Should disadvantage be reduced to a single number?

22nd June 2023 by Timo Hannay [link]

Update 22nd June 2023: See also this associated commentary in Schools Week.

Spoiler alert: this post will answer the question posed in the title with a "no". We suggest a more nuanced approach that takes into account the distribution of socioeconomic deprivation in each school, the different types of deprivation that exist and the wide variety of local contexts in which superficially similar schools can find themselves.

To summarise:

• The most common measure of in-school disadvantage is the proportion of pupils who are eligible for the Pupil Premium (roughly equivalent to the proportion who are eligible for free school meals, or FSM). However, even schools with identical or near-identical values for this metric can exist in widely divergent local contexts with very different socioeconomic characteristics.
• One might argue that this is at least partly because school populations do not necessarily reflect those of their surrounding neighbourhoods. But analyses of the deprivation levels of pupils' own home postcodes give similar results. In other words, variations in local context are also reflected in each school's pupil population, even for schools that notionally experience the same level of socioeconomic disadvantage.
• Moreover, deprivation comes in a variety of forms. In addition to income deprivation, on which Pupil Premium and FSM measures are based, the UK government also provides deprivation metrics for crime, health, housing and the environment, among other factors. These can also be relevant to schools and often correlate only weakly, if at all, with the more commonly used income-based measures, so deserve separate recognition.

Dimensions of deprivation

Before diving into the data, it might be useful to review the different types of deprivation recognised by the government – specifically, the Department for Levelling Up, Housing and Communities. Their most general deprivation indicator is the Index of Multiple Deprivation (IMD), which combines the following seven components (with weightings in brackets):

• Income (22.5%): The proportion of people experiencing deprivation relating to low income. There are also separate sub-indicators for income deprivation affecting children (IDACI) and that affecting older people (IDAOPI). The IDACI is similar to the FSM measure used in schools, but also includes children of pre-school age.
• Employment (22.5%): The proportion of working-age people involuntarily excluded from the labour market.
• Education (13.5%): The lack of attainment and skills in the local population.
• Health (13.5%): The risk of premature death and impaired quality of life through poor physical or mental health.
• Crime (9.3%): The risk of personal and material victimisation.
• Housing (9.3%): The physical and financial accessibility of housing and local services.
• Environment (9.3%): The quality of the indoor and outdoor local environment.

(For further details, see this useful infographic as well as the other documents provided on this page.)

In addition, the Office for Students creates an indicator called POLAR4, which is a measure of the higher-education participation rate among young people. It is not a component of the IMD, and it is a measure of advantage rather than disadvantage (in the sense that high values are good), but is also an interesting metric, so we include it here.

Figure 1 is a correlation matrix for all 11 of these indicators, showing how they relate to one another statistically. Positive correlations are in blue and negative ones in red, while the size and intensity of the dots correspond to their strength. (The blue diagonal is simply a consequence of the fact that each indicator correlates perfectly with itself.)

Unsurprisingly, the IMD itself correlates most strongly with it's major components: income, employment, education and health. It also correlates with crime. And of course the various income measures correlate with each other. Housing and environmental deprivation hardly correlate with anything else at all, including each other. And for reasons already mentioned, POLAR4 (participation in higher education) correlates negatively with most other factors. Hover over the dots in the figure to see actual correlation coefficients (which range from 1 for perfect correlations to -1 for perfect inverse correlations, and 0 for uncorrelated factors.)

Figure 2 shows how different local areas (specifically, Lower layer Super Output Areas, or LSOAs) are distributed with respect to these different indicators. Looking at 'Income' against 'IMD' for the 'National sample' (a random selection of about 10% of all LSOAs, just for ease of display), it is obvious that the correlation is very strong. Nevertheless, for any given IMD value there is still considerable variation in the Income metric. This is even more true for 'Health', 'Crime', 'Housing' and 'Higher education'. There are also some big regional differences. For example, see the distribution of 'Housing' deprivation in the North East compared to London.

(Use the menus below to explore other correlations and hover over the dots to see corresponding values. Click on the figure legend to turn on or off individual regions. You can also click on the dots in the correlation matrix in Figure 1 above to view the corresponding plot in Figure 2 below.)

This still leaves many other measures of deprivation, from the Social Mobility Index and wellbeing to travel times and house prices. We have looked at some of these in the past (eg, here and here) and hope to do more with them in the future. But for our present purposes the above list seems like plenty to be getting on with.

Same difference

What of schools? Figure 3 shows a scatter plot composed of mainstream state primary schools in England, with the percentage of pupils who are eligible for the Pupil Premium (PP) on the horizontal axis and a variety of local deprivation measures on the vertical axis. IMD shows a positive relationship with PP eligibility: schools with higher proportions of PP children also tend to be in neighbourhoods with higher IMD scores. But the correlation is far from perfect: schools with the same PP proportion are often in neighbourhoods with very different IMD scores. (For the statistically minded, the coefficient of determination, known as r², is 0.38.)

