Writing for the Tes, I highlight some issues with a recent systematic review about the impact of guided play.
Although the review has many strengths, there are three issues that limit what we can conclude from it.
First, the underying studies are poor, and not much is done to account for this issue.
Second, the definitions used for free play, guided play, and direct instruction are muddled, including the aggregation of business-as-usual with direct instruction. This threatens the research team’s conclusions.
Third, using just 17 studies, the team conduct 12 separate meta-analyses. On closer inspection, the way that the studies are combined is even more questionable.
One of my earliest encounters with social and emotional learning as a teacher came in the early 2010s when I removed a faded poster from the mouldy corner of my new classroom.
I was reminded of this experience when Stuart Locke, chief executive of a trust, tweeted his shock that the Education Endowment Foundation advocated social and emotional learning (EEF, 2019b). Stuart based his argument on his own experiences as a school leader during the 2000s and a critical review of some underlying theories (Craig, 2007).
Given this, I decided to look at the evidence for SEL, unsure of what I would find.
When thinking about how strong the evidence is for a given issue, I find it helpful first to imagine what evidence would answer our questions. Two broad questions I have about SEL:
Is investing in SEL cost-effective compared to alternatives?
What are the best ways of improving SEL?
We would ideally have multiple recent studies comparing different SEL programmes to answer these questions. These studies would be conducted to the highest standards, like the EEF’s evaluation standards (EEF, 2017, 2018). Ideally, the array of programmes compared would include currently popular programmes and those with a promising theoretical basis. These programmes would also vary in intensity to inform decisions about dosage.
Crucially, the research would look at a broad array of outcomes, including potential negative side-effects (Zhao, 2017). Such effects matter because there is an opportunity cost to any programme. These evaluations would not only look at the immediate impact but would track important outcomes through school and even better into later life. This is important given the bold claims made for SEL programmes and the plausible argument that it takes some time for the impact to feed through into academic outcomes.
The evaluations would not be limited to comparing different SEL programmes. We would even have studies comparing the most promising SEL programmes to other promising programmes such as one-to-one tuition to understand the relative cost-effectiveness of the programmes. Finally, the evaluations would provide insights into the factors influencing programme implementation (Humphrey et al., 2016b, 2016a).
Any researcher reading this may smile at my naïve optimism. Spoiler: the available evidence does not come close to this. No area of education has evidence like this. Therefore, we must make sense of incomplete evidence.
A history lesson
Before we look at the available evidence for SEL, I want to briefly trace its history based on my somewhat rapid reading of various research and policy documents.
A widely used definition of SEL is that it refers to the process through which children learn to understand and manage emotions, set and achieve positive goals, feel and show empathy for others, establish and maintain positive relationships, and make responsible decisions (EEF, 2019b).
CASEL, a US-based SEL advocacy organisation, identify five core competencies: self-awareness, self-management, social awareness, relationship skills, and responsible decision-making (CASEL, 2022). A challenge with the definition of SEL is that it is slippery. This can lead to what psychologists call the jingle-jangle fallacy. The jingle fallacy occurs when we assume that two things are the same because they have the same names; the jangle fallacy occurs when two almost identical things are taken to be different because they have different names.
Interest in social and emotional learning has a long history, both in academic research and in the working lives of teachers who recognise that their responsibilities extend beyond ensuring that every pupil learns to read and write. In England, the last significant investment in social and emotional learning happened in the late 2000s and was led by Jean Gross CBE (DfE, 2007). By 2010, around 90% of primary schools and 70% of secondary schools used the approach (Humphrey et al., 2010). The programme was called the social and emotional aspects of learning (SEAL) and focused on five dimensions different from those identified by CASEL but with significant overlap.
In 2010, the DfE published an evaluation of the SEAL programme (Humphrey et al., 2010). Unfortunately, the evaluation design was not suitable to make strong claims about the programme’s impact. Before this evaluation, there were five other evaluations of the SEAL programme, including one by Ofsted (2007), which helped to pilot the approach.
In 2010, the coalition government came to power, and the national strategies stopped. Nonetheless, the interest in social and emotional learning arguably remains as a 2019 survey of primary school leaders found that it remained a very high priority for them. However, there were reasonable concerns about the representativeness of the respondents (Wigelsworth, Eccles, et al., 2020).
In the past decade, organisations interested in evidence-based policy have published reports concerning social and emotional learning. Here are twelve.
In 2011, an overarching review of the national strategies was published (DfE, 2011).
In 2012, NICE published guidance on social and emotional wellbeing in the early years (NICE, 2012).
In 2013, the EEF and Cabinet Office published a report on the impact of non-cognitive skills on the outcomes for young people (Gutman & Schoon, 2013)
In 2015, the Social Mobility Commission, Cabinet Office, and Early Intervention Foundation published a series of reports concerning the long-term effects of SEL on adult life, evidence about programmes, and policymakers’ perspectives (EIF, 2015).
In 2015, the OECD published a report on the power of social and emotional skills (OECD, 2015).
In 2017, the US-based Aspen Institute published a scientific consensus statement concerning SEL (Jones & Kahn, 2017).
In 2018, the DfE began publishing findings from the international early learning and child wellbeing (IELS) study in England, including SEL measures (DfE, 2018).
In 2019, the EEF published a guidance report setting out key recommendations for improving social and emotional learning (EEF, 2019b).
In 2020, the EEF published the results of a school survey and an evidence review that supported the 2019 guidance report (Wigelsworth, Eccles, et al., 2020; Wigelsworth, Verity, et al., 2020).
In 2021, the Early Intervention Foundation published a systematic review concerning adolescent mental health, including sections on SEL (Clarke et al., 2021).
In 2021, the EEF updated its Teaching and Learning Toolkit, which includes a strand on social and emotional learning (EEF, 2021).
In 2021, the Education Policy Institute published an evidence review of SEL and recommended more investment, particularly given the pandemic (Gedikoglu, 2021).
To make sense of this array of evidence, we need to group it. There are many ways to do this, but I want to focus on three: theory, associations, and experiments.
