Income, housing and private school access in Britain

This is an Accepted Manuscript of an article published by Taylor & Francis in Education Economics on 18 Jan 2021, available online: http://www.tandfonline.com/10.1080/09645292.2021.1874878 .

1         Introduction

This paper studies access to Britain’s private schools over recent decades. The relevance of this issue lies in its potential implications for educational inequality and, hence, social and economic mobility. In Britain, private schools[1] propel a narrow sector of the population through high-quality education towards successful careers in elite positions (Elliot Major and Machin, 2018).

In some countries, private schools have little or no resource advantage over state-run schools – an example is Germany (DESTATIS, 2016). Moreover, in many countries only small or zero effects of private schooling on student academic outcomes are reported (e.g. Elder and Jepsen, 2014, Pianta and Asari, 2018,  for the US; Nghiem et al., 2015 for primary schooling in Australia; Agasisti et al. (2016) for Italy). However, Britain is distinctive: the socio-economic divide between private and state school children is among the highest in the developed world (OECD, 2012, Table B2.3); and its private schools, funded by exceptionally high fees, have a very substantial resource advantage over state schools (Green and Kynaston, 2019). Significant effects on educational attainments at both primary and secondary levels have been found over the past 50 years (e.g. Halsey et al., 1980; Feinstein and Symons, 1999; Dearden et al., 2002; Sullivan and Heath, 2003; Malacova, 2007; Parsons et al., 2017; Henderson et al., 2020), consistent with the evidence of the effects of large resource gaps and peer effects on academic achievement (e.g., Sacerdote, 2011; Lavy et al., 2012; Antecol et al., 2016).

There are considerable labour market advantages associated with Britain’s private schooling (Macmillan et al., 2015; Crawford et al., 2016; Green et al., 2017; Belfield et al., 2018). Recruitment to the upper echelons of power in British business, politics, administration and media continues to be tied to private schools (Reeves et al., 2017; Sutton Trust and Social Mobility Commission, 2019). The labour market premium for private schooling increased for earlier generations growing up in the 1960s to 1980s (Green et al., 2012). Considering the persistent ability of private schools to access high-ranking universities (Jerrim et al., 2016) and that the higher education premium has grown more dispersed (Lindley and McIntosh, 2015; Naylor et al., 2016; Lindley and Machin, 2016), there is little reason to expect a break in the trend. Many private schools have become more internationalist in their orientation, both admitting more international students and channelling more British pupils into pathways towards a global upper echelon of universities; such a global outlook has the potential to open up worldwide job market opportunities with high earnings expectations (e.g. Weenink, 2008; Kenway and Fahey, 2014).

Even though only about 9 per cent of adults have ever been to a private school, the British private school sector is, therefore, a significant pathway through which some families obtain long-term advantages for their children. An appreciation of the, possibly changing, extent to which participation in Britain’s private education is unequal and exclusive to those with higher income and wealth should thus provide a highly relevant contribution to the understanding of educational and economic inequality. While high fees make it self-evident that Britain’s private schools are more accessible by high-income families, we know little about the extent to which participation is income-concentrated, or how that concentration might be changing, or the extent and distribution of bursary provision, or how the role of family income and wealth may affect access.

This paper contributes to the, as yet small literature on private school participation in several ways. First, we document the concentration of private school participation in family income and how it has changed since the 1990s. We find that attendance is highly income-concentrated, but not so much that there are not some in the low-income deciles attending private school. We also find that the concentration ratio is quite stable over time, though there is a modest drop in the last period analysed (2011-2017). Second, we also document that, for many, regional private school fees have become unaffordable out of family income; the change is starker for those on lower incomes. The presence of low-income families at private schools, despite the lack of affordability, point to either bursaries or family wealth. Third, therefore, we also document the distribution of bursaries. We find that the distribution of bursaries is overall progressive, their value linked to financial needs, though with no evidence of substantive change; yet most low-income private school families do not receive bursaries.

The low frequency of bursaries and declining affordability suggest an important role of longer-term financial family resources. In aggregate, the private wealth-income ratio in 21st century Britain is substantially higher than in the 1980s and 1990s despite the losses following the 2008 financial crisis (Atkison, 2018; Piketty and Zucman, 2014). Our fourth contribution, therefore, is to estimate the marginal effects of family income using a pseudo-panel with cells defined by the region of residence and by grouped income deciles. Our models find a substantive, potentially non-linear income effect. It is positive and substantial for those in the top income decile, and insignificant for those below.  Fifth, we estimate the marginal effects of projected housing wealth (a large fraction of total wealth) on private school access among home-owning families. We find a significant positive association of home value with private school access, which is, as the income effect, stronger for families in the top income decile.  

In Section 2 we describe the British system of private schooling and consider potential expectations about trends in the income concentration of participation, the role of family resources in private school choice and the distribution of fee-assistance through bursaries and scholarships. Sections 3 describes our empirical model and the primary data source, the Family Resources Survey. Our findings are presented in Sections 4 and 5. In our conclusion, Section 6, we note the implications of our findings for the schools’ policies on bursaries and for the government’s policies on social mobility.

2         Access to Private Schooling in Britain

2.1        The British private school sector

One in eleven schools in Britain is private. Notwithstanding some heterogeneity, private schools are generally considered elite institutions, reflecting both the social composition of their pupil intake, their resources and the destinations of their alumni (Maxwell and Aggleton, 2015; Reeves et al., 2017). On average, the fee income from families with children in private day schools exceeds the income from the state provided to maintained schools by a factor of around 2.5 or more, depending on education stage; when supplemented by private schools’ endowments of physical and financial wealth, the per-pupil resource gap becomes at least three (Green and Kynaston, 2019). 

With these resources, British private schools can specialise in small class sizes, with a pupil-teacher ratio, at 8.5, less than half that of state-maintained schools in 2016 (ISC, 2019, p. 25), and offer a broader education including more sports and a much wider range of extra-curricular activities. Facilities are typically generous and well-equipped, although there is some heterogeneity across private schools in respect of resources, as well as, for example, in their size, pedagogical approaches, boarding facilities or religious denomination. This private-state resource gap is typically much higher than in many other countries, though data are scarce. In Germany, for example, where schools have a legal obligation to accommodate families regardless of income, there is a financial resource gap of
-5.3 per cent at upper-secondary level (DESTATIS, 2016). In Australia, the resource advantage is 30 per cent for non-sectarian private schools and -5 per cent for private religious schools (ACARA, 2019), though there is considerable fee variation linked to their pupils’ social background (Lye and Hirschberg, 2017). The United States has in recent years become another exception where non-sectarian private schools, though relatively small in number, deploy a teacher-pupil ratio twice that in public schools, while the Catholic private schools have a small resource advantage (Broughman et al., 2019).

Private schools are managed and governed autonomously, though three-quarters of them are registered as charities and must thus comply with charity law; doing so qualifies them for a small amount of public subsidy through tax relief. Though they are not required to follow the National Curriculum, their pupils usually enter the same national exams that state secondary school pupils take, these being the gateway to university and other post-secondary education. Since the start of the 1980s, private schools’ resources have transformed. With the cessation of the Assisted Places Scheme that ran from 1981 till 1997 (Whitty et al., 1998), neither schools nor potential pupils receive direct public subsidies except for small government programmes targeted at vulnerable children in foster care or those qualifying for the UK’s Music & Dance Scheme in specialist schools.

There are notable regional variations in the availability of private schools. In England’s North East, for example, there were 1.2 private schools per 10,000 pupils while in the South East and Greater London the figures were 3.6 and 3.8 private schools, respectively (DfE, 2018). In Wales, there were 1.5 private schools per 10,000 pupils in 2017/2018 (Welsh Government, 2018) and in Scotland, the figure stood at 1.4 private schools per 10,000 pupils (Scottish Government, 2018). Although boarding is an option for the very affluent, region-specific differences in the provision of private education will likely correlate participation.

Since the 1980s, the average participation rate in private education has remained stable at around 7 per cent of all school pupils in England (DfE, 2018). Yet private schools now operate in a different environment from the one they faced in 1980. The external economic and social context has changed substantially through wider access to higher education, rising ‘value’ of education credentials, increased income and wealth inequality, and the emergence of a global elite in a globally connected world. Approximately 5 percent of the pupils in Britain’s private schools have non-British parents living abroad (Ryan and Sibieta, 2011).

2.2        Inequality of access and the determinants of private school choice

Broader economic developments are likely to have affected the extent and distribution of private school participation in contrasting ways since the 1990s: the pull of the (aforementioned) potentially increasing benefits; the increasing costs; changes in the provision of means-tested bursaries; changes in income and wealth inequality and in the relationship between family income and wealth – especially given increases in housing wealth and potential wealth transfers within the extended family.

The costs of private schooling have risen considerably (Ryan and Sibieta, 2011; Parsons et al., 2017). By 2018 the average annual fee before any extras had reached £14,280 for day school and £33,684 for boarding school: in real terms about 60 per cent above the figures for 2000, and three times the 1980 fees. Meanwhile, while income inequality has fallen a little since its peak in 2001[2], the rising wealth-income ratio may have enabled some lower-income families to gain better access to private school than before. Fee reductions for some students are also potentially important. Across a range of countries, access to means-tested student aid has influenced the distribution of college enrolment depending on the degree to which credit constraints bind (Nielsen et al., 2010; Fack and Grenet, 2015). In Britain, the Assisted Places Scheme (Whitty et al., 1998) – through which the state had funded, at its height in 1998, 8.5 per cent of places at private schools – was ended in 1997 and fully phased out by 2004. Spurred by the requirements of the Charities Act 2006, provision of means-tested bursaries expanded. The Independent Schools Council (ISC), the umbrella association of UK private school groups covering about 80 per cent of private school pupils, reported in 2019 that around 28 per cent of pupils received some financial assistance from the schools, up from 19 per cent in 1998. Fee discounts, typically small reductions for siblings or for teachers’ children, were the most common form of financial support (13.2 per cent of students), followed by scholarships (10.8 per cent) and means-tested bursaries (7.8 per cent) (ISC, 2019).  Only one per cent of pupils attend for free. Schools provide discounts for a range of reasons, including to influence the school’s academic standing and associated peer effects, price discrimination to increase revenue, and a measure of corporate social responsibility (Wilde et al., 2016). While scholarships are allocated based on academic or cultural/sporting abilities, the award of means-tested bursaries often also entails academic or other ability selection. In total, the bursaries amount to about four per cent of private school turnover. 

With these developments pulling in different directions, it is not possible to state whether access to private school would have widened or narrowed over time. Either way, the developments also point to the need for a good understanding of the effects of income and wealth on private school choice. Two earlier studies of private school choice in Britain have found that current income enters non-linearly; that other socio-economic background indicators are also important; and that demand is sensitive to price with an estimated elasticity of -0.26 for private school choice at age 7 (Blundell et al., 2010; Dearden et al., 2011). Complementing these, sociological research has found a mix of motivations and some ambivalence among parents choosing private schooling; emphasising roles both for traditional attitudes and allegiances and for an instrumental approach to choosing private schooling as a way to gain access to high-ranking universities and rewarding careers (Fox, 1985; Ball, 1997; West et al., 1998; Foskett and Hemsley-Brown, 2003). The socially exclusive peer group is seen as a positive attraction for some families and a negative factor for others. Anders et al (2020) demonstrate the importance of family values, while also showing the influence of location and of permanent income.

Educational choices are likely to be taken for a long horizon, given the costs of switching between sectors. In Britain, school sector choices generally happen at key ages: seven, eleven (sometimes 13) and 16. Given the magnitude of the investments, the choice of school sector will be affected by long-term financial family resources (Chevalier et al., 2013; Carneiro and Ginja, 2016). Wealth gains may influence school choice either through life-cycle wealth effects or by relaxing credit constraints through additional collateral (Cooper, 2013). In general, there is a positive association of household net worth with total educational achievement in Britain (McKnight and Karagiannaki, 2013).  About two-thirds of total private wealth (except pensions) is held in property (ONS, 2018). Home value gains may thus have the potential to influence household decisions, including around educational choice. In general, housing wealth has been found to correlate with consumption, individual health, fertility, and labour supply (Dettling and Kearney, 2014; Fichera and Gathergood, 2016; Burrows, 2018; Disney and Gathergood, 2018). Lovenheim and Reynolds (2013) find that housing wealth gains change students’ chances to access flagship universities in the US. Although with notable regional heterogeneity, residential real estate prices have more than doubled in real terms since 2000 across England (HM Land Registry, 2018b). At the same time, equity withdrawal expanded, particularly over the early 2000s (Reinold, 2011). It, therefore, appears appropriate to consider the role of family wealth either in property or financial assets in addition to income to study private school access.

In light of the above, our empirical description and analysis address these questions: How has the family income concentration of private school attendance been changing in Britain? Who receives bursaries: how large are they and are they progressive in their distribution? What are the effects of family income? What is the correlation of housing wealth with private school access?

3         Data and Variables

3.1        Data

Our primary data comes from the Family Resources Survey (FRS) and the Households Below Average Income (HBAI) programme. The FRS is a continuous series of annual cross-sectional random probability surveys of private households in the UK. Since its launch in 1994, it has collected information on household composition, economic activity, income and broader financial resources from all adults in around 25,000 households up until 2010 and a sample of around 20,000 respondent households per year after that (DWP, 2018a, 2018b). We consider data from survey waves since 1997 for families in Great Britain (excluding Northern Ireland) with dependent children aged 5-15 years (the years of compulsory primary and secondary schooling). We further restrict the sample to families with family heads in the age bracket 25-64 years to avoid confounding from transitions into and out of the labour market.

As part of the household grid, FRS queries survey participants “What type of school or college does [name] attend?” From the responses, we derive private school attendance as a dummy that distinguishes children in private schools (defined as any school where at least some pupils pay fees) from children in state-maintained schools. Because the focus is on mainstream primary and secondary schools, we remove observations on special schools and on those few who have moved on to non-advanced further education.[3] In all, the analytical sample includes information about 154,529 children from 96,915 families.

