Understanding school and community deprivation – Part 1: Secondary schools

31st August 2026 by Timo Hannay [link]

The Pupil Premium (PP) is an enormously significant policy that provides extra support to children from poorer families and the schools that serve them. But as we have discussed before, it is less useful as a guide to the circumstances faced by those schools: even those with identical PP rates can experience very different local situations.

We have therefore proposed a complementary approach that involves grouping schools not just by the PP rates, or even by geographical region, but by using a range of deprivation indicators (crime, health, income and so on) together with a simple machine-learning algorithm to group schools with similar socioeconomic characteristics into statistical clusters. This can provide a clearer view of the most common combinations of challenges faced by schools without having to treat each one as a unique case, which would obviously be impractical for policy development. This suggestion has since garnered some attention. With a new UK prime minister and cabinet now in place and a renewed focus on national inequalities, we thought this a good time to update our analysis. Our particular thanks to the Gatsby Charitable Foundation for supporting this work.

This post focuses on mainstream state secondary schools in England – we report separately on primary schools below. The updated analysis resulted in six clusters with the following broad socioeconomic, educational and workforce characteristics:

1. Affluent Suburban (685 schools): Commuter-belt suburbs and small towns, mostly inland with long journeys to work. The least deprived cluster on most measures. Strong attainment and the lowest NEET (not in education, employment or training) destinations, though academic progress is modest given the relative affluence. The oldest, most stable teaching workforce and the fullest schools.
2. Affluent High Housing Cost (275 schools): A combination of affluent London and the well-off countryside, united by costly housing and relatively high environmental deprivation (due to either crime and pollution or poor-quality housing stock). The highest rates for participation in higher education (HE) and the strongest academic progress at school, with the lowest absence and suspensions. But also signs of staff turnover and difficulty in recruitment.
3. Suburban Middle England (889 schools): The largest cluster, incorporating suburban locations nationwide, and middling on almost every deprivation indicator, though with below-average HE participation and many pupils going into further education (FE) or employment. Educationally average too, though with relatively high absence and exclusions. Low uptake of GCSE languages and of A-level science, technology, engineering and maths (STEM) subjects.
4. Urban London (365 schools): Overwhelmingly urban and London-based, with deprivation amid prosperity and short commute times. This cluster over-performs on many education indicators, matching much more affluent clusters in terms of progress and showing the highest HE progression rates. Pupils are ethnically diverse, while teachers are relatively young, with the lowest staff retention rates and relatively high agency staff spend.
5. Poor Suburban and Coastal (738 schools): The most coastal cluster, incorporating struggling seaside communities as well as smaller towns in the North and Midlands. Education, health and employment deprivation are all high, though barriers to housing are low as property is cheap. This cluster has the lowest HE participation rate, with weak academic attainment and progress, high absence, high rates of special educational needs (SEN) and the largest post-school flow into FE.
6. Poor Urban (494 schools): Big, mostly northern cities outside London. The most deprived cluster on almost every indicator. The worst attainment and progress, with high suspensions and NEET rates. Also the youngest teachers with the highest sickness rates, vacancy rates and agency spend. Pupils show the highest rate of SEN support, but also the lowest rate of Education, Health and Care (EHC) plans.

It is important to note that the labels we have associated with these clusters are merely shorthand descriptions provided for convenience and do not purport to describe every school in each group. Also, bear in mind that these clusters were generated using only local deprivation indicators – see below for more details – their educational characteristics, including pupils, staff, attendance, academic outcomes and post-school destinations were outputs from the process rather than inputs. Thus school characteristics and educational outcomes are reflections of local socioeconomic circumstances, and this goes beyond the simple economic dichotomy of affluence versus poverty.

To appreciate this more fully it is perhaps useful to also summarise the ways in which key educational themes are reflected across these clusters (again, as outputs summarising their characteristics, not as inputs used to define the clusters in the first place). In particular, which ones vary straightforwardly by income deprivation and which seem to be influenced by other factors reflecting different types of place?