Theory is perhaps the most complicated. To save my own embarrassment, I will simply point out that social and emotional learning programmes have diverse theoretical underpinnings, and these have varying levels of evidential support. Some are – to use a technical term – a bit whacky, while others are more compelling. A helpful review of some of the theory, particularly comparing different programmes, comes from an EEF commissioned review (Wigelsworth, Verity, et al., 2020). I also recommend this more polemical piece (Craig, 2007).
The next group of studies are those that look for associations or correlations. These studies come in many different flavours, including cohort studies that follow a group of people throughout their lives like the Millennium Cohort Study (EIF, 2015). The studies are united in that they look for patterns between SEL and other outcomes. Still, they share a common limitation: it is hard to identify what causes what. These studies can highlight areas for further investigation, but we should not attach too much weight to them. Obligatory XKCD reference.
Experiments can test causal claims by estimating what would have happened without the intervention and comparing this to what we observe. Experiments are fundamental to science, as many things seem promising when we look at just theory and associations, but when investigated through rigorous experiments are found not to work (Goldacre, 2015).
There are four recent meta-analyses, which have included experiments (Mahoney et al., 2018). These meta-analyses have been influential in the findings from most of the reports listed above. The strength of meta-analysis, when based on a systematic review, is that it reduces the risk of bias from cherry-picking the evidence (Torgerson et al., 2017). It also allows us to combine lots of small studies, which may individually be too small to detect important effects. Plus, high-quality meta-analysis can help make sense of the variation between studies by identifying factors associated with these differences. To be clear, these are just associations, so they need to be interpreted very cautiously, but they can provide important insights for future research and practitioners interested in best bets.
Unfortunately, the meta-analyses include some pretty rubbish studies. This is a problem because the claims from some of these studies may be wrong. False. Incorrect. Mistaken. Researchers disagree on the best way of dealing with studies of varying quality. At the risk of gross oversimplification, some let almost anything in (Hattie, 2008), others apply stringent criteria and end up with few studies to review (Slavin, 1986), while others set minimum standards, but then try to take account of research quality within the analysis (Higgins, 2016).
If you looked at the twelve reports highlighted above and the rosy picture they paint, you would be forgiven for thinking that there must be a lot of evidence concerning SEL. Indeed, there is quite a lot of evidence, but the problem is that it is not all very good. Take one of the most widely cited programmes, PATHS, for which a recent focused review by the What Works Clearinghouse (think US-based EEF) found 35 studies of which:
22 were ineligible for review
11 did not meet their quality standards
2 met the standards without reservations
Using the two studies that did meet the standards, the reviewers concluded that PATHS had no discernible effects on academic achievement, student social interaction, observed individual behaviour, or student emotional status (WWC, 2021).
Unpacking the Toolkit
To get into the detail, I have looked closely at just the nine studies included in the EEF’s Toolkit strand on SEL with primary aged children since 2010 (EEF, 2021). The date range is arbitrary, but I have picked the most recent studies because they are likely the best and most relevant – the Toolkit also contains studies from before 2010 and studies with older pupils. I chose primary because the EEF’s guidance report focuses on primary too. Note sampling studies from the Toolkit like this avoids bias since the Toolkit itself is based on systematic searches. The forest plot below summarises the effects from the included studies. The evidence looks broadly positive because most of the boxes are to the right of the red line. Note that multiple effects were reported in two studies hence 11 effects, but nine studies for review.
It is always tempting to begin to make sense of studies by looking at the impact, as we just did. But I hope to convince you we should start by looking at the methods. The EEF communicates the security of a finding through padlocks on a scale from 0-5, with five padlocks being the most secure (EEF, 2019a). Of the nine studies, two are EEF-funded studies, but for the remaining seven, I have estimated the padlocks using the EEF’s criteria.
Except for the two EEF-funded studies, the studies got either zero or one padlock. The Manchester (2015) study received the highest security rating and is a very good study: we can have high confidence in the conclusion. The Sloan (2018) study got just two padlocks but is quite compelling, all things considered. Despite being a fairly weak study by the EEF’s standards, it is still far better than the other studies.
The limitations of the remaining studies are diverse, but recurring themes include:
High attrition – when lots of participants are randomised but then not included in the final analysis, this effectively ruins the point of randomisation (IES, 2017a).
Few cases randomised – multiple studies only randomised a few classrooms, and the number of cases randomised has a big impact on the security of a finding (Gorard, 2013).
Poor randomisation – the protocols for randomisation are often not specified, and it is not always possible to assess the integrity of the randomisation process (IES, 2017b)
Self-reported outcomes – most studies used self-reported outcomes from pupils or teachers, which are associated with inflated effect sizes (Cheung & Slavin, 2016). The EEF’s studies have also consistently shown that teacher perceptions of impact are poor predictors of the findings from evaluations (Stringer, 2019).
Unusual or complex analysis choices – many studies include unusual analysis choices that are not well justified, like dichotomising outcome variables (Altman & Royston, 2006). Further, the analyses are often complex, and without pre-specification, this gives lots of ‘researcher degrees of freedom’ (Simmons et al., 2011).
Incomplete reporting – the quality of reporting is often vague about essential details. It is difficult to properly assess the findings’ security or get a clear understanding of the exact nature of the intervention (Hoffmann et al., 2014; Montgomery et al., 2018).
Social threats to validity – where classes within a school are allocated to different conditions, there is a risk of social threats to validity, like resentful demoralisation, which were not guarded against or monitored (Shadish et al., 2002).
The SEL guidance report
Stuart’s focus was originally drawn to the Improving Social and Emotional Learning in Primary Schools guidance report (EEF, 2019b). A plank of the evidence base for this guidance report was the EEF’s Teaching and Learning Toolkit. At the time, the toolkit rated the strand as having moderate impact for moderate cost, based on extensive evidence (EEF, 2019b). Since the major relaunch of the Toolkit in 2021, the estimated cost and impact for the SEL strand have remained the same, but the security was reduced to ‘very limited evidence’ (EEF, 2021). The relaunch involved looking inside the separate meta-analyses that made up the earlier Toolkit and getting a better handle on the individual studies (TES, 2021). In the case of the SEL strand, it appears to have highlighted the relative weakness of the underlying studies.