The situation becomes even more interesting if we look separately at the components of the IMD:

• Income deprivation (r²=0.40), especially that relating to families with children (r²=0.44), shows a relatively clear positive correlation. So does employment (r²=0.37).
• But the correlations for health (r²=0.30), education (r²=0.29) and crime (r²=0.26) are weaker.
• And those for housing and the environment (both r²=0.01) are essentially absent altogether.

POLAR4 also correlates with PP, albeit negatively (ie, it tends to be lower in the neighbourhoods of schools with higher proportions of PP children). But once again the relationship is far from clear-cut (r²=0.14).

(Use the menu below to select a local deprivation measure. Hover over the graph to see individual school details; hover over the line to see linear regression statistics.)

Figure 4 shows the same analysis for secondary schools. Once again, correlations are modest for the IMD and income deprivation (both r²=0.35), and a bit clearer when limited to families with children (r²=0.38), but weaker for employment (r²=0.26), crime (r²=0.25), health (r²=0.23), education (r²=0.18), the environment (r²=0.12) and housing (r²=0.02). POLAR4 shows a very weak negative correlation (r²=0.04).

(Use the menu below to select a local deprivation measure. Hover over the graph to see individual school details; hover over the line to see linear regression statistics.)

Identically individual

So much for the national picture, what does this variation look like from the point of view of individual schools? Figure 5 shows all the mainstream state primary schools that reported exactly 20.0% of their children as eligible for the Pupil Premium in the latest available data (2022-23), which is about average¹. From the simplistic point of view of PP measures, these schools are not merely similar but literally indistinguishable.

The horizontal bars show the breakdown of each school's local area into deprivation bands, with Band A (red) corresponding to the most deprived 2.5% of all postcodes in England, Band B (orange) to the next 5% and so on. Band G (purple) corresponds to the least deprived 62.5% of all postcodes in England. The bar at the top labelled 'ENGLAND' shows these national bands for reference.

The first thing to notice is that these 'identically' disadvantaged schools exist in a wide range of neighbourhoods, from those with very high proportions of poor households (top) to those with very few (bottom). The second thing to notice is that the ordering depends greatly on which measure of deprivation you choose. Compare, for example, income deprivation affecting children with employment, education, health, crime, housing the environment and participation in higher education. It is true that some schools tend to come near the top of most of these lists and others more often appear close to the bottom, but plenty of schools show very different rankings depending on the metric chosen.

(Use the menu below to select a local deprivation measure. Hover over the graph to see individual school details.)

Figure 6 shows a similar analysis for secondary schools. In this case we have slightly extended the range of PP values: it is no longer exactly 20.0%, but anywhere from 19.7% to 20.3%². This is because there are fewer secondary schools and the extra wiggle room provides us with a suitable number to analyse; in the overall scheme of things, they remain almost identical on this measure. Once again we see a wide range of local contexts and once again the ordering of schools varies according to which deprivation metric we choose: income deprivation affecting children, employment, education, health, crime, housing the environment and participation in higher education.

(Use the menu below to select a local deprivation measure. Hover over the graph to see individual school details.)

Pupil diversity

All of this shows that schools deemed similar, or even identical, on the PP measure can exist in very different local contexts. But what about the pupil populations of the schools themselves? These are not necessarily representative of their local populations and so might vary much less. Fortunately we can get a handle on this too. One source of school funding is calculated based on the IDACI band of each pupil's home postcode, so by looking at the amounts of money allocated to a school we can infer the number of pupils in each band³.

Figure 7 shows this analysis for the same group of primary schools we saw in Figure 5 above (omitting eight schools for which we could not infer the relevant pupil information). Here, too, we see a wide range of band distributions across schools that are notionally identical with respect to PP eligibility. In other words, there is a wide variation in the socioeconomic mix of the neighbourhoods in which these pupils live.

Figure 8 shows the same analysis for the secondary schools that we saw in Figure 6 above (omitting one school for which we could not infer the relevant pupil information). Once again, we see a wide range of socioeconomic contexts for the home addresses of pupils attending each school.

Making a difference

Even if the deprivation profiles of these schools differ, does this have any practical consequences? We believe it does. Educating children in areas with poor health outcomes is – or at least should be – different from doing so in areas where health is generally good. The same goes for crime, the environment, participation in higher education and so on. Each of these tells us something different, and they are all distinct from the school's PP measure.

To take one simple example, Figure 9 shows the relationship between PP and various post-18 destination measures. There is a correlation (albeit an imperfect one) with any sustained destination⁴ (ie, not NEET), but essentially none for the proportions of pupils continuing in education, going on to higher education or entering into apprenticeships.

(Use the menu below to select a destinations measure. Hover over the graph to see individual school details; hover over the line to see linear regression statistics.)

Figure 10 shows the same post-school destinations data against the POLAR4 measure of participation in higher education. In contrast to PP, there is no significant correlation with the proportions of students going on to any sustained destination, but there are (admittedly weak) correlations with the proportions continuing in education, going on to higher education or entering into apprenticeships. True, the local POLAR4 measure is composed in part of data from students who attended the very school in question, so these effects are somewhat tautological. But not entirely so: in most cases they incorporate leavers from many other local schools. Also, POLAR4 data refer to students from about 10-15 years ago, so they represent the longer-term characteristics of each local area, not recent outcomes from individual schools.