• Academic attainment and progress: Attainment 8 scores (reflecting absolute GCSE performance) show a simple overall relationship, with more affluent clusters outdoing poorer ones. But Progress 8 (which takes into account pupil attainment at primary school and is therefore a fairer measure of school effectiveness) is more complicated, with Affluent Suburban (Cluster 1) schools under-performing and Urban London (Cluster 4) ones over-performing. This should be of particular concern to schools in the first group.
• Curriculum and subject participation: Pupils taking triple science (ie, separate biology, chemistry and physics exams) at GCSE vary by income deprivation, as do entry rates for A-level physics. But other patterns here deviate from this simple trend: GCSE foreign language entry rates are exceptionally high among Affluent High Housing Cost (Cluster 2) schools and Urban London (Cluster 4) schools, but low elsewhere. The same is broadly true for A-level maths and chemistry, while Urban London schools stand alone when it comes to A-level computing. Why such disparities and what could be done about them?
• Attendance and exclusions: In general, these rise with increasing income deprivation, though Suburban Middle England (Cluster 3) schools tend to show somewhat elevated absence and exclusion rates.
• SEND and EHC plans: These also vary according to income deprivation, but with a curious twist – they do so in opposite directions. As a result, more affluent clusters tend to have higher proportions of pupils on formal EHC plans but lower proportions receiving informal SEN support. The reverse is true of poorer clusters. This points to disparities not so much in the incidence of pupils with special needs, but in the way that they are diagnosed and supported, which also ought to be of concern.
• Staffing and workforce stability: Poorer clusters tend to have younger teachers, higher rates of sickness leave and higher vacancy rates. Beyond these simple trends, staff leaving for another state school, as well as spending on supply teachers, both tend to be high in Urban London (Cluster 4) and Poor Urban (Cluster 6) schools. Meanwhile, teachers leaving the state system are most prevalent among Affluent High Housing Cost (Cluster 2) and Urban London (Cluster 4) schools, which also spend the most on staff development. These results point to different staffing and recruitment support needs across the various clusters.
• School capacity and admissions: School occupancy rates are generally lower for poorer clusters. The pattern for admissions is more complicated with Affluent High Housing Cost (Cluster 2) schools showing the lowest success rate for first-choice applicants. This suggests distinct challenges in either managing pupil roll decline or in satisfying parental preferences.
• Destinations: NEET outcomes at age 16 and 18 both vary by income deprivation. In contrast, FE destinations at age 16 split the country in two, being high for Clusters 3, 5 and 6 but low for other groups of schools. Conversely, Clusters 1, 2 and 4 show high rates of progression to HE at age 18. Clusters 2 and 4 also show low proportions of 18-year-olds going on to employment destinations. Assuming that individual talent and propensities are evenly distributed, why such large differences in educational and career paths?

With that long but hopefully useful summary, the rest of this post will describe the methods and results in greater detail.

 

Technical tweaks

The overall approach applied here was very similar to our original 2024 analysis: for each school, we calculated local-area averages for deprivation indicators concerning crime, education, employment, environment, health, housing and income. We also added an indicator for higher education participation rate. A k-means machine-learning algorithm was then used to cluster schools, before analysing each cluster's socioeconomic, geographical and educational characteristics. As before, the income deprivation metric we used was the Income Deprivation Affecting Children Index (IDACI) since this most closely reflects the circumstances of families with school-age children. The main methodological differences since last time were:

• Postcode data were from 2026 (not 2024).
• Deprivation indicators were from 2025 (not 2019).
• HE participation rates came from the Office for Students TUNDRA data (not their POLAR4 data set). TUNDRA uses a more recent cohort (students who left school in 2016 versus 2014), doesn't rely on population estimates, includes only state schools, and provides finer geographical resolution than POLAR4.

Although we fed eight different measures for each school into the k-means clustering algorithm, some of these are correlated. In such situations it's common to reduce the number of dimensions to a smaller number of statistically derived 'principal components' that capture most of the relevant information in fewer values. After experimenting, we came to the conclusion that two principal components would be optimal. These capture 77% of the overall variance, and both dimensions correspond to identifiable attributes in the real world, namely income deprivation (Principal Component 1) and housing cost/environment (Principal Component 2). Furthermore, compared to clustering on all eight features, using two principle components matched more schools (85%) than using three principle components (80%).