Being evidence-informed is not about always being right. It is about making the best possible decisions with the available evidence. And as the evidence changes, we change our minds. For what it is worth, my view is that given the strong interest among teachers in social and emotional learning, it is right for organisations like the EEF to help schools make sense of the evidence – even when that evidence is relatively thin.
This rapid deep dive into the research about SEL, has also given me a necessary reminder that from time-to-time it is necessary to go back to original sources, rather than only relying on summaries. For instance, the EEF’s recent cognitive science review found just four studies focusing on retrieval practice that received an overall rating of high, which I know many people are surprised to learn given the current interest in using it (Perry et al., 2021).
I’ll give the final word to medical statistician Professor Doug Altman: we need less research, better research, and research done for the right reasons (Altman, 1994).
Ashdown, D. M., & Bernard, M. E. (2012). Can Explicit Instruction in Social and Emotional Learning Skills Benefit the Social-Emotional Development, Well-being, and Academic Achievement of Young Children? Early Childhood Education Journal, 39(6), 397–405. https://doi.org/10.1007/S10643-011-0481-X/TABLES/2
Bavarian, N., Lewis, K. M., Dubois, D. L., Acock, A., Vuchinich, S., Silverthorn, N., Snyder, F. J., Day, J., Ji, P., & Flay, B. R. (2013). Using social-emotional and character development to improve academic outcomes: a matched-pair, cluster-randomized controlled trial in low-income, urban schools. The Journal of School Health, 83(11), 771–779. https://doi.org/10.1111/JOSH.12093
Brackett, M. A., Rivers, S. E., Reyes, M. R., & Salovey, P. (2012). Enhancing academic performance and social and emotional competence with the RULER feeling words curriculum. Learning and Individual Differences, 22(2), 218–224. https://doi.org/10.1016/J.LINDIF.2010.10.002
Goldacre, B. (2015). Commentary: randomized trials of controversial social interventions: slow progress in 50 years. International Journal of Epidemiology, 44(1), 19–22. https://doi.org/10.1093/ije/dyv005
Gorard, S. (2013). Research design: creating robust approaches for the social sciences (1st ed.). SAGE.
Hattie, J. (2008). Visible learning: a synthesis of over 800 meta-analyses relating to achievement. Routledge.
Higgins, S. (2016). Meta-synthesis and comparative meta-analysis of education research findings: some risks and benefits. Review of Education, 4(1), 31–53. https://doi.org/10.1002/rev3.3067
Hoffmann, T. C., Glasziou, P. P., Boutron, I., Milne, R., Perera, R., Moher, D., Altman, D. G., Barbour, V., Macdonald, H., Johnston, M., Lamb, S. E., Dixon-Woods, M., McCulloch, P., Wyatt, J. C., Chan, A.-W., & Michie, S. (2014). Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide. BMJ (Clinical Research Ed.), 348, g1687. https://doi.org/10.1136/BMJ.G1687
Jones, S. M., Brown, J. L., Hoglund, W. L. G., & Aber, J. L. (2010). A School-Randomized Clinical Trial of an Integrated Social-Emotional Learning and Literacy Intervention: Impacts After 1 School Year. Journal of Consulting and Clinical Psychology, 78(6), 829–842. https://doi.org/10.1037/a0021383
Montgomery, P., Grant, S., Mayo-Wilson, E., Macdonald, G., Michie, S., Hopewell, S., Moher, D., & CONSORT-SPI Group. (2018). Reporting randomised trials of social and psychological interventions: the CONSORT-SPI 2018 Extension. Trials, 19(1), 407. https://doi.org/10.1186/s13063-018-2733-1
Morris, P., Millenky, M., Raver, C. C., & Jones, S. M. (2013). Does a Preschool Social and Emotional Learning Intervention Pay Off for Classroom Instruction and Children’s Behavior and Academic Skills? Evidence From the Foundations of Learning Project. Early Education and Development, 24(7), 1020. https://doi.org/10.1080/10409289.2013.825187
Schonfeld, D. J., Adams, R. E., Fredstrom, B. K., Weissberg, R. P., Gilman, R., Voyce, C., Tomlin, R., & Speese-Linehan, D. (2014). Cluster-randomized trial demonstrating impact on academic achievement of elementary social-emotional learning. School Psychology Quarterly : The Official Journal of the Division of School Psychology, American Psychological Association, 30(3), 406–420. https://doi.org/10.1037/SPQ0000099
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalised causal inference. Houghton Miffin.
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological Science, 22(11), 1359–1366. https://doi.org/10.1177/0956797611417632
Snyder, F., Flay, B., Vuchinich, S., Acock, A., Washburn, I., Beets, M., & Li, K. K. (2010). Impact of a social-emotional and character development program on school-level indicators of academic achievement, absenteeism, and disciplinary outcomes: A matched-pair, cluster randomized, controlled trial. Journal of Research on Educational Effectiveness, 3(1), 26. https://doi.org/10.1080/19345740903353436
Torgerson, C., Hall, J., & Lewis-Light, K. (2017). Systematic reviews. In R. Coe, M. Waring, L. Hedges, & J. Arthur (Eds.), Research methods and methodologies in education (2nd ed., pp. 166–179). SAGE.
The Telegraph has reported the DfE’s free service to promote vacancies as a threat to the business model of organisations like TES, which has existed since 1910. The DfE’s service aims to save schools £75 million each year. I’m sceptical that schools spend so much money, so I decided to see if I could find it.
Today, TES is a web of companies ultimately owned by the US-based Providence Equity Partners. TES reported a turnover of just under £100 million last year. If all their income came from subscribers like me who pay £54 a year, they would have 1.9 million subscribers. This is equivalent to every teacher in England having four subscriptions!
To understand an organisation’s priorities, it is helpful to look at the staff they employ. TES are not really in the business of publishing a magazine, as the table below shows. Just one in ten of their team are editorial – most work in sales and marketing.
£58 million. That was the income from advertising vacancies last year. Most of this, £42 million, came from a subscription model where schools pay for unlimited advertising. TES has successfully transitioned more and more schools away from one-off adverts towards this model over the past few years.
