of the OECD’s Programme of International Student Assessment (PISA), which are cross-national assessments of student achievement at age 15, will enable us to set the UK’s achievements in a broader comparative perspective. The OECD’s 10-yearly Programme of International Assessment of Adult Competencies (PIAAC) is one of the very few data sources with information on respondents’ skills, not just their paper qualifications, and again enables us to compare the UK rigorously with other countries.

The measurement framework could in principle be extended to include measures of the consequences of social mobility. For example, the impact of social mobility on wellbeing, economic growth or social cohesion. Similarly, it could be extended in the other direction to look at what might be called the ‘drivers of the drivers’, such as the causes of youth unemployment.

Mobility outcomes

The first section of the new index covers 5 dimensions of intergenerational social mobility in the adult population. Social mobility has been a major topic of research both in sociology and economics, with sociologists tending to focus on occupational class mobility and economists focusing on income mobility. While we expect rates of occupational class mobility and income mobility to have a lot in common, social class and income are conceptually distinct and we would not necessarily expect findings about each to be exact replicas of the other. Indeed, there is one notable example where trends in income and class mobility in Britain appear to have followed distinctly different trajectories in the late 20th century.^{[footnote 10]}

We also add intergenerational educational mobility to the measurement framework. Educational mobility is a relative newcomer to mobility research, but inequalities in educational attainment among parents are highly consequential for their children’s life chances, in some respects even more influential than either parental occupation or income.^{[footnote 11]} Although closely related to occupation and income, education can also be seen as a distinct dimension of stratification. It reflects cultural differences and distinctions, not just the distribution of material advantage and disadvantage, and has been described as one of the ‘big 4’ dimensions of intergenerational transmission of advantage.^{[footnote 12]} Given the rapid changes in the distribution of education over the last 50 years, educational mobility potentially could reveal rather different patterns from either class or income mobility. It has also become a mainstay of cross-national studies by bodies such as those of the World Bank and the OECD.^{[footnote 13]}

Another newcomer is housing mobility. Housing is a major component of wealth for many people and so patterns of housing mobility – moving from one tenure to another, such as from renting to home ownership – will give us some clues about wealth mobility.^{[footnote 14]} Housing mobility is also of considerable interest in its own right, given the importance of housing for wellbeing, security, and asset growth. Housing conditions, such as overcrowding, are also important for children’s life chances, affecting their health and education, and is strongly correlated with housing tenure (being much rarer among owner-occupiers)

Recent research has also demonstrated that intergenerational housing mobility follows rather different geographical patterns from the other types of mobility.^{[footnote 15]} It  becomes an important factor for a more rounded and holistic understanding of social mobility.

Finally, wealth – the stock of assets that a person holds as distinct from their stream of income. This has recently been highlighted as an important factor perpetuating inequality across the generations and has a significant influence on a person’s mobility chances in other domains.^{[footnote 16]} Wealth is more unequally distributed than the other dimensions, with a substantial proportion of the population having no wealth at all.^{[footnote 17]} Wealth can also be directly transmitted from parents to children in a way that is simply not possible in the case of education, income or, except in a few cases, occupation. However, it is particularly challenging to secure suitable data for studying wealth mobility in Britain.

Absolute and relative mobility measures

In examining these different types of mobility, the measurement framework distinguishes between absolute and relative mobility. Both types are important to consider as they provide different perspectives on mobility. As noted earlier, the concept of absolute mobility captures whether people are doing better, worse, or the same as their parents and can be measured, for example, by the proportion of people who have higher real incomes than their parents did at a similar stage of the life cycle. Relative mobility, in contrast, compares the mobility chances of people coming from different social origins. It focuses on the extent to which people who grew up in more advantaged homes end up with higher incomes or occupational positions than their peers who had grown up in less advantaged homes. Measures of relative mobility can be thought of as describing the strength of the intrinsic association between parents’ and adult children’s positions. Economists, for example, measure relative income mobility by computing the correlation between parents’ and children’s percentile positions within the income distribution.

Research suggests that rates of absolute mobility are typically driven by changes in the rate of economic growth and in the shape of the occupational structure or the distribution of income. For example, during a period of rapid economic growth, there tends to be increasing numbers of higher-level vacancies to be filled – ‘increasing room at the top’. So higher percentages of the population will experience upward mobility (and a lower percentage will experience downward mobility). Similarly, in a period of sustained economic growth, we can expect that a larger proportion of the population will have higher real incomes than their own parents had in the past. Absolute rates of upward mobility will, therefore, tend to increase, and rates of downward mobility will decline.

However, even if there is an overall increase in upward mobility, it might still be the case that people from higher socio-economic backgrounds remain at the front of the queue for top positions, and that those from lower socio-economic backgrounds remain at the back of the queue. Relative mobility captures this latter idea of people’s positions in the queue while absolute mobility focuses on the proportions who have actually moved. For example, in the case of the expansion of higher education in the latter decades of the 20th century, there was a large increase in the number of ‘first generation’ university students. But the children of more affluent parents were at the front of the queue and took even more advantage of the new opportunities than did the children of less affluent parents. So while children from poorer families did improve their chances of getting to university, children from affluent families did even better. As a result, the higher education participation gaps between rich and poor children actually widened (equivalent to a decline in relative mobility) at the same time as absolute upward educational mobility increased.^{[footnote 18]}

A further important distinction between the 2 concepts is that expanding ‘room at the top’ can increase upward mobility without any increase in downward mobility. Successful measures to increase relative mobility, in contrast, will lead to increases both in upward and in downward mobility. In effect, attempts to increase relative mobility involve a ‘zero-sum game’ with both winners and losers (and the potential therefore for pushback from potential losers) whereas attempts to increase absolute upward mobility involve a ‘positive-sum game’, where winners can predominate.

However, while absolute and relative mobility are conceptually distinct concepts, they may in practice be related (although, as we showed with educational mobility, this may not always hold true). So, cross-national comparisons of trends in absolute and relative rates of mobility suggest that periods when rates of upward mobility have increased have also exhibited greater relative mobility.^{[footnote 19]} Similarly, in the case of geographical differences, recent research in Britain and the US has demonstrated that areas with higher absolute rates of upward mobility also have greater relative mobility.^{[footnote 20]} In particular, improved chances for absolute upward mobility might well drive improvements in relative mobility. For example, when there is a slack labour market with many applicants for each vacancy, employers may hire on tighter criteria. But in good times employers may become less risk averse and may become more open to recruiting applicants from less-favoured backgrounds. Similarly, in good times, affluent families may be relatively relaxed about their children’s success but may strive harder to find competitive advantage when prospects are poorer. Running in the opposite causal direction, increased relative mobility might lead to a more dynamic economy (‘diversity pays’) and for this reason increase opportunities for absolute upward mobility. In terms of policies, this has the implication that targeting absolute versus relative mobility is not necessarily an ‘either/or’ choice: specific policies (for example ‘levelling up’ disadvantaged parts of the country) might actually impact on both absolute and relative mobility simultaneously.

We next turn to more detailed accounts of the indicators in the measurement framework for measuring these 5 different types of mobility.

Mobility outcome 1: Intergenerational occupational class mobility

The mainstay of research on social class mobility has been the large-scale representative national survey, such as the Office for National Statistics’s Labour Force Survey (LFS) and, previously, the General Household Survey (GHS). Respondents are typically asked to provide the information necessary to measure their own current occupational class position (which can then be coded into the National Statistics – Socio-Economic Classification). And they are also asked to report the same kind of information about the occupational class position that their parent(s) held at the time that the respondent was aged 14. In this way the survey captures a measure of parental class position at the time the respondent was growing up. While there is an undoubted risk of ‘recall bias’ when adult children are asked to remember what their parents’ occupations or education had been in the past, methodological research has suggested that their reports are reasonably accurate, the biggest problem being missing data.^{[footnote 21]}

Other data sources can be used which avoid these issues of recall bias, and they are included in the measurement framework too. These sources include the panel studies such as the 1970 British Cohort Study and the linked census data of the Longitudinal Study. However, these studies are not available on a regular basis in the way that the LFSs are, and therefore less suitable for regular monitoring of class mobility. Therefore, we regard them as supplements to the LFS.

Representative national surveys such as the LFS, containing the required questions about respondents’ and their parents’ occupational positions, have been conducted irregularly in Britain since 1949. There have been some issues of comparability between the methods employed in these studies, but a series of surveys from the 1970s onwards up to the most recent LFS allow us to track trends in occupational class mobility over the last 5 decades.

It is also possible to compare class mobility in Britain with that in other European countries, using the harmonised data in the European Social Survey (ESS), although the ESS has relatively small sample sizes and hence large confidence intervals.^{[footnote 22]} Comparative research using larger national samples has also been carried out (although these tend to involve difficulties with measurement harmonisation).^{[footnote 23]}

Mobility outcome 2: Educational mobility

Educational mobility can be studied in exactly the same way as occupational mobility, using large-scale representative surveys with questions asking respondents to recall the educational levels that their parents had reached. Since both occupational class and educational level can be classified into categorical schemas, exactly the same statistical techniques can be used too.

While educational mobility has been little studied in Britain, it has become a mainstay of cross-national research by the OECD and World Bank on social mobility. In Britain there is a scarcity of suitable data since large-scale surveys such as the LFS do not measure parental education. However, all the main UK panel studies do contain data on parental education and we will, in future, illustrate what can be done using the British Household Panel Study and the UK Household Longitudinal Study (UKHLS).^{[footnote 24]}

Comparative research is also feasible for educational mobility, indeed more so than for occupational mobility. This is largely because the OECD’s PIAAC involves large-scale surveys (with around 5,000 respondents in each country) in over 40 developed countries and includes standardised questions (and harmonised coding using the International Standard Classification of Education classification) of respondents’ and parental highest qualifications. The second cycle of this programme is currently underway and the results will be available in 2024.