3.2        Measuring income and home value

FRS is the primary data source for income conditions of British households. FRS includes detailed questions about families’ current income from all sources.  As part of the HBAI programme, the Department for Work and Pensions (DWP) re-weights and imputes very high incomes to compensate for under-representation and under-reporting in the raw data at the top of the income distributions. Although not without issues, the procedure goes some way towards reconciling the survey data with data from income tax records (Burkhauser et al., 2018a, 2018b). All monetary variables are in 2017/18 prices. Under the assumption that all family members benefit equally from total disposable income, we equivalise total net income by dividing it by the square root of family size, i.e. the combined number of adults and dependent children in the family.

In addition to current family income, FRS asks a set of questions about households’ housing tenure, the number of bedrooms in the accommodation, the type of accommodation (detached, semi-detached, terraced, flat, others), the council tax band for owner-occupiers, and, for mortgaged home-owners (86% of all owner-occupiers in the survey), the purchase price and purchase year. Early waves of FRS up until 1995/96 included a question for mortgaged home-owner to estimate the price their property would fetch if they were to sell. Data quality concern let to the removal of the survey question from wave 1997/98 onward. Thus, to examine the role of housing wealth for private school access, we impute a plausible current home value for all owner-occupier families from the information in the survey. We estimate a hedonic Poisson regression model of the initial purchase price on the number of bedrooms, type of accommodation, council tax band, and purchase year in the sample of mortgagors separately for each government region over the survey waves 1997/98-2017/18. The estimated coefficients are used to predict the current property value for each owner-occupier household in FRS from 1997 onward.  Averaged by survey year, government region, and property type, there is a very strong correlation (r=0.98) of the projected home value with average property prices published by the HM Land Registry (HM Land Registry, 2018a). For the subsequent analysis, we deflate the projected home value to 2017/18 prices.

Furthermore, FRS also includes information on financial awards (scholarship, bursary, grant or similar award) for education for children in secondary education or above. From this information, we construct an indicator variable whether a child in lower secondary education receives any bursary/ scholarship and if so, the total annual value deflated to 2017/18 prices.

Besides these variables, we also use information about the region of residence, the number of adults in the family, the age of the family head, housing tenure, the socio-economic class of the parents, and the parents’ age when leaving full-time education.

4          Empirical Model

To test the relationship of financial family resources with private school access, we draw on a set of common assumptions which imply a structural relationship between both (e.g. Becker and Tomes, 1994; Acemoglu and Pischke, 2001; Solon, 2004). Parents care about their children’s future life chances and decide whether to enrol their children at a private school. The decision is assumed to be non-separable over time, reflecting typical school choices at the beginning of primary and secondary school, respectively. Because credit markets are imperfect, parents cannot borrow against their children’s future human capital returns. Since there is no public loan scheme that could help to secure credit, private school access will depend on families’ private access to financial resources through income and capital markets. For wealthy, high-income families who are not credit constrained in their decision to send their children to private school, additional financial resources may not affect the private education decision; except if private education has characteristics of a consumption good. However, for families who are credit constrained a rise in lifetime income or wealth is predicted to increase the likelihood of investing in private education. In addition, we assume that there is a third group of income-poor parents who face an affordability constraint, namely that lifetime consumption is constrained to be above a certain minimum. For all in this group, the decision is the corner solution of non-participation, even in response to marginal increases in income or wealth. Additional wealth can raise private school access by relaxing credit constraints through a collateral effect or by easing the lifetime budget constraints.

Our econometric model builds on these assumptions:

(1)

Demand for private school education for children of family i, , is a function of the natural log of permanent family resources,  (income or home value), a set of control variables with an influence on financial family resources and private school access, region-by-year effects, , and a family-specific taste shifter for private education, , which can include expected returns from private school education. The region-by-year effects control for common factors that influence all families within regions and years. The effect of financial family resources on private school access is given by . The term  is unobserved and possibly correlated with financial family resources  and school choice. Thus, a cross-sectional analysis will potentially lead to biased estimates.

Moreover, while private school choice will depend on longer-term resources such as permanent income, survey data measures current income. The gap between measured current and family permanent income can be either transitory shocks to family income or measurement error. If ignored, this gap will attenuate the estimated income effect.

To solve both issues, we construct cohorts that we track through repeated cross-sections to estimate a fixed-effects model (Deaton, 1985). By aggregating the family decision to the cohort level, we derive the following model:

(2)

where is the proportion of privately educated children for all  children in family cohort  at time . Similarly,   are the average permanent family resources of cohort  at time , respectively. Because region and period effects are assumed to be the same for all cohorts who live in region  at time ,  remains unchanged.

Since the families who form the cohorts in the sample will differ for each year,  is time-variant. Following the literature around pseudo-panels, we can treat  as an unknown fixed parameter with  if sample sizes for each cohort are sufficiently large (e.g. Deaton 1985; Verbeek 2008). Similarly, with sufficiently large numbers of observations within cohorts, the measurement error in family resources approaches zero if there is no time-variant cohort component to measurement errors (Antman and McKenzie, 2007). For income,  is, therefore, closer to the concept of permanent income,  will pick up common trends that affect all families within a region.

Families are grouped into mutually exclusive cohorts, c, based on family income decile and region of residence. In pseudo-panel analyses, there is a trade-off between cohort size and the number of observations in the panel. We distinguish between families with incomes in the tenth decile, in the seventh to the ninth decile,  in the fourth to the sixth decile, and families with incomes at or below the third decile across eight British regions (North East and Yorkshire, North-West, the Midlands, East of England, Greater London, South East, South West, and the combined devolved nations of Scotland and Wales). Income deciles were derived in the sample of all non-pensioner households in Great Britain separately for each survey wave. The derived 32 cohorts are tracked over the 21 cross-sections of FRS 1997-2017. Table A1 in the online appendix gives mean cohort sizes. 

We arrive at the following baseline estimation model:

(3)

The model identifies family resource effects, for income and for home value, on school access through variations in  conditional on income rank, region and region-period effects. The family income effect, , is identified from differential income growth over the income distribution within regions. Likewise, the home value effect  is identified from the differential growth of home values across family income ranks within regions.  As in Acemoglu and Pischke (2001), families’ rank in the income distribution is thought to control for their unobservable characteristics. Because private school fees are high and credit markets are imperfect, we will also test if  or differ across grouped income deciles.

Like with any other method, there are some caveats. First, pseudo panels can be sensitive to the definition of the constituting cohorts. Thus, we also look at cohorts-based on region and socio-economic class instead of income rank. Second, the within estimator can be imprecise without enough time variation in the dependent and independent variables. Although there is within variation in cell-average income, home values, and private school participation, it is conceivable that with finite samples, some cohorts contribute little information to the parameter estimates. To assess the importance of unobserved heterogeneity and sources of variation,  we also estimate more restrictive models with   and pooled period effects Third, while the within estimator conditions out time-invariant unobserved characteristics, estimates will be biased in the presence of omitted time-variant variables that affect family resources and private school demand. For example, the analysis of home values on private school participation is limited to the self-selected group of homeowners. If the percentage of home-owning families with children remained the same within cohorts, the fixed effects would condition out the unknown selection mechanism. However, homeownership has started to slip after 2007 from 68 per cent to 58 per cent, which suggests possible changes in group selection as well.  Ideally, we would instrument for self-selection. In the absence of a convincingly exogenous time-varying instrument, the set of time-variant control variables,   , can help to assess the magnitude and direction of the bias. But even if the parameter estimate for    is biased upward, it can still document how changing access to permanent private financial resources may facilitate private school access.

5         Documenting the Income Concentration of Private School Participation and Bursary Provision.

The main aim of this section is to document, for the first time, the extent to which private school participation is family income concentrated, and whether this concentration is changing. In addition, we describe the extent of bursary provision and its correlates; if large enough and income-progressive, bursaries might be expected to modify the link between family resources and private school participation.

5.1        Income Concentration

Figure 1 plots participation in private schooling after 1997 against the percentile rank of net equivalised weekly family income for families with children in the 5-15 age range. It shows the extent to which participation is especially skewed at the very top of the income distribution. At the 100th percentile, half of the children go to private school. At the 95th percentile, however, this proportion is much lower, with only 15 per cent of children in the private sector. While still much greater than the average, it is striking that only a minority of the affluent families in the top 5 per cent are paying for private education.[4]

Figure 1 also shows that there is a non-zero – albeit very low – private-school participation among low- or middle-income families, for whom full fees would be difficult or impossible to cover out of current income alone.

A key question for policy discourse is whether access to private school has been widening. Table 1 summarises the income concentration of children in the private sector over the periods 1997-2003, 2004-2010 and 2011-2017. The top panel reports the participation rate in private education across the income distribution. Consistent with Figure 1, around a quarter of five children in the top-income decile went to a private school. This proportion and the overall participation rates have remained very stable since 1997. Only the bottom third of the income distribution increased their participation slightly from 1.2 per cent to 2.1 per cent over the study period. Despite this improvement, children from top-decile families remained around twelve times more likely to access private education than those with family incomes below the 4th decile in 2017.

Figure 1: Income concentration of private school participation, 1997-2017

Note: Participation of 5-15-year-olds in private schools in Great Britain by equivalised net family income percentile rank across all non-pensioner households. Equivalised using the square root of family size. The chart is restricted to families with income at or above the 5th percentile.

Target: Families with 5-15-year-old children in Great Britain. Source: HBAI, 2018. FRS 1997/98-2017/18.

The bottom panel gives the income concentration index of private school attendance. This index encapsulates the inequality of private school uptake over the whole distribution.[5] Index values can range from -1 (all private school concentrated among the poorest) to +1 (all concentrated among the richest). A value of 0 would indicate the absence of income inequalities in private school choice. The trend in the index summarises the descriptive findings in the rest of the table. The estimates confirm the substantial pro-rich income concentration of private school participation. This income concentration remained remarkably stable until the early 2010s then shows a statistically significant drop in the latest period.

Table 1: Private school participation across the net income distribution of families with children at private school and concentration index, 1997-2017

 1997-20171997-20032004-20102011-2017
 Income Rank (decile)
10th0.264
(0.007)
0.268
(0.011)
0.265
(0.011)
0.260
(0.013)
7th – 9th0.055
(0.002)
0.059
(0.003)
0.056
(0.003)
0.051
(0.003)
4th – 6th 0.017
(0.001)
0.018
(0.001)
0.018
(0.001)
0.017
(0.002)
<4th0.015
(0.001)
0.012
(0.001)
0.014
(0.001)
0.021
(0.002)
 Concentration Index
 0.528
(0.007)
0.560
(0.011)
0.542
(0.012)
0.486
(0.013)

Weighted concentration index of private school attendance for families in Britain with children aged 5-15 years between 1994 and 2017.  The proportion of children in private schools by net equivalised family income rank across all non-pensioner households in Great Britain. Standard errors in parentheses.

Table 2: Family financial resources and the relative costs of private education relative, 1997-2017

  Income RankTotal
<4th4th – 6th7th – 9th10th
Equivalised Weekly Net Income (in £)1997142
(0.8)
260
(0.9)
419
(2.0)
1007
(42.7)
293
(3.7)
2017194
(1.9)
321
(1.2)
515
(3.3)
1279
(93.5)
405
(10.6)
Growth rate (%)1.83
(0.1)  
1.19
(0.0)  
1.14
(0.0)  
1.35
(0.5)  
1.92
(0.2)
Projected home value (in £1,000)199794.9
(2.2)
102.9
(1.7)
142.4
(2.4)
244.1
(9.7)
124.3
(1.5)
2017242.4
(9.4)
250.4
(7.2)
338.7
(8.4)
582.1
(28.9)
321.6
(6.4)
Growth rate (%)2.55
(0.1)
7.17
(0.4)
6.89
(0.4)
6.92
(0.8)
7.94
(0.3)
The ratio of annual average regional day school fees to mean equivalised net family income (in %)199796.3
(0.60)
53.2
(0.23)
33.4
(0.19)
14.2
(0.60)
47.3
(0.59)
2017130.0
(1.74)
79.4
(0.48)
50.5
(0.42)
21.3
(1.58)
63.4
(1.66)
Δ33.6
(1.84)
26.3
(0.53)
17.1
(0.46)
7.1
(1.69)
16.1
(1.76)
The ratio of annual average regional day school fees to mean projected home value wealth (in %)19977.5
(0.17)
7.0
(0.11)
5.1
(0.08)
3.0
(0.11)
5.8
(0.07)
20175.3
(0.18)
5.2
(0.13)
4.0
(0.09)
2.4
(0.11)
4.1
(0.07)
Δ-2.2
(0.25)
-1.8
(0.17)
-1.1
(0.12)
-0.6
(0.16)
-1.7
(0.10)

Mean equivalised weekly net family income in 2017/2018 prices. Mean projected home value based on the accommodation type, the number of bedrooms, and region and survey year. Home values for owner-occupiers. Regional average day school fees from annual ICS census 1997/2017.  Standard errors in parentheses.

The comparative stability in both the rate and the distribution of private school participation contrasts with trends in the ‘affordability’ of private education out of family income. The first panel of Table 2 shows the rises of incomes between 1997 and 2017, while the second panel shows the rises in home value, at each income decile. Overall, the growth rate was about two per cent per year. Within income groups, family income rose fastest at the bottom and the top of the income distribution. Home value increased substantially more than family incomes. Average home value expanded 2.6-fold from £124,300 in 1997 to £321,600 in 2017, an increase of almost eight per cent per year. Top-income families saw the largest absolute rise in projected home value, but the relative gains were fairly even across the top two-thirds of the family income distribution. The smaller figure for lower-income families could indicate some degree of spatial segregation. Nonetheless, the average home value among low-income owner-occupiers in 2017 approached figure for top-income families two decades earlier. Given this strong growth, housing wealth gains may have contributed to family resources for private education among homeowners.