TES has a smart business model. The ‘attract’ part is generating the cash to fund the strategic acquisition of other digital services like Educare, which it acquired in 2019 for £12 million. TES aims to expand these new businesses to their existing customer networks.
School leaders will ultimately decide if they want to pay TES and other recruiters so much when there’s a good free option from the DfE. I’ve previously described how I think trusts should lead the way in transitioning towards the DfE’s Teaching Vacancies site.
If trust leaders need another incentive, they might like to know that the highest-paid director at TES received average emoluments of £450k in each of the past five years (£728, £236, £485, £310, £494k).
1. TES reports that they have 15,000 subscribers, but note that multiple readers use each subscription, suggesting that some are institutional subscribers who pay more. If they all paid £54 a year like me, TES would generate almost £800,000 income through this approach, which should mean a magazine is still viable.
2. Not all of the £58 million comes from schools in England. The accounts do not breakdown by country, but reports from previous years indicate that England remains their most important market, followed by Australia.
My attention was recently drawn to the DfE’s vacancies website by Stuart Locke who was outraged by the cost of advertising vacancies. The DfE website was created with the goal of saving schools money and making it easier for jobseekers to find roles in schools.
What is the purpose of the site?
DfE analysis estimated that the education system spends around £75 million each year advertising vacancies. There’s no link to the original research, which makes me sceptical, but this works out at about £3,000 per school every year which seems plausible.
The second aim is to improve the experience for job seekers by creating a single central record. Such systems exist in other sectors, including the NHS. This would save teachers time and lead to a healthier labour market.
A bonus of creating a single central record is that it would open some useful, low-cost research opportunities to better understand the education labour market.
How is it being used?
The data on the portal is not available to download, but it is possible to scrape it, which I did using rvest. This works well for most of the data, but it is not always possible to match the data with other records, but this only happened in about five percent of cases. This mainly happens when a role is advertised at the trust, rather than school level.
There are around 1,600 roles currently advertised on the site. Note that this includes teaching and non-teaching roles in schools.
I looked at the 25 largest trusts to see how the number of adverts on the DfE’s website compares to the number advertised on each trust’s own site, which I examined manually. I have three takeaways from the graph below.
There are lots of adverts missing from the DfE’s single central record.
Adding these missing roles would nearly double the DfE’s single central record.
Trusts should lead the way
The largest trusts are major employers, which is why I think they should lead the way in creating a single central record of vacancies. Given the number of vacancies that large trusts advertise, they can help to rapidly expand the DfE site so that it becomes the go to place for employers and employees alike.
A single central record of vacancies won’t fix everything, but it is a practical, easy step to create a better education system.
In an exclusive for the Sunday Times, Ofsted’s Amanda Spielman revealed that she anticipates the number of Outstanding schools to be roughly halved.
Many of these Outstanding schools have not been inspected for a long time because former Secretary of State for Education Michael Gove introduced an exemption. In effect, this created a one-way road to Outstanding for schools since it was still possible for schools to become Outstanding, but once there it was rare to go back.
To understand the point of Ofsted, I want to explore five ways – or mechanisms – that might lead to improvements.
1. Identifying best practice
Some people argue that Ofsted has a role in identifying the highest performing schools so that other schools can learn from them.
This mechanism relies on some demanding assumptions, including that we can (1) accurately identify the best schools; (2) disentangle the specific practices that make these schools successful; and (3) that this best practice is actually applicable to the context of other schools that might seek to imitate them.
2. Supporting parental choice
The logic here is that parents use Ofsted reports to move their children away from lower rated schools towards higher rated schools. Short-term, this gets more children into better schools. Longer-term, the less effective schools may close, while the higher-rated schools may expand.
This mechanism relies on high-quality, comparable information. Can you spot the problem? The mixed picture of reports under the old and new framework makes this a really difficult task – one that I suspect even the most engaged parents would find hard. If we think this mechanism is important, then perhaps we should invest in more frequent inspections so that parents have better information.
Personally, I’m sceptical about the potential of this mechanism. I worry about the accuracy and comparability of the reports. Also, the potential of this mechanism is limited by the fact that it can only really work when pupils transition between phases since so few pupils move schools midway through a phase and even if they are moving this probably comes with significant downsides such as breaking up friendship groups. Further, the potential of this mechanism is much more limited in rural areas where there is less realistic choice between schools. Finally, I worry about the fairness of this mechanism – what about the pupils left behind?
Given the downgrading of many Outstanding schools I also cannot help but wonder if this mechanism might have been acting in reverse – how many pupils sent to ‘Outstanding’ schools in the past decade might have gone to a different school had it been re-inspected?
3. Identifying schools that need more support
Targeting additional resources and support where it is most needed makes a lot of sense. If we have accurate information about which schools would most benefit from support, then it is simple enough to then prioritise these schools.
Of course, for this mechanism to work, we need to correctly identify schools most in need and we need to have additional support that is genuinely useful to them.
4. Identifying areas for improvement
Ofsted’s reports identify key areas for improvement. This is potentially useful advice that schools can then focus on to improve further.
I’m sceptical about the potential of this mechanism alone because in my experience Ofsted rarely tells schools things that they do not already know.
5. Understanding the state of the nation
Ofsted have extraordinary insights into the state of the nation’s schools. Rather than supporting individual schools, this information could be used to tailor education policies by providing a vital feedback loop.
To get the most from this mechanism, it would be great to see Ofsted opening up their data for researchers to explore in a suitably anonymised manner.
Caveats and conclusions
I have not mentioned the role of Ofsted in safeguarding. Most people agree that we need a regulator doing this. But there is less consensus once focus goes from ‘food hygiene’ to ‘Michelin Guide’, to extend Amanda Spielman’s analogy.
I think it’s useful focusing on mechanisms and not just activities. It also worth considering cost-effectiveness – are there cheaper ways of activating these mechanisms? For instance, I’ve been really impressed by how Teacher Tapp have given rich insights into the state of the nation’s schools on a tiny budget. For context, Ofsted’s annual budget is more than the £125 million given to the EEF to spend over 15 years.