Mobility outcome 3: Housing mobility

Housing mobility can be studied in the same kind of way and with the same statistical methods as occupational and educational mobility. Parental housing is similar to parental education and occupation in that respondents can recall reasonably accurately what the housing situation of their household was when they were growing up. Suitable questions can therefore, at least in principle, be included in representative national surveys such as the LFS. Housing mobility does not therefore require long-term panel studies or linked administrative data in the way that parental income does. However, there is in practice a problem of limited data availability. To the best of our knowledge they are only included in the Wealth and Assets Survey (WAS), a biennial survey starting in 2006 and covering Great Britain (excluding Northern Ireland). Housing mobility can also be measured using the birth cohort studies and using the linked census data of the ONS Longitudinal Study.

One recent study of housing mobility in Britain using the WAS and the National Child Development Study (NCDS) and Britain’s birth cohort studies has shown starkly different patterns of change over time from any of the trends seen in previous sections of this review. Firstly, with respect to absolute mobility, the study found a substantial decline in upward housing mobility. Among people born in the late 1950s, 74% owned their own home even though their parents had not been home-owners. This fell to 49% of people born 20 years later in the late 1970s. This decline in chances of upward housing mobility was accompanied by a large decline in relative housing mobility: home ownership by the children became much more tightly linked to the home ownership of their parents than it had done previously, the odds ratio increasing from 1.6 to 5.4.^{[footnote 25]}

To the best of our knowledge, there are no harmonised cross-national studies of housing mobility, so we are unable to set the British patterns of mobility within an international perspective.

Mobility outcome 4: Income mobility

The measurement of income mobility involves very different challenges from those involved in the measurement of occupational or educational mobility. Economists have not generally used sources like the LFSs which have been the mainstay for research on class mobility.^{[footnote 26]} This is because respondents’ knowledge of their parents’ income when they were growing up are likely to be inaccurate, particularly as income tends to be quite volatile from year-to-year, more so than with occupations. As a result economists have preferred to use long-term panel studies where members of families have been followed up over time. In such panel-studies, parents of teenage children might be asked directly about their income in, say, 1974, and then their children would be followed up and be directly asked about their own income 30 or so years later, in 2004, when they (the children) had reached mid-career themselves. The parents’ income is adjusted for inflation in order to be comparable with those of their children.

Unfortunately there are only a few long-term panel studies in Britain which enable this kind of exercise to be conducted (notably the 1958 NCDS birth cohort study, the 1970 Birth Cohort Study, and the 2001 Millennium Birth Cohort Study), and they also suffer from considerable attrition. A stronger method has been used by American researchers who have been able to link the tax records of parents (from the time children were growing up) with the adult children’s tax records. Ideally Britain would develop a similar dataset based on linked tax records.

In Britain, these long-term panel studies have been the primary source used by economists to measure relative income mobility. Recently, however, a mixed method has been developed for studying absolute income mobility. This has enabled economists to chart trends in absolute rates over time analogous to the sociologists’ measure of the absolute rate of upward class mobility. In the case of income mobility, the upward mobility rate is defined as the percentage of children whose real family income at a given age was higher than their parents’ family income had been at the same age. These mixed methods involve the combination of results from panel studies and ongoing cross-section surveys similar to the LFS, and enable researchers to produce regular estimates of absolute mobility rates suitable for a monitoring exercise. However, the methods do involve a strong assumption (about the invariance of relative mobility over time) and should be regarded as second best to the gold standard of linked parent and child tax records that has been used in the US. However, checks with data from countries where the new method can be compared with gold-standard data linkage indicates that the new method provides a reasonable approximation.^{[footnote 27]}

The recent British research using these methods combined results from a long-term panel study (the 1970 British Cohort Study) with annual data from the series of Family Resources and Family Expenditure Survey (FRS and FES) to estimate absolute income mobility of adult children at age 30 across birth cohorts born from 1964 to 1987 (so reaching age 30 from 1994 to 2017). The method, termed the ‘copula and marginals’ method, takes the parent and child income matrix (the ‘copula’) which can be derived from the 1970 British Cohort Study and combines this with the information in the FRS and FES about the changing distributions of parents’ and of children’s real incomes (the ‘marginals’).^{[footnote 28]}

The results for Britain show that there was an increase in absolute upward mobility among the earlier cohorts peaking among those born in the mid-1970s and then declining among the most recent birth cohorts.

For measuring relative income mobility in Britain, the only suitable data at present are long-term panel studies such as the 1958 NCDS Birth Cohort Study and the 1970 British Cohort study described previously. The members of these birth cohorts are still being followed up as they grow older, and the most recent data from them enables us to observe incomes of people in mid-career. This is important because, not only does income tend to increase in the course of a person’s labour market career, but the strength of intergenerational persistence is markedly greater when observed in mid-career than when observed earlier. This is analogous to the phenomenon of ‘counter mobility’ described in the context of intergenerational occupational mobility.

A number of studies have used these birth cohort data to look at levels and trends in relative income mobility. The most recent study has been able to estimate a son’s lifetime economic mobility by drawing on the income data reported by sons from ages 26 to 42 years. This study showed that intergenerational persistence was markedly greater among sons in the 1970 birth cohort (a rank correlation of around 0.31) than it had been in the 1958 birth cohort (a rank correlation of 0.20).^{[footnote 29]} (The study focused on sons because many daughters have intermittent labour market careers, reflecting pregnancy and maternity).

The Millennium Cohort study, following up young people born between 2000 and 2002, is a successor to the 1958 and 1970 birth cohorts, and will in due course enable economists to look at relative income mobility across a wider range of cohorts.

Another panel study that has been used to study relative income mobility is the (UKHLS). This is based on a different design from the birth cohort studies, however, and therefore does not yield strictly comparable data. One recent study using the UKHLS data to study household income mobility covered cohorts born from 1973 to 1991, so covering more recent birth cohorts.^{[footnote 30]} The study found a rank correlation of 0.30, closely in line with that found for the 1970 birth cohort. However, methodological differences in the techniques used mean that comparison between this study and the birth cohort studies are somewhat hazardous.

Regular monitoring of relative income mobility appears to be rather problematic at present, given the limited data sources available in Britain. The Millennium Cohort Study will fill the gap in the future but it will also be important to ensure that new long-term panel studies are commissioned in order to maintain the series.

There have been a number of studies comparing rates of both absolute and relative income mobility between affluent countries. These studies have often had to use data based on very different methodologies in contrast with the harmonised surveys of the ESS and PIAAC. However, harmonised national panel studies offer a promising way forward.

Mobility outcome 5: Wealth mobility

The study of wealth mobility involves similar challenges to those involved with the study of income mobility. The strongest methodology is to use linked administrative data, as has been possible in some Nordic countries which maintain detailed population registers. A second-best method would be to use long-term panel studies like those used to study income mobility. Unfortunately detailed information on parental wealth has not been available in any of the British panel studies.

However, Britain is fortunate to have the WAS, a biennial survey started in 2006 covering Great Britain (not Northern Ireland). This gives detailed information on the wealth held by sample members, and it also includes some recall information on parental circumstances such as their occupation, education and housing (but not their wealth). So instead of directly measuring parental wealth, researchers have been able to use this other parental information to derive a proxy for parental wealth.^{[footnote 31]} In essence, the method is to use data from older respondents in the survey to establish the pattern of relationships between education, occupation, housing tenure and wealth. This empirically-established pattern is then used to impute the wealth of parents with given education, occupation and housing tenure. This method naturally has some limitations: in particular it ignores variations in wealth between people who have similar mixes of education, occupation and housing tenure. In this particular case there is also the problem that the age of parents was not ascertained in the questionnaire, and so issues of life-cycle bias are present. These limitations mean that the estimates of intergenerational transmission of wealth are likely to be downwardly biased. Nevertheless, this is the best that can be done with the current data available for Britain, and offers the prospect of being able to monitor future trends in intergenerational wealth transmission.

The estimates that this approach has yielded for Britain are essentially equivalent to the measures of relative mobility estimated for income mobility. As it happens, the estimate for wealth turns out to be very similar to recent estimates of relative income mobility, namely around 0.30 (based on the rank-rank correlation, which appears to be relatively robust to problems such as life-cycle bias). It is very possible that the ‘true’ correlation will be somewhat higher than this.

Summary of final mobility outcomes

Measuring mobility outcomes is challenging, especially in the case of income mobility where we have to rely on a rather small number of long-term panel studies for appropriate measures of parental income. While the inclusion in the LFSs of detailed questions on parental occupations has been a huge asset to the study of occupational mobility, and the WAS is a great asset to the study of housing mobility, there remain serious data gaps with respect to both education and wealth mobility.

There are also complex technical issues, for example with respect to recall bias, life-cycle bias, attrition and attenuation that need to be addressed when analysing the available data.

The intermediate outcomes

The second main component of the index consists of ‘intermediate outcomes’ over the course of young people’s school careers, transitions into work, and their early careers in the labour market. Here we are interested in the extent to which young people from different social backgrounds acquire the educational skills and qualifications and labour market experiences that will enhance their chances of upward mobility (or reduce their chances of downward mobility) subsequently. We term these intermediate outcomes since they are likely to be highly influential in people’s eventual social mobility. These intermediate outcomes are to be measured
annually and are the core of the annual social mobility index.

Selection of outcomes

We take a life-course approach covering:

Intermediate outcome	Age
1 – The years of compulsory schooling	5 to 16
2 – Routes into work	16 to 29
3 – Work in early adulthood	25 to 29
4 – Career progression	35 to 44

In principle, the framework could be extended to include later stages of the life course too, such as early retirement. It could also be extended to include the pre-school years, and could potentially include other intermediate outcomes such as young people’s chances of buying their own homes or of acquiring pension wealth through membership of an occupational pension scheme. The Commission will keep these potential additions under review.