The third panel then shows the substantively declining affordability of school fees relative to net family income: from 14 per cent to 21 per cent at the top decile, and from 96 per cent to 130 per cent for those in the bottom third. Indeed, by 2017, only those in the top decile could send children to private school without spending more than half their income per equivalised family member. The last panel shows, however, that affordability in relation to home value, for owner-occupiers, improved significantly.

If housing wealth were associated with private school access, we would expect lower-income families with privately educated children to own relatively more valuable property. Table 3 breaks down the ratio of projected home value over annual equivalised net family income by income decile for children in the private sector and in state schools for owner-occupiers. As expected, the figures show a substantially higher housing wealth-to-income ratio for children in the private school sector. The private-state gap is especially large for families at the bottom of the income distribution; which further indicates that families with children at private schools may be able to draw on other resources than income.

Table 3: Ratio of family housing wealth to annual equivalised net income by children’s school type in Britain, 1997-2017

  Income RankTotal
<4th4th – 6th7th – 9th10th
Home value to income ratioState20.1
(0.17)
12.0
(0.06)
10.5
(0.05)
7.2
(0.13)
11.2
(0.06)
Private59.4
(3.56)
24.1
(0.93)
16.6
(0.41)
7.8
(0.26)
10.7
(0.29)
Δ39.3
(3.56)
12.1
(0.93)
6.0
(0.41)
0.6
(0.28)
-0.5
(0.30)

The ratio of mean projected family home value wealth to mean annual equivalised net family income across 96,812 children in owner-occupier families in Britain in FRS, 1997-2017. Robust Standard errors in parentheses

5.2        Bursaries and scholarships

Bursaries or scholarships may have offset some of the fee rises. About 4 per cent of school turnover is spent overall on bursaries, while 1 per cent of pupils receive full bursaries and go free (e.g., ISC, 2018, 2019). We know, however, little about the distribution of financial support among private school families. Using the FRS data, Table 4 shows the proportion of pupils in lower secondary education at private schools who receive bursaries or scholarships and the average value of financial support, split by family income deciles. Throughout the period, around 15 out of every 100 pupils received direct financial support. Significantly, for all those below the top decile, a large majority – four out of five children – receive no grants or bursaries. Evidently, this source of funds does not account for private school participation in this group. The value of financial support was around £4,900 in 2011-2017, little changed from earlier periods, and thus a smaller fraction of the fees.

Table 4: Frequencies and value of bursaries/scholarships by income rank for the periods 1997/2003, 2004/2010 and 2011/2017

 Income rank1997-20032004-20102011-2017Δ
% in private secondary schools in receipt of grants/ bursaries10th 9.4
(1.6)
13.2
(1.9)
10.0
(2.0)
0.6
(2.8)
7th – 9th13.2 (1.9)14.8 (2.1)18.3 (2.7)5.1 (3.3)
≤6th22.9
(2.5)
17.0
(2.5)
20.1
(2.7)
-2.8
(3.7)
Total14.8
(1.1)
14.8
(1.2)
15.3
(1.5)
0.5
(1.9)
The average annual value of grants/ bursaries per supported student (in £ at 2017/2018 prices)10th 4,091
(822.1)
3,822
(572.5)
2,992
(587.0)
-1099
(1010.1)
7th – 9th4,512
(593.2)
4,479
(623.7)
5,095
(724.5)
584
(936.3)
≤6th5,470
(595.7)
5,887
(815.0)
6,240
(928.2)
770
(1102.9)
Total4,863
(378.6)
4,687
(397.1)
4,935
(489.6)
71
(618.9)

Note: Base population comprises students aged 11-15 years; standard errors in parentheses.

Both the allocation and the value of bursaries are, as expected, progressive with respect to parents’ incomes. Nonetheless, even at the 10th decile, around ten per cent of students received financial aid. It is conceivable that these are scholarships, that eligibility is not only tested against income, or that schools are not fully informed about parent’s financial circumstances. Relatively large standard errors make it difficult to draw confident conclusions about the changing degree of progressiveness. Nevertheless, there is a suggestion of a decline in the allocation of bursaries to low-income families in the middle period, which follows the cessation of the Assisted Places Scheme, and recovery thereafter. 

To analyse who received bursaries or grants we ran separate multivariate analyses (see Table A2 in the online appendix). These show that economic needs matter more for the amount awarded than for the likelihood of receiving an award. Higher levels of material deprivation and a family head who is out of work are individually and jointly associated with the annual value of bursaries. Consistent with the descriptive patterns, these associations have not changed significantly over time. Overall, the data could not support claims that the private school sector has widened access for students from low-income families through more generous financial support.

6         Analysis of Income and Wealth Effects

Given that the affordability of fees out of income is in question for all but the top- income decile, and that bursaries distribution and values are far too small a scale to explain much of the participation, we now turn to analyse the effects of cell-wide average family resources in the pseudo-panel described above. 

6.1        Family income effects

First, we consider family income effects in the whole sample; Table 5 reports our main findings. In column (1), we condition only on a set of period dummies. The coefficient of the family income effect is 0.122.  However, this estimate does not account for family background effects related to income rank and regional differences. In column (2), we thus add panel fixed effects (income rank-period dummies). The estimated income effect drops slightly to 0.118, which indicates that private access may be more about the money rather than relative income rank. 

Column (3) give the headline estimate based on equation (3). Despite panel fixed effects and the controls for region-period effects, the coefficient of log mean family income is substantial at 0.116. In other words, a 10 per cent increase in family income is associated with 1.16 point rise of private school participation. Out of the 24-point participation gap between children from the top decile and the bottom third of the income distribution, income can roughly explain 22 points (0.116×LN(1,279/194)).

Our conceptual framework suggests that income effects may be heterogeneous across the income distribution. Therefore, column (4) estimates income effects for each income bracket. According to the estimates, marginal income effects on private school participation are largest for the most affluent families in the 10th top decile, where the estimate is 0.118. For families below the top income decile, increases in cell-average income might not raise private school access. The coefficients are near zero and statistically insignificant, albeit with large standard errors. Although an F-test of equal income effects across income-deciles does not reject the null hypothesis (F(3,31)=1.88, p=0.153), there is overall no evidence of a very affluent group that are unaffected by income – a finding similar to that of Acemoglu and Pischke (2001) in respect to college attendance in the US. Adding a set of time-variant control variables to the estimation model in columns 5 and 6 increases the estimated effect of income slightly and leads to a rejection of a homogenous effect across the income distribution, but does not alter any of the substantive conclusions.

Further robustness tests with a pseudo-panel based on socio-economic class instead of income deciles lead to similar conclusions: Income affects private school access, and the point estimates are larger for families at the top of the socio-economic hierarchy. Table A3 in the appendix summarises the key findings.

Table 5: Income effects on the probability of attending private school from pseudo-panel fixed effects regressions, 1997-2017

 (1)(2)(3)(4)(5)(6)
Ln mean family income0.122***
(0.016)
0.118**
(0.038)
0.116**
(0.033)


0.121***
(0.032)


Ln mean family income #      
(10th income decile)





0.118**
(0.034)


0.123***
(0.033)
(8th – 9th income decile)





0.026
(0.062)


-0.007
(0.070)
(6th – 7th income decile)





-0.006
(0.050)


-0.032
(0.054)
 (≤5th income decile)





0.040
(0.036)


0.034
(0.044)
FENOYESYESYESYESYES
PeriodYESYESYESYESYESYES
Region # PeriodNONOYESYESYESYES
Time-variant controlsNONONONOYESYES
R20.6400.8250.8630.8640.8690.870
F(constant income effect) – p-value   0.153 0.029

Estimates from a pseudo-panel of 672 cells of eight aggregate regions, four grouped income decile of pupils’ family income, and 21 survey years. The dependent variable is the proportion of students aged 5-15 years in private schools within cells derived from the FRS, 1997/1998-2017/2018. The underlying population are all children aged 5-15 years in Great Britain. Columns (5) and (6) control for the with cell proportion of family heads under 35 years, the proportion over 55 years, the share of single parents and the proportion of families where at least one parent left full-time education at the age of 21 or higher. Heteroscedasticity and autocorrelation robust standard errors in parentheses. Clusters at region-income ranks level. * p < 0.05, ** p < 0.01, *** p < 0.001.

6.2        Housing wealth

Given that our descriptive analysis earlier pointed at a potential role of housing wealth for private school access,   we next assess how changes in housing wealth, through home price changes, correlate with private school participation. This is done for owner-occupier families, for whom we imputed current home value.

Table 6 presents coefficients for the relationship of imputed home value with private school attendance over the period 1997-2017. Column (3) has the key finding from an estimation of equation (3): a 10 per cent appreciation of projected home value for owner-occupier families raised private school participation by 1.84 points. Compared with column (1), the addition of panel fixed effects and region-period effects did not reduce the estimated home housing wealth.

Columns (4) estimates separate home value effects by family income band. As before, income deciles are based on the family income distribution across all non-pensioner households in Great Britain in each wave. The housing wealth effect is again highest for families at the top of the income distribution. But unlike for family income, there is evidence for a positive housing wealth effect across the distribution. The inclusion of time-variant controls, in columns (5) and (6) reduce the estimates slightly without changing any of the substantive points.

Robustness tests in the alternative pseudo panels confirm the positive and substantial effect of home value on private school access for owner-occupier, especially at the top of the socio-economic class distribution, respectively. However, the headline estimate is about half the size at 0.093 points (Column 3, Table A4 in the online appendix).

Table 6: The effects of home value on the probability of attending private school from pseudo-panel fixed effects regressions, 1997-2017

 (1)(2)(3)(4)(5)(6)
Ln mean home value0.175***
(0.028)
0.094**
(0.028)
0.184***
(0.026)


0.157***
(0.029)


Ln mean home value #      
(10th income decile)





0.211***
(0.030)


0.186***
(0.032)
(8th – 9th income decile)





0.180***
(0.029)


0.150***
(0.032)
(6th -7th income decile)





0.179***
(0.031)


0.150***
(0.033)
 (≤5th income decile)





0.179***
(0.026)


0.147***
(0.028)
FENOYESYESYESYESYES
PeriodYESYESYESYESYESYES
Region # PeriodNONOYESYESYESYES
Time-variant controlsNONONONOYESYES
R20.5350.8060.3250.8610.8620.863
F(constant home value effect) – p-value   0.088 0.048

Results from a pseudo-panel of 672 cells from 8 regions, four grouped income decile of pupils’ family income and 21 survey years. The dependent variable is the proportion of students aged 5-15 years in private schools within cells derived from the FRS, 1997/1998-2017/2018. The underlying population are all children aged 5-15 years from owner-occupier families. Columns (5) and (6) controls in addition for the proportion of family heads under 35 years, the proportion over 55 years, the share of single parents, and the proportion of families where at least one parent left full-time education at the age of 21 or higher. Heteroscedasticity and autocorrelation robust standard errors in parentheses. Clusters at region-income ranks level. * p < 0.05, ** p < 0.01, *** p < 0.001.

Finally, Table 7 reports the estimated income and home value effects from a combined model of family resources in the sample of homeowners.  The table reports pseudo-panel estimates based on income bands in columns (1) to (5) and based on socio-economic class in columns (6) and (7). The income effect in the pooled pseudo-panel for this group (column 1) is similar to the estimate in column (1) of Table 6. With the addition of home value (column 2), panel fixed effects (column 3) and period-region dummies (column 4), the income effect drops to 0.036 and fails to reach statistical significance at common levels. By contrast, the estimated effect of home value remains substantial. It appears that it is the access to permanent wealth, measured by home value, rather than current family income that determines private school access in the group of owner-occupier families. The robustness tests in columns  (5) and (6) supports this conclusion.

Table 7: The effects of home value and income on the probability of attending private school from pseudo-panel fixed effects regressions, 1997-2017

 (1) Income bands(2) Income bands(3) Income bands(4) Income bands(5) Income bands(6) NS-SEC(7) NS-SEC
Ln mean home value 0.098***
(0.024)
0.084**
(0.028)
0.174***
(0.027)
0.145***
(0.029)
0.084***
(0.017)
0.083***
(0.016)
Ln mean family income0.112***
(0.017)
0.070***
(0.010)
0.049*
(0.020)
0.036
(0.019)
0.039*
(0.017)
0.019
(0.013)
0.020
(0.013)
FENONOYESYESYESYESYES
PeriodYESYESYESYESYESYESYES
Region # PeriodNONONOYESYESYESYES
Time-variant controlsNONONONOYESNOYES
N67267267267267211751175
R20.5660.6560.8090.8610.8630.6690.670

Pseudo-panel estimates. Columns (1) to (5) are based on 672 cells from eight regions, four grouped income deciles and 21 survey years. Columns (5) and (6) use seven socio-economic classes and eight regions to follow 56 cohorts over 21 years. The dependent variable is the proportion of students aged 5-15 years in private schools within cells derived from the FRS, 1997/1998-2017/2018. The underlying population are all children aged 5-15 years from property-owning families with a mortgage. Heteroscedasticity and autocorrelation robust standard errors in parentheses. Clusters at region-income ranks level. * p < 0.05, ** p < 0.01, *** p < 0.001.

7         Conclusions

Access to private schools in Britain matters, given the success these schools have in incubating future elites, and the ongoing academic and public discourse surrounding social mobility. The fees have risen much more than income, to the extent that the costs for just one child have become more than half of net family income per family member for all but the highest income decile, while the ratio of housing wealth to fees has changed little. The degree of income concentration is high, despite a small fall in the recent period: the income concentration index is 0.49 over 2011-2017. For comparison, access to regular after school activities such as sports clubs is associated with an income concentration index of 0.08. A small proportion of non-income-rich households do attend private school. Bursaries and grants, though income progressive and related to need, are relatively low in value and distributed to only one in five of families outside the top decile; they cannot, therefore, account for more than a minor share of the participation of these non-income-rich families. However, among homeowners, non-rich families with privately educated children have much greater housing wealth than families with children in state schools. Other factors, including parental values and location (distance from good quality state schooling), are also found to be significant determinants of private school choice (Anders et al, 2020); being steady over time, these factors are likely to affect the choices of those able to afford the fees through either income or wealth.