The bottom line is that these measures generally tell us something different to the PP metric. POLAR4 is a case that relates directly to educational outcomes that are already measured (ie, post-school destinations). Other metrics such as health, environment and crime are more likely to relate to outcomes that are not currently measured, at least not in any routine or systematic way, such wellbeing. Perhaps they should be. One might argue that if they don't correlate with exam results (which, on the whole, they don't, at least not to the degree that income measures do) then they shouldn't really matter to schools or to anyone else concerned with education. With apologies for the pun, that would to take an extremely impoverished view of what it takes to raise a child.

Equal but not the same

By reducing school-level disadvantage to a one-dimensional ranking, we risk overlooking the different forms of deprivation that children experience, and the qualitatively distinct social contexts in which schools operate. For example, schools might need to act differently depending on whether the most prevalent forms of deprivation in their area relate to crime, housing, health, the environment or participation in higher education. In addition, those (including us) who might compare one school against another superficially similar one need to recognise that when disadvantage is boiled down to a single number it often simplifies away a range of contextual dissimilarities that might be important.

Even if we look only at factors relating to income, all-or-nothing threshold indicators like PP eligibility risk missing important nuance. For example, non-PP children living in areas of high income deprivation are surely more likely to be poor even if they don't technically qualify for PP funding. Conversely, the experiences of poor families living in relatively affluent neighbourhoods may be quite different to those who are surrounded by other families of similar socioeconomic status, necessitating different forms of support.

None of this is intended to diminish the importance of the Pupil Premium as an indictor of disadvantage, as a source of vital income for schools or as a means of restoring a semblance of educational equality, only to say that on its own it provides no more than a partial picture. Neither is it our purpose to try and spell out all of the educational and welfare implications of the different forms of deprivation described here. Rather, we are drawing attention to the fact that the world of social and educational disadvantage is more complicated than analysts like us usually allow. With Ofsted's one-word rating system coming under fire and SchoolDash's own consumer-facing service, The Schools Guide, offering an alternative to simplistic school league tables, perhaps it's time for all of us to more fully recognise the diversity, as well as the similarities, across our schools. After all, we have the data.

To that end, SchoolDash Insights now includes in its 'Schools' section (subscription or trial account required) contextual analyses of the local area and pupil population of each school. We hope that this will make it easier to appreciate the kinds of nuances described above, not just across the few dozen schools mentioned here, but for tens of thousands of schools across the country.

We welcome questions and comments – please write to us at [email protected]. And if you would like us to write to you when we have new analyses to share then please sign up for our free monthly-ish newsletter.

Footnotes:

1. Full list of primary schools analysed for Figures 5 and 7 (click to see corresponding SchoolDash profile): Abbey Primary School, All Saints CofE VC Infants School, Ashwater Primary School, Black Torrington Church of England Primary School, Blakeney Primary School, Burnopfield Primary School, Byfield School, Church Lench CofE First School, Cromford Church of England Primary School, Crosscanonby St John's CofE School, Defford-Cum-Besford CofE School, Duchy of Lancaster Methwold CofE Primary School, East Herrington Primary Academy, Elm Park Primary School, Fatfield Academy, Garvestone Community Primary School, Grange First School, Hemington Primary School, Hylands Primary School, Kielder Primary School and Nursery, Kirkby Thore School, Knowl Hill Church of England Primary Academy, Manor Infants' School/Manor Longbridge, Meadowside Primary School, Mereside Church of England Primary Academy, Micklands Primary School, Milton St John's CofE Primary School, Mousehole School, Newton Regis CofE Primary School, Our Lady Queen of Martyrs Catholic Primary School, Esh Winning, Prudhoe Castle First School, Red Row First School, Ridgeway Infant School, Riverside Junior School, Russell Street School, Sawtry Junior Academy, Sexton's Manor Community Primary School, Silloth Primary School, Spreyton School, St James Church of England Primary and Nursery School, St John the Evangelist Catholic Primary School, St Joseph's RC Primary School, St Keverne Primary School, St Michael's Church of England Academy, St Michael's CofE Primary School, Great Lever, St Monica's Catholic Primary School, St Patrick's Catholic Primary School, St Thomas's CofE Primary School, Steam Mills Primary School, Stogumber CofE Primary School, Stratton Primary School, Terrington St Clement Community School, The Cambridge Primary School, Thurlby Community Primary Academy, Tilstock CofE Primary and Nursery, Upholland Roby Mill CofE Voluntary Aided Primary School, Wellhouse Junior and Infant School, Weston Primary School, Whybridge Infant School, Willesborough Junior School, Wintringham Primary Academy, Woodbank Primary School, Woodside Academy, Wrightington Mossy Lea Primary School, Wyberton Primary Academy.
2. Full list of secondary schools analysed for Figures 6 and 8 (click to see corresponding SchoolDash profile): Abbeyfield School, All Saints Catholic Voluntary Academy, Anthony Gell School, Arnold Hill Spencer Academy, Bingley Grammar School, Bishop Fox's School, Blackminster Middle School, Bourne Academy, Bourne Community College, Budehaven Community School, Caedmon College Whitby, Carterton Community College, Church Stretton School, Coombe Girls' School, Corby Technical School, Emerson Park Academy, Epsom and Ewell High School, Exmouth Community College, Fairfax, Gildredge House, Golborne High School, Honiton Community College, Honley High School, King Edward VI Handsworth School, Kingstone High School, Lakelands Academy, Longcroft School and Sixth Form College, Mossley Hollins High School, Nunnery Wood High School, Oldfield School, Outwood Academy Valley, Penryn College, Preston Muslim Girls High School, Roding Valley High School, Rushey Mead Academy, Sir William Romney's School, Southborough High School, St Alban's Catholic High School, St Clement's High School, St Gabriel's RC High School, a Voluntary Academy, St John Payne Catholic School, Chelmsford, St Joseph's Catholic Middle School, Hexham, St Mark's West Essex Catholic School, St Mary's Church of England Middle School, St Peter's Church of England Middle School, St Wilfrid's Church of England Academy, Tauheedul Islam Boys' High School, Teign School, The Angmering School, The Bulmershe School, The Deanery CE Academy, The Douay Martyrs Catholic School, The Ferrers School, The Henry Box School, The Maplesden Noakes School, The Purbeck School, The Reach Free School, The Regis School, The Swan School, Valentines High School, West Hatch High School.
3. In practice it's not quite that straightforward. Although this funding line follows a national formula, some local authorities choose to allocate the funds differently, making the corresponding pupil numbers a bit harder to calculate for certain schools and effectively impossible for a few of them. The latter group is necessarily excluded from this analysis.
4. 'Any sustained destination' means that the student stayed in education or employment for at least two terms after leaving school, typically at age 18.