We include a range of ages because social disparities emerge early on, and tend to increase later in life.^{[footnote 32]} What’s more, a consensus seems to be emerging that all stages of childhood are important for mobility.^{[footnote 33]}

The transition years cover the period when some young people leave education and move into training, apprenticeships, work or, more worryingly for their future career prospects, into unemployment or economic inactivity (although in England continuing participation in education, at least on a part-time basis, is now legally required until age 18). These years are in many respects critical for young people’s future progress in the labour market and for their chances of social mobility. There is powerful evidence of long-term scarring effects of youth unemployment on future career trajectories. At the other pole, many young people will move to higher education, which has much higher returns in the labour market than do lower level qualifications such as GCSEs in England, Wales and Northern Ireland or Nationals in Scotland (though returns to education also vary depending on the institution that a person attended and the subjects studied).^{[footnote 34]}

The kinds of job that one obtains after leaving full-time education will also be influential for subsequent careers, as some types of jobs offer greater prospects for acquiring skills or form part of a regular career ladder. However, first jobs are not wholly determinative of later outcomes. Social disparities in labour market outcomes tend to be relatively small at the start of the career and then progressively widen. This is reflected in the way that the association between parental incomes and children’s incomes becomes stronger at older ages.^{[footnote 35]} Therefore, we include career progression until the mid-career phase (sometimes termed ‘occupational maturity’) in the measurement framework.

Across all these career stages we are interested in the extent of the differences between people from different socio-economic backgrounds, and in particular the outcomes for lower socio-economic groups. In the following sections, we look at each of these stages in turn and describe the indicators to be included in the measurement framework.

Intermediate outcome 1: The years of compulsory schooling (aged 5 to 16)

There has been a very large literature on the importance of different stages of schooling, with powerful theoretical arguments that investments in preschool education are likely to be the most influential. However, the empirical evidence is rather mixed and on balance we think it is safest to assume that there is no one ‘critical’ period and that all stages of education are important. Therefore, we follow the official government approach, dividing education into early years (age 2 to 5 years), primary school, with assessments taken at around age 11 years (key stage 2), and compulsory secondary schooling with formal GCSE assessment at around age 16 years in England, Wales and Northern Ireland (key stage 4).

We must emphasise that education is a devolved responsibility and that Scotland has long had a distinct educational system and does not use the key stage framework. However, there are formal examinations – National 1 to 5 (usually taken at age 15 to 16 years), Highers, and Advanced Highers – in Scotland which have broadly similar roles to GCSEs and A levels in England, Wales and Northern Ireland. Scotland differs from the rest of the UK, however, in having 6 rather than 7 years of secondary education, and 4-year rather than 3-year first degree courses (and with free tuition for Scottish students at Scottish universities). Northern Ireland also differs from England, Scotland and Wales in having a predominantly selective system of secondary education. These institutional differences (especially with respect to examinations) limit comparability, although they also provide opportunities to explore the effects of different policies.

One further major issue with the educational outcomes in the measurement framework is that of ‘grade inflation’. This was highlighted during the COVID-19 pandemic when the regular externally-assessed GCSE and A Level examinations were replaced by teacher assessments. This means that there may be a lack of equivalence over time in the meaning of the exam results, and potentially of their relationships with social background too. There is also Ofqual’s new National Reference Test, which has measured performance in English (reading and writing) and in maths annually since 2017, using a standardised approach and designed to ensure equivalence over time. However, this test is intended to inform the setting of grade boundaries at GCSE at the national level, and results by social background do not appear to be available.^{[footnote 36]} The framework includes a new measure developed by government statisticians to address this issue as well as direct measures of educational skills and competencies, not simply of paper qualifications.

Selection of indicators

As with the drivers, the ideal is to have regular, annual measures of these indicators for the UK as a whole, for the 4 nations as well as for regions and local authorities. We also need measures of social background, ideally ones which are consistent across the different indicators, enabling one to compare the evolution of social inequalities across the life course.

There is in all 4 nations a great deal of administrative data on educational outcomes. For example, in England we can use data on the Early Years Foundation Stage Profile, on attainments at age 11 years (key stage 2), and a range of measures of attainments at age 16 years (key stage 4). However, because education is a devolved responsibility there is, to the best of our knowledge, no harmonised dataset covering the whole of the UK. In Wales and in Northern Ireland there is data at 11 and 16 along the same lines as those for England. In all 3 countries receipt of FSM (which roughly equates to a measure of poverty) is the only available measure of social background, but eligibility for FSM is a devolved matter, so the requirements are slightly different in each case. In Scotland, a completely different area-based measure of social background is used, the Scottish Index of Multiple Deprivation (SIMD). We cannot therefore use administrative sources to compare across the UK as a whole. Also, as noted earlier, FSM is not an ideal measure because there is a great deal of heterogeneity within the non-FSM population (roughly 75% of young people). This makes FSM and its derivatives a poor measure for making comparisons between areas because the extent of heterogeneity will vary between areas. In addition, due to the transitional protections covering FSM eligibility as we move from old-style multiple benefits to Universal Credit, there is a greatly increased number of children eligible for FSM. This means that the average child on FSM today is probably not as disadvantaged as the average child on FSM 10 years ago. So this may contribute to closing the measured gap, even with no underlying change in the pattern of achievement.

Other sources such as the Next Steps panel and Millennium Cohort Studies data provide a much more fine-grained measure of social origin and would therefore be preferable, but are not available annually. They also show substantially larger inequalities than the administrative data do.

The absence of harmonised measures across the UK is especially unfortunate as the differences between the education systems of the 4 nations could give policymakers important insights into the effectiveness of their respective educational policies. To our knowledge, the only regular harmonised dataset covering the whole of the UK with a consistent measure of social background is one constructed by the OECD as part of PISA. This measures the achievements in maths, literacy and science of students aged 15 years, which we use as our indicator of skills.^{[footnote 37]} There are also one-off academic studies such as the Millennium Cohort Study (covering the whole of the UK), and for England the Next Steps panel study, which have fine-grained measures of social background. The Growing up in England study (GUIE), which links Census 2011 data with school results, also includes fine-grained measures of social background and will permit in-depth analysis of social inequalities in English education.

The disadvantage gap index is a new measure, developed by government statisticians in anticipation of changes to assessment methods at age 16 years (key stage 4).^{[footnote 38]} It is based on the FSM that is the only background measure currently included in English administrative data on school education.^{[footnote 39]} However, it differs from previous measures used by the Department for Education (DfE) in that it is a ‘positional’ measure. In effect, the index measures how disadvantaged (FSM eligible) and non-disadvantaged students differ with respect to their positions in rankings of performance at GCSE. The index ranks pupils in England and asks whether poorer pupils typically rank lower than other pupils. A disadvantage gap of zero would indicate that pupils from disadvantaged backgrounds perform equally as well as pupils from other backgrounds.

The gap index is designed to be resilient to changes in grading systems and assessment methods. While the absolute differences (in English and Maths GCSE grades) may differ between years, perhaps because of changes in examining methods or because of grade inflation, the gap index measures results in terms of the rankings of poorer pupils and non-poorer pupils and therefore it offers greater comparability between years.

Intermediate outcome 2: Routes into work – the transition years (aged 16 to 29)

The paths that young people take after the end of compulsory schooling have major implications for their future careers. The end of compulsory schooling is a major transition point for many young people, although with the gradual increase in numbers staying on in education after the age of 16 years (especially after the COVID-19 pandemic limited opportunities for young people in the labour market),^{[footnote 40]} the key transition will tend to move upwards to age 18 years, as it has already done in many other developed countries. In England (but not in Northern Ireland, Scotland or Wales) the participation age has already been raised to 18 years, with young people aged 16 and 17 years required to continue in education or training either full-time or part-time (in conjunction with employment, internship or voluntary work).

Lack of training or work experience after the age of 16 years is particularly disadvantageous for young people’s subsequent career and earnings, while staying on to take A levels, which opens the route to higher education, is particularly advantageous. It is important to monitor disparities between young people from different backgrounds in the routes that they take, and whether the gaps are widening or narrowing.

It is also important to monitor disparities not just with respect to the initial paths that people take but also their eventual educational achievements, both their paper qualifications and the skills that they have acquired. While much previous research has emphasised the importance of qualifications such as degrees and professional training for subsequent careers and earnings, more recent research has suggested that skills, not just paper qualifications, are important too (especially given the concerns, confirmed by the Wolf Report, that many paper qualifications have little content or value in the labour market and that practical training on the job may be more valuable).^{[footnote 41]}

Selection of indicators

There is good administrative data for England on destinations after age 16 years (key stage 4), although only using the binary advantage or disadvantage measure based on receipt of FSM. There is also data for Scotland on destinations, but using a measure of area deprivation (the SIMD) which is not comparable with the English measures based on FSM.

We can also use the LFSs, which cover the whole of the UK in a consistent way and have data on whether young people are continuing with their education, have taken up employment, or are not in education, employment or training (NEET).

Data on entry to higher education are more problematic. The Universities and Colleges Admissions Service provides UK-wide statistics on applications and acceptances at each higher education (HE) institution, but uses an area-based measure of social background (POLAR and TUNDRA) rather than a standard measure based on parental circumstances. It is not at all clear that area-based measures tap the same concept that standard measures of parental occupation, education or income tap, and they are also vulnerable to the ecological fallacy (making inferences about individuals based on the groups they belong to).^{[footnote 42]} The Higher Education Statistics Agency also provides data on HE students and their outcomes, but while it provides statistics on gender, ethnicity and disability, it does not provide any data on social background. The Longitudinal Education Outcomes (LEO) linked administrative dataset, in contrast, does provide suitable data and measures of social background, but only for England.

The only suitable UK-wide data covering higher education is that from the LFS, which we use for the indicator of highest qualification achieved. The LFS can be disaggregated by country and region, but not by local authorities. It does, however, include a granular measure of parental socio-economic position.

Direct measures of skills, as opposed to participation or qualifications, have rarely been collected in the UK (although there were once direct measures of literacy, which have since been discontinued). However, the UK does participate in the OECD’s PIAAC. This was first conducted in 2012 and a second round is planned for 2022, and will be included in the index as soon as the data becomes available. The only social background measure in PIAAC, however, is parental education, not parental occupation or income.