Which mechanisms do you think are most promising? Are there other mechanisms? Are there better ways of achieving these mechanisms? Are there more cost-effective ways?
The ITT market review has the potential to make a dramatic difference to the future of the teaching profession and in turn the life experiences of young people. I’ve previously written about how the review could succeed by removing less effective providers from the market and replacing them with better ones.
In this post, I want to examine another mechanism: programme development. The consultation published earlier this year describes a number of activities, including more training for mentors, but I want to unpick the details of the mechanisms to help clarify thinking and focus our attention on the most important details.
There are a number of lenses we can use to look at programme development, including:
Curriculum, pedagogyª, assessment – each of these has the potential to improve the programme.
What trainees will do differently – this is a useful lens because it brings us closer to thinking about trainees, rather than just activities. Relatedly, Prof Steve Higgins invites us to think about three key ways to improve pupil learning; we can imrove learning by getting pupils to work harder, for longer, or more effectively or efficiently.
Behaviour change – ultimately, the market review is trying to change the behaviour of people, including programme providers, partner schools and of course trainee teachers. Therefore, it is also useful to use the capability, opportunity, motivation model of behaviour change (COM-B).
Ease of implementation – we need to recognise that ITT programmes have quite complex ‘delivery chains’ involving different partners. When considering the ease of implementation – and crucially scalability – it might help to consider where in the delivery chain changes in behaviour need to take place. Changes at the start of the delivery chain, such as to the core programme curriculum, are likely easier to make compared to those at the end such as changes within the placement schools.
(ªBe gone foul pedants, I’m not calling it andragogy.)
With these four lenses in hand, let’s consider how the market review might support the development of the programme.
The curriculum is as good a place as any to start, but first I’d like to emphasise that ITT programmes are complex – many different actors need to act in a coordinated manner – and this is perhaps felt most acutely when it comes to the curriculum. Instead of advocating the teaching of particular things, I’d like to highlight three specific mechanisms that could lead to change.
First, prioritising the most important learning, for instance, I am yet to find a trainee who would not benefit from more focused subject knowledge development. You can insert your own pet project or peeve here too.
Second, reducing the redundancy, or duplication, by cutting down on the overlap of input from different actors. For instance, in my experience, it is common for different actors to present models that are functionally similar, but different. There are lots of different models concerning how best to scaffold learning and different actors may introduce their personal favourite. Of course, there are sometimes sound reasons for presenting different models since the similarities and differences can help us to appreciate deeper structures, but where this variation is arbitrary it is just adding to the noise of an already challenging year for trainee teachers.
Third, sequencing is often the difference between a good and a great curriculum. Improved sequencing can help to optimise learning either by ensuring that trainees progressively develop their understanding and practice, or by ensuring that as trainees encounter new ideas, they also have the opportunity to apply them. The EEF’s professional development guidance highlights four mechanisms: building knowledge, motivating staff, developing teaching practice, and embedding practice. A challenge for ITT providers – given the logistics of school placements – is that there is often quite a gap between building trainee knowledge and providing opportunities – particularly involving pupils – for them to apply this knowledge.
Depending on the level of abstraction that we think about the programme, there are different mechanisms. At the highest level, it is instructive to think about trainees working longer, harder, more effectively or efficiently. I suspect we are at the limit of what can be achieved by getting trainees to work longer hours – short of extending the programme length. The market review consultation recommends a minimum length of 38 weeks so assuming trainees work a 40 hour week, we need to decide what is the best way for them to spend their 1,520 hours?
Turning away from the curriculum, how might we improve the effectiveness and efficiency of our teaching methods? Here are some of the areas that I would explore.
Can we make it easier for trainees to access brilliant content? High-quality textbooks tightly aligned with the programme content would be a very useful and scalable resource. Having to comb through lots of different reports not tailored specifically for programmes is a real inefficiency.
Do we want trainees to spend so much time engaging with primary research? It’s definitely a useful skill to develop, but the best way to be able to independently and critically engage with primary research is not to simply be thrown into it. It’s not that this is an inherently bad idea, just that it has a high opportunity cost.
How do we make better use of trainee’s self-directed study? I suspect giving access to better resources – particularly for subject knowledge development, is an easy win. There may also be merit in helping to develop better study habits.
Do we really need trainees to complete an independent research project? I think trainees should engage more with research, but as users, not producers. My starting point would be helping trainees to recognise different types of claims, and assessing the rigour and relevance of the supporting evidence. This is not too technical, and it is fundamental to building the research literacy of the profession. For the purists who cannot let go of the individual research project, I would point to the need for greater scaffolding – managing an entirely new research project just has too many moving parts for trainees become proficient in any of them. One way doing research could be scaffolded is through some micro-trials similar to the work of the EEF’s teacher-choices trials, or WhatWorked. There is a growing body of evidence from other fields that replications can be a useful teaching tool and generate useable knowledge. These do not have to be limited to trials, but could also include common data collection instruments and aggregating data. This would allow the systematic accumulation of knowledge from a large and interested workforce. The forthcoming Institute of Teaching could help to coordinate this kind of work.
Can we cut down on time trainees spend on other things? Travelling between venues, waiting around, and cutting and preparing resources all seem like areas that could be optimised. The gains here might not be big, but they are probably quite easy to achieve.
Can we improve the quality of mentoring? The market review consultation focuses quite a bit on mentoring. I agree that this is probably a really promising mechanism, but it is also probably really hard to do – particularly at scale.
Some of the ideas listed above are easier to do than others, and will have different impacts. When considering the ease of implementation – and crucially scalability – it might help to consider where in the delivery chain changes in behaviour need to take place. Changes at the start of the delivery chain, such as to the core programme curriculum, are likely easier to make compared to those at the end such as changes within the placement schools. Through this lens it becomes obvious that it’s considerably harder to improve the quality of mentoring provided by thousands of mentors compared to investing in providing high-quality, structured information in textbooks, for instance.