Following these descriptive findings, we have deployed pseudo-panel methods to estimate that the income effect of a 10 per cent income rise on private school access is a 1.16 percentage point rise in participation; outside the top-income decile, the estimated effect is statistically insignificant. The impact of a 10 per cent rise in housing wealth is for a 1.84 percentage point rise in participation for homeowners. Considering the low participation outside the top-income decile, these effects are substantial.

These results contribute further understanding to debates about social mobility trends in Britain (Elliot Major and Machin, 2018). The exclusiveness of access, and its comparative stability over more than two decades of social and economic change, is arguably one reason for the lack of progress over social mobility. The data on bursaries and scholarships have shown how little they can have contributed to any change. The importance of the strong link between wealth and private schooling underpins other recent research showing how family wealth affects success in adult life (Fagereng et al., 2018). 

While these results are the first to show the importance of family wealth for access to high-quality education in Britain, they may underestimate the role of family resources. With our data, we have only been able to examine one, albeit substantial, component of family wealth. Other elements, such as the wealth of family relatives, especially grandparents, are likely to be very important. Family resource surveys typically do not observe transfers within the extended family, and one recommendation for the future is that survey designers consider ways in which this information could be collected. In the case, resources within the extended family have gained in importance, the here estimated home value effect is likely to proxy some of it.

The estimates imply that the future demand for private schooling will be linked to trends in wealth, as well as the growth of top incomes. If relative wealth inequality continues to increase as it has done in recent decades, our findings point to, if anything, a prospect of even more exclusivity in private school access. That trend will be exacerbated if, in particular, homeownership continues its downward trend while home values continue to rise. In that light, we can conclude that policies to improve social mobility should include as a key objective the opening up of private school access to wider sections of the population. Externally-imposed reforms such as removing the schools’ charitable status, taxing school fees, focusing ‘contextual admissions’ to universities on school-type, or integrating private schools either partially or entirely with the state sector aim to reduce demand.[6] Internally driven reform focuses largely on an expansion of means-tested bursaries. Our analysis supports that hitherto bursaries have been income progressive, though too small and scarce to affect the overall exclusivity substantially. Though rising wealth inequality and fees might make this an uphill struggle for fund-raisers, greatly expanded bursary provision not linked to academic selection could make a difference. Means-tested bursaries would need to expand considerably in reach and scale, and the selection criteria should include a strong focus on family wealth, not just income. Whether such internal-driven change is feasible remains uncertain.

References

ACARA (2019) National Report on Schooling in Australia 2017. Sydney. Available at: http://www.acara.edu.au.

Acemoglu, D. and Pischke, J.-S. (2001) ‘Changes in the wage structure, family income, and children’s education’, European Economic Review, 45(4–6), pp. 890–904.

Anders J, Green F, Henderson M, Henseke G. (2020) Determinants of Private School Participation: all about the money? British Educational Reserach Journal. Online.

Angrist, J., Lavy, V. and Schlosser, A. (2010) ‘Multiple Experiments for the Causal Link between the Quantity and Quality of Children’, Journal of Labor Economics, 28(4), pp. 773–824.

Antecol, H., Eren, O. and Ozbeklik, S. (2016) ‘Peer Effects in Disadvantaged Primary Schools: Evidence from a Randomized Experiment’, Journal of Human Resources. University of Wisconsin Press, 51(1), pp. 95–132.

Antman, F. and McKenzie, D. J. (2007) ‘Earnings Mobility and Measurement Error: A Pseudo-Panel Approach’, Economic Development and Cultural Change, 56(1), pp. 125–161.

Atkinson, A. B. (2018). Wealth and inheritance in Britain from 1896 to the present. The Journal of Economic Inequality, 16(2), 137-169.

Ball, S. J. (1997) ‘On the cusp: Parents choosing between state and private schools in the UK: Action within an economy of symbolic goods’, International Journal of Inclusive Education, 1(1), pp. 1–17.

Becker, G. S. and Lewis, H. G. (1973) ‘On the Interaction between the Quantity and Quality of Children’, Journal of Political Economy. University of Chicago Press, 81(2), pp. S279–S288. Available at: http://www.jstor.org/stable/1840425.

Becker, G. S. and Tomes, N. (1994) ‘Human Capital and the Rise and Fall of Families’, in Becker, G. S. (ed.) Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education. 3rd edn. pp. 257–298.

Belfield, C., Britton, J., Buscha, F., Dearden, L., Dickson, M., Van Der Erve, L., Sibieta, L., Vignoles, A., Walker, I. and Zhu, Y. (2018). The relative labour market returns to different degrees. IfS Research report: June 2018.

Blundell, R., Dearden, L. and Sibieta, L. (2010) ‘The demand for private schooling in England: The impact of price and quality’, IFS working papers, 10(21).

Brewer, M., Etheridge, B. and O’Dea, C. (2017) ‘Why are Households that Report the Lowest Incomes So Well-off?’, Economic Journal, 127(605), pp. F24–F49.

Broughman, S.P., Kincel, B., and Peterson, J. (2019) Characteristics of Private Schools in the United States: Results From the 2017–18 Private School Universe Survey First Look (NCES 2019-071). U.S. Department of Education. Washington, DC: National Center for Education Statistics.

Burkhauser, R.V., Hérault, N., Jenkins, S.P. and Wilkins, R. (2018a) ‘Survey Under‐Coverage of Top Incomes and Estimation of Inequality: What is the Role of the UK’s SPI Adjustment?’,  Fiscal Studies, 39(2), pp.213-240.

Burkhauser, R.V., Hérault, N., Jenkins, S.P. and Wilkins, R. (2017) ‘Top incomes and inequality in the UK: reconciling estimates from household survey and tax return data’, Oxford Economic Papers, 70(2), pp.301-326.

Burrows, V. (2018) ‘The Impact of House Prices on Consumption in the UK: a New Perspective’, Economica, 85(337), pp. 92–123.

Carneiro, P. and Ginja, R. (2016) ‘Partial insurance and investments in children’. The Economic Journal, 126(596), pp.F66-F95.

Chevalier, A., Harmon, C., O’Sullivan, V. and Walker, I. ( 2013)  ‘The impact of parental income and education on the schooling of their children’, IZA Journal of Labor Economics, 2(1), p.8.

Cooper, D. (2013) ‘House price fluctuations: The role of housing wealth as borrowing collateral’, Review of Economics and Statistics, 95(4), pp. 1183–1197.

Crawford, C.  Gregg, P., Macmillan, L., Vignoles, A. and Wyness G. (2016) ‘Higher education, career opportunities, and intergenerational inequality’, Oxford Review of Economic Policy, 32(4), pp. 553–575.

Dearden, L., Ferri, J. and Meghir, C. (2002) ‘The Effect of School Quality on Educational Attainment and Wages’, Review of Economics and Statistics, 84(1), pp. 1–20.

Dearden, L., Ryan, C. and Sibieta, L. (2011) ‘What Determines Private School Choice? A Comparison between the United Kingdom and Australia’, Australian Economic Review, 44(3), pp. 308–320.

Deaton, A. (1985) ‘Panel data from time series of cross-sections’, Journal of Econometrics. North-Holland, 30(1–2), pp. 109–126.

DESTATIS (2016) Finanzen der Schulen. Schulen in freier Trägerschaft und Schulen des Gesundheitswesens 2013. Wiesbaden: Statistisches Bundesamt. Available at: http://www.destatis.de.

Dettling, L. J. and Kearney, M. S. (2014) ‘House prices and birth rates: The impact of the real estate market on the decision to have a baby’, Journal of Public Economics. North-Holland, 110, pp. 82–100.

DfE (2018) Schools, Pupils, and their Characteristics: January 2018. London. Available at: https://www.gov.uk/government/statistics/schools-pupils-and-their-characteristics-january-2018.

Disney, R. and Gathergood, J. (2018) ‘House Prices, Wealth Effects and Labour Supply’, Economica, 85(339), pp. 449–478.

DWP (2018a) Family Resources Survey: Background note and methodology. Available at: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/692141/family-resources-survey-2016-17-background-note-and-methodology.pdf.

DWP (2018b) Households Below Average Income (HBAI) Quality and Methodology Information Report. Available at: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/691919/households-below-average-income-quality-methodology-2016-2017.pdf.

Elder, T. and Jepsen, C. (2014) ‘Are Catholic primary schools more effective than public primary schools?’, Journal of Urban Economics, 80, pp. 28–38.

Fack, G. and Grenet, J. (2015) ‘Improving college access and success for low-income students: Evidence from a large need-based grant program’, American Economic Journal: Applied Economics, 7(2), pp. 1–34.

Fagereng, A. Mogstad, M. and Rønning, M. (2018) ‘Why Do Wealthy Parents Have Wealthy Children?’, CESifo Working Paper, No. 6955.

Feinstein, L., and Symons, J. (1999). Attainment in Secondary School. Oxford Economic Papers, 55(2), 300-321.

Fichera, E. and Gathergood, J. (2016) ‘Do wealth shocks affect health? New evidence from the housing boom’, Health Economics, 25(Suppl 2), pp. 57–69.

Foskett, N. and Hemsley-Brown, J. (2003) ‘Economic aspirations, cultural replication and social dilemmas–interpreting parental choice of British private schools’, in Walford, G. (ed.) British Private Schools: Research on Policy and Practice. London: Routledge, pp. 188–201.

Fox, I. (1985) Private Schools and Public Issues: The Parents’ View. London: Macmillan Press.

Green, F. and Kynaston, D. (2019) Engines of Privilege: Britain’s Private School Problem. London: Bloomsbury.

Green, F. Machin, S., Murphy, R. and Zhu, Y. (2012) ‘The Changing Economic Advantage from Private Schools’, Economica, 79(316), pp. 658–679.

Green, F., Henseke, G. and Vignoles, A. (2017) ‘Private schooling and labour market outcomes’, British Educational Research Journal, 43(1).

Halsey, A. H., Heath, A. F. and Ridge, J. M. (1980) Origins and destinations : family, class, and education in modern Britain. Clarendon Press.

Henderson, M., Anders J., Green, F. and Henseke G. (2020). “Private Schooling, Subject Choice, Upper Secondary Attainment and Progression to University.” Oxford Review of Education 46(295-312).

HM Land Registry (2018a) ‘About the UK House Price Index’, pp. 1–19. Available at: https://www.gov.uk/government/publications/about-the-uk-house-price-index/about-the-uk-house-price-index.

HM Land Registry (2018b) UK House Price Index England: May 2018, UK House Price Index. Available at: https://www.gov.uk/government/publications/uk-house-price-index-england-may-2018/uk-house-price-index-england-may-2018.

HM Land Registry (2019) Comparing house price indices in the UK. Available at: https://www.gov.uk/government/publications/about-the-uk-house-price-index/comparing-house-price-indices-in-the-uk.

ISC (2018) ISC Census and Annual Report 2018. Available at: https://www.isc.co.uk/media/4890/isc_census_2018_report.pdf.

ISC (2019) ISC Census and Annual Report 2019. Available at: https://www.isc.co.uk/media/5479/isc_census_2019_report.pdf.

Jerrim, J., Parker, P., Chmielewski, A. and Anders, J. (2016) ‘Private schooling, educational transitions, and early labour market outcomes: Evidence from three Anglophone countries’, European Sociological Review, 32(2), pp. 280–294.

Kenway, J. and Fahey, J. (2014) ‘Staying ahead of the game: the globalising practices of elite schools’, Globalisation, Societies and Education. Taylor & Francis, 12(2), pp. 177–195.

Lavy, V., Silva, O. and Weinhardt, F. (2012) ‘The Good, the Bad, and the Average: Evidence on Ability Peer Effects in Schools’, Journal of Labor Economics, 30(2), pp. 367–414.

Lindley, J. and Machin, S. (2016) ‘The Rising Postgraduate Wage Premium’, Economica, 83(330), pp. 281–306.

Lindley, J. and McIntosh, S. (2015) ‘Growth in within graduate wage inequality: The role of subjects, cognitive skill dispersion and occupational concentration’, Labour Economics. Elsevier B.V., 37, pp. 101–111.

Lovenheim, M. F. and Reynolds, C. L. (2013) ‘The Effect of Housing Wealth on College Choice: Evidence from the Housing Boom’, Journal of Human Resources, 48(1), pp. 1–35.

Lye, J., & Hirschberg, J. (2017). Secondary school fee inflation: an analysis of private high schools in Victoria, Australia. Education Economics, 25(5), 482-500.

Macmillan, L., Tyler, C. and Vignoles, A. (2015) ‘Who gets the top jobs? the role of family background and networks in recent graduates’ access to high-status professions’, Journal of Social Policy, 44(3), pp. 487–515.

Elliot Major, L. and Machin, S. (2018) Social mobility and its enemies. Penguin Random House.

Malacova, E. (2007) ‘Effect of single‐sex education on progress in GCSE’, Oxford Review of Education.  Routledge , 33(2), pp. 233–259.

Maxwell, C. and Aggleton, P. (eds) (2015) Elite Education: Inernational Perspectives. London: Routledge.

McKnight, A. and Karagiannaki, E. (2013) ‘The wealth effect: how parental wealth and own asset-holdings predict future advantage.’ in Wealth in the UK: Distribution, Accumulation, and Policy, Oxford: Oxford University Press, p.119.

Nationwide (2015) Nationwide House Price Index Methodology Background to Nationwide House Price Information. Available at: https://www.nationwide.co.uk/-/media/MainSite/documents/about/house-price-index/nationwide-hpi-methodology.pdf.