Intermediate outcome 3: Work in early adulthood (aged 25 to 29)

A person’s initial steps up the occupational ladder have major implications for their subsequent career. The labour market has been polarising in recent years into ‘good’ jobs with formal contracts, higher earnings, formal training, low risks of unemployment and regular career ladders in contrast to ‘precarious’ jobs with short-term contracts, agency working, wages often at or below the minimum wage, little training beyond that needed to carry out the tasks, high risks of unemployment and little in the way of potential advancement. Sorting into these different types of jobs is largely based on qualifications, but social origins also make a difference even after taking account of qualification levels. In addition migrants, especially those with little English, are particularly likely to find themselves in the precarious labour market.

To cover early labour market experiences we include indicators of unemployment, occupational level, and earnings among young people. We select those aged 25 to 29 years in order to cover young people who have gone through higher education and include the 4 classic outcomes of economic activity, unemployment, occupational level, and earnings.

Selection of indicators

For all of these intermediate outcomes we can use the LFS, as we did with NEET, permitting disaggregation to a regional level. The LEO linked administrative dataset covers economic activity and earnings and permits disaggregation to a local authority level, but only for England.

It is also possible to estimate, from the same LFS data, the net inequalities after taking account of, for example, qualifications achieved. A full analysis taking into account previous career stages, such as experiences of being NEET, however, requires panel data such as the UKHLS which takes into account previous stages of each respondent’s career. (The LFS only records current job position, not previous positions held by respondents, and is not able to factor in, for example, whether someone had been NEET 5 years earlier).

Intermediate outcome 4: Career progression (aged 35 to 44)

Just as socio-economic disparities tend to grow over the educational career, so there is evidence that they grow over the course of a career in the labour market. For example, the strength of the association between parents’ and children’s income increases over the course of the latter’s career, while sociologists have also drawn attention to the phenomenon of ‘counter mobility’ in the case of class mobility. So young people from professional and managerial backgrounds may experience downward mobility and take lower-level jobs at the beginning of their career but then move back up subsequently, outstripping their contemporaries who had started in the same jobs but were from more disadvantaged origins. There are also important variations in career progression between men and women, and among ethnic minorities, which we will turn to in the next section.

Selection of indicators

In order to investigate career progression, we need to use panel studies such as the UKHLS which follow the same people over their careers instead of the cross-section surveys like the LFS. For example, the LFS asks respondents for their current earnings at the time the survey was conducted, but it does not have room to ask about earnings earlier in the respondents’ careers and so cannot give us estimates of career progression. In contrast, the UKHLS covers the whole of the UK and follows up the same people over their life course with repeated interviews every year. These repeated interviews with the same people mean that accurate measures of career progression can be obtained. The UKHLS sample is also refreshed each year with the addition of new sample members, so that it can yield annual measures.

While there are a number of other panel studies of this kind (for example, birth cohorts of children born in 1946, 1958, and 2000), they do not permit annual readings of career progression. The measurement framework therefore uses the UKHLS for measuring career progression with respect to the acquisition of further training and qualifications, occupation and income.

We also include an indicator of the class pay gap. The class pay gap refers to the difference in pay between people from different social class backgrounds who have gained access to the same occupations.^{[footnote 43]} It appears to be strongly related to career progression, and to widen at older ages. Previous analysis of the class pay gap has primarily focused on elite occupations, although (like the gender pay gap) it can potentially apply to any occupation. Unlike the other measures of career progression, it can be estimated from cross-sectional surveys such as the LFS.

Summary of intermediate outcomes

The measurement framework focuses on differences in outcomes over the earlier stages of the life course, covering the years of compulsory schooling, the transition years, early adulthood, and career progression. There is a substantial body of theory and of empirical research showing that differences at later stages are shaped by differences at earlier stages, and are likely to have causal impacts on achieved rates of absolute and relative mobility.

We would expect the intermediate outcomes to influence the direction of future causal and policy analysis. For example, to the extent that young people from different social backgrounds experience different educational outcomes, what are the unmeasured factors that might influence this? What has mediated the relationship between background and attainment? There is also value in monitoring overall trends in the intermediate outcomes without splitting by socio-economic background. This may yield valuable information on whether conditions in the UK are favourable to mobility, and may also inform the future selection of drivers. Much of this will also be informed by pooled analysis of overlapping personal characteristics and geography, but further analysis is unfortunately hampered by the limitations of the background data.

Our selection of intermediate outcomes has been limited by the availability of reliable data. As we note, there are many surprising gaps in the current availability of data, limiting the Commission’s ability to report and advise the government accurately. There is a lack of granular data on parental circumstances, and, in the case of data resources such as the LFS or UKHLS which do cover the whole of the UK, insufficient sample sizes to permit analysis of local areas. There is also a lack of regular harmonised data across the UK for the educational indicators.

For example, to the best of our knowledge, there is no UK-wide, regularly-updated database relating young people’s educational achievements to their social backgrounds. The closest we can get is a half-measure in England, FSM eligibility, which captures roughly the poorest 15% of students. This forces us to divide pupils into 2 groups that broadly correlate with higher and lower income and occupational-class background. But subtle differences are lost in these groupings. Within-group variation is also likely to be significant, and, even more seriously, the amount of within-group variation may itself vary across the UK. This makes FSM eligibility problematic for geographical analysis. Neither does the measure capture the family’s income over time. So we can’t, for example, distinguish between children from families that have permanently low incomes from ones with short periods of low income.

We will look carefully at what can be done to improve this. We are committed to pursuing independent research looking at how the government can improve data on social mobility that will enhance not only this index but countless other areas. Therefore, we expect the index to evolve over time.

Intermediate outcomes: pooled 5-yearly analysis by protected characteristics and geography

A major interest of the Commission is whether differences in social mobility chances are broadly uniform across different parts of the UK and across different groups within the population. This is an important contribution to the Levelling Up agenda. Are social mobility chances, for example, the same in Scotland, Wales and Northern Ireland as they are in England? Are they the same for women as for men? For the main ethnic and religious groups in the UK, and for disabled people or those who have other protected characteristics? In considering these questions, it is worthwhile distinguishing 3 distinct issues. The first is whether one particular group, for example disabled people, have systematically poorer chances than do their peers without disabilities from similar social backgrounds. In other words, is there an added penalty for disabled people?

A second important issue is whether these group differences apply equally at different stages of the life cycle. For example, we know that young women have overtaken young men in their educational achievements. Have they also overtaken young men on entering the labour market and in their career progression? Or does this gap close or reverse when it comes to the labour market? We should also emphasise that, just as with the analysis of career progression in the previous section of this report, questions about different stages of the life cycle really need a panel study which follows up the same people over time. For example, comparing women now in late careers with those currently in the early stages of their career could be misleading given the great changes that have taken place in women’s education over the last 30 years.

And a third issue is whether the extent of class disparities are the same within different subgroups of the population. For example, it has been suggested that a selective system of education, with students separated at age 11 years into those attending grammar schools and those attaining non-selective secondary schools, tends to worsen social disparities in education.^{[footnote 44]} This then raises the question as to whether perhaps Northern Ireland, which has retained a selective system, has greater class inequalities within its educational system than do Wales or Scotland.

The measurement framework is designed to be able to address such questions. To conduct this analysis, we need larger samples since we are distinguishing a larger number of cells in the underlying tables. So we pool 5-years’ worth of LFS data when addressing these questions, and do not show trends over time since the LFS only started asking questions about parental background in 2014. However, providing the LFS continues to ask these questions on parental background, it will be possible to look at trends over time in future reports.

There are a very large number of potential analyses that can be carried out using the LFS. For each of the intermediate outcomes and indicators described in the previous section, it is possible to carry out analyses showing the interplay between social background, age, gender, ethnicity, religion, disability, and marriage and civil partnership.

In principle, the framework should be extended to cover all the ‘protected characteristics’ defined in the 2010 Equality Act, namely age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex and sexual orientation. The LFS covers all of these apart from gender reassignment, although for some characteristics the numbers are too small to permit precise estimation. Administrative datasets such as the DfE’s LEO dataset also permit analyses of overlapping characteristics for England, although only using the FSM-based measure of disadvantage and only for a limited range of protected characteristics.

Summary of overlapping characteristics and geographic measures

The overlap between socio-economic background and a range of ‘protected’ and geographic characteristics is an important feature of the index. These intersections can take a variety of forms. They could take the form of ‘double disadvantages’, as in the case of gender, social background and earnings. Or they could take the form of greater social-background differences in one context than another, as in the case of social background, educational qualifications and place. Or they could be apparent at one stage of the life course but not at others, as in the case of ethnicity, social background and transitions after the end of compulsory schooling.

While the LFSs have enabled us to explore a number of intersections, many important gaps remain. In particular, we cannot explore in as much detail as we would wish intersections with places using the LFS. Sample sizes limit the extent to which analyses of overlapping characteristics are possible even at a regional level, and analyses at a local authority level are out of the question using the LFS. The census and administrative datasets allow for some analyses of overlapping characteristics at a local authority level, but the 2021 to 2022 census does not contain a satisfactory measure of socio-economic background, while most administrative databases only have the limited FSM-based measures of socio-economic background. (The linked censuses of the ONS Longitudinal Study will however provide good measures of socio-economic background in due course).

Additional in-depth analysis will be needed to understand the mechanisms that generate the intersections and to develop appropriate policy responses.

Drivers of social mobility

A broad range of theories have been put forward by researchers and policy-makers suggesting why rates of social mobility change, or fail to change, and what governments could do to promote change in the future. The World Economic Forum (WEF) for example lists 10 pillars which it describes as the drivers or enablers of relative intergenerational mobility, namely health, education access, quality and equity, lifelong learning, technology access, work opportunities, fair wages and employment conditions, social protection and inclusive institutions.^{[footnote 45]} Our theory of change draws on some of the WEF’s ideas, but is based more securely on the most recent evidence about the causal role played by these drivers in promoting social mobility in developed countries such as Britain, both nationally and locally.

We focus therefore on the following drivers of social mobility:

• conditions of childhood
• educational opportunities and quality of schooling
• work opportunities for young people
• social capital and connections

Further details of the indicators and available data are provided in Annex B.