The longer-term impact of COVID-19 on pupil attainment and wellbeing

22nd June 2023 by Timo Hannay [link]

Update 27th June 2023: See also this coverage in Tes.

In a final-for-now instalment of our collaboration with Hodder Education, Nottingham Trent University and the Nuffield Foundation, we have been looked again at academic attainment and wellbeing among primary-school children before, during and since the pandemic.

The resulting white paper is out today. Key findings include:

• Children in all years of primary school remain approximately 2 months behind in grammar, punctuation and spelling (GPS)
• Children in Key Stage 1 are approximately 1 month behind in reading and maths
• Children in Key Stage 2 during autumn 2022 appear to have made up previous losses in reading and maths
• Schools with high levels of disadvantage showed the largest drops in attainment over the course of the pandemic-related school closures
• The disadvantage gap between children eligible for the Pupil Premium and their peers remained large for all subjects and year groups, and increased each autumn between 2020 and 2022 for Year 6 in reading and maths
• And much more...

Please read the full report here.

As always, we welcome your feedback and suggestions: [email protected]

Out of the ordinary: School recruiting in 2022/23

8th June 2023 by Timo Hannay [link]

Update 8th June 2023: See also this blog post from Teacher Tapp as well as coverage by Schools Week and Tes.

As England's education system emerges blinking from peak teacher-hiring season, welcome to another annual roundup of school recruiting activity. As before, we have teamed up with with the Gatsby Foundation, who have provided generous funding, and Teacher Tapp. Our approach combines SchoolDash's vacancies data (gathered nightly from school websites) with Teacher Tapp's insights into the opinions and intentions of teachers and school leaders (derived from their unique online surveys). This helps us to get a handle not only on what's happening, but also on the reasons why and on what we might expect next.

If the approach is familiar, the results bear much less similarity to previous years – in this as in so many other respects, we live in unusual times. We encourage you to read our full joint report. The analysis below is for readers who want to dive further into the school vacancies and headteacher appointments data.

To summarise the main findings presented here:

• The numbers of secondary school teacher vacancy adverts have been higher during the current academic year (2022/23) than last year (2021/22) or the most recent pre-pandemic year (2018/19). Comparing similar periods and across all subjects, there have so far been 12% more adverts than last year and 28% more than before the pandemic. The biggest rises were seen in Technology, Humanities and the Languages, which were up by around 40%-50% compared to pre-pandemic levels. All subject areas showed increases of more than 10%.
• The latest year-on-year increases have been driven in large part by higher activity at the start of the current school year – ie, in autumn and winter 2022 – which is usually a less active time of year. Activity this spring, which is typically high season, peaked somewhat earlier than usual, in late April rather than early May. This may have been in part because of the extra mid-May national holiday for the coronation. The last week of April 2023 saw the highest number of adverts observed since we began tracking them in 2017.
• There was considerable regional variation, with the North West and South West showing the largest proportional increases compared to before the pandemic, while the North East and Yorkshire and the Humber showed the smallest increases.
• The rise in the number of vacancy adverts for school technicians has been greater still. 2022/23 has so far seen an increase of 46% compared to the equivalent period during the most recent pre-pandemic year, though it was down 12% on last year's unprecedented highs.
• New headteacher appointments have also risen and are already higher for the 2022/23 academic year than for any recent year, even with a month still to go before the end of term and nearly three months before the official end of the school year.