Naylor, R., Smith, J. and Telhaj, S. (2016) ‘Graduate returns, degree class premia and higher education expansion in the UK’, Oxford Economic Papers, 68(2), pp. 525–545.

Nghiem, H., Nguyen, H., Khanam, R. and Connelly, L. (2015) ‘Does school type affect cognitive and non-cognitive development in children? Evidence from Australian primary schools’, Labour Economics, 33, pp. 55–65.

Nielsen, H. S., Sørensen, T. and Taber, C. (2010) ‘Estimating the Effect of Student Aid on College Enrollment: Evidence from a Government Grant Policy Reform’, American Economic Journal: Economic Policy, 2(2), pp. 185–215.

O’Donnell, O., O’Neill, S., Van Ourti, T. and Walsh, B. (2016) ‘conindex: Estimation of concentration indices’, The Stata journal, 16(1), pp. 112-138.

OECD (2012) Public and Private Schools: How Management and Funding Relate to their Socio-economic Profile, Paris: OECD Publishing.

ONS (2018) Wealth in Great Britain Wave 5: 2014 to 2016. Available at: https://www.ons.gov.uk/peoplepopulationandcommunity/personalandhouseholdfinances/incomeandwealth/bulletins/wealthingreatbritainwave5/2014to2016/previous/v2.

Parsons, S., Green, F. , Ploubidis, G., Sullivan, A. and Wiggins, D. (2017) ‘The influence of private primary schooling on children’s learning: Evidence from three generations of children living in the UK’, British Educational Research Journal, 43(5), pp. 823–847.

Pianta R. C. and Ansari, A (2018) ‘Does Attendance in Private Schools Predict Student Outcomes at Age 15? Evidence From a Longitudinal Study’. Educational Researcher 2018, 47(7):419-434.

Piketty, T. and Zucman, G. (2014) ‘Capital is Back: Wealth-Income Ratios in Rich Countries 1700–2010’, The Quarterly Journal of Economics. Narnia, 129(3), pp. 1255–1310.

Reeves, A., Friedman, S., Rahal, C. and Flemmen, M. (2017) ‘The Decline and Persistence of the Old Boy: Private Schools and Elite Recruitment 1897 to 2016’, American Sociological Review, 82(6), pp. 1139–1166.

Reinold, K. (2011). Housing equity withdrawal since the financial crisis. Bank of England Quarterly Bulletin, Q2.

Ryan C. and Sibieta L. (2011) A Comparison of Private Schooling in the United Kingdom and Australia. The Australian Economic Review, 44(3): 295-307.

Sacerdote, B. (2011) Peer Effects in Education: How might they work, how big are they and how much do we know Thus Far?, Handbook of the Economics of Education. Elsevier.

Scottish Government (2018) Summary statistics for schools in Scotland. Available at: https://www2.gov.scot/Topics/Statistics/Browse/School-Education/Summarystatsforschools.

Solon, G. (2004) ‘A model of intergenerational mobility variation over time and place’, in Corak, M. (ed.) Generational income mobility in North America and Europe. Cambridge, New York: Cambridge University Press, pp. 38–47.

Sullivan, A. and Heath, A. F. (2003) ‘Intakes and Examination Results at State and Private schools’, in Walford, G. (ed.) British Private Schools: Research on policy and practice. London: Woburn Press, pp. 77–104.

Sutton Trust and Social Mobility Commission (2019) Elitist Britain 2019: The educational backgrounds of Britain’s leading people. London. Available at: https://www.gov.uk/government/publications/elitist-britain-2019.

Verbeek, M. (2008) ‘Pseudo-Panels and Repeated Cross-Sections’, in Mátyás, L. and Sevestre, P. (eds) The Econometrics of Panel Data. Springer, Berlin, Heidelberg, pp. 369–383.

Weenink, D. (2008) ‘Cosmopolitanism as a form of capital: Parents preparing their children for a globalizing world’, Sociology, 42(6), pp. 1089–1106.

Welsh Government (2018) School Census Results, 2018.

West, A., Noden, P., Edge, A., David M. and Davies J. (1998) ‘Choices and Expectations at Primary and Secondary Stages in the State and Private Sectors’, Educational Studies, 24(1), pp. 45–60

Whitty, G., Power, S. and Edwards, T. (1998) ‘The assisted places scheme: Its impact and its role in privatization and marketization’, Journal of Education Policy, 13(2), pp. 237–250.

Wilde, R. J., Green, F., Taylor-Gooby, P. and Wiborg, S. (2016) ‘Private Schools and the Provision of “Public Benefit”’, Journal of Social Policy, 45(2), pp. 305–323.


[1] Throughout this paper ‘private schools’ are fee-paying schools; elsewhere these are also variously termed ‘independent schools’, ‘prep schools’ at primary level, and confusingly ‘public schools’ in the case of the more prestigious secondary schools.

[2] https://preview.tinyurl.com/zo5m7du

[3] While the school type question has remained unchanged throughout, the wording of the response option underwent a minor change, from “Any PRIVATE school (prep or secondary)” before the 2002/2003 wave to “Any PRIVATE/Independent school (prep, primary, secondary, City Technology Colleges)” thereafter. This coincided with a significant increase in the estimated private school participation rate from 3.4 percent in 2001 to 4.4 percent between in 2002. Yet a simple pooled OLS regression of private school participation on income decile interacted with period dummies with data from 2001 and 2002, revealed no significant period-income rank interaction effects. The null hypothesis of no joined differences could not be rejected at p=0.70 (F=0.71). We judge that this change is unlikely to seriously bias the trend analysis of private school access. To check, we find that restricting the analysis to the years after 2002 does not make any significant difference to the findings.

[4] The chart is restricted to families with income at or above the 5th percentile. Income data becomes more unreliable at the very bottom of the distribution (Brewer et al. 2016).  

[5] The concentration index is a bivariate generalisation of the Gini coefficient, and there is a theoretical literature that examines the properties of different definitions and their relation to different concepts of inequality (e.g., O’Donnell et al., 2016).

[6] https://www.privateschoolreform.co.uk/our-publications.

Teachers Under Pressure

It must be exhilarating, if challenging, to set out for the first time on a teaching career in Britain’s schools. But, from eye-witness reports in recent years, for some new recruits the strains are not long arriving. Now, as term restarts, the chaos surrounding the pandemic can only be adding to the pressures that teachers have laboured under for a long time.

The stats suggest that dissatisfaction is not confined to an unhappy few. In England, among the newly qualified teachers in 2014, some 14 percent had left after a year; after five years, a third had gone. It seems quite a waste. Teacher retention has been declining for some while, and had fallen yet again in 2019 — despite attempts to stem the tide.

What is it about the job of teaching nowadays that lies behind this trend?

Two answers are commonly given to this question: workload and pay. Workload is typically interpreted in terms of working hours, and one can point to the long hours worked during term-time, compared to other professions and compared to teachers in other countries. However, the simple fact of long work hours does not explain why retention has become an increasingly urgent problem, since teachers’ term-time working hours have been relatively high for decades and have not been noticeably increasing. Similarly, teachers’ pay has not notably fallen in the long term behind that of many alternative jobs in the professions.

My recent paper suggests a different possible explanation. Put briefly, two particular aspects of teachers’ job quality seem to be deteriorating: “work intensity” and “task discretion”.

Work intensity means the “rate of physical and/or mental input to work tasks performed during the working day”. Intensification of work happens when, for example, more tasks have to be done in the same time (those tea breaks and pauses are squeezed) or when the tasks themselves become more intensive, or when they are made to take longer because of frequent interruptions. I used data from the Skills and Employment Surveys, a nationally representative survey series which naturally includes many teachers, carried out every five years or so. Because high work intensity is manifested in a variety of ways, it is usually best to use multiple measures.

The diagram below presents one of the measures used: the percent of teachers who “strongly agree” that their job requires them to work very hard. It shows a striking increase in the work intensity of teachers over the years, and especially between 2012 and 2017, a period during which spending per pupil was declining. By 2017, the work intensity of teachers had become higher than any other major occupation in Britain. Indeed, some 90 percent of teachers report that their job requires them to work very hard, as compared with just 44 percent for all other workers combined.

Another measure is the percent of jobs where the work involves ‘working at very high speed’ at least three quarters of the time: this proportion increased from 16.1 percent of teachers in 1992 to 57.9 percent in 2017.

Task discretion means the extent to which employees have an influence over the tasks they do at work. I used an index which summarised how much influence teachers personally had over the tasks they do, how they did them, the pace at which they did them and the quality standards they worked to. As the diagram shows, while work intensity was increasing, the level of task discretion was declining. To illustrate what’s behind that downward slide, in 1992 some 73% of teachers reported that they had a great deal of influence over how they performed their tasks; by 2017, this proportion had come down to 36%. In short, teachers’ sense of control over their work seems to have diminished, and notably more than for other professionals.

Having both high work intensity and reduced control or influence over your work indicates high strain in a job, a classic source of stress. The two diagrams in tandem suggests that there are likely to be many more teachers in high strain jobs in 2017 than back in 1992.

According to the same data, other aspects of teachers’ job quality – such as ‘working time quality’ (including total hours worked), job security and pay did not show any major changes over the years, at anything like the rate that work intensity rose or task discretion fell.

My paper does not show why work intensity has risen, nor why task discretion has been falling, but we can speculate. One straightforward explanation sometimes given for work intensification in public sector jobs is especially salient for the recent decade: that a decline in funding means that teachers have more pupils to teach, along with more marking and associated work. An explanation for the decline in task discretion may lie in the increasing bureaucratic and formal requirements of accountability that teachers must nowadays demonstrate. Yet while these explanations are plausible, the full story will surely be more complex.

While the paper does not aim to prove the link to the problem of teacher retention, the declining discretion and sharply rising work intensity – while hours themselves have not changed much – should prompt a sharpened focus on what needs addressing in the management of teachers. It is these factors that have moulded a good part of the changes in well-being and job satisfaction since around 2006.

The enormous extra strains of teaching in the shadow of COVID-19 have come on top of the long-term trends. For pre-school teachers especially there is the additional health insecurity of working face to face during the latest lockdown, with no vaccine protection. Yet teaching remains a career choice with the potential for deep intrinsic rewards. Is it too much to hope that the recovery we look for in 2021 might be accompanied by a serious re-think of how teachers’ jobs should be designed, monitored and supported, so as to build on that promise?

————————— Francis Green “British Teachers’ Declining Job Quality: Evidence from the Skills and Employment Survey”. Oxford Review of Education, online. Published 20/1/2020.

———————————————————————————————————————————–

This is the final accepted version of the paper submitted to the publisher:

British Teachers’ Declining Job Quality: Evidence from the Skills and Employment Survey

Francis Green

LLAKES Centre, UCL Institute of Education

Key words: work intensity, task discretion, working time, well-being, pay, retention

Address for correspondence: Francis Green, UCL Institute of Education, Bedford Way, London WC1H 0AL

orcid.org/0000-0002-6786-5012

Published by Oxford Review of Education, online 8 January 2021

Funding Acknowledgement:

For the Skills and Employment Survey series, the primary funder has been the Economic and Social Research Council (ESRC). Additional funding for individual surveys came from the Department for Education (and its predecessors), the UK Learning and Skills Council, the UK Commission for Employment and Skills, FutureSkills Scotland, the East Midlands Development Agency, the Department for Employment and Learning (Northern Ireland), Highlands and Islands Enterprise, the Wales Institute of Social and Economic Research, Cardiff University, the Welsh government. A consortium of 27 funders including the Leverhulme Trust and the Employment Department funded the 1992 survey. The surveys were designed and led by Duncan Gallie, Francis Green and Alan Felstead, with later contributions to individual waves from David Ashton, Bryn Davies, Ying Zhou, Hande Inanc and Golo Henseke.

1. Introduction

Teachers’ jobs are commonly described by the trends in their pay and working hours. A teacher’s job quality, however, like that of any other worker, also includes other significant components – both other extrinsic factors such as promotion prospects, and intrinsic factors such as the degree of task discretion (OECD, 2014). In this paper I use the British Skills and Employment Survey (SES) series to provide a wider perspective on teachers’ job quality, covering several components. While these data have been extensively used for the analysis of job quality, skills utilisation and employment relations in general, it is hitherto an untapped data source for understanding teachers’ jobs and their work-related well-being.

Teachers’ job quality is an important concern, not only from the perspective of existing and prospective teachers, but also from that of education ministers tasked with ensuring a nation’s supply of teachers. Since at least 2011, schools in England have experienced a falling rate of teacher retention, the effect of which is more serious at a time of increasing demand for teachers with rising pupil numbers (Lynch et al., 2016; Sims, 2017; Worth and Van den Brande, 2019; DfE, 2018a; Foster 2019). While retention rates vary according to subject specialisms and region, in Britain recent public discourse surrounding the cause of declining teacher retention has focused on teachers’ workloads and relative pay. The conundrum which, in part, motivates this paper is that neither work hours – commonly taken as the measure of workload – nor pay offer a potentially satisfactory account as to why retention rates have declined. British teachers’ working hours are long when compared with other countries (Sellen, 2016), but do not exhibit an upward long term trend; nor is there a rise in evening or weekend working (Allen et al., 2019). The pay of teachers covered by the Pay Review Board declined in real terms between 2005 and 2015, but the fall was not significant once the changing demographic composition of the teacher workforce was controlled for; nor did teachers’ salaries fall especially fast when compared with those of other professions (Bryson and Forth, 2017).

There are relatively few sources around the world of individual, nation-wide survey data on jobs and work which permit identification of a subset of teachers. SES is one of these. My intention is to contribute to the discourse surrounding teacher supply in Britain, but at the same time to demonstrate the utility of application of such surveys for specific populous occupations anywhere, providing that good sampling methods are deployed and that sufficiently disaggregated occupation codes are included. I find that in several domains there has been a decline in teachers’ job quality in Britain, and that the trends are consistent with changes in teachers’ work-related well-being. Given earlier findings that link work-related well-being with employee turnover, the findings are consistent with the emergence of increasing teacher wastage in recent years.