Other possible drivers

Ours is certainly not an exhaustive list of the possible drivers of social mobility and a long list of additional factors have been advanced – such as health and health inequalities, opportunities for geographical mobility (or internal migration), digital connectivity, or good transport connections. However, the evidence that these additional factors have a causal influence on rates of mobility is inconclusive or lacking. As a result, the framework is focused on drivers for which there is a convincing evidence base that they can have a direct causal influence on rates of social mobility, and for which data is available for regular monitoring. We have also excluded what could be called ‘the drivers of the drivers’. We prefer to keep the index focused and manageable at this stage, rather than to attempt to encompass the whole of social and economic policy. As new evidence emerges, we would expect the list of drivers to be refreshed.

Driver 1: Conditions of childhood

We naturally start at the beginning of life – the situation in which children grow up. A central issue here is the inequality between families in their mobility-relevant resources. The basic theory is that the larger the distance between each step of the ladder, the harder it will be for those at the bottom to climb up. Children’s social mobility chances are affected by whether their parents have more, or fewer, resources than other parents have to help them climb up the ladder.^{[footnote 46] [footnote 47] [footnote 48]} In turn, children’s educational success is likely to influence their subsequent careers.

This theory that financial inequality between families is a powerful barrier to improved relative income mobility has been popularised by the idea of the Great Gatsby Curve.^{[footnote 49]} The curve shows that, the more income inequality there is in a country, the less relative mobility there is. It is not wholly clear that this relationship is a causal one, but American experiments giving additional financial resources to poorer families have proved to have an impact on their children’s educational success.^{[footnote 50]} There is also strong American evidence that the extent of economic inequality within a local area (commuting zone or county) has a substantial causal impact on the area’s absolute rate of income mobility.^{[footnote 51]} In the British context, poverty (which is essentially a reflection of inequality at the lower end of the income distribution) has been seen as particularly important.

However, the family resources which may be relevant for mobility are not just economic ones. There can also be important differences in the educational and cultural resources that families have, and these educational resources may be as important as the economic ones for children’s success within the education system, or in obtaining good jobs subsequently. Certainly, having parents who have been through HE themselves is a great asset when trying to navigate the complex British HE system. Parental education can be even more important than parental occupation or income for children’s educational attainment.^{[footnote 52]} The distribution of parenting skills and cultural capital are likely to be important too.^{[footnote 53]} Sociologists have also argued that, in line with the psychological theory that losses outweigh gains, parents from advantaged backgrounds will be particularly keen to protect their children against downward educational mobility.^{[footnote 54]}

The unequal distributions of financial and cultural resources between families are undoubtedly major drivers of inequalities between children in their educational attainments and, given the strong links between educational attainment and labour market outcomes, are likely to be important for their achievements in the labour market too. However, while the unequal distribution of financial and cultural resources is a major reason for the persistence of advantage across generations, educational systems can to a greater or lesser extent mitigate the impact of these inequalities. Equality of opportunity in access to education has long been seen as a major driver of improved social mobility. Comparative research has demonstrated that the extent of educational opportunities and their quality can be important drivers of a country’s or local area’s rates of social mobility (both absolute and relative).^{[footnote 55]}

Selection of indicators

There is good administrative data across the UK on the percentage of children living in relative poverty (the Department for Work and Pensions’ Households Below Average Income statistics). There is also good data on earnings inequality in the different parts of the UK, although not specifically of income inequality among parents of school-age children (the ONS’s Annual Survey of Hours and Earnings, ASHE). Both of these indicators can be disaggregated to regional and local authority levels. In contrast, there is no standard data source on educational inequality between families, but a measure based on parental education can be constructed from the UKHLS. A limitation, though, is that sample sizes are not large enough to disaggregate to local authority levels. A range of more direct measures of parental engagement with their children, such as time spent with children, how often parents read to children or take them out to sporting activities, are available in various panel studies (for example the Millennium Cohort Study), but are not available on an annual basis.

What else did we consider?

This list could be expanded to include additional indicators such as child health. Access to high-quality pre-school childcare is another potential addition and has been shown to be associated with children’s development. Overcrowded housing also affects children’s subsequent progress, although research has suggested that housing tenure itself is not important once we take into account other drivers such as poverty.^{[footnote 56]} Given that these potential additions are correlated with indicators such as childhood poverty that are already included, we do not add them at this stage but will keep them under review.

Driver 2: Educational opportunities and quality of schooling

We have distinguished between 2 different sorts of educational drivers and interventions, namely those focused on expansion of educational opportunities and those focused on the equity of their distribution. Educational expansion over time has occurred in all developed countries through the progressive raising of the school-leaving age and the increasing provision of higher and further education after the age of compulsory schooling. Expansion of opportunities will increase absolute rates of upward educational mobility, and may also be important in increasing relative mobility. For example, while the legislation applies to all families irrespective of their social background, raising the school-leaving age has made more difference to disadvantaged groups, since the children from advantaged families typically already stay on longer at school. Rigorous studies in the UK and in Germany have both shown that raising the school-leaving age to 16 years reduced class inequalities in educational achievement, and in Germany it also had knock-on effects on relative occupational mobility too. The British researchers, however, were unable to find any impact on subsequent occupational mobility although it had positive impacts on educational mobility.^{[footnote 57]}

Equalising opportunities for young people from different backgrounds has also been a major focus of government interventions, such as comprehensive reorganisation in Britain. Both cross-national research and cross-area research in Britain has suggested that selective systems of education where children are selected for different educational tracks at the end of primary school tend to be associated with greater class inequalities in children’s educational achievements.^{[footnote 58]} Other reforms designed to equalise opportunities have included greater funding for schools serving disadvantaged areas (for example through the pupil premium) and reform of selection procedures for oversubscribed schools and universities (for example, contextual admissions), although the jury is still out on the effectiveness of these reforms.^{[footnote 59]}

It is likely that both expansion and improved equity are causally important for social mobility. A major piece of cross-national research on European countries found that: ‘Considering the broad picture, taking each country over the whole period we have studied, we find no cases in which social fluidity (that is, relative social mobility) increased without either an equalising effect of educational expansion or equalisation in the relationship between origins and education, or both.’^{[footnote 60]}

Selection of indicators

Measures of equality of opportunity in education are covered among the ‘intermediate outcomes’ in the next part of the measurement framework. In this section, therefore, we focus on the quality of education provided, and the opportunities for access to different forms of post-16 education (both of which were found to be important for explaining area differences in social mobility in the USA).

Our focus here is on the opportunities and the quality of the education provided for young people, rather than on, say, the standards that young people themselves have achieved. The standards achieved will in part be a result of the conditions of childhood and will therefore be partly a consequence rather than purely a driver.

It is difficult to obtain good data for these 4 indicators of opportunity and quality in UK education. Ideally we would construct measures, as American researchers have been able to, recording the number of college places per head in the relevant catchment area (recognising that the higher up the education ladder we go, the larger the geographical catchment area). To measure opportunities in the UK, we have to rely on rates of participation in post-16 education and training and in HE. Participation is a reasonable proxy for opportunity when measuring opportunities at the national level, where reforms such as the creation of new universities following the Robbins Report in the 1960s clearly provided greatly increased opportunities for young people to move on to higher education. Participation is, however, a less good proxy for measuring opportunities at a local level. This is because low take-up within an area may simply reflect the social composition of the area (such as childhood poverty) and the number of suitably-qualified applicants rather than the actual opportunities offered by the government. There is also the added complication that young people from advantaged backgrounds are much more likely to choose to move further afield for HE than are young people from less advantaged backgrounds.^{[footnote 61]}

Measuring the quality of the available schooling or HE is also challenging. The American research which demonstrated the causal role of school quality used output measures such as test scores and drop-out rates, after controlling for parental income. These are somewhat analogous to the measures of schools’ ‘added value’ – the difference between pupils’ attainment when they start and when they leave – which have been developed for key stage 5 (aged 16 to 18 years) in England. It is important to control for social background, or for prior attainment in the case of ‘added value’, because otherwise we risk confusing the quality of the education provided by the school with the quality of the intake to the school. Another possible indicator is the assessment of school quality published by Ofsted for England, and these were included in the original index. However, these assessments appear to have a low correlation with value-added measures, and are not available for the other countries of the UK in the same form.^{[footnote 62]} While the value-added measures have been criticised, they do seem to be the most suitable indicator available.^{[footnote 63]}

Assessments of the quality of HE are also challenging. There are a number of rankings of HE quality, but these rankings tend to take into account research activity as well as the teaching quality.^{[footnote 64]} They also fail to take account of inputs and confuse the quality of the institution’s performance with the quality of the intake. More useful for our purposes are the Higher Education Statistics Agency’s (HESA) performance indicators, such as drop-out rates from each institution. Crucially, HESA also provides benchmarks, based on the subjects studied and the prior academic qualifications of entrants. These benchmarks enable us to compare actual performance against expected performance based on the intake to each university. This data provides measures broadly equivalent to the value-added measures for schools and represents the best available measure for assessing university quality, and can be disaggregated to a national, regional and local area.

Further work with the education departments in all UK nations, the HESA and the ONS on developing better, UK-wide harmonised measures of educational quality and opportunity would be hugely beneficial.