The k-means algorithm also requires us to decide on the number of clusters into which to divide our sample. This is often done with the help of an 'elbow plot' of the kind shown in Figure 1. It demonstrates how well the clusters represent their constituent schools as we increase the number of clusters. This uses a statistical measure known as 'inertia' for which lower values indicate a better fit. Sometimes the inertia reduces rapidly up to a certain number of clusters, with little improvement above this. In our case we could plausibly choose two, four or more clusters. We decided to use a higher number in order to more fully represent the range of school circumstance, settling on six. (The questions of how many clusters to use, and how to decide this, were explored in greater detail in our previous analysis.)

(Hover over the graph to see corresponding data values.)

Figure 2 shows how state secondary schools in England are distributed with respect to the two principal components described above (horizontal and vertical axes), and how they were clustered by the k-means algorithm (colours). Note that schools are separated more by Principal Component 1 (horizontal axis) than by Principal Component 2 (vertical axis), which is to be expected given the greater statistical influence of the former. Also, the distribution is relatively smooth rather than falling into natural groups, which is consistent with the similarly smooth elbow plot in Figure 1. To repeat an analogy from our previous report, grouping schools in this way is more like slicing pizza (which you could reasonably cut in a range of different places) than breaking off dough balls (which are naturally lumpy). This doesn't make the clusters arbitrary, but it does mean that we shouldn't read too much significance into which side of a particular boundary any individual school lies.

Naturally given that they are based on new data, these results differ somewhat from our previous six-cluster analysis. Fully 42% of schools moved from one cluster to another, and some clusters (most notably Cluster 2 in the present analysis) changed in character and therefore in name. But there are many similarities too: the definitions and statistical impacts of the principal components are very close, each of the previous clusters has an obvious corresponding cluster in this analysis, and 58% of schools that were present in both analyses remained in the same cluster. Of those schools that moved, about three-quarters were located close to boundaries. Only 173 of the 3,211 secondary schools that appeared in both analyses (5.4%) moved from one cluster to another despite lying well within their original 2024 cluster.

(Click on the legend to turn individual clusters on or off; double-click to show one cluster on its own. Hover over the dots to see corresponding school information.)

 

Cluster characteristics

So much for the clustering process, what real-world characteristics do these clusters have, and where are their constituent schools located?

An answer to the first question is provided in Figure 3, which shows the numbers of secondary schools in each cluster along with the their mean deprivation scores for crime, education, employment, environment, health, housing and income, as well as their higher education participation rates. In general, things get worse as the cluster number increases from left to right. This is not a coincidence: we deliberately numbered the clusters, roughly speaking, from least to most deprived. Even so, some indicators buck this trend to a greater or lesser degree. See environment, housing, income and higher education participation rates (which of course also runs in the opposite direction, with high values being good).

Though not used as an input to the clustering process, it is also interesting to see how urbanisation, distance to the coast and commuting times vary by cluster. This speaks to their geographical locations, which we explore further below.

(Use the menu below to switch between indicators>. Hover over the columns to see corresponding data values.)

Figure 4 shows the regional distributions of secondary schools by cluster. It is clear from this that (a) alw every region is represented in every cluster, and (b) the proportions nevertheless vary enormously, with London's domination of Clusters 2 and 4 (dark blue columns) most conspicuous. Other regions also show distinctive, if less extreme, patterns. (Show all regions again.)

(Click on the legend to turn individual regions on or off; double-click to show one region on its own. Hover over the columns to see corresponding values.)

Also relevant to the government's devolution agenda has been the establishment of strategic authorities (SAs), often (though not necessarily) under the leadership of an elected mayor. These have their roots in the creation of the Greater Manchester Combined Authority in 2011, but have since grown in number and this year took on the official "strategic authority" designation. Figure 5 shows the cluster composition of the whole of England (top bar) and the 19 English strategic authorities that have so far been created or announced. The bar at the bottom shows the distribution of clusters for schools not located in any SA. It is clear that (a) individual SAs contain a wide range of locality types, (b) their compositions also vary greatly between each other, and (c) as a whole, SAs tend to represent higher proportions of deprived areas.

(Click on the legend to turn individual clusters on or off; double-click to show one cluster on its own. Hover over the bars to see corresponding values.)