2. Job quality and Well-being

In 1992, when the data begin, there were still five years of the Thatcher/Major years of government to run. In these years neo-liberal reforms of education were already being introduced, while spending per pupil stagnated; the pupil-teacher ratio (PTR) had been increasing since 1990. With the New Labour government of Blair, the first few years were fiscally constrained. By 2001, however, education spending was beginning to rise again, and the PTR to fall; expenditure increased by more than 5% a year through the 2000s (Belfield and Sibieta, 2016). Then, following the financial crisis and the subsequent severe fiscal austerity, real-terms spending per pupil declined by 8% in England between 2009-10 and 2017-18,[1] though rather less in Wales[2] and less still in Scotland. Broadly reflecting that spending path, the PTR in England bottomed out at 17.7 in 2012, rising to 18.9 by 2018[3]; in Wales, the PTR was rising from 2010[4]; in Scotland, the PTR began increasing as soon as the financial crash hit in 2008, but levelled out from 2012.[5] The period since 2010 also saw the extension of independent governance for schools in England that converted to become ‘academies’ and the inauguration of autonomous ‘free schools’. With an ongoing regime of regular inspection, intensified competition for pupils through performance league tables and changing performance indicators, the external pressures on schools to succeed despite their constrained budgets were sustained.

In these circumstances, substantive changes in the experience of working in teaching could be expected. Many of these changes may be characterised with the use of three related concepts: job quality, work-related well-being and job satisfaction. A multi-dimensional concept, job quality is defined as those objective job characteristics which affect how people’s needs are met through work (Green, 2006). These characteristics have been classified into a set of domains (e.g. Muñoz de Bustillo et al., 2011; Green and Mostafa, 2012; OECD, 2014). The classification adopted by the European Parliament sorts characteristics into seven domains: four ‘intrinsic’ domains – work intensity, skills and discretion, social support and physical working conditions, and three extrinsic domains – pay, prospects (including security), working time quality (including hours) (European Parliament, 2016/2017).

Job quality is conceptually distinguished from work-related well-being – that is, the hedonic or eudaemonic well-being that is specifically associated with work life. Indicators of work-related well-being can be measures of affect or a measure such as job satisfaction that involve cognitive assessment (Warr, 2007, p.28).Yet all theories of job quality predict a positive relationship between job quality and well-being. In economics, for example, the utility of a job is a direct function of its pay and working conditions. Leading theories within occupational psychology are the job demands and control theory (Karasek, 1979), the Effort-Reward Imbalance theory (Siegrist, 1996) and the Job Demands-Resources model (Bakker and Demerouti, 2007). While these theoretical approaches expect positive relationships between job quality and health or well-being, they hypothesise that the effects are non-linear, and that job quality domains interact in their effects. For example, in the demands-control model, a job with high work intensity is expected to have a more detrimental effect on mental well-being if it also allows the worker little latitude or task discretion (that is, little control).

Job satisfaction is perhaps the most widely used measure of work-related well-being, both for large generic worker samples and for individual workplaces. However, this cognitive assessment is typically thought to incorporate comparison with past or other potential jobs for the individual subject. It is generally a useful predictor of turnover (Green, 2010; Brown et al., 2012). For a teacher, then, job satisfaction would not be solely dependent on conditions in the subject’s current school.

A considerable body of evidence supports the theorised connections between job quality and well-being. In the context of teaching, Ferguson et al. (2012) found that workload is one of the key predictors of depression, anxiety and low job satisfaction in northern Ontario. Torres (2016) finds an inverse link between teachers’ perceptions of workload manageability and decisions to quit from US Charter Management Organisations, while Huyghebaert et al. (2018) report that work intensity among teachers in France is associated positively with emotional exhaustion and presenteeism, and negatively with job satisfaction and (self-rated) performance. Wang et al (2018) found that school principals’ job satisfaction, also, was negatively influenced by work intensification. In Britain, Barmby (2006) found that workload and marking are the two most frequently cited reasons given for considering leaving. Bryson et al. (2019) find substantive and significant correlations between indicators of intrinsic job quality and both “job contentment” and job satisfaction. These studies of teachers are just part of a growing literature filling out our knowledge of the relationships between job quality and worker well-being across multiple occupations (Eurofound, 2019).

Drivers and Trends in Teachers’ Job Quality

The drivers of job quality can be external to an organisation – macroeconomic or aggregate factors such as economic growth, technological advances, fiscal expansion or contraction (especially relevant for public sector jobs), or changes in the regulation of jobs – or internal, linked to management, job design and control. In the case of teachers, changing spending per pupil, consequent changes in the pupil-teacher ratio and other alterations in educational regulation are key external factors. Given the changes outlined above, a decline in job quality might be expected during the 1990s, and in the period of fiscal austerity following the financial crisis. By contrast, in the early 2000s when teachers were benefiting from more resources and educating fewer pupils, one might expect to see rising job quality.

Existing evidence on teachers’ changing job quality in Britain is largely focused on pay and hours. Pay is important for teachers’ retention (Hendricks, 2014; Sims, 2017); yet, as noted above, relative teachers’ pay has not obviously deteriorated. The debate about the impact of teachers’ workload is expressed in hours (e.g. Higton et al., 2017). And yet, English teachers’ hours problem has been characterised as their excess over international norms and their failure to decrease, rather than as a secular increase (Sellen, 2016; Allen et al., 2019; Worth et al., 2018; Worth and Van den Brande (2019).

There is relatively little direct evidence on other key elements of teachers’ job quality. Exceptions are Sellen (2016), who shows that British teachers in their early years have relatively low levels of training, and Sims (2020) who finds that teacher job satisfaction is significantly correlated with student discipline, hours, leadership support and scope for career progression and support from headteachers. Qualitative research compares some British teachers’ jobs with those in Finland (Webb et al., 2004) and with Norway (Lloyd and Payne, 2012). There are no studies, to my knowledge, of job quality trends among British teachers, other than those which examine wages or hours.

Nor are there studies of trends in British teachers’ affective well-being. We do, however, learn something about trends in job satisfaction: between 2010 and 2017 there is stability in teachers’ job dissatisfaction (Hillary et al., 2018; Worth and Van den Brande, 2019); but between 2013 and 2018 there is rising job dissatisfaction among lower-secondary teachers (Jerrim and Sims, 2019).

In light of the above, in the analysis that follows, using the SES data series I address three research questions: a) What is the long-run trend in teachers’ job quality in Britain? b) What are the trends in indicators of teachers’ well-being at work? c) Is the well-being trend consistent with the trend in job quality?

3. Data on Teachers in the Skills and Employment Survey (SES)

The SES (Felstead et al., 2019) series provides information on several facets of job quality that can contribute to our understanding of teachers’ jobs in Britain. The SES collected data from working adults in England, Wales and Scotland who were interviewed in their own homes. Relevant information on job quality starts in 1992, with subsequent waves in 1997, 2001, 2006, 2012 and 2017. In each case the sample was drawn using random probability principles subject to stratification based on socio-economic indicators. One eligible respondent per address was randomly selected for interview, and the response rate ranged from 72% in 1992 to 49% in 2012. For each survey, weights were computed to take into account the differential probabilities of sample selection, the over-sampling of certain areas and some small response rate variations between groups (defined by sex, age and occupation). All of the analyses that follow use these weights.

Though designed for all workers, school teachers are covered in sufficient numbers with the statistical power to afford informative estimates of trends in teachers’ job quality. Teachers are identified as those with SOC codes 2314, 2315 and 2316, covering teachers in nursery, primary, secondary and special schools. From 2006 all persons aged 20 to 65 were eligible, but before 2006 SES covered only those up to 60. The probability for a sample respondent to be a teacher depends neither on the year of the survey nor on region. For consistency the base sample analysed here comprises those aged 20 to 60 living in Britain, excluding the Scottish Highlands, in total 857 teachers. Of these, 72% were female; 86% were in England, 4% in Wales and 9% in Scotland; 45% were aged 40 and under, and the proportion in the private sector was 13%. These proportions are relatively close to estimates from the Quarterly Labour Force Survey (QLFS), even though QLFS uses a different sampling methodology from SES (QLFS is a household, predominantly telephone-based survey).

Job Quality Indicators in SES

Work Intensity may be defined as ‘the rate of physical and/or mental input to work tasks performed during the working day’ (Green 2001, 56). Made up of several elements, it is experienced in complex and differentiated ways by workers, not least by teachers (Ballet and Kelchtermans, 2009; CooperGibson Research (2018). Multiple indicators are called for to tap different aspects. Measurement relies on workers’ self-reports and has no obvious metric unit equivalent to, for example, working hours. To measure work intensity generically, I use three variables. First, respondents were asked whether they agreed with the statement “My job requires that I work very hard”, using a 4-point scale (“strongly agree/ agree/ disagree/ strongly disagree”). Second, they were asked “How often does your work involve working at very high speed”; they could answer against a 7-point frequency scale. Third, using the same scale jobholders were asked “How often does your work involve working to tight deadlines”. These three variables are positively correlated. I define an index, Required Work Intensity, to be the first principal component. As validation, this index correlates as expected with well-being, and with a measure of how many factors control work effort; among all workers, those who work in more intensive jobs receive a compensating pay premium, as economics predicts (Green et al., 2006).

Work intensification, within any group, is defined as the change in Required Work Intensity between survey waves.

Skills and Discretion is measured in SES with four variables. First, ‘task discretion’  includes measures which assess how much influence people report: deciding what tasks they are to do, how the tasks are done, how hard they work and the quality standards to which they work. The response options are ‘a great deal’, ‘a fair amount’ , ‘not much ‘ and ‘none at all’. A summary index was constructed and normalised to the range 1-4, with 4 indicating ‘high discretion’. Second, ‘organisational participation’ is a 0/1 dummy variable, where ‘1’ indicates that respondents reported that they would have ‘quite a lot’ or ‘a great deal’ of say in any change in their place of work which affected how their job is done. Third, training participation measures participation in any of several types of training activity in the past year. Fourth, ‘training quality’ is a dummy variable indicating whether the most recent spell of training received had ‘improved their skills a lot’. In addition, the demand-control model implies that high work strain stems from having low discretion and high work intensity in a job. To capture this possibility I computed a dummy variable for high work strain, equal to one if respondents both strongly agreed that their job required them to work very hard and reported task discretion below 2.0 .

Pay and Prospects are measured using three variables: Hourly Pay (in 2015 prices, before tax), High Security (dummy for whether respondents feel that there is no chance, or that it is very unlikely, that they will lose their job and become unemployed in the coming year), and High Promotion Expectation (dummy for whether respondents report a 75% chance or more of being promoted in the next 5 years in their present organisation).

Working Time Quality is measured with three variables: Usual Weekly Hours (including overtime whether paid or not; full-time workers), Schedule Rigidity (dummy for respondents disagree/strongly disagree that they can decide start or finish times of their work), and Time-Off Rigidity (dummy for somewhat or very difficult to take time off during work hours for family or personal matters).

Trends in the two remaining job quality domains delineated by the European Parliament’s classification – namely, physical working conditions and social support – are not covered in the SES series.

Well-being Indicators in SES

From 2001 onwards the SES datacontain Warr’s two scales for affective well-being: ‘Depression®Enthusiasm’ and ‘Anxiety®Contentment’. These capture the structure of feelings and emotions arising from either work or non-work settings (Warr, 1990; 2007); I abbreviate these henceforth as Enthusiasm and Contentment. These two dimensions affect quitting and absenteeism, and both are related to mental health indicators, organisational commitment and job satisfaction (Green 2010; Warr 2007).

SES also collects data about a further specific aspect of work-related well-being (encompassing also travel-to-work): respondents are asked how often they ‘come home from work exhausted’. I create a dummy variable equal to one for those who answer at the top two points (‘often’ or ‘always’) of the 5-point response scale.

Finally, respondents state how satisfied they are with their jobs, both overall and according to fourteen facets.

4. Findings

The analysis that follows is primarily descriptive, using the above indicators. I use other professional occupations (those, other than teachers, coded as in SOC Major Group 2) as a point of comparison for teachers’ jobs in each wave. I also use regression modelling for two purposes: to assess the significance of the time trend in each aspect of teachers’ job quality, after including controls for demographic and other changes in workforce composition; second, to decompose the observed changes in teacher well-being into those that can, in a statistical sense, be ‘accounted for’ by simultaneous observed changes in job quality indicators, and those that cannot.

I now turn to the first research question and investigate what has been the trend in teachers’ job quality.

4a Trends in Teachers’ Job Quality

Work Intensity

Teachers’ work intensity (Table 1a) was in all years higher than that of other professionals, and moreover for all other occupations together (the population mean for the Required Work Intensity index is zero). Indeed, no other occupations, for which there is a large enough sample to make the comparison, work as intensively as do teachers: their nearest rivals are Health and Social Services Managers, and Legal Professionals.

Teachers’ work intensity has increased since 1992 in two stages. The proportion reporting that they are required to work very hard rose from 54% to 82% between 1992 and 2001; similarly, the proportion reporting that they worked at very high speed at least three quarters of the time rose from 16% to 51% over the same period. There followed a period of reprieve during the 2000s. The Work Intensity Index dipped between 2001 and 2006 then rose back to nearly the same level in 2012 as 2001. Between 2012 and 2017 the index again soared. By 2017 a remarkable 90% strongly agreed that their job requires them to work very hard. This proportion compares with just 44% for all other occupations, and 52% for Other Professional occupations. In sum, taking the whole period, work intensity has followed a statistically significant upward trend for each indicator and for the Required Work Intensity Index. No other large occupation has shown anything like this degree of work intensification.