To find out more, read on. Subscribers to SchoolDash Insights can also see the latest live vacancies and appointments data in the Recruitment and Headteachers sections.

Teacher turnover

Figure 1 provides a long-term perspective on teacher recruitment activity, showing the numbers of vacancy adverts found on school and college websites during each week from September 2017 to May 2023. (See Footnote 1 for a brief explanation of how the data were gathered.) Annual seasonal cycles are clearly visible, with two pre-pandemic years that showed very little variation, followed by two much less active mid-pandemic years, and now two post-pandemic years in which overall activity has been much higher than usual.

As well as showing all subjects together, the data have also been divided into eight broad subject areas, each of which show slightly different patterns: Arts, English, Humanities, Languages, Mathematics, Science, Technology and other subjects. We will explore these in more detail below.

(Use the menu below to select a subject area to view. Hover over the lines to see corresponding data values.)

Figure 2 shows the same data, but overlays each academic year using lines of different colours in order to make it easier to compare one year with another. In general, there is a repeating annual pattern in which activity is low during the autumn and winter, then rises after New Year, reaching a peak in April and May before declining again during the summer break. Until the COVID-19 pandemic, this seasonality in the weekly data had been extremely consistent from year to year, with 2018/19 (green line) providing a good representative example. However, in 2019/20 (blue line) the usual April-May peak was absent due to school closures in spring and summer 2020. Activity remained depressed in 2020/21 (red line), especially when schools were closed again in early 2021. Activity rebounded in 2021/22 (purple line), rising to higher-than-normal levels in spring 2022. This elevated activity was sustained through the 2022/23 academic year (black line), reaching a new weekly record in late April 2023.

Looking again at all years together, it is informative to compare them using cumulative data, which makes annual differences easier to see. Using 2018/19 (green) as a reference, there was clearly a big reduction in 2019/20 (blue), followed by the further decline in 2020/21 (red) and the bounce back to higher-than-normal levels in 2021/22 (purple). So far in 2022/23 (black), activity has been higher still and this looks set to become the most active year since we began tracking school recruitment in 2017/18.

It is also interesting to see differences by subject (again, using cumulative data across all years). So far this year, all subjects are running ahead of previous years, though on a proportional basis Technology, Humanities and Languages have shown the biggest increases relative to pre-pandemic norms (compare the green and black lines).

(Use the menus below to view weekly or cumulative data, and to select different subject areas. Click on the figure legend to hide or view individual academic years. Hover over the lines to see corresponding data values.)

Another way to view these changes, in both proportional and absolute terms, is provided in Figure 3. This summarises the changes seen so far this year compared to the equivalent periods last year (red columns) and in the most recent pre-pandemic year (blue). Looking at percentage changes for both years together, all subjects showed year-on-year increases in the range 8%-14%, but because 2021/22 was also an exceptional year, this amounted to increases relative to pre-pandemic levels of anywhere between 11% (for Mathematics) and 52% (for Technology).

In terms of absolute numbers of adverts, Science and Humanities showed the largest year-on-year increases (over 700 each; see red columns), while Humanities and Technology have shown the largest increases since before the pandemic (around 1,500-2,000 each; blue columns). In total, we have so far found about 4,000 more adverts than the during the same period last year and well over 8,000 more than in the most recent pre-pandemic year.

(Use the menu below to switch between percentage changes and changes in numbers of adverts. Hover over the columns to see corresponding data values and numbers of advertisements.)

Figure 4 shows the percentage change in numbers of adverts between 2022/23 (ie, the current academic year so far) and the equivalent period in 2018/19 (the most recent pre-pandemic year), broken down by type of school. Note that this omits positions advertised by multi-school federations or trusts because it is harder to associate these vacancies with specific schools. These kinds of advertisements constitute a small but steadily increasing proportion of the total, so the changes shown here will tend to understate increases (or, equivalently, overstate decreases) in overall activity. They are nevertheless helpful in determining relative trends between different school types and locations.

Across all subjects, two regions – the North East and Yorkshire and the Humber – showed the smallest changes, while the South West and the North West showed the largest increases. Schools located in more affluent areas tended to show greater increases, but so did those with poorer pupil populations and more children with low prior attainment, as well as non-selective and rural schools.

(Use the menus below to explore different school groups and subjects. Hover over the columns to see corresponding data values and numbers of advertisements.)

A SySTEMic view

Figure 5 takes a closer look at selected STEM subjects, which have been of particular interest to us over the years. We have already seen above that Mathematics didn't increase last year relative to pre-pandemic levels, but has increased this year. What about Science? Using the cumulative data in the figure below, we can view separately the trends for specialist science teachers in different disciplines with those for generalist science teachers. This shows that the numbers of adverts for Biology, Chemistry and Physics teachers grew last year relative to before the pandemic (compare the green and purple lines), but have not increased any further this year (black line). In contrast, generalist Science teacher adverts did not increase last year, but have grown this year. It is tempting to speculate that this recent shift towards recruiting generalist science teachers might be a sign that schools are experiencing difficultly, or at least lack of confidence, hiring specialist science teachers, but we're not able to confirm this from these data alone.