Finally, let’s examine assessment. An assessment is a process for making an inference, so what inferences do we need to make as part of an ITT programme? I think there are four types for each prospective teacher:
1. Are they suitable to join our programme?
2. What are the best next steps in their development?
3. Are they on track to achieve Qualified Teacher Status (QTS)?
4. Have they met the Teachers’ Standards to recommend QTS?
The first and final inference are both high stakes for prospective teachers. The Teachers’ Standards are the basis for the final inference – but what is there to support the first inference? From sampling the DfE’s Apply Service, it is evident that there is quite diverse practice between providers – how might we support providers to improve the validity and fairness of these assessments? Assessment is always tricky, but it is worth stepping back to appreciate how hard the first assessment is – we are trying to make predictions about some potentially very underdeveloped capabilities. What are the best predictors of future teaching quality? How can we most effectively select them? How do we account for the fact that some candidates have more direct experience of teaching than others?
The second and third inference are about how we can optimise the development of each trainee, and also how we identify trainees who may need some additional support. This support might be linked directly to their teaching, or it may concern wider aspects needed to complete their programme such as their personal organisation. Getting these assessments right can help to increase the effectiveness and efficiency of the programme – and in turn the rate of each trainee’s development.
Assessment is difficult so it would almost definitely be helpful to have some common instruments to support each of these inferences. For instance, what about some assessments of participants’ subject knowledge conducted at multiple points in the programme. These could provide a valuable external benchmark, and also be used diagnostically to support each trainee’s development. Done right, they could also be motivating. Longer-term, this could provide a valuable feedback loop for programme development. Common assessments at application could also help shift accountability of ITT providers onto the value that they add to their trainees, rather than just selecting high-potential trainees.
I’ve used focused on the mechanisms that might lead to improvements in ITT provision. We can think of these mechanisms with different levels of abstraction and I have offered four lenses to support this: curriculum, assessment, and pedagogy; what trainees will do differently; ease of implementation; and behaviour change.
My overriding thought is that there is certainly the potential for all ITT providers to further improve their programmes using a range of these potential mechanisms and others. However, improvement will not be easy and the DfE will need to focus on capability, opportunity, and motivation. In other words, support and time are necessary to realise some of these mechanisms. Therefore, is it worth thinking again about the proposed timescale? Including what happens once providers have been reaccredited?
The ITT Market Review aims to ensure consistently high-quality training in a more efficient and effective market. It is currently out for consultation and could dramatically reshape how we prepare teachers in England.
The main recommendation is that all providers should implement a set of new quality requirements and that an accreditation process should ensure that providers can meet these requirements.
The identification of placement schools and ensuring that these placements are aligned with the training curriculum
The identification and training of mentors, including introducing the new role of lead mentors
The design and use of a detailed assessment framework
Processes for quality assurance throughout the programme
The structures and partnerships needed to deliver a programme and hold partners accountable for the quality of their work
An expectation that courses last at least 38 weeks, with at least 28 weeks in schools
How could the review succeed?
Instead of focusing on the nature of the proposals – should courses last at least 38 weeks? Is the Core Content Framework appropriate? – I want to analyse the proposals in their own terms: are they likely to achieve their stated goals? To do this, it helps to think about potential mechanisms that could lead to improvements and potential support factors and unintended effects.
Mechanism 1: Removing less effective providers
Less effective providers could be removed from the market, and this could raise average quality. This could happen in multiple ways: providers might decide it is all too much not re-apply, which is problematic if they are a strong provider. Second, some providers may merge. Third, some providers may try but fail to meet the requirements. Time will tell how many of the 240 accredited ITT providers fit into each category.
How do we accurately assess the quality of provision?
Getting this right is fundamental if removing less effective providers is a crucial mechanism for strengthening the market. However, we should consider for every less effective provider we remove, how many strong providers we are willing to sacrifice because of the fundamental trade-off between false positives and negatives in any selection process.
The distribution of provider quality and how the assessment is done will influence the relative trade-off between sensitivity and specificity. Do we know the distribution of provider quality? My hunch is that most providers are similar, but there are long tails of stronger and weaker providers. If this is the case, do we draw the line to chop off the tail of weaker providers, or do we cut into the body of similar providers?
The second consideration is how to judge provider quality. The consultation offers a high-level process on page 29 involving a desk-based exercise with providers responsible for submitting evidence. But who will apply the quality requirements to the evidence submitted? Civil servants supported by some expert input? This might work well for some aspects, such as assessing quality assurance processes, but the heart of the reforms – the curriculum – is much harder to assess.
To maximise the accuracy of judgements, it makes sense to do it in phases: an initial separation of those that very clearly do or do not meet the criteria and then a more intensive stage for those that might meet the requirements. Otherwise, an appeal mechanism might be wise. Using a phased approach could improve the assessments’ accuracy while making the most of everyone’s finite resources.
While still thinking about the distribution of provider quality, it is worth asking if there is enough meaningful variation. If most providers are pretty similar, then at best, we can only make relatively minor improvements by removing the least effective providers. There might be more meaningful variation at the subject level or even at the level of individual tutors. If true, and we could accurately measure this variation, this would hint at a very different kind of market review (licences for ITT tutors, anyone? No?). For context, eight providers each developed over 500 teachers last year, and UCL almost reached 1,600 – should we look at a more granular level for these providers?
The less effective providers are gone; what next?
We now need to replace the capacity we removed by introducing new providers or expanding the remaining higher-quality providers. Removing lots of less effective providers is a promising sign the mechanism is working, but it poses a challenge: can we bring in new capacity that is – quite a bit – better? This may depend on how much we lose: is it 5, 15 or even 50 per cent?
Do we know if new providers will join? It would probably be wise to determine if this is likely – and the potential quality – before removing existing providers. The quality requirements set a high bar for new entrants, so a rush of new providers seems unlikely. That said, some Trusts and Teaching School hubs may come forwards – especially if given the grant funding for the set-up work advocated in the review. Other providers like the ECF and NPQ providers not already involved with ITT, including Ambition Institute, may consider applying.
Expanding existing strong providers seems desirable and straightforward enough, but we should heed the warnings from countless unsuccessful efforts of scaling promising ideas. Spotting barriers to scalability – before you hit them – is often tricky. Sir David Carter’s observation that when Trusts grow to the point that the CEO can no longer line manage all of the headteachers, a scalability barrier has been reached – new systems and processes are needed to continue operating effectively.