Figure 6 shows more precise locations for the schools in each cluster, which might be summarised as follows (click on the cluster numbers to change the map):

• Cluster 1: Affluent suburbs and commuter belts around London and other metropolitan centres.
• Cluster 2: A mixture of affluent places in London and some better-off rural locations.
• Cluster 3: A range of suburban locations all over England.
• Cluster 4: Less affluent parts of London along with certain other urban centres.
• Cluster 5: Poorer suburban and coastal communities.
• Cluster 6: Poorer urban locations, especially in the midlands and the north.

(Use the menu to select a cluster. Use the map controls to pan and zoom. Hover over the dots to see corresponding school names.)

With that, we can finally summarise what we know about the socioeconomic and geographical characteristics of each cluster, and assign them some slightly more meaningful names, if only impressionistic ones. (Once again, click on the cluster names to change the map above.)

1. Affluent Suburban: 685 schools. The most affluent cluster on every measure, with low housing barriers, good living environment and high participation in higher education (50.5%). Predominantly suburbs and small towns with a substantial rural fringe. Mostly inland (median 48km from coast) and the longest travel times to major employment centres (median 32 mins; 53.4% over 30 mins). Classic commuter belt. (Corresponds to "Affluent Suburban." in our previous analysis.)

2. Affluent High Housing Cost: 275 schools. Composed of both Greater London (55%) and rural (36%). Affluent London neighbourhoods and affluent countryside united by high environment and housing barriers with otherwise low deprivation and the highest participation in higher education (53.9%). Both types of location within this cluster show high levels of environmental deprivation, but for different reasons: areas in London score badly on outdoor environment (air quality and traffic), while the rural areas score badly on indoor environment (poor-quality housing stock). (Previously "Affluent Urban.")

3. Suburban Middle England: 889 schools. The largest cluster and middling on most things, including all deprivation indicators, though participation in higher education is below average (40.3%). Modest share of coastal locations. (Previously "Suburban.")

4. Urban London: 365 schools. The most urban cluster, with deprivation amid prosperity. 72% Greater London and 93% urban. High IDACI (0.47) with expensive housing and poor living environment but near-average education and employment deprivation and high TUNDRA (51.1%), and the shortest travel times to major employment centres (median 23 mins). (Previously "Urban.")

5. Poor Suburban and Coastal: 738 secondary schools. Deprived based on education, health and employment indicators, and with the lowest participation in higher education of any cluster (36.9%), but also low housing barriers (ie, relatively cheap property prices) . Northern locations account for 48% and 28% are within 5km of the sea (median distance 27km), making this the most coastal cluster of all. It captures struggling coastal towns along with smaller towns in the North and Midlands. (Previously "Poor Suburban.")

6. Poor Urban: 494 schools. Big cities outside London. The most deprived cluster on every indicator except housing cost: education 38.2, IDACI 0.53, crime +0.74, health +0.95. Also low participation in higher education (37.7%). Fully 91% of locations are urban, with 71% in the North and 0% in London. Well-connected to employment centres (median travel time 23 mins) and with a notable coastal minority (23%) composed of port cities. (Previously "Poor Urban.")

Figure 7 shows the educational characteristics of the secondary schools in these clusters. Some attributes, such as competition for school places, proportions of qualified teachers or Ofsted ratings show relatively little variation across clusters. Others, such as attainment, persistent absence, proportions of older teachers, teacher sickness rates or NEET destinations, track closely with income deprivation. But others still show more complex cluster-to-cluster variation. These include progress, foreign-language GCSEs, staff development spend and FE or HE destinations.

Figure 8 shows some further indicators by cluster, this time focusing on STEM subjects. The proportions of GCSE students taking triple science, as well as A-level students taking further maths or physics, vary mosty by income deprivation. But the proportions taking chemistry or maths vary by urbanisation rather than income, and for computing Urban London (Cluster 4) schools lead the pack.

Now we can summarise these secondary school clusters in terms of their educational characteristics and challenges:

1. Affluent suburban: Very good attainment outcomes (Attainment 8 50.9) and lowest NEET (KS4 3.9%, KS5 7.8%), but arguably underperforms in terms of progress given its affluence (Progress 8 +0.18). Relatively low SEN but high EHC. The oldest and most stable teacher workforce (25.2% over 50, lowest agency spend, low staff development spend). Also the fullest schools. But KS5 HE progression is only average (54.1%) with a high proportion of 18-year-olds going into employment. Also low take-up of GCSE foreign languages and A-level chemistry.