Skills and Discretion

Table 1b shows that teachers have also gone through two periods of declining task discretion since 1992. First, in the 1992-2001 period, teachers’ task discretion declined by 0.46 points; this decline was part of a generalised decline across all professional and other occupations in the 1990s, though with different timing in the public and private sectors (Gallie et al., 2004). Second, teachers’ discretion declined further after 2012. All four elements of the discretion index declined, but to illustrate with one of the responses that lie behind this fall: the proportion of teachers in the SES sample who reported that they had “a great deal of influence” over deciding how to do their tasks was 48% in 2012, but 31% in 2017.

High work strain – the combined indicator of low task discretion and high work intensity – showed a remarkable increase over the long term: the proportion of teachers working under high strain has gone from virtually none in 1992 to 21.3% in 2012 and to 27.3% in 2017. Taken over the whole period, teachers are nearly twice as likely as other professionals to be working under high strain (16.2% compared with 8.5%).

There has also been a dramatic decline in the extent to which teachers report that they can participate in decision-making in their schools. Back in 1992, some 45% of teachers reported that they had either “quite a lot” or “a great deal” of say over organisational changes; this proportion had reduced by 2017 to just 20%. This decline has occurred within both primary and secondary sectors, and cannot be simply associated with governance changes stemming from the move to academy control (primarily, after 2010).

Echoing the concerns of Sellen (2016), who held that good quality training was being crowded out by job demands, the SES data here reveal that the overall participation rate in training has declined from 92% in 2006 to 86% in 2017. The perceived training quality reported by those that did receive training did not compensate. Indeed the 1estimate of training quality fell: the proportion who said that their recent training spell raised their skills a lot was 41% in 2006 and 31% in 2017.

Pay and Prospects

As shown in Table 1c, pay has no long term trend since 1997. It rises up to a peak in the 2006 wave; by 2017 it had fallen in real terms by 13% since that peak. After conditioning on year, sex, age, age squared, school sector, and hours of work, this post-2006 decline remains significant. Meanwhile, teachers’ pay relative to other professional workers fell from 80% in 2006 to 71% in 2017.

While current pay is a central indicator of teachers’ ability to meet their material needs, also relevant both financially and psychologically are the prospects for future pay. Teachers had a notably better chance of promotion in 2012 than they had in 1992. They have always experienced relatively high job security, not only compared with average workers but also in comparison with other professional workers (see Table 1c). There is no notable trend in perceived job security since 1997; but the advantage over other professional workers disappeared in 2017, with other professionals catching up with teachers’ high security.

Working Time Quality

Finally, Table 1d confirms that there is no major long-term trend in working hours for full-time teachers, consistent with Allen et al (2019). The table also brings in other aspects of Working Time Quality. Scheduling Rigidity, where employees have no control over the start and finish of work times, is much worse for teachers than for other professionals in all three waves for which there is information. There is no significant trend either in the extent of teachers’ Scheduling Rigidity or in the gap with Other Professionals. Then, in the final two waves SES also covered another negative aspect of time quality. Time-Off Rigidity is also much worse for teachers than for Other Professionals, and has increased in recent years: by 2017 over three quarters (78%) report that it is somewhat or very difficult to take time off during work hours, up from 63% in 2012.

4b Trends in Teachers’ Work-Related Well-being

Table 2 shows the trends in teachers’ work-related well-being. There is a long-term decline in teachers’ well-being according to two of the measures. According to the Contentment indicator, there was no significant change in well-being over 2001 to 2012, but it fell considerably between 2012 and 2017. The fall in this period ( -0.249) amounts to 0.41 of the standard deviation of the Contentment index. In the same interval, the proportion often or always coming home from work exhausted rose from 75% to 84%, after having oscillated in earlier years. In every year, teachers’ Contentment score is substantively below that of other professional occupations; and a very much higher proportion of teachers, than of other professionals, report coming home from work exhausted: by 2017 the gap had reached 85% versus 45%).

There is more of a distinct U-shape in the trajectory of the Enthusiasm scale among teachers. There was a considerable, statistically significant rise, from 2001 to 2006 (+0.26, i.e. 0.28 of its standard deviation). This scale remained high in 2012 but collapsed between then and 2017 by 0.28.

To what extent can the dynamic path of well-being be accounted for by changes in job quality? There are no instrumental variables in the data with which to set up a causal model of well-being. Nevertheless, in the light of previous literature linking job quality to well-being (noted above) it is of interest to model how strongly the cross-sectional variation of each well-being indicator is associated with cross-sectional variation in job quality, and then ask whether the change in well-being over time can be predicted from the change in job quality, under the naïve assumption that the cross-sectional association is an unbiased estimate of the time-series association. This conventional method of decomposition is presented in Table 3. For each scale, the coefficient in the first model gives the raw change and its significance of each part of the U-shape: the rise between 2001 and 2006, then the fall between 2006 and 2017. The second model shows the same coefficient, after controlling for job quality. Comparing the coefficient between the two columns shows the extent to which the well-being change is accounted for. The full regression estimates are presented in the online Appendix.

The table shows that, in each case, more than half of the decline in well-being between 2006 and 2017 is accounted for by associated decreases in job quality. For example, the Contentment of teachers fell by 0.25, but if hypothetically there had been no change in job quality the associated decline in well-being is only 0.06 and statistically insignificant. More detailed analysis shows that the job quality variables that are most strongly associated with the declines in well-being are the Required Work Intensity index, Task Discretion and their interaction: together, comparing the 2006 and 2017 levels, these variables account for 30% of the decline in Enthusiasm, 78% of the decline in Contentment and 37% of the rise in the proportion of teachers returning home exhausted at the end of the day.[6] In contrast, working hours contributed little to the changes in well-being.

The earlier rise in Enthusiasm between 2001 and 2006 is not accounted for by any of the observed measures. Whatever may have caused this increase, it is not something that observably varied between individuals.

4c. Two Extensions: Differences by Sector and by Nation, and Facets of Job Satisfaction

To this point I have treated all teachers together. Given the differential education spending and policies between Britain’s nations, and the resource gaps between the private and state sectors, it is of interest to test whether job quality differs significantly along these two dimensions. I do this by pooling waves. However, even with the clustered random sampling methodology and some oversampling of the smaller nations in some years, the underlying sample numbers are quite low, implying that only substantial differences in the underlying populations could expect to be detected with sufficient statistical power.

First, teachers in Scotland report much lower work intensity than elsewhere in Britain (by 0.39 on the Work Intensity Index, equal to 0.34 of a standard deviation) and higher job security. Second, private school teachers systematically report lower levels of work intensity (by 0.33 of a standard deviation); in contrast, the training that they receive is self-reported to be of lower quality. There are no other, statistically significant observed job quality differences between the sectors and nations. Moreover, no differences in the trends between nations or sectors are detected.

Corroborative evidence of the importance of work intensity and discretion is found in respondents’ job satisfaction about multiple job facets (see Appendix). Compared with other professional occupations, teachers are more satisfied with their promotion prospects, their job security, the opportunity to use their abilities, the ability and efficiency of management and the variety in their work; but they are much less satisfied with the amount of work they have to do and with the hours work and with their influence over their jobs. Over time, teachers in the years subsequent to 2006 reported job satisfaction reductions in nine facets out of fourteen, with other facets unchanged. The facets of job satisfaction that fell most were in the ‘hours worked’ and in ‘the amount of work’. Meanwhile, the proportion of teachers expressing a desire for much more influence over how they do their job rose from 15% to 24%.

5. Discussion

These findings contribute a broader general characterisation of the problem concerning the job quality of teachers in one country, Britain, beyond just pay and hours. On the positive side, the prospects of promotion for teachers increased notably, so that by 2012 the chances of being promoted were as high as for other professional occupations; while the job security of teachers has remained high throughout. Consistent with Allen et al. (2019), I also confirm that teachers’ job quality has not shown a deterioration in terms of hours worked. Nor has there been a decline in real pay, taken over the whole sweep of available SES data since 1997. Real pay  rose to its highest in the 2006 wave.

Between then and 2017, however, real pay declined, somewhat faster than the pay of other professionals. This contrast with the study by Bryson and Forth (2017), which found no significant downward pay trend between 2005 and 2015, might be accounted for by methodological differences. Their analysis is based on the government’s Annual Survey of Hours and Earnings, which uses formal working hours, and only covers teachers within the remit of Pay Review Boards (excluding private school teachers and academy school teachers (potentially including some school managers); the pay data are employer-reported with a much larger sample.

The most striking aspect of job quality revealed by the SES data is teachers’ high work intensity and high work intensification. Compared to other professional workers and all other occupations, teachers work more intensively during their work hours, and their work intensity has risen to unprecedented levels. The issue is highlighted by the following stark statistic: nine out of ten teachers in 2017 “strongly agreed” with the statement that their job “required (them) to work very hard”. That compares with only a half of workers in other professional jobs, and with 44 percent for the whole population; over time it compares with only 54 percent of teachers back in 1992.

The SES data also reveals, for the first time, declines in other key elements of job quality: their influence over what they have to do, their methods and the quality standards to which they must conform fell sharply in the 1990s and then again between 2012 and 2017; their organisational participation has fallen away remarkably since the start of the 1990s; their continuing training participation has declined in the years since 2006 when SES began tracking it consistently, and the quality of that training did not make up for the declining quantity. Their working time quality became, if anything, also slightly worse, with decreased flexibility with respect to time off for dealing with personal or family matters.

These changes in job quality – especially the work intensification in combination with the decline in task discretion – account in part for some of the changes observed in teachers’ well-being. During the early 2000s an increase was recorded according to Warr’s Enthusiasm scale, but particularly since 2012 there has been a fall in well-being according to both of Warr’s scales. The proportion reporting that they come home from work exhausted also reached an unprecedented level (85%) in 2017. Meanwhile, most facets of teacher job satisfaction have fallen since 2006,  a deterioration consistent with the downward trend in lower-secondary teachers’ job satisfaction between 2013 and 2018 recorded by Jerrim and Sims (2019).

The limitations of these findings should be noted. For a start, physical working conditions and social support are not covered in SES, while other domains of job quality are not fully covered. The SES sample sizes are large enough to determine trends for teachers as a whole, but not for detailed breakdowns concerning trends in sectors, levels and localities. A further limitation is that the modelling of the connection between job quality and well-being is cross-sectional, as is typical of standard decomposition techniques; the ‘accounting’ for change derives from these associations, rather than from causal modelling of how job design affects well-being. Finally, while the relevance of the declining job quality for retention can be imputed from theory and from cited research, the SES affords no new direct evidence surrounding potential links between job quality and teacher retention.

My findings do not provide direct evidence about reasons for the decline in teachers’ job quality; nor have I attempted to unpack the factors constituting work intensification for teachers. One external factor – the overall spending per pupil – has been advanced in previous research as an explanation for job quality decline among teachers in Tasmania (Easthope and Easthope, 2000). In a similar vein, it is suggestive that the period of relative stability in many aspects of job quality and of well-being in the 2000s coincided with rising state education expenditures, while the strongest work intensification and the steepest declines in, for example, task discretion and organisational participation occurred during periods of declining spending. One mechanism could simply be that, with less to spend, each teacher has more pupils to serve in multiple ways. Another could be that the funding squeeze has been accompanied by a growing culture of accountability and performance management (Perryman and Calvert, 2020; Allen and Sims, 2018), experienced as greater work pressure and as a loss of opportunities for creative expression.

The potential role of internal factors, including school management, is recognised by the response of government to the teacher retention problem in recent years. The issue of ‘workload’ is prominent in this discourse. Workload is the totality of the tasks to be performed in a job, hard to measure generically since it depends on the specifics of the job. Unfortunately workload is commonly but inaccurately proxied by working hours. The workload and the allocated work time together determine the work intensity, which comprises one of the intrinsic domains of job quality. In October 2014 England’s Department for Education (DfE) launched a ‘Workload Challenge’. Subsequent reports on marking, planning and resources and data management were produced, with policy recommendations for workload reduction. Yet the response was low (CooperGibson, 2018, p.48) and the DfE announced a strategy to support teacher retention (DfE, 2018b): an intention to simplify the system of accountability for headteachers with the aim of reducing teachers’ workloads, an increase in early career professional development, pay incentives, and an encouragement to headteachers to allow more flexible working (part-time and off-site working) in order to promote improved working time quality.

Like many other professions, teachers’ jobs have been seriously disrupted by the 2020 pandemic, including school closures and home-working. Any improvement in teachers’ job quality that can be achieved in a post-COVID-19 environment should be beneficial, not only from the perspective of teachers, but also for schools and the pupils whose education depends so much on the quality of teaching. Just as, more widely, improving job quality can improve productivity (Irvine, 2020), so also teachers with better jobs may become better teachers. What my findings add to the discourse surrounding teacher dynamics is that key to success for any improvement strategy will be whether reforms can reverse the tide of falling autonomy in teachers’ daily tasks, their decreasing sense of participation in a school’s decision-making and, above all, the intensification of their work.

References

Allen, R., A. Benhenda, J. Jerrim and S. Sims (September 2019). New evidence on teachers’ working hours in England. An empirical analysis of four datasets. UCL Institute of Education.

Allen, R. and S. Sims (2018) The Teacher Gap, Abingdon, Routledge.

Bakker, A. B. and E. Demerouti (2007). “The Job Demands-Resources model: state of the art.” Journal of Managerial Psychology 22(3): 309-328.

Ballet, K. and G. Kelchtermans (2009). “Struggling with workload: Primary teachers’ experience of intensification.”Teaching and Teacher Education 25(8): 1150-1157.

Barmby, P. (2006). Improving teacher recruitment and retention: the importance of workload and pupil behaviour. Educational Research, 48(3), 247-265. 

Belfield, C. and L. Sibieta (2016). Long-Run Trends in School Spending in England. London, Institute for Fiscal Studies, IFS Report R115.