Figure 5 also separates out Computing (which in the analyses above is part of the booming Technology subject area). Using cumulative data again, we can see the huge proportional increase last year relative to pre-pandemic levels (green and purple lines), followed by a further increase in the current academic year (black line).

(Use the menus below to view weekly or cumulative data, and to select different subject areas. Click on the figure legend to hide or view individual academic years. Hover over the lines to see corresponding data values.)

Figure 6 quantifies these changes further. In proportional terms, the different types of science teacher position are up by between 12% and 20% relative to pre-pandemic levels (blue columns), but for specialist Biology, Chemistry and Physics positions almost none of this was due to increases in the current academic year (red columns), while for general Science teacher positions almost all of the change was due to growth seen this year. In contrast, Computing has shown a huge increase since before the pandemic (+68%), a considerable proportion of which (19 percentage points) has been due to a rise in recruiting activity during the current academic year.

In terms of absolute numbers of adverts, general Science positions and Computing positions have shown broadly comparable increases of around 750-850 additional adverts so far this year compared to the equivalent period before the pandemic, but the latter has risen from a much lower initial baseline level, hence the larger percentage increases.

(Use the menu below to switch between percentage changes and changes in numbers of adverts. Hover over the columns to see corresponding data values and numbers of advertisements.)

Primary concerns

In recent months we have also begun to track teacher vacancy adverts issued by primary schools. This work is conducted in collaboration with Teach First and the summary data below is reproduced with their kind permission. Although it does not extend back to before the pandemic, we do now have just over a 12 months' worth of data, allowing simple year-on-year comparisons, as shown in Figure 7. Across all positions, primary schools show similar seasonality to secondary schools, but there is no obvious sign of substantial increases between spring 2022 and spring 2023. There are some differences between Early Years, Key Stage 1, Key Stage 2 and generic Primary Teacher positions, but we wouldn't want to read too much into these at this early stage.

(Use the menu below to select a role to view. Hover over the lines to see corresponding data values.)

Technical talent

What of support staff? In addition to teacher vacancies, we also track those for secondary school technicians, the six-year trend for which is shown in Figure 8. Looking first across all subjects, the patterns here are less clear-cut than those seen for teachers in Figure 1. This is partly because the seasonality is different (technician recruitment traditionally peaks in June and September, either side of the summer break) and partly because the mid-pandemic crash and post-pandemic rebound have been even more extreme for technicians than for teachers. Note also that the absolute numbers of technician vacancies are much lower than those for teachers, so they are also more susceptible to statistical variations, especially when looking at individual subject areas, which in this case are broken down into Arts, Science, Technology and Other.

(Use the menu below to select a subject area to view. Hover over the lines to see corresponding data values.)

Figure 9 shows each of the last five years of technician vacancy advert data overlaid for comparison. The most recent pre-pandemic year, 2018/19 (green line), displays the usual seasonality, with peaks at the beginning and end of the school year in September and June. By comparison, spring and summer recruiting was much reduced in 2019/20 (blue line). This continued into the first part of 2020/21 (red line), but then activity took off in summer 2021 and has since continued at unusually high levels throughout 2021/22 (purple line) and 2022/23 (black line). Given that technician recruiting usually peaks right at the end of the school year, it will be interesting to see how June and July 2023 unfold.

Looking again at all years together, the cumulative data show just how exceptional the last two schools years have been, not only compared to the pandemic-affected years, but also compared to pre-pandemic norms. Although this year is currently tracking slightly below last year, it is nevertheless showing extremely high activity by any normal pre-pandemic standard.

(Use the menus below to view weekly or cumulative data, and to select different subject areas. Click on the figure legend to hide or view individual academic years. Hover over the lines to see corresponding data values.)

Figure 10 shows exactly how the percentage changes have vary by subject area, comparing activity so far this year with the equivalent period last year (red columns) and with the most recent pre-pandemic year (blue columns). Looking again at both years together, all subjects showed increases of about 40%-60% compared to pre-pandemic baselines, though they were down by around 5%-20% compared to last year's exceptionally high levels.

In terms of numbers of adverts, Technology and Science showed the largest increases relative to pre-pandemic levels (about 400-500 each; see blue columns), but also the largest decrease from last year (around 150-250 each; red columns). In total, we have so far found over 1,100 more technician adverts than the during the equivalent period before the pandemic, but almost 500 fewer than last year.

(Use the menu below to switch between percentage changes and changes in numbers of adverts. Hover over the columns to see corresponding data values and numbers of advertisements.)

Figure 11 shows the percentage change in numbers of technician adverts between the current academic year and the most recent pre-pandemic year, broken down by type of school. As for teachers in Figure 4, this analysis omits positions advertised by multi-school federations or trusts and so will tend to understate increases (or overstate decreases) in overall activity. But it is still useful in understanding relative trends between different types of schools.

Across all subjects, the regions that showed the smallest change were Yorkshire and the Humber, the East of England, the North East and the East Midlands. By far the highest was the North West. As for teachers, non-selective and rural schools tended to show greater increases, as did schools located in more affluent areas. Other trends tended to be more ambiguous.