What are the barriers for an ITT provider? The brilliant people and the delicate partnerships with placement schools that have often developed over several years are challenging to scale. No doubt there are many more.
Before we forget, what about those providers that merged to get through the application process? How do we ensure the best practice is embedded across their work? Again, this isn’t easy, and we will likely have to base the judgements on providers’ plans rather than the actual implementation, given the timeline. Nonetheless, it seems likely that money and time will help. An analogy is the Trust Capacity Fund that provides additional funding to expanding Trusts for focused capacity building.
If we think that removing less effective providers is an effective mechanism for the ITT Market Review, then we should:
Purposefully design and implement the selection process
Plan for how to replace the removed capacity
Ensure that time and money are not undue obstacles
Consider phasing the approach
In part two, I explore another mechanism – programme development – that the ITT Market Review might use to achieve its goals.
Research tries to answer questions. The range of education research questions is vast: why do some pupils truant? What is the best way to teach fractions? Which pupils are most likely to fall behind at school? Is there a link between the A-levels pupils study and their later earnings in life?
Despite the bewildering array of questions, education research questions can be put into three main groups.
Description. Aims to find out what is happening, like how many teachers are there in England? What is the average KS2 SAT score in Sunderland?
Association. Aims to find patterns between two or more things, like do pupils eligible for free school meals do worse at GCSE than their more affluent peers?
Causation. Aims to answer if one thing causes another, like does investing in one-to-one tuition improve GCSE history outcomes?
The research question determines the method
A really boring argument is what is the best type of research. Historically, education has been plagued with debates about the merits of qualitative versus quantitative research.
A useful mantra is questions first, methods second. Quite simply some methods are better suited to answer some questions than others. A good attempt to communicate this comes from the Alliance for Useful Evidence’s report, ‘What Counts As Good Evidence?’
Have a go at classifying these questions into the three categories of description, association, or causation.
How many teachers join the profession each year in England?
What percentage of children have no breakfast?
How well on SATS do children do who have no breakfast?
Does running a breakfast club improve pupils’ SATS scores?
How prevalent is bullying in England’s schools?
Are anti-bullying interventions effective at stopping bullying?
Does reading to dogs improve pupils’ reading?
Is it feasible to have a snake as a class pet?
Is there a link between school attendance and pupil wellbeing?
Does marking work more often improve science results?
Effect sizes are a popular way of communicating research findings. They can move beyond binary discussions about whether something ‘works’ or not and illuminate the magnitude of differences.
Famous examples of effect sizes include:
The Teaching and Learning Toolkit’s months’ additional progress
Hattie’s dials and supposed ‘hinge point’ of 0.4
Like anything, it is possible to use effect sizes more or less effectively. Still, considering these four questions will ensure intelligent use.
What type of effect size is it?
There are two fundamentally different uses of effect sizes. One communicates information about an association; the other focuses on interventions. Confusing the two effect sizes leads to the classic statistical mistake of confusing correlation with causation.
Understanding the strength of associations, or correlations, is important. It is often the first step to learning more about phenomena. For instance, knowing that there is a strong association between parental engagement and educational achievement is illuminating. However, this association is very different from the causal claim that improving parental engagement can improve school achievement (See & Gorard, 2013). Causal effect sizes are more common in education; we will focus on them with the remaining questions.
How did the overall study influence the effect size?
It is tempting to think that effect sizes tell us something absolute about a specific intervention. They do not. A better way to think of effect sizes is as properties of the entire study. This does not make effect sizes useless, but they need more judgement to make sense of them than it may first appear.
Let’s look at the effect sizes from three EEF-funded trials (Dimova et al., 2020; Speckesser et al., 2018; Torgerson et al., 2014):
All these programmes seem compelling, and Using Self-Regulation to Improve Writing appears the best. These are the two obvious – and I think incorrect – conclusions that we might draw. These studies helpfully illustrate the importance of looking at the whole study when deciding the meaning of any effect size.
1. Some outcomes are easier to improve than others.
The more closely aligned an outcome is to the intervention, the bigger the effects we can expect (Slavin & Madden, 2011). So we would expect a programme focusing on algebra to report larger effects for algebra than for mathematics overall. This is critical to know when appraising outcomes that have been designed by the developers of interventions. In extreme cases, assessments may focus on topics that only the intervention group have been taught!
There’s also reason to think that some subjects may be easier to improve than others. For instance, writing interventions tend to report huge effects (Graham, McKeown, Kiuhara, & Harris, 2012). is there something about writing that makes it easier to improve?
2. If the pupils are very similar, the effects are larger.
To illuminate one reason, consider that around 13 per cent of children in the UK have undiagnosed vision difficulties (Thurston, 2014). Only those children with vision difficulties can possibly benefit from any intervention to provide glasses. If you wanted to show your intervention was effective, you would do everything possible to ensure that only children who could benefit were included in the study. Other pupils dilute the benefits.
3. Effects tend to be larger with younger children.
Young children tend to make rapid gains in their learning. I find it extraordinary how quickly young children learn to read, for example.
A more subtle interpretation I’ve heard Professor Peter Tymms advocate is to think about how deep into a subject pupils have reached. This may explain the large effects in writing interventions. In my experience, the teaching of writing is typically much less systematic than reading. Perhaps many pupils are simply not very deep into learning to write so make rapid early gains when writing is given more focus.
4. More rigorous evaluations produce smaller effects.
A review of over 600 effect sizes found that random allocation to treatment conditions is associated with smaller effects (Cheung & Slavin, 2016). Effects also tend to be smaller when action is taken to reduce bias, like the use of independent evaluations (Wolf, Morrison, Inns, Slavin, & Risman, 2020). This is probably why most EEF-funded trials – with their exacting standards (EEF, 2017) – find smaller effects than the earlier research summarised in the Teaching and Learning Toolkit.