Driver 3: Work opportunities for young people

Theory and research show that reducing inequality of opportunity within education is not sufficient for equalising mobility chances in the labour market. A great deal of research has demonstrated that, even among young people with similar educational achievements, those from more advantaged backgrounds fare better in the labour market and that, even within the same occupations, they earn substantially more than their equally qualified peers from less advantaged backgrounds.^{[footnote 65]} Also, local labour-market conditions appear to be important for social mobility over and above the educational profile of the local population.^{[footnote 66]}

The demand for labour, especially for high-skilled professional and managerial work, has played a prominent role in sociological accounts of the drivers of absolute rates of upward mobility. Nationally, the expansion of professional and managerial work has meant that there is ‘increasing room at the top’ and has enabled greater upward mobility from disadvantaged backgrounds.^{[footnote 67]} Similarly, local areas with more dynamic economies and greater demand for high-skilled labour tend to show higher rates of both absolute and relative mobility.^{[footnote 68]} This research also suggests that the more socially mobile areas are ones with higher ‘returns to education’ – that is to say, areas where highly qualified people are in demand and can command higher wages. This may be as a result of more knowledge-intensive industries in the area.^{[footnote 69]}

A related theory is that recruitment to some types of job is more ‘meritocratic’ in the sense that recruitment relies more on applicants’ skills and qualifications than on their social background. Research has suggested that social background plays a lesser role in the graduate labour market than it does in lower-level labour markets.^{[footnote 70]} Similarly research on the class pay gap (the earnings gap between people with the same qualifications but different social backgrounds) in Britain has shown that these gaps are smaller in newer or more technically-oriented fields such as engineering than in traditional ‘gentlemanly’ professions, such as architecture, law and medicine, which perhaps rely more on subjective measures of suitability for the profession.^{[footnote 71]}

Selection of indicators

Both nationally and locally, the demand for labour at the time someone starts to look for a job is likely to have a major impact on their occupational starting point (or unemployment) and is likely to have long-term implications for their career, given for example the ‘scarring effects’ of early unemployment or low-level work on future earnings.^{[footnote 72]} As previously discussed, work opportunities are important for understanding changing prospects for mobility both over time and across the different areas of the UK.

The most direct measure of work opportunities is the number of job vacancies. The ONS has conducted a regular survey of vacancies across the UK since 2003, and this gives an overall measure of demand for labour.^{[footnote 73]} However, the survey is not designed to be disaggregated either by country or by region, which limits its usefulness. Nor does it identify ‘entry level’ jobs, which are the most relevant for understanding the current opportunities for young people leaving education. We therefore include additional proxy indicators of work opportunities for young people, namely the rate of youth unemployment, the type of employment taken by young people, and their earnings. Youth unemployment rates, for example, tend to be much higher than those of older workers. They will also tend to be more sensitive to changes in the demand for labour (since many older workers will have relatively secure long-term jobs), and are, therefore, more suitable for annual measures than are the overall rates of unemployment.

Unemployment among young people can be estimated from the LFSs, both over time and across the 4 nations and the English regions. To obtain data broken down by local authority, we have to turn to administrative data on the claimant count. The ONS publishes experimental data on claimant counts, but these are overall figures and cannot be disaggregated by age group. However, this is the only data available at a local level. In general, surveyed unemployment is preferred to the claimant count measure, because the criteria for being able to claim change over time, affecting the measure.

We are also interested in the level of work available, not just the rate of employment, so we include indicators of the percentage of young people taking up professional, managerial or technical work. For this indicator we can again use the LFS.

Median earnings of young people are another indicator of the overall demand for labour, and are available from the ASHE at a local authority level. An important supplementary indicator is the ‘returns to education’. This is the earnings premium which graduates obtain in comparison with non-graduates. This can be estimated from the LFS, but only at a national or regional level. We hope to add this to future reports.

We should also note the complicating factor of geographical mobility. Young people who are unable to obtain work locally will tend to look further afield, while in some professional and higher administrative/managerial positions there will be national rather than local labour markets. If we wish to understand local labour market conditions, therefore, we need to distinguish ‘movers’ from ‘stayers’.

Improved indicators, particularly of vacancies would be a valuable step forward, but this component of the framework is relatively well supplied with indicators that between them allow for national, regional and local authority comparisons.

Driver 4: Social capital and connections

Aside from the economic aspects of the labour market, social factors can potentially be important too. A long tradition of research has suggested that social capital – social connections and the trusting relationships that derive from them – can provide the social infrastructure for a more dynamic economy and society.^{[footnote 74]} The theory behind this idea is that high levels of ‘generalised trust’ within a community reduces ‘transaction costs’, making it easier for people to do business with each other. High levels of social capital are a resource for the community as a whole. The American research on the drivers of upward mobility in different local areas found a strong causal impact of social capital. It also found that high rates of violent crime within an area reduced the likelihood of upward mobility.^{[footnote 75]} Some social milieus therefore appear to be more or less favourable to social mobility.

Selection of indicators

We regard this as a largely ‘experimental’ section of the measurement framework. The role of social capital and connections in promoting social mobility is less well-understood than the role of the labour market, at least in the UK. However, a long tradition of research has suggested that social capital can provide the social infrastructure for a more dynamic economy and society. The important American research on differences between areas of the USA convincingly demonstrated that the extent of social capital in an area has a causal impact on the area’s rate of upward income mobility, as does violent crime. While one should be cautious about generalising from an American context to the British one, these are potentially important topics for inclusion among the drivers. They do however need to be backed up by further research in the UK.

Social capital could also be relevant to fostering entrepreneurship, and has been emphasised in the literature on entrepreneurship within ethnic communities.^{[footnote 76]} Other social factors have also been suggested as important for innovation and entrepreneurship, notably having a high proportion of highly educated young people with the digital skills needed for contemporary start-ups.

Physical connectivity, in the form of transport links, has also been suggested as one source of variations in mobility chances between local areas. While highly plausible in the context of developing countries, where there are large rural-urban differences and the distances can be much larger, it is less clear what implications transport links have for social mobility in highly developed countries such as Britain. This could, however, be particularly important for parts of Scotland, Wales and Northern Ireland and needs to be kept under review, with rigorous research investigating the causal relationship between transport links and social mobility.

Social capital is a multidimensional concept. In its most recent review of social capital in the UK, the ONS distinguished 4 broad aspects of social capital – personal relationships, social network supports, civic engagement, and trust and cooperative norms – with several indicators for each aspect.^{[footnote 77]} However, there are relatively few regular, nationally-representative datasets which include these indicators, the main ones drawn on by the ONS being the The Department for Digital, Culture, Media and Sport (DCMS)’s Community Life Survey (which only covers England), the Economic and Social Research Council-funded UKHLS, and the biennial European Social Survey. The ESS contains questions which directly tap into the key ideas of the concept of social capital, namely social connections and generalised trust, and covers the whole of the UK. However, the ESS has rather too small a sample size to enable comparisons across the UK, and estimates of change over time are likely to have very large confidence intervals. We therefore use the UKHLS, which has included measures of civic engagement (the same key indicator as used in the American research) in the measurement framework. The UKHLS can be disaggregated to a regional level but not to a local level. The ONS report shows only one indicator that can be disaggregated to a local level, namely election turnout. For this year, we have relied on data from the OECD, including Eurostat’s European Union Statistics on Income and Living Conditions and the Gallup World Poll, to explore satisfaction with personal relationships and feelings of safety when walking home at night, respectively.

Summary of the drivers

The measurement framework focuses on 4 main drivers of social mobility covering conditions of childhood, opportunities and quality of the education provided, demand for young people’s labour and their opportunities in the labour market, and, experimentally, the role of social capital and connections. There is a substantial body of theory and of empirical research showing that these 4 drivers are likely to have causal impacts on rates of absolute and relative mobility.

We could also extend the scope of the drivers by including the ‘drivers of the drivers’, such as the causes of school effectiveness, or of youth unemployment. The Commission will be keeping this addition under review, and will also be seeking to strengthen the understanding of causal relationships.

For the empirical indicators, our aim is to have coverage of the UK as a whole (given the Commission’s statutory duty to monitor progress in the UK, not just in England), but also to be able to drill down to national, regional and local levels, and to do so on an annual basis. These are challenging requirements, although there are some administrative datasets which permit us to meet all these requirements in the cases of the conditions of childhood and of work opportunities for young people.

A particular problem, however, is with the educational drivers of social mobility. The devolved nature of education in the UK has meant that there is no harmonised set of administrative schooling data covering all 4 nations of the UK. To be sure, the LFSs do cover the whole of the UK, but sample sizes restrict the extent to which we can disaggregate to a local level.

More work is needed on the social capital driver, both with regard to data and to our understanding of the causal links with social mobility.

Future work on the index

The index provides a basis for monitoring both the prospects for social mobility in the future, and final social mobility outcomes. By including intermediate outcomes, such as social disparities in young people’s educational attainments and routes into work, we can detect where things are currently getting better or worse. By including analyses of overlapping characteristics, we can detect whether particular subgroups or areas of the country are particularly at risk of adverse outcomes. And by including the likely drivers of social mobility, we can see where there are potential problems that could cause adverse future outcomes.

Granularity of occupational classes

Our annual report for 2022 follows our previous work in using the ONS’s 3-part occupational class schema: professional and/or managerial, intermediate, and working class. We would like to improve on this in future work. The professional and/or managerial class hides great variation in income, while the working class currently excludes those who are long-term unemployed. A more granular breakdown of occupational classes would be of considerable analytical benefit.

Benchmarking the UK

For identifying potential problems we have to decide on benchmarks against which to make comparisons. A natural approach is to ask whether one particular country, region, or local authority is performing significantly worse than would be expected given the overall figures for the UK as a whole (or for England as a whole in the case of regional comparisons within England). International comparisons can also be used to establish benchmarks for identifying where the UK’s performance is significantly worse than would be expected given the outcomes in similar affluent democracies, such as members of the OECD. At a minimum, however, such benchmarking exercises require that the data be properly harmonised. We have drawn attention to topics, such as the UK’s education statistics, where harmonised data across the UK is not currently available.

We have also emphasised that monitoring and benchmarking exercises of the sort that the measurement framework permits cannot be the final word on whether there is a real problem or not. There can always, in theory, be perfectly innocent explanations for any discrepancy between the observed statistic and the benchmark. For example, a rapid increase in childhood poverty in one of the English regions could be a result of selective patterns of in- or out-migration – ; patterns of selective migration have regularly been found to be a major complicating factor when making area comparisons.^{[footnote 78]} This is why we talk of warning signs or flashing lights. The warning sign needs to be followed up by more detailed investigation to check the cause of the problem. Monitoring can indicate where there are concerns that need to be followed up but it cannot on its own explain the disparity from the benchmark.

Comparisons with a benchmark can also work in the opposite direction, indicating where one particular country or area is performing significantly better than might be expected.