2. Affluent High Housing Cost: Highest progress (Progress 8 +0.30), with lowest absence, suspensions and sickness. Relatively low SEN but high EHC. High GCSE foreign languages and A-level maths entry rates, as well as HE destinations (60.0%). But also signs of elevated staff turnover, with high rates of temporarily filled positions and teachers leaving the state system.

3. Suburban Middle England: Average on many measures, but low take-up of languages at GCSE and relatively high absences and exclusions. Low proportions taking A-level maths and chemistry. High proportions of KS4 FE and KS5 employment destinations.

4. Urban London: High levels of income deprivation (PP 37.7%), but over-performs in terms of attainment and progress (Progress 8 is the same as Affluent Suburban). Relatively low exclusions. Highest HE progression of all (63.2%). Only 22.3% White British pupils. Teachers are relatively young, with the lowest retention rate and high agency spend. High uptake of GCSE foreign languages but low entry rates for GCSE triple science.

5. Poor Suburban and Coastal: Many adverse indicators, including low attainment and progress, along with high absence rates. Low EHC but high SEN. Relatively young teachers, with high sickness rates too. Low uptake of GCSE foreign languages, GCSE triple science and most A-level STEM subjects. Highest KS4 FE destinations (44.7%).

6. Poor Urban: The worst performing cluster on many measures, with the lowest attainment and progress. High exclusions (37.1%), KS4 NEET (8.7%) and KS5 NEET (14.7%). The youngest teachers (only 16.8% over 50), with the highest sickness rates, vacancy rates and agency spend (£3,553 per teacher). The lowest incidence of EHC plans but the highest incidence of SEN support. The lowest proportion of school-leavers going to HE. Also low uptake of GCSE foreign languages, GCSE triple science and most A-level STEM subjects, though chemistry entry rates are middling.

 

Cluster characteristics

We hope that this analysis has provided insights into the range of different situations in which secondary schools in England find themselves. These are neither completely unique to each school nor as simple as saying that all schools with the same Pupil Premium metric, or located in the same geographical region, can be considered truly similar for the purposes of policies or interventions. This is our attempt to provide local nuance while keep the major national trends clearly in view.

Of course, the various correlations presented here do not establish causes, which remain open for debate. Furthermore, this is far from the only way to cluster schools. We could have used other input data, a different algorithm and a higher or lower number of clusters. So treat this as a type of approach that can be fine-tuned for particular research questions.

A similar account of primary schools follows below. In the meantime, our thanks once again to the Gatsby Charitable Foundation for supporting this work. As ever, we welcome your thoughts too – please write to us at: [email protected].

Understanding school and community deprivation – Part 2: Primary schools

31st August 2026 by Timo Hannay [link]

This post about primary schools follows on from our accompanying analysis of secondary schools. Please refer to that previous post for fuller background and descriptions of the methods used. Due to their different characteristics and catchment area sizes, we conducted the clustering analyses separately for primary and secondary schools (with the small proportion of all-through schools appearing in both analyses).

In summary, we find the following primary school clusters:

1. Affluent Suburban (3,301 schools): Similar to the secondary school cluster of the same name. Leads on most Key Stage 2 (KS2) academic measures, with the lowest proportions of Pupil Premium (PP) and Special Educational Needs (SEN) pupils, but also the most over-subscribed schools.
2. Affluent Rural (1,856 schools): Far more rural and much less London-centric than the corresponding secondary school cluster (which was called 'Affluent High Housing Cost'), with low PP but no more than average attainment. Small under-occupied schools with high proportions of white British pupils. By far the oldest teachers, with a high proportions leaving the state system, possibly into retirement.
3. Suburban Middle England (4,069 schools): As for the secondary school cluster of the same name, this can be considered a baseline that's close to the national mean on most measures.
4. Urban London and Expensive Coastal (1,948 schools): Compared to the corresponding secondary school cluster ('Urban London'), this adds expensive coastal areas to the London core and displays a milder version of the same trends. Good academic performance that is close to 'Affluent Suburban' levels despite high levels of disadvantage. Diverse pupil population with high Education, Health and Care (EHC) plan rates. Falling pupil rolls with signs of teacher recruitment challenges.
5. Poor Suburban and Coastal (3,460 schools): This primary cluster shares both the geography and educational patterns of the corresponding secondary school cluster. Weak attainment, high absence and high SEN , though less extreme than for the 'Poor Urban' cluster below.
6. Poor Urban (2,312 schools): Occupies the same midland and northern cities as the corresponding secondary school cluster and is also weakest in academic attainment, with the highest PP and SEN rates, worst absence and youngest teachers.