Brown, A., A. Charlwood and D. A. Spencer (2012). “Not all that it might seem: why job satisfaction is worth studying despite it being a poor summary measure of job quality.” Work Employment and Society 26(6): 1007-1018.

Bryson, A. and J. Forth (2017). Wage Growth in Pay Review Body Occupations. London, Report to the Office of Manpower Economics.

Bryson, A., Stokes, L. & Wilkinson, D. (2019) Who is better off? Wellbeing and commitment among staff in schools and elsewhere, Education Economics, 27:5, 488-506

CooperGibson Research (2018). Exploring teacher workload: qualitative research. London, DfE.

DfE (2018a) “Analysis of Teacher supply, retention and mobility”. DfE September 2018.

DfE (2018b). Teacher Recruitment and Retention Strategy. DfE 2018.

DfE (2018c) Trends in School Spending: 2002-03 to 2016-17. August 2018.

Easthope, C. and G. Easthope (2000). “Intensification, extension and complexity of teachers’ workload.” British Journal of Sociology of Education 21(1): 43-58.

Eurofound (2019). Working Conditions and Workers’ Health. Luxembourg, Publications Office of the European Union.

European Parliament (2014-2019). European Parliament Resolution of 13 September 2016 on creating labour market conditions favourable for work-life balance (2016/2017(INI)).

Felstead, A., Gallie, D., Green, F., Henseke, G. (2019). Skills and Employment Surveys Series Dataset, 1986, 1992, 1997, 2001, 2006, 2012 and 2017. [data collection]. UK Data Service. SN: 8589, http://doi.org/10.5255/UKDA-SN-8589-1

Ferguson, K., L. Frost and D. Hall (2012). “Predicting Teacher Anxiety, Depression, and Job Satisfaction.” Journal of Teaching and Learning 8(1).

Foster, D. (2019). Teacher Recruitment and Retention in England. London, House of Commons Library, Briefing Paper 7222.

Gallie, D., A. Felstead and F. Green (2004). “Changing patterns of task discretion in Britain.” Work Employment and Society 18(2): 243-266.

Green, F. (2001). “It’s been a hard day’s night: The concentration and intensification of work in late twentieth-century Britain.”British Journal of Industrial Relations 39(1): 53-80.

Green, F. (2006). Demanding Work. The Paradox of Job Quality in the Affluent Economy. Woodstock, Princeton University Press.

Green, F. (2010). “Well-being, job satisfaction and labour mobility.”Labour Economics 17(6): 897-903.

Green, F. and T. Mostafa (2012) Trends in Job Quality in Europe. A Report Based On The Fifth European Working Conditions Survey. Luxembourg: Publications Office of the European Union.

Hendricks, M. D. (2014). “Does it pay to pay teachers more? Evidence from Texas.” Journal of Public Economics 109(January): 50-63.

Higton, J., S. Leonardi, N. Richards, A. Choudhoury, N. Sofroniou and D. Owen (2017). Teacher Workload Survey  2016. London, DfE.

Hillary, J., Andrade, J. and Worth, J. (2018). Teacher Retention and Turnover Research. Research Update 4: How Do Teachers Compare To Nurses And Police Officers? Slough: NFER.

Huyghebaert, T., N. Gillet, N. Beltou, F. Tellier and E. Fouquereau (2018). “Effects of workload on teachers’ functioning: A moderated mediation model including sleeping problems and overcommitment.”Stress and Health 34(5): 601-611.

Irvine, G., Ed. (2020). Can Good Work Solve the Productivity Puzzle? London, Carnegie Trust and RSA.

Jerrim, J. and Sims, S. (2019) The Teaching and Learning International Survey (TALIS) 2018, London, DFE.

Karasek, R. A. (1979). “Job demands, job decision latitude, and mental strain: Implications for job design.” Administrative Science Quarterly 24: 285-308.

Lloyd, C. and J. Payne (2012). “Delivering better forms of work organization: Comparing vocational teachers in England, Wales and Norway”. Economic and Industrial Democracy 33(1): 29-49.

Lynch, S., Worth, J., Bamford, S. and Wespieser, K. (2016). Engaging Teachers: NFER Analysis of Teacher Retention Slough: NFER.

Muñoz de Bustillo, R., E. Fernández-Macías, J.-I. Antón and F. Esteve (2011). Measuring More than Money:The Social Economics of Job Quality. Cheltenham, Edward Elgar.

OECD (2014), “How good is your job? Measuring and assessing job quality”, Chapter 3, OECD Employment Outlook 2014, OECD Publishing, Paris.

Perryman, J. and G. Calvert (2020). “What motivates people to teach, and why do they leave? Accountability, performativity and teacher retention.” British Journal of Educational Studies. 68 (1): 3-23. DOI: 10.1080/00071005.2019.1589417

Sellen, P. (2016) Teacher workload and professional development in England’s secondary schools: insights from TALIS. Education Policy Institute. London.

Siegrist, J. (1996). “Adverse Health Effects of High Effort/Low Reward Conditions.” Journal of Occupational Health Psychology 1(1): 27-41.

Sims, S. (2017). What Happens When You Pay Shortage-Subject Teachers More Money? Simulating The Effect Of Early-C Areer Salary Supplements On Teacher Supply In England. London, The Gatsby Charitable Foundation.

Sims, S. (2020). “Modelling the relationships between teacher working conditions, job satisfaction and workplace mobility.” British Educational Research Journal 64 (2): 301-320.

Torres, A. C. (2016). “Is This Work Sustainable? Teacher Turnover and Perceptions of Workload in Charter Management Organizations.”Urban Education 51(8): 891-914.

Wang, F., K. Pollock and C. Hauseman (2018). “School Principals’ Job Satisfaction: The Effects of Work Intensification.” Canadian Journal of Educational Administration and Policy(185): 73-90.

Warr, P. (1990). The measurement of well-being and other aspects of mental health. Journal of Occupational Psychology, 63: 193-210.

Warr, P. (2007). Work, Happiness, and Unhappiness. London: Lawrence Erlbaum Associates.

Webb, R., G. Vulliamy, S. Hämäläinen, A. Sarja, E. Kimonen and R. Nevalainen (2004) Pressures, rewards and teacher retention: a comparative study of primary teaching in England and Finland, Scandinavian Journal of Educational Research, 48:2, 169-188.

Worth, J., Lynch, S., Hillary, J., Rennie, C. and Andrade, J. (2018). Teacher Workforce Dynamics in England. Slough: NFER.

Worth, J. and Van den Brande, J. (2019). Teacher Labour Market in England: Annual Report 2019. Slough: NFER.

Table 1a.   Work Intensity, 1992-2017

Strongly agree “job requires I work very hard” (% of jobs)Work involves working at very high speed at least ¾ of the time    (% of jobs)Work involves working to tight deadlines at least ¾ of the time    (% of jobs)Work Intensity Index
1992
Teacher53.616.1nana
Other Professional35.99.8nana
1997
Teacher77.1nanana
Other Professional43.0nanana
2001
Teacher82.550.467.20.60
Other Professional42.634.059.60.05
2006
Teacher75.851.056.90.48
Other Professional42.030.864.30.01
2012
Teacher79.741.773.30.57
Other Professional50.835.864.20.23
2017
Teacher89.757.973.71.05
Other Professional51.835.359.70.18
     
Annual trend for teachersϮ+ 0.15 **+ 0.24 **+ 0.12 **+ 0.13 **
n829734628618

Base: those in work aged 20-60, all waves with available data pooled.

Ϯ Beta coefficient on year when underlying variable regressed with weights on year, sex, age, age squared, education stage, hours of work, among teachers; significance of year coefficient: * p < 0.05, ** p < 0.01.

Table 1b.   Skills and Discretion, 1992-2017

Task DiscretionOrganisational Participationa  (% of jobs)Training Participation in Past Year (% of jobs)Training Qualityb
1992 
Teacher2.6845.2 
Other Professional2.5045.8 
1997 
Teacher2.48  
Other Professional2.50  
2001  
Teacher2.2240.2  
Other Professional2.2942.7  
2006  
Teacher2.2731.091.640.6
Other Professional2.3434.581.236.0
2012  
Teacher2.2630.288.538.6
Other Professional2.2924.879.036.5
2017  
Teacher2.1320.186.031.4
Other Professional2.2837.883.348.7
     
Annual trend for teachersϮ– 0.23 **– 0.17 **-.0.12 *– 0.06
n829734460407

Base: those in work aged 20-60, all waves with available data pooled.

a. “Quite a lot”, or “a great deal” of say in changes affecting how job is done

b. For those participating, percent for whom the most recent training episode “improved skills a lot”.

Ϯ Beta coefficient on year when underlying variable regressed with weights on year, sex, age, age squared, education stage, hours of work, among teachers; significance of year coefficient: * p < 0.05, ** p < 0.01.

Table 1c   Pay and Prospects

Hourly Pay (2015 £)High Job Securitya (% of workers)High Promotion Expectationb (% of workers)
1992
Teacher15.2
Other Professional34.3
1997
Teacher14.781
Other Professional19.367.2
2001
Teacher17.791.8
Other Professional20.882.4
2006
Teacher19.689.124.8
Other Professional24.582.730.9
2012
Teacher16.679.732.5
Other Professional24.670.631
2017
Teacher15.485.7
Other Professional21.688.3
    
Annual trend for teachersϮ+ 0.01+ 0.02+ 0.05 *
n625704439

Base: those in work aged 20-60, all waves with available data pooled.

a. no chance, or very unlikely to lose job and become unemployed in the coming year.

b. at least a 75% chance of being promoted in the next 5 years in present organisation.

Ϯ Beta coefficient on year when underlying variable regressed with weights on year, sex, age, age squared, education stage, hours of work, among teachers; significance of year coefficient: * p < 0.05, ** p < 0.01.

Table 1d  Working Time Quality Indicators

Usual weekly hours incl. overtime (full-time workers)Schedule Rigiditya (% of workers)Time-Off Rigidityb (% of workers)
1997 
Teacher45.9
Other Professional45.1
2001 
Teacher48.2
Other Professional45.5
2006 
Teacher45.378.9
Other Professional41.437.1
2012 
Teacher45.181.263.4
Other Professional41.330.217.3
2017 
Teacher46.274.678.1
Other Professional41.330.020.1
    
Annual trend for teachersϮ– 0.03– 0.03+ 0.04 *
n551459229

Base: those in work aged 20-60, all waves with available data pooled.

a. Disagree or strongly disagree can decide start and finish times

b. Somewhat or very difficult to take time off during work hours for family/personal matters

Ϯ Beta coefficient on year when underlying variable regressed with weights on year, sex, age, age squared, education stage, among teachers; significance of year coefficient: * p < 0.05, ** p < 0.01.

Table 2  Work-Related Well-Being

 EnthusiasmContentmentOften or always comes home from work exhausted (% of workers)
1997
Teacher72.0
Other Professional43.7
2001
Teacher0.050-0.43477.3
Other Professional-0.009-0.23645.5
2006
Teacher0.265-0.38672.2
Other Professional0.104-0.08137.0
2012
Teacher0.262-0.40975.3
Other Professional-0.014-0.3340.2
2017
Teacher-0.019-0.65884.9
Other Professional-0.043-0.19945.0
    
Annual trend for teachersϮ– 0.02– 0.09 *+ 0.05
n630630715

Base: those in work aged 20-60, all waves with available data pooled.

Ϯ Beta coefficient on year when underlying variable regressed with weights on year, sex, age, age squared, education stage, hours of work, among teachers; significance of year coefficient: * p < 0.05, ** p < 0.01..

Table 3 Accounting for Teachers’ Well-being Changes, 2001 to 2006, and 2006 to 2017.

 (1)(2)(3)(4)(5)(6)
Period of changeEnthusiasm ScaleEnthusiasm ScaleContentment ScaleContentment ScaleHigh ExhaustionHigh Exhaustion
2001 to 20060.226*
(2.32)
0.266**
(2.95)
0.0530
(0.53)
0.0486
(0.53)
– 0.0498
(- 1.15)
– 0.0268
(- 0.66)
2006 to 2017-0.225*
(-2.14)
-0.109
(-1.11)
-0.248*
(-2.30)
-0.0642
(-0.64)
0.121**
(2.64)
0.0546
(1.24)
Job QualityNoYesNoYesNoYes
n601601601601618618

t statistics in parentheses: * p < 0.05, ** p < 0.01

Coefficients are the estimated change over each period. The job quality variables included in models (2), (4) and (6) for each period of change are: Work Intensity Index, Task Discretion Index, the interaction between work intensity and task discretion, organisational participation, high job security, working hours. All models control for age and gender. Full regression estimates are shown in the Appendix.


[1]  https://www.ifs.org.uk/election/2019/article/school-spending. A DfE report shows expenditure flatlining in the 2010s, rather than declining, with spending per pupil at the same level in 2016-17 as it had been in 2011-12 (DfE, 2018c).

[2] https://sibietaeconed.files.wordpress.com/2019/04/school-spending-in-wales-090419-2.pdf

[3] National statistics, School Workforce in England 2011 https://www.gov.uk/government/statistics/school-workforce-in-england-november-2011; National Statistics, School Workforce in England 2018 https://www.gov.uk/government/statistics/school-workforce-in-england-november-2018 .

[4]https://statswales.gov.wales/Catalogue/Education-and-Skills/Schools-and-Teachers/Schools-Censushttps://gov.wales/sites/default/files/statistics-and-research/2019-05/school-census-results-2014.pdf; https://gov.wales/schools-census-results-january-2019.

[5] https://www2.gov.scot/Topics/Statistics/Browse/School-Education/Summarystatsforschools. Note that Scotland’s education system has long been structurally distinct from that of England and Wales, and that since the time of New Labour the disposition of education spending has been the responsibility of the devolved national administrations of Scotland and Wales.

[6] These percentages are computed using a linear Oaxaca decomposition analysis, pooling the 2006 and 2017 waves.