Further work

As well as maintaining and developing the data resources needed to monitor drivers and outcomes, a priority is to carry out further in-depth work to understand the causal relations between drivers and outcomes. Establishing causation is notoriously difficult in the social sciences, but our review of the literature has found a few notable examples of research designs (taking advantage, for example, of ‘natural experiments’ such as raising the school-leaving age) that provide convincing tests of whether the observed relationships are causal or not. The ONS Longitudinal Study, which links data across the 10-yearly censuses, is likely to provide opportunities for research of this kind (especially when linked with other administrative data resources). This could be particularly valuable for understanding the drivers of area differences in opportunities and outcomes.

1. David Glass, ‘Social mobility in Britain’, Published on CAMBRIDGE.ORG; John Goldthorpe, ‘Social mobility and class structure in modern Britain’,1980. Published on SEMANTIC SCHOLAR.COM; Anthony Heath, ‘Social mobility’, 1981. Published on CAMBRIDGE.ORG. ↩

2. Social Mobility Commission, ‘The long shadow of deprivation: differences in opportunities’, 2020. Published on GOV.UK. ↩

3. The World Bank, ‘Fair progress? Economic mobility across generations around the world’, 2018. Published on WORLDBANK.ORG; The Organisation for Economic Co-operation and Development, ‘A broken social elevator. How to promote social mobility’, 2010. Published on OECD.ILIBRARY.ORG. ↩

4. Carys Roberts, Grace Blakely and Luke Murphy, ‘A wealth of difference: reforming taxation of wealth’, 2018. Published on IPPR.ORG. ↩

5. Abigail McKnight, ‘Estimates of the asset-effect: the search for a causal effect of assets on adult health and employment outcomes’, 2011. Published on STICERD.LSE.AC.UK; Abigail McKnight and Eleni Karagiannaki, ‘The wealth effect: how parental wealth and own asset-holdings predict future advantage’, in John Hills, Francesca Bastagli, Frank Cowell, Howard Glennerster, ‘Wealth in the UK’. Published on OXFORDUNIVERSITYPRESSSCHOLARSHIP.COM ↩

6. Wojciech Kopczuk, ‘What do we know about the evolution of top wealth shares in the United States?’, 2014. Published on NATIONAL BUREAU OF ECONOMIC RESEARCH.ORG. World Inequality Lab, ‘World inequality report’, 2018. Published on WIDWORLD; Bert Brys and others, ‘Tax design for inclusive economic growth’, 2016. Published on OECD.ILIBRARY.ORG. ↩

7. Carlotta Balestra and Richard Tonkin, ‘Inequalities in household wealth across OECD countries’, 2018. Published on OECD.ILIBRARY.ORG. ↩

8. Anthony Heath and others, ‘Social progress in Britain’, 2018. Published on OXFORDUNIVERSITYPRESS.COM. ↩

9. Robert Putnam, ‘Our kids: the American dream in crisis’, 2016. Published on ACADEMIC.OUP.COM. ↩

10. Jo Blanden and others, ‘Intergenerational persistence in income and social class: the impact of within-group inequality’, 2013. Published on ROYAL STATISTICAL SOCIETY ONLINELIBRARY.WILEY.COM; David Cox, Michelle Jackson and Shiwei Lu, ‘On square ordinal contingency tables: A comparison of social class and income mobility for the same individuals’, 2009. Published on ROYAL STATISTICAL SOCIETY ONLINELIBRARY.WILEY.COM; John Goldthorpe, ‘Understanding – and misunderstanding – social mobility in Britain: the entry of the economists, the confusion of politicians and the limits of educational policy’, 2013. Published on CAMBRIDGE.ORG. ↩

11. Erzsébet Bukodi and others, ‘Primary and secondary effects of social origins on educational attainment: new findings for England’, 2021. Published on ONLINELIBRARY.WILEY.COM. ↩

12. Martin Hällsten and Max Thaning, ‘Wealth as one of the ‘big 4’ SES dimensions in intergenerational transmission’, 2021. Published on ACADEMIC.OUP.COM. ↩

13. The World Bank, ‘Fair progress? Economic mobility across generations around the world’, 2018. Published on WORLDBANK.ORG; The Organisation for Economic Co-operation and Development, ‘A broken social elevator. How to promote social mobility’, 2010. Published on OECD.ILIBRARY.ORG. ↩

14. Jo Blanden and others, ‘Trends in intergenerational home ownership and wealth transmission’, 2021. Published on CENTRE FOR ECONOMIC PERFORMANCE LSE.AC.UK. ↩

15. Brian Bell and others, ‘Where is the land of Hope and Glory? The geography of intergenerational mobility in England and Wales’, 2019. Published on CENTRE FOR ECONOMIC PERFORMANCE LSE.AC.UK; Jo Blanden  and others,‘Trends in intergenerational home ownership and wealth transmission’, 2021. Published on CENTRE FOR ECONOMIC PERFORMANCE LSE.AC.UK. ↩

16. Martin Hällsten and Max Thaning, ‘Wealth as one of the ‘big 4’ SES dimensions in intergenerational transmission’, 2021. Published on ACADEMIC.OUP.COM; Ricky Kanabar and Paul Gregg, ‘Intergenerational wealth transmission in Great Britain’, 2021. Published on RESEARCHPORTAL. BATH.AC.UK; Alex Davenport and others, ‘Why do wealthy parents have wealthy children?’, 2021. Published on IFS.ORG.UK. ↩

17. Rowena Crawford and others, ‘Household wealth in Great Britain: distribution, composition and changes 2006-12’, 2016. Published on IFS.ORG.UK. ↩

18. Jo Blanden and others, ‘Educational inequality and intergenerational mobility,’ in Stephen Machin and Anna Vignoles (editors) ‘What’s the good of education? The economics of education in the UK’, 2005. Published on PRESS.PRINCETON.EDU. ↩

19. Richard Breen and Walter Müller, ‘Education and intergenerational social mobility in Europe and the United States’, 2020. Published on STANFORD.UNIVERSITYPRESSSCHOLARSHIP.COM. ↩

20. Brian Bell and others, ‘Where is the land of Hope and Glory? The geography of intergenerational mobility in England and Wales’, 2019. Published on CENTRE FOR ECONOMIC PERFORMANCE LSE.AC.UK; Raj Chetty and others, ‘Where is the land of opportunity? The geography of intergenerational mobility in the United States’, 2014. Published on NATIONAL BUREAU OF ECONOMIC RESEARCH.ORG. ↩

21. David Batty and others, ‘Accuracy of adults’ recall of childhood social class: findings from the Aberdeen Children of the 1950s  study’, 2005. Published on BMJ.COM; Nanna Lien and others,  ‘Adolescents’ proxy reports of parents’ socio-economic status: How valid are they?’, 2001. Published on BMJ.COM. ↩

22. Erzsébet Bukodi and others, ‘Intergenerational mobility in Europe: a new account’, 2019. Published on ACADEMIC.OUP.COM; Eurofound, ‘Social Mobility in the EU’, 2017. Published on EUROFOUND.EUROPA.EU. ↩

23. Richard Breen, ‘Social mobility in Europe’,  2004. Published on UNIVERSITYPRESSSCHOLARSHIP.COM. ↩

24. A previous British study using the BHPS is: John Ermisch and Lindsay Richards, ‘Trends in educational mobility in the UK’, 2016. Published on CSI.NUFF.OX.AC.UK. ↩

25. Jo Blanden and others, ‘Trends in intergenerational home ownership and wealth transmission’, 2021. Published on CENTRE FOR ECONOMIC PERFORMANCE LSE.AC.UK. ↩

26. As a second best economists have sometimes imputed income on the basis of recall data from surveys on parental occupation and education. See Social Mobility Commission, ‘Social mobility, the class pay gap and intergenerational worklessness: new insights from the Labour Force Survey’, 2017. Published on GOV.UK; John Jerrim and others, ‘2-sample, 2-stage least squares (TSTL) estimates of earnings mobility: how consistent are they?’, 2016. Published on SEMANTICSCHOLAR.ORG. ↩

27. Robert Manduca and others, ‘Trends in absolute income mobility in North America and Europe’, 2020. Published on IZA.ORG. ↩

28. The method is described by Raj Chetty and others in, ‘The fading American dream: trends in absolute income mobility since 1940’, 2017. Published on NATIONAL BUREAU OF ECONOMIC RESEARCH.ORG. Analogous methods have also been used by sociologists, see for example Andrea Tyree, ‘Mobility ratios and association in mobility tables’, 1973. Published on JSTOR.ORG. ↩

29. Paul Gregg and others, ‘Moving towards estimating sons’ lifetime intergenerational economic mobility’, 2017. Published on ONLINELIBRARY.WILEY.COM. This study presents a range of estimates in order to take account of various biases, such as attenuation bias, life cycle bias, and spells of worklessness. We report the rank correlations, rather than the elasticity, because rank correlations appear to be much less affected by these biases. The correlations reported above are from table 6. ↩

30. Bertha Rohenkohl, ‘Intergenerational income mobility in the UK: new evidence using the BHPS and Understanding Society’, 2019. Published on STUDOCO.COM. Rohenkohl was able to average incomes across several years, obtaining estimates of ‘permanent income’. However, her study looked at household income rather than individual income, and included both sons and daughters. ↩

31. Ricky Kanabar and Paul Gregg, ‘Intergenerational wealth transmission in Great Britain’, 2021. Published on RESEARCHPORTAL. BATH.AC.UK. ↩

32. Alice Sullivan and others, ‘Pathways from origins to destinations: stability and change in the roles of cognition, private schools and educational attainment’, 2020. Published on ONLINELIBRARY.WILEY.COM;  Matt Dickson and others, ‘Early, late or never? When does parental education impact child outcomes?’, 2016. Published on ONLINELIBRARY.WILEY.COM. ↩

33. Raj Chetty and Nathaniel Hendren, ‘The impacts of neighborhoods on intergenerational mobility II: county-level estimates’, 2018. Published on ACADEMIC.OUP.COM. ↩