And here is a brief summary by theme:

• Academic attainment and progress: Affluent Rural schools (Cluster 2) tend to underdeliver, providing average attainment despite low Pupil Premium, though reading and science are stronger. Poor Suburban and Coastal schools (Cluster 5) are consistently adverse and Poor Urban schools (Cluster 6) are weakest across the board.
• Attendance and exclusions: These generally rise with increasing poverty, from 4.6% in Cluster 1 (Affluent Suburban) to 6.1% in Cluster 6 (Poor Urban), and persistent absence from 9.4% to 17.5%, with Cluster 5 (Poor Suburban and Coastal) in between (5.6% and 14.7%, respectively). Cluster 6 also has the highest average suspension rate, at 3.81 per 100 pupils.
• SEND and EHC plans: SEN also increase with rising income deprivation, from 12.7% in Cluster 1 (Affluent Suburban) to 17.5% in Cluster 6 (Poor Urban), with Cluster 5 (Poor Suburban and Coastal) just behind at 16.6%. EHC plans do not: the highest rate is in Cluster 4 (Urban London and Expensive Coastal, 4.3%).
• Staffing and workforce stability: Teacher age varies sharply by cluster, with 32.7% over 50 in Affluent Rural schools (Cluster 2) versus 19.2% in Poor Urban schools (Cluster 6). Retention is slightly below average (81.4%) in Cluster 2 (Affluent Rural) – possibly a combination of retirements and small-school effects. Cluster 4 (Urban London and Expensive Coastal) shows the clearest signs of recruitment strain, with the lowest share of qualified teachers (96.7%), the highest agency spend (£4,804 per teacher) and the most temporarily-filled vacancies. Cluster 6 (Poor Urban) also spends heavily on agency supply.
• School capacity and admissions: Schools in Cluster 1 (Affluent Suburban) appear to be the most oversubscribed, with the lowest first-choice success rate (93.7%). Occupancy is lowest in Cluster 2 (Affluent Rural, 79.7%) and Cluster 4 (Urban London and Expensive Coastal, 83.9%), suggesting falling rolls, caused by small rural schools in the former case and London demographics in the latter.

 

Primary colours

Figure 1 shows the 'elbow' plot for primary school clusters. As for secondary schools, there is no obvious breakpoint in the line where the gradient changes suddenly, so we used the same middling value of six clusters.

(Hover over the graph to see corresponding data values.)

Figure 2 shows how these primary schools were distributed and clustered. The overall arrangement looks very similar to that for secondary schools, though as we shall see, this hides some significant underlying differences in the primary school clusters.

(Click on the legend to turn individual clusters on or off; double-click to show one cluster on its own. Hover over the dots to see corresponding school information.)

The characteristics of these primary school clusters are shown in Figure 3. Bear in mind that primary schools have smaller catchment areas than secondary schools, so hyper-local characteristics have more effect. They are also more likely to be located in lightly populated areas. Even so, they show very similar patterns to secondary schools except that Cluster 2 for primary schools is less urban, has lower crime and has much longer commuting times than Cluster 2 for secondary schools.

(Use the menu below to switch between indicators>. Hover over the columns to see corresponding data values.)

Figure 4 shows how England's regions are represented in each primary school cluster. Again, these look very similar to the secondary school clusters, except for Cluster 2, which has a lot less London and a lot more East of England and South West. (Show all regions again.)

[Question: Should we switch this around to show cluster composition by region?]

Figure 5 show the mix of primary school clusters in each of England's statutory authorities (plus those schools not in an SA). Once again, this is very similar to the secondary school patterns except for Cluster 2, which has less presence in the Greater London Authority and more in Cumbria, Devon and Torbay, and York and North Yorkshire.