(Use the menus below to explore different school groups and subjects. Hover over the columns to see corresponding data values and numbers of advertisements.)

Headcounts

We do not track headteacher vacancies, but we do monitor headteacher appointments using data from the Department for Education (DfE). They publish the names of school leaders and update this information daily when informed of new appointments. If anything, this is therefore a lagging indicator – unlike vacancy advertisements, which are of course leading indicators of new appointments.

Figure 12 shows the number of headteacher changes for each month over the last four academic years, and for the current 2022/23 academic year to date. (We automatically omit typo corrections, changes in surnames and similar updates, so the numbers shown are putative changes in people, not just edits to the database.) The main annual peak is in September, with slightly raised turnover also evident at the beginning of the spring and summer terms, in January and April. So far this year, all of those months have shown unusually high levels of headteacher turnover. The result is that 2022/23 has already seen more new headteacher appointments than any recent year – even without including the further new appointments that will presumable be registered in June-August 2023. This overall increase is more evident among primary schools than secondary schools.

(Use the menu below to switch between all schools, primary schools and secondary schools. Click on the figure legend to hide or view individual academic years. Hover over the columns to see corresponding data values.)

How much hire?

Why such unusually high teacher turnover? Some of the increase, especially that seen in spring and autumn 2022, was very likely caused, at least in part, by teachers who delayed switching jobs during the pandemic. However, the higher levels of activity have continued into spring 2023, fully two years after the last nationwide school closures. With ongoing industrial action among school staff, government teacher training targets routinely missed and a generally tight UK labour market, it's tempting to assume that there are wider factors at play. To put it another way, there aren't enough teachers to go around.

This wider interpretation is supported by the picture among technicians, who can more easily switch sectors and have seen ever larger increases in vacancy rates. So far these show no signs of abating. Headteachers, too, seem to be on the move, though this follows two years of somewhat reduced turnover, so might yet prove to be no more than a post-pandemic unwinding of delayed career moves and retirements.

We remain in uncharted territory and at this stage it's anyone's guess whether things will now gradually return to something like the pre-pandemic normal or whether the current trends will turn out to be a new kind of normal. However these trends unfold, we will continue to follow them with interest.

Please read our full joint report with the Gatsby Foundation and Teacher Tapp here, which also includes lots of teacher survey data. SchoolDash Insights users can also follow the latest weekly updates in school staffing using the Recruitment, Headteachers and Staff sections. Non-subscribers are welcome to sign up for a free trial, and everyone who finds our analyses remotely interesting is invited to sign up for our free monthly-ish newsletter. We also welcome your feedback to: [email protected].

Footnotes:

1. All data were gathered using an automatic process that visits school websites every night and extracts information about any new vacancies it finds there This process does not capture all vacant positions because: (a) not all positions are advertised on school websites, (b) even when they are, they are not necessarily presented in a way that can be automatically indexed, and (c) websites are sometimes unresponsive or otherwise unavailable. The data presented should therefore be thought of as being based not on a comprehensive list of all vacancies but on a subset. However, positions have been detected for well over 90% of schools and these are broadly representative of the overall population of schools.

Are free schools filling school provision cold spots?

6th June 2023 by Timo Hannay [link]

Our recent analysis of school provision cold spots included a somewhat throwaway remark at the end which suggested that the free schools programme might not be the answer to this particular systemic shortcoming because it tends to favour places where people have the wherewithal to set up a new school, and these won't necessarily coincide with the places where they are most needed. This post looks into that claim in a bit more detail.

Free for all?

Figure 1 shows the locations of all the free schools opened since 2018. Red dots indicate primary school, blue dots secondary schools and green dots all-through schools (ie, those serving both primary and secondary pupils). There are 205 in all, of which 50 opened in 2018, 54 in 2019, 36 in 2020, another 36 in 2021, 28 in 2022 and just one so far in 2023 (which isn't surprising since most new schools open in September). Click here to see all years again.. To put this in perspective, there are over 20,000 mainstream state primary and secondary schools in England, so these amount to an addition of only about 1% – though of course even a single new school can have a big impact in the community where it is located.

(Use the menu below to switch between years. Hover over the dots to see school names and years of opening.)

The map above already gives us a sense that these new schools are not evenly distributed around the country. For one thing, secondary schools seem to be mostly in or near big conurbations, while primary schools appear more likely to have arisen in more suburban or rural areas. In addition, there seem to more new schools in the Midlands and the South than in the North. On the other hand, existing schools (and of course pupils themselves) are not evenly distributed either and we need to take this into account.

Figure 2 helps to do this. It shows the regional distribution of existing schools (left-hand column) alongside that of the new free schools (right-hand column). Across all phases, new free schools are indeed skewed towards the South (lower segments) and away from the North (upper segments). This is even more true when we look at primary schools alone, while secondary schools show a relative bias towards London (purple segment). All-through schools are particularly skewed towards southern regions (though with none at all in London). However, bear in mind that the sample size here is small: there have been only 17 new all-through schools since 2018.