34. Jack Britton and others, ‘How English domiciled graduate earnings vary with gender, institution attended, subject and socio-economic background’, 2016. Published on IFS.ORG.UK. ↩

35. Paul Gregg and others, ‘Moving towards estimating sons’ lifetime intergenerational economic mobility’, 2017. Published on ONLINELIBRARY.WILEY.COM. ↩

36. UK Government, ‘National Reference Test: information’, 2016. Published on GOV.UK. ↩

37. But note that an Education Policy Institute study based on the Millennium Cohort Study gives rather different results from PISA. See Luke Sibieta and Joshua Fullard, ‘The evolution of cognitive skills during childhood across the UK’, 2021. Published on EPI.ORG.UK. ↩

38. Department for Education, ‘Measuring disadvantaged pupils’ attainment gaps over time (updated)’, 2014. Published on GOV.UK. ↩

39. Pupils are defined as disadvantaged if they are known to have been eligible for free school meals 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. ↩

40. Social Mobility Commission, ‘State of the nation 2021: social mobility and the pandemic’, 2021. Published on GOV.UK. ↩

41. Department for Education, ‘Review of vocational education: the Wolf report’,2011. Published on GOV.UK; Eric Hanushek and Ludger Woessmann, ‘The knowledge capital of nations: education and the economics of growth’, 2015. Published on MITPRESS.MIT.EDU; Eric Hanushek and Ludger Woessmann, ‘The role of cognitive skills in economic development’, 2008. Published on AEAWEB.ORG. ↩

42. Recent critiques show that the POLAR area-based measure does not perform well as a measure of parental circumstances. See Vikki Boliver and others, ‘Using contextual data to widen access to higher education’, 2021. Published on TAYLOR AND FRANCIS.ONLINE.COM; John Jerrim, ‘Measuring socio-economic background using administrative data. What is the best proxy available?’, 2020. Published on UCL.AC.UK. ↩

43. Sam Friedman and Daniel Laurison, ‘The class ceiling: why it pays to be privileged’, 2020. Published on SAGEPUB.COM. ↩

44. Social Mobility Commission, ‘The long shadow of deprivation: differences in opportunities’, 2020. Published on GOV.UK. ↩

45. World Economic Forum, The global social mobility report 2020: equality, opportunity and a new economic imperative’. 2020. Published on WEFORUM.ORG. ↩

46. UK Government, ‘Understanding living costs while studying at university or college’, 2021. Published on GOV.UK. ↩

47. Sutton Trust, ‘Entry into grammar schools for disadvantaged pupils in England’, 2013. Published on SUTTONTRUST.COM. ↩

48. Department for Education, ‘House prices and schools: do houses close to the best-performing schools cost more?’, 2017. Published on GOV.UK. ↩

49. Alan Krueger, ‘The rise and consequences of inequality in the United States’, 2012. Published on OBAMAWHITEHOUSE.ARCHIVES.GOV. ↩

50. Kerris Cooper and Kitty Stewart, ‘Does household income affect children’s outcomes? A systematic review of the evidence’,2020. Published on EPRINTS.LSE.AC.UK. ↩

51. See the work of Raj Chetty and Nathaniel Hendren who were able to use an ingenious comparison of movers and stayers to distinguish selection effects from causal effects: Raj Chetty and Nathaniel Hendren, ‘The Impacts of neighborhoods on intergenerational mobility I: childhood exposure effects’, 2016, and ‘The impacts of neighborhoods on intergenerational mobility II: county-level estimates’, 2016. Published on NATIONAL BUREAU OF ECONOMIC RESEARCH.ORG. We should emphasise, however, that the differences in the social and economic environments of the US and UK mean that we cannot straightforwardly assume that US results will apply to the UK. ↩

52. Erzsébet Bukodi and others,‘Primary and secondary effects of social origins on educational attainment: new findings for England’, 2021. Published on ONLINELIBRARY.WILEY.COM; Martin Hällsten and Max Thaning, ‘Wealth as one of the ‘big 4’ SES dimensions in intergenerational transmission’, 2021. Published on ACADEMIC.OUP.COM. ↩

53. Social Mobility Commission,‘The childhood origins of social mobility: socio-economic inequalities and changing opportunities’, 2016. Published on GOV.UK. ↩

54. Richard Breen and John Goldthorpe, ‘Explaining educational differentials: towards a formal rational action theory’, 1997. Published on JOURNALS.SAGEPUB.COM. ↩

55. Richard Breen and Walter Müller, ‘Education and intergenerational social mobility in Europe and the United States’, 2020. Published on STANFORD.UNIVERSITYPRESSSCHOLARSHIP.COM; Raj Chetty and Nathaniel Hendren, ‘The impacts of neighborhoods on intergenerational mobility II: county-level estimates’, 2016. Published on NATIONAL BUREAU OF ECONOMIC RESEARCH.ORG. ↩

56. Amy Clair, ‘Housing: an under-explored influence on children’s well-being and becoming’, 2019. Published on LINK.SPRINGER.COM. ↩

57. On Britain, see Franz Buscha and Patrick Sturgis, ‘Increasing inter-generational social mobility: is educational expansion the answer?’, 2015. Published on ONLINELIBRARY.WILEY.COM. On Germany, see Bastian Betthӓuser, ‘Fostering equality of opportunity? Compulsory schooling reform and social mobility in Germany’, 2017. Published on ACADEMIC.OUP.COM. ↩

58. Jan Stuhler, ‘A review of intergenerational mobility and its drivers’, 2018. Published on JANSTUHLER.COM; Herman Van de Werfhorst, ‘Early tracking and social inequality in educational attainment: educational reforms in 21 European countries’, 2020. Published on JOURNALS.UCHICAGO.EDU; Social Mobility Commission, ‘The long shadow of deprivation: differences in opportunities’, 2020. Published on GOV.UK. ↩

59. David Robinson and Viola Salvestrini, ‘The Impact of interventions for widening access to higher education. A review of the evidence’, 2020. Published on EPI.ORG.UK. ↩

60. Richard Breen and Walter Müller, ‘Education and intergenerational social mobility in Europe and the United States’, 2020. Published on STANFORD.UNIVERSITYPRESSSCHOLARSHIP.COM. ↩

61. The Sutton Trust, ‘Home and away: social, ethnic and spatial inequalities in student mobility’, 2018. Published on SUTTONTRUST.COM. ↩

62. Sophie Von Stumm and others, ‘School quality rankings are weak predictors of students’ achievements and well-being’, 2020. Published on ACAMH.ONLINELIBRARY.WILEY.COM. ↩

63. Largely because of the unreliability of the measures for individual schools. However, this unreliability should be reduced if we aggregate up to local authorities. See Thomas Perry, ‘English value-added measures: examining the limitations of school performance measurement’, 2016. Published on BERA-JOURNALS.ONLINELIBRARY.WILEY.COM. ↩

64. QS Top Universities, ‘QS world University rankings’, 2022. Published on TOPUNIVERSITIES.COM. ↩

65. Social Mobility Commission, ‘Social mobility, the class pay gap and intergenerational worklessness: new insights from the Labour Force Survey’, 2017. Published on GOV.UK; Sam Friedman and Daniel Laurison, ‘The class ceiling: why it pays to be privileged’, 2020. Published on SAGEPUB.COM. ↩

66. Social Mobility Commission, ‘The long shadow of deprivation: differences in opportunities’, 2020. Published on GOV.UK. ↩

67. John Goldthorpe, ‘Social class mobility in modern Britain: changing structure, constant process’, 2016. Published on THEBRITISHACADEMY.AC.UK. ↩

68. Social Mobility Commission, ‘The long shadow of deprivation: differences in opportunities’, 2020. Published on GOV.UK. ↩

69. But note that classic treatments of returns to education suggested that increased returns would be associated with lower relative mobility. See for example, Gary Solon, ‘A model of intergenerational mobility variation over time and place’, in Miles Corak (editor), ‘Generational income mobility in North America and Europe’, 2009. Published on CAMBRIDGE.ORG. ↩

70. Michael Hout, ‘Status, autonomy, and training in occupational mobility’, 1984. Published on JOURNALS.UCHICAGO.EDU; John Goldthorpe and Michelle Jackson, ‘Education-based meritocracy: the barriers to its realisation’, 2008, in Annette Lareau and Dalton Conley (editors), ‘Social class: how does it work?’, 2008. Published on JSTOR.ORG. ↩

71. Sam Friedman and Daniel Laurison, ‘The class ceiling: why it pays to be privileged’, 2020. Published on SAGEPUB.COM. Compare the suggestion in The Long Shadow of Deprivation that what marks out areas with greater relative mobility is that they are ones where education plays a more important role in recruitment – in other words, areas which are more ‘meritocratic’. ↩

72. Wiji Arulampalam and others, ‘Unemployment scarring’, 2001. Published on OUP.COM; Paul Gregg and Emma Tominey, ‘The wage scar from male youth unemployment’, 2005. Published on SCIENCEDIRECT.COM. ↩

73. Office for National Statistics, ‘Vacancies and jobs in the UK: September 2021’, 2021. Published on ONS.GOV.UK. ↩

74. Robert Putnam and others, ‘Making democracy work: civic traditions in modern Italy’ 1994. Published on PRESS.PRINCETON.EDU. ↩

75. Raj Chetty and Nathaniel Hendren, ‘The impacts of neighborhoods on intergenerational mobility II: county-level estimates’, 2018. Published on ACADEMIC.OUP.COM. ↩

76. Monder Ram, ‘Enterprise support and minority ethnic firms’, 1998. Published on TAYLOR AND FRANCIS.ONLINE; Monder Ram and Trevor Jones, ‘Ethnic minority business in the UK: a review of research and policy developments’, 2008. Published on JOURNALS SAGEPUB.COM. ↩

77. Office for National Statistics, ‘Social capital in the UK: 2020’, 2020. Published on ONS.GOV.UK. ↩

78. See for example the discussion by Elias Einio and Henry Overman, ‘The (displacement) effects of spatially targeted enterprise initiatives’, 2016. Published on EPRINTS.LSE.AC.UK. ↩