Figure 6 shows exact locations for the primary schools in each cluster, which can be summarised as follows:

• Cluster 1: Affluent suburbs around London and other metropolitan centres. (Similar to the corresponding secondary school cluster.)
• Cluster 2: Mostly better-off rural locations. (Far lower London representation than in the corresponding secondary school cluster.)
• Cluster 3: A range of suburban locations all over England. (Similar to corresponding secondary school cluster.)
• Cluster 4: London plus expensive coastal areas. (More coastal representation than in the corresponding secondary cluster.)
• Cluster 5: Poorer suburban and coastal communities. (Similar to the corresponding secondary school cluster.)
• Cluster 6: Poorer urban locations, especially in the midlands and the north. (Similar to the corresponding secondary school cluster.)

(Use the menu to select a cluster. Use the map controls to pan and zoom. Hover over the dots to see corresponding school names.)

The educational characteristics for these primary school clusters are shown in Figure 7. In general, attainment gaps between clusters are narrower than for secondary schools (about 0.6 of the school-level standard deviations compared to about 0.9 for Attainment 8), which is consistent with disadvantage gaps widening through the school years. The gap between girls and boys (5–7 points) holds in every cluster, but is largest in Cluster 4. Trends for teacher age and agency spend broadly align with income deprivation, but problems with low occupancy predominate in Affluent Rural (Cluster 2) and Urban London and Expensive Coastal (Cluster 4), not in the deprived clusters. First-choice success rates are relatively high (94–96%) everywhere, with Affluent Suburban (Cluster 1) the only place where they dip slightly. Ofsted ratings are also broadly consistent everywhere.

Putting all this together, we can summarise the educational characteristics of the primary school clusters as follows:

1. Affluent Suburban: The academic leader on essentially every KS2 measure. The lowest Pupil Premium (12.1%), SEN (12.7%) and absence/persistent absence (4.6%/9.4%). But also the lowest first-choice success rate (93.7%), suggesting the most oversubscribed primaries in England.

2. Affluent Rural: Despite low Pupil Premium (13.6%), attainment is merely average, though reading and science are somewhat better. Also the lowest occupancy rates of any cluster (79.7%), suggesting small rural schools and falling pupil rolls. By far the oldest teacher workforce (32.7% over 50), the highest White British pupil share (86.4%) and lowest EAL (5.3%). Teacher retention is slightly below average (81.4%) with a high proportion leaving the state system, possibly caused by small-school and retirement effects.

3. Suburban Middle England: Close to the national mean on most measures. As for secondary schools, this can be considered a baseline cluster.

4. Urban London and Expensive Coastal: Similar to the corresponding secondary school pattern of good performance despite high levels of disadvantage, but less extreme. Pupil Premium (28.3%) is above average but so is attainment, which is close to 'Affluent Suburban' levels. Extremes of diversity (EAL 35.0%, White British 42.1%) and, unlike the secondary school cluster, the highest EHC rate (4.3%). Lowest share of qualified teachers (96.7%), highest agency supply spend (£4,804 per teacher), highest temporarily-filled vacancies, and high proportions leaving the system suggest difficulties in recruiting. The low occupancy rate (83.9%) suggests falling rolls.

5. Poor Suburban and Coastal: Consistently adverse in terms of attainment, with high SEN (16.6%), absence (5.6%) and persistent absence (14.7%). But as for secondary schools, these trends are less extreme than for Cluster 6.

6. Poor Urban: Highest PP (41.1%), and weakest attainment on every measure. Also worst absence (6.1%), persistent absence (17.5%) and suspensions (3.81 per 100 pupils), with the highest rate SEN (17.5%). Also the youngest teachers (only 19.2% over 50) with high agency supply spend.

We hope that this post usefully supplements our accompanying analysis of secondary schools, showing both the many similarities and some important differences between the phases. As for secondary schools, the range of different situations in which primary schools find themselves are neither completely unique nor as simple as saying that all schools with similar levels of income deprivation – or those that happen to be geographically close – can be assumed to be similar for policy purposes.

If you have thoughts to share, we would love to hear them: [email protected].