(Use the menu to switch between all schools, primary schools, all-through schools and secondary schools. Click on the figure legend to turn individual regions on or off; double-click to show one region on its own. Hover over the graph to see corresponding values.)

But school provision cold spots occur across the country, not just in certain regions. Another approach is to distinguish between urban areas (where cold spots are are virtually absent) and rural ones (where most of them are located). This breakdown is shown in Figure 3.

Across all phases, new free schools are very similarly distributed to schools as a whole. This suggests that recently opened free schools have not resulted in any significant shifts in the proportions of urban, suburban or rural schools. However, there are some differences by phase: primary schools skew towards suburban areas, secondary schools towards urban and rural areas, and all-through schools (with their admittedly small sample size) towards rural areas alone. In other words, recently opened new free schools appear to have added slightly to primary provision in suburban areas, and to secondary provision in urban and rural areas, but across all phases the balance is essentially unchanged.

(Use the menu to switch between all schools, primary schools, all-through schools and secondary schools. Click on the figure legend to turn individual classifications on or off; double-click to show one on its own. Hover over the graph to see corresponding values.)

Going the distance

These urban/rural classifications are indicative, but in the end cold spots are about distances to the nearest school and ideally we should address this directly. That requires us to decide exactly what kinds of schools we want to include. Our previous analysis allowed readers to choose which schools types to omit: independent (ie, private, fee-charging) schools, selective (ie, grammar) schools, faith schools, oversubscribed schools and/or schools with low Ofsted ratings. Which of these to include or exclude is a subjective judgement on which reasonable people might disagree depending on their individual expectations and circumstances. But for the purpose of the analysis that follows we are going to limit ourselves to mainstream, non-selective, co-educational (ie, mixed-sex) state schools with Ofsted ratings (where there is one) of 'Good' or 'Outstanding'. In other words, we are including faith schools and those that are over-subscribed. But we are excluding single-sex schools on the basis that they are not open to all pupils¹.

Figure 4 shows the distributions of local areas across England – specifically, Lower layer Super Output Areas, or LSOAs, of which there are just under 33,000 across the country. The blue columns show the distribution of all areas across England based on their distance from the nearest eligible primary school (excluding, of course, the recently opened free schools that are the subject of our analysis). The red columns show the distribution for local areas in the new primary free schools are located. It is clear that these new free schools do in fact tend to be in neighbourhoods that are a bit further away than average from an existing school. But this is an extremely local effect: areas within 0.6km of an existing primary school are a bit less likely to provide the site for a new free school, while anything further away than this is slightly more likely. In effect, this indicates that new free schools aren't usually placed right next to existing schools, but there's little evidence here that any of them have been located in areas with no nearby primary provision at all (admittedly a very small proportion of all areas).

To look at it another way, the median LSOA in England is just over 0.4km away from the nearest primary school, while the median LSOA in which a new primary free school has been located is just under 0.5km away from an existing primary school. So from the point of view of reducing average distances to primary schools, the positioning of new free schools has been little better than random.

(Hover over the columns to see corresponding values; click on the figure legend to show or hide each data set.)

Figure 5 shows the same analysis for secondary schools. These are arguably more consequential because, as shown in our earlier analysis, cold spots are more prevalent among secondary schools than among primary schools. Here again, we see that the red columns representing new free schools tend to be shifted to the right compared to the blue columns representing all secondary schools. But the transition point at which red columns become higher than the blue ones is only about 2km. So here too, we see that new free schools tend not to be located very close to existing secondary schools, but there's no evidence that any of them have been placed in areas that are very distant from the nearest school – ie, there are no red columns on the far right-hand side of the figure.

In terms of actual distances, the median LSOA is just over 1.2km away from the nearest secondary school, while the median LSOA in which a new secondary free school has been located is just over 1.3km away from an existing secondary school. Again, not much difference.

(Hover over the columns to see corresponding values; click on the figure legend to show or hide each data set.)

None of this is to say that the free schools established since 2018 should not have been placed where they are – if the relevant groups felt motivated enough to put in the huge amount of work involved and the Department for Education was satisfied enough to approve and pay for their establishment then it's fair to assume that there are good reasons for them to exist. But to return to the original question, it's not obvious that groups able to set up new free schools are located in all of the places where they might be required. The evidence presented here suggests that the free school programme is not currently helping to eliminate the worst cold spots in school provision, nor is it promoting regional levelling-up.

We welcome your thoughts; please send them to: [email protected]. And if you would like to hear about more analyses like this then sign up for our free monthly-ish newsletter. SchoolDash Insights subscribers can further explore the school locations and characteristics, as well as benchmark individual schools. Non-subscribers are welcome to request a trial account or a demo. Help for families in finding and choosing a school is available at The Schools Guide.

Footnotes:

1. It may of course be the case that some neighbourhoods are served by a pair of single-sex schools and in these cases we may be overestimating the distance to the nearest school available to a local girl or boy. However, this affects only secondary schools (there are very few or no single-sex state primary schools) and by excluding grammar schools we are in any case leaving out a substantial proportion of single-sex state schools. In the end, this restriction makes very little difference to the overall results.