In this article we continue our exploration of inclusion data across England by focusing on exclusions, one of the key components of our inclusion index.
Practices towards pupil exclusions have come under scrutiny recently, in part through an increased focus on off-rolling practices, and also around speculation linking them to a rise in knife crime. However, the overall rate of exclusions has increased each year for the last four years in both primary and secondary schools. Exclusions have been found to have a lasting impact on the mental health of students and can impact their relationship and trust with teachers and the education system. We also know from our analysis for schools, MATs and local authorities that certain groups are disproportionately excluded. These include certain ethnic groups (most notably black Caribbean), disadvantaged pupils and pupils with special educational needs and disabilities (SEND). The national average fixed-term exclusion rate in the most recent year’s published data (2016/17) is 4.76% but exclusion rates are around five times higher for SEND pupils than for those without . We also know from our recent work on the Post-16 SEND Review for the Mayor of London that the practice of off-rolling is particularly prevalent for 14 and 15 year old pupils with SEND, especially those with a lower level of prior attainment that are likely to negatively impact school GCSE results.
For these reasons, exclusions indicators are key part of our inclusion index, making up one third of the overall score. This article explores the exclusion data we have used at a regional and local authority level.
Exclusion rates are broken down into permanent and fixed-term exclusions. In our inclusion index, we include both these types of exclusions for SEN support pupils and for EHCP pupils separately. The rates shown are in percentage terms and can be interpreted as the number of exclusions per 100 pupils. This gives us four indicators and we then rank local authorities (LAs) according to how low their exclusion rates are relative to other LAs. This rank for the latest three year’s of data is then converted into a score out of 100 where 100 is the best possible inclusion score.
In this article, we will provide an overview of the current landscape around exclusions and then explore factors that contribute to exclusions of SEND pupils, including primary SEN type and levels of deprivation.
The interactive map below allows you to explore SEND exclusions data for each local authority in England. Hover over a local authority on the map to see its average exclusion data over the last three years. As mentioned above, the exclusions score is calculated out of 100 and the rates shown when hovering refer to the number of exclusions per 100 pupils.
Figure 1: Interactive map of exclusions profile
We looked at patterns of exclusion rates across local authorities and over time to see if there were any trends and noticed a couple of things, discussed below.
Figure 2: Overall exclusion score by year and region
This regional pattern is broadly similar to the overall inclusion scores found in the first article in this Inclusion Series. Noticeably, relative to other regions, the North East, North West and the South West have performed increasingly worse for overall exclusions over the three year period in contrast to East Midlands and Outer London which have been improving. In 2016/17, the average fixed term exclusion rate for EHCP pupils for LAs in Outer London was 12.7%, and 9.7% for those on SEN support, the lowest regional average in the country. Yorkshire and the Humber had the second highest fixed term exclusion rate at 17.6% for those with EHCPs and the highest fixed term exclusion rates for those on SEN support at 19.3%. Interestingly, Inner London had the biggest gap in fixed term exclusion rates between those with and without EHCPs in 2017, with the rate higher for EHCP pupils by 5.3% points.
LA Level Analysis
Drilling down further into the LA level figures, we pulled out the top 10 and bottom 10 LAs for their performance across exclusions overall and took a look at the underlying indicators to see if there were any patterns. The figures shown are averages across the three year period.
Figure 3: Top 10 LAs
Figure 4: Bottom 10 LAs
It is interesting that in the list of the top ten, York, Stockton-on-Tees and Bracknell Forest actually have quite high fixed term exclusion rates for their SEN support cohort; in fact their rates are broadly in line with four of the bottom 10 LAs figures. However, their permanent exclusion rates are particularly low which improves their overall exclusion scores in our index. Looking at figure 4, what is noticeable is that there is more of a gap between the top ten and bottom ten LAs for permanent exclusion rates than there is for fixed term rates. The average permanent exclusion rate for EHCP pupils in the top 10 LAs was 0.04%  compared to the average in the bottom 10 which was 0.22%, five and a half times higher. In contrast, the average fixed term exclusion rate for EHCP pupils in top ten was 7.9% compared to 23.6% in the top ten, just under three times higher.
When drilling a bit more into the data, we found that this was primarily due to the small number of permanent exclusions in schools and the general low percentage rates; it makes the data quite volatile. This prompted further analysis into just fixed term exclusions; the most notable difference we found is that Yorkshire and the Humber’s average exclusion score is now the lowest as a region despite coming up about mid table in figure 2.
Figure 5: Fixed term exclusions scores by region
This made us curious about why Yorkshire and the Humber appears to have such high fixed term exclusion rates. We wanted to look at a couple of factors that may be driving these high exclusion rates and feeding into a low inclusion score. Based on our experience with this sort of work, we explored (1) the link between primary needs and exclusion rates and (2) the link between exclusion rates and deprivation.
Primary Needs Analysis
LAs with a higher proportion of SEMH (social, emotional and mental health) pupils are correlated with having higher fixed term exclusion rates. This is particularly true for pupils in Yorkshire and the Humber as a region.