5. Scale matters
A frustrating finding in many research fields is that as programmes get larger, effects get smaller. One likely reason is fidelity. A fantastic music teacher who has laboured to create a new intervention is likely much better at delivering it than her colleagues. Even if she trained her colleagues, they would likely remain less skilled and motivated to make it work. Our music teacher is an example of super realisation bias that can distort small scale research studies.
Returning to our three EEF-funded studies, it becomes clear that our initial assumption that IPEELL was the most promising programme may be wrong. My attempt at calibrating each study against the five issues is shown below. The green arrows indicate we should consider mentally ‘raising’ the effect size. In contrast, the red arrows suggest ‘lowering’ the reported effect sizes.
This mental recalibration is imprecise, but accepting the uncertainty may be useful.
How meaningful is the difference?
Education is awash with wild claims. Lots of organisations promise their work is transformational. Perhaps it is, but the findings from rigorous evaluations suggest that most things do not make much difference. A striking fact is that just a quarter of EEF-funded trials report a positive impact.
Historically, some researchers have sought to give benchmarks to guide interpretations of studies. Although they are alluring, they’re not very helpful. A famous example is Hattie’s ‘hinge point’ of 0.4, which was the average from his Visible Learning project (Hattie, 2008). However, the included studies’ low quality inflates the average; the contrast with the more modest effect sizes from rigorous evaluations is clear-cut. However, it does highlight the absurdity of trying to compare effect sizes with universal benchmarks.
The graphic below presents multiple representations of the difference found in the Nuffield Early Language Intervention (+3 months’ additional progress) between the intervention and control groups. I created it using this fantastic resource. I recommend using it as the multiple representations and interactive format help develop a more intuitive feeling for effect sizes.
How cost-effective is it?
Thinking about cost often changes what looks like the best bets. Cheap, low impact initiatives may be more cost-effective than higher impact, but more intensive projects. An excellent example is the low impact and ultra-low-cost of texting parents about their children’s learning (Miller et al., 2016).
It is also vital to think through different definitions of cost. In school, time is often the most precious resource.
Effect sizes are imperfect but used well they have much to offer. Remember to ask:
What type of effect size is it?
How did the overall study influence the effect size?
Simpson, A. (2018). Princesses are bigger than elephants: Effect size as a category error in evidence-based education. British Educational Research Journal, 44(5), 897–913. https://doi.org/10.1002/berj.3474
Graham, S., McKeown, D., Kiuhara, S., & Harris, K. R. (2012). A meta-analysis of writing instruction for students in the elementary grades. Journal of Educational Psychology, 104(4), 879–896. https://doi.org/10.1037/a0029185
Hattie, J. (2008). Visible learning: a synthesis of over 800 meta-analyses relating to achievement. Abingdon: Routledge.
Thurston, A. (2014). The Potential Impact of Undiagnosed Vision Impairment on Reading Development in the Early Years of School. International Journal of Disability, Development and Education, 61(2), 152–164. https://doi.org/10.1080/1034912X.2014.905060
Wolf, R., Morrison, J., Inns, A., Slavin, R., & Risman, K. (2020). Average Effect Sizes in Developer-Commissioned and Independent Evaluations. Journal of Research on Educational Effectiveness, 13(2), 428–447. https://doi.org/10.1080/19345747.2020.1726537
Schools are hotbeds of innovation. In my role supporting schools to develop more evidence-informed practice, I always admire teachers’ creativity and dedication. However, I also see colleagues trying to do too many things, including things likely to have limited impact based on the best available evidence.
A clear message from the Education Endowment Foundation’s popular resource on putting evidence to work is that schools should do fewer things better (EEF, 2019). This includes stopping things that are less effective in order to release the capacity to do even better things. In my experience, these messages are beginning to take hold; they also feature prominently in the new national professional qualifications.
At a system level, I think we should do more to stop ineffective initiatives. The Department for Education (DfE) is increasingly good at scaling up initiatives with promise, such as the Nuffield early language intervention (NELI), which, according to multiple rigorous evaluations, has improved children’s communication and language (Dimova et al., 2020).
What about ineffective programmes?
A recent evaluation of Achievement for All’s flagship programme – used by around 10 per cent of schools in England – provides a fascinating case study (Humphrey et al., 2020). The evaluation was concerning: it found children in the control schools did considerably better than their peers in schools using the intervention. The study received the EEF’s highest security rating of five padlocks based on the randomised design, large scale, low dropout and low risk of wider threats to validity. This is on top of the EEF’s exacting standards, involving independent evaluation and pre-specifying the analysis to reduce ‘researcher degrees of freedom’ (EEF, 2017; Gehlbach & Robinson, 2018).
In short, we can be very confident in the headline: children in the Achievement for All schools made two months’ less progress in reading, on average, compared to children in schools that did not receive the programme.
What happened after the evaluation?
The EEF (2020) published helpful guidance for schools currently using the programme, and Achievement for All published a blog (Blandford, n.d.) essentially rejecting the negative evaluation – yet many schools continue to use the programme.
The contrast is stark: when programmes are evaluated with promising results, they are expanded; when evaluations are less positive, there are limited consequences.
What if we actively stopped ineffective interventions?
If we assume that the findings from the evaluations of programmes such as Achievement for All generalise to the wider population of schools already using the programme – a quite reasonable assumption – then investing in stopping it is an excellent investment.
A bold option is to simply pay organisations to stop offering ineffective programmes – think ‘golden goodbyes’. The government, or a brave charity, could purchase the intellectual property, thank the staff for their service, provide generous redundancy payments, and concede that the organisation’s mission is best achieved by stopping a harmful intervention.
If that feels too strong, what about simply alerting the schools still using the programme and supporting them to review whether the programme is working as intended in their own school. Remember, for Achievement for All, this is around 1 in 10 of England’s schools. New adopters of ineffective programmes could be discouraged by maintaining a list of ‘not very promising projects’ to mirror the EEF’s ‘promising projects’ tool, though we may need a better name.
These ideas scratch the surface of what is possible, but I think there is a strong case for using both positive and negative findings to shape education policy and practice.
Finally, there is an ethical dimension: is it right to do so little when we have compelling evidence that certain programmes are ineffective?
This post was originally published in the BERA Blog