We know from our SEND work anecdotally that the SEMH (social emotional and mental health) cohort tends to experience the highest levels of exclusions out of SEND students and we wanted to see if this was also backed up by the available data. To give some background, SEMH is a primary need type under the umbrella of SEND and in 2018 they made up nearly 17% of the SEND cohort across primary, secondary and special schools in the UK. We calculated the proportion of total pupils identified with SEMH in each LA across school phases and looked for correlation patterns against exclusion rates. The first thing that we found was that there was a relationship between higher proportions of SEMH pupils and higher fixed term exclusion rates.
When looking at 2017 data, we split local authorities into three groups according to their deprivation scores. The average fixed term exclusion rates for SEND pupils in LAs with a low proportion of SEMH pupils was 14.7% in contrast to 16.2% for LAs with a high proportion of SEMH pupils. This is a 1.5% point gap.
The second thing we noticed was that when analysing the data split by region, the relationships did seem to differ quite significantly. The relationship appears to be especially strong in Yorkshire and the Humber (i.e. the more SEMH students there are, the more the fixed term exclusion rates increase), though these results should be treated with caution as there are only 15 LA data points.
Figure 6: Relationship between SEMH and exclusion rates in Yorkshire and the Humber (data averaged over three years)
We also noticed that LAs in Yorkshire and the Humber had very low EHCP identification rates: in 2018, 26% of all SEND pupils were given an EHCP, which is the lowest regional average. It is possible then that some SEMH pupils are not given an EHCP, for whatever reason, but are then excluded from school.
Impact of Deprivation
At Mime, we find analysis of the SEMH cohort to be interesting as it is so tightly linked to social and demographic factors, which is different to needs like visual impairment and physical disabilities. This motivated our desire to look at some contextual factors that may be influencing SEMH and exclusion rates and therefore driving our inclusion score.
More deprived LAs tend to have higher exclusion rates and slightly higher proportions of SEMH pupils
When looking at the data, we found the general trend that more deprived areas were correlated with higher fixed term exclusion rates for both those with and those without an EHCP. We split local authorities into three groups according to their deprivation scores. In the most deprived LAs in 2017, an average of 15.4% of SEND pupils without an EHCP were excluded. This compares to 13.5% for the most affluent LAs, a 1.9% point difference. In 2018, the most deprived LAs had an average of 2.45% of the total pupil cohort diagnosed with SEMH, in contrast to 2.04% of the total school population in the least deprived LAs. Though small percentages, this does amount to a large number of pupils over the whole school population.
Recent research by Oxford University suggested that certain ethnic groups (black Caribbean and white and black Caribbean) are over represented in the SEMH cohort, something we have also found to be true in some of the analysis work we have done on SEND needs in London. What we do know from public data sources is that exclusion rates are higher for certain ethnic groups: based on 2017 published data, Gypsy Roma pupils faced the highest fixed term exclusion rates at 16.8%. Black Caribbeans came in third at 9.2% followed by white and black Caribbeans with 9.1%. Black Caribbean pupils also faced the highest average permanent exclusion rate with an average of 0.21%. To put this into context, the ethnic group with the lowest exclusion rates was the Chinese cohort with 0.26% of them receiving fixed term exclusions and 0.00% receiving permanent exclusions.
This analysis reinforces the point made earlier that contextual factors do seem to influence exclusion rates and also the diagnosis of SEMH. Knowing that exclusions can be detrimental to a child’s education and self-confidence, LAs should therefore be very careful (as many already are) about over-excluding certain groups and checking unconscious biases are not fuelling these actions.
What does this analysis mean for local authorities?
- If you have higher levels of SEMH in your SEND cohort you are more likely to see higher levels of exclusions
- If your LA is more deprived you are also likely to see higher exclusion rates as well as higher SEMH rates
- LAs could implement targeted policies (as many have already done) to focus on certain groups when trying to reduce exclusion rates
 Recent studies done by the University of Exeter have found a ‘bi-directional association’ between psychological distress and exclusion. Many people in education have also argued that it is a counterproductive disciplinary tool as it teaches a subset of children who already dislike school that if they misbehave they can avoid school even more.
Ford, TJ, et al. The Relationship between Exclusion from School and Mental Health: a Secondary Analysis of the British Child and Adolescent Mental Health Surveys 2004 and 2007. 6 July 2017. https://ore.exeter.ac.uk/repository/bitstream/handle/10871/28337/Psychological%20medicine%20revision%2023%20June%202017.pdf?sequence=1
 Note that these averages are not weighted to the population of the LAs
 This proportion in 2018 is highest in secondary schools (19%) followed by primary (16%) and then special schools (13%). This need type replaced BESD (Behavioural, emotional and social difficulties)
 We did some clustering analysis and, though the sample sizes were too small to draw any conclusions for certain, we did observe that the relationship was particularly strong in deprived areas of Yorkshire
 Interestingly, the gap is much less noticeable for those with EHCPs and also permanent exclusion rates, though this may be due to many LAs having 0 for this indicator