Do retention payments for early career teachers work?

Policy implications of evidence from new NFER analysis

One of the Government’s key priorities is ‘recruiting 6,500 teachers’, which is focussed on teachers in secondary schools, special schools and colleges. The expected falling rolls in primary means it has not been included in this target. Our previous analysis has suggested this may be challenging to achieve because of sustained under-recruitment, although there are green shoots of post-pandemic recovery in the latest recruitment and retention trends. 

Achieving the target could also be costly, depending on what policy mix is deployed to deliver it. While the Spending Review appeared somewhat generous towards schools compared to many other Government departments, the overall spending envelope is not large considering the funding allocated to deal with other challenges such as SEND reform. 

Therefore, Government will need to spend any money allocated to delivering 6,500 teachers very efficiently. One way to do that is to focus the scarce resource that is available towards measures that are targeted to the phases and subjects in greatest need, rather than universally. That approach would imply that policies such as bursaries and retention payments are likely to be key, and preferred to above-inflation pay rises for all. 

The evidence base on early career retention payments (ECRPs) is steadily accumulating and today we have published new analysis of the impact they have on teacher retention. The research, funded by the Nuffield Foundation, explores the impacts of all five schemes that have been piloted since 2018: maths and physics retention payment (MPRP), teacher student loan reimbursement (TSLR), maths phased bursary (MPB), early career payment (ECP) and levelling up premium (LUP).  

These schemes all targeted additional payments to in-service early career teachers (in their first five years, with loan reimbursements paid to teachers in their first ten years) teaching shortage secondary subjects (including maths, physics, chemistry, computing, languages and biology). Some payments were linked to certain local areas and schools. The targeted retention incentive (TRI) is too recent to be able to evaluate yet. 

As the Government prepares its business plan for delivering 6,500 teachers, potentially including the design of a successor scheme to the TRI, here are four key findings from our new research that can guide that work: 

1) Retention payments are probably an effective tool, but our new evidence is less conclusive than previous studies 

Previous impact evaluations of retention payments such as MPRP and MPB suggested that retention payments are associated with reductions in the number of teachers leaving. Previous MPRP and MPB evaluations found statistically significant impacts and reasonably large effects that echoed findings from some studies in the United States. These found positive impacts of similar schemes on teacher retention, although other US studies have shown little or no retention impact. 

Our analysis generally finds impacts that are smaller than these previous estimates. Our findings for MPB are very similar to the DfE evaluation. However, our estimated impact for MPRP is considerably lower than the previous evaluation and not statistically significant (which means we have less confidence than we would like that there is a genuine impact). We investigated the reasons for this difference further and found that both estimates are sensitive to seemingly arbitrary and innocuous aspects of how the analysis is conducted. This suggests there is an underlying level of uncertainty that means the findings from any one study should be interpreted cautiously and within the context of the wider literature. 

Our analysis provides the first estimate of how effective the levelling up premium – payments of up to £3,000 per year introduced in 2022 – has been in improving retention. While it is associated with a reduction in the leaving rate by around one percentage point, the impact is not statistically significant. 

Our estimates of the impact of the ECP and TSLR schemes suggest that the schemes may have been associated with a slight worsening in retention, although neither are statistically significant. Estimating the overall impact of the retention payments, we find the impact is small (leaving rate reduction of 0.7 percentage points) and not statistically significant. 

Overall, this means our study provides some weak evidence that retention payments improve retention, but it is far from conclusive. Viewed within the wider literature, it suggests retention payments may be an important tool for improving teacher retention, but the evidence is somewhat mixed. 

2) Retention payments add most value once the bursary policy is maxed out 

We explore the implications of our analysis and other studies of the impact of retention payments and bursaries for overall value for money. We simulate what impact on teacher numbers might be seen over the long term from spending the same amount of additional money on either bursaries or retention payments. 

The simulation analysis finds that bursaries are considerably more cost effective than retention payments. This is backed up by the MPB evaluation, which found that reducing bursaries and increasing retention payments led to a net loss in overall teacher numbers. 

However, this doesn’t imply that retention payments should be stopped. Bursaries for maths, physics, computing and chemistry are just below the level of the teacher starting salary, a level that DfE has been reluctant to exceed in the past (meaning it is the maximum in practice, although it could be exceeded). The findings therefore imply instead that retention payments can provide additional scope for improving retention and teacher supply as part of a wider strategy and where they are focussed on subjects that already have maximum bursaries. 

3) We found no compelling evidence of groups for whom retention payments are more effective that should guide future policy design 

Our analysis also looks at whether teachers with certain characteristics or working in certain school types or areas are more responsive to retention payments, aiming to glean insights that could improve future scheme design and value for money. 

However, our analysis finds that there is no conclusive evidence of any one scheme or model for retention payment schemes being more effective than any other. We also find no evidence of significant variation of impact by characteristics that is likely to be relevant for the design of future retention payment schemes to maximise their effectiveness. 

This means that policy design can instead be guided by other relevant factors. We believe these should include: 

• contextual factors, such as whether recruitment and retention are a challenge for particular subjects. 
• alignment with other policy approaches, such as the level of bursaries. As noted above, our evidence suggests that retention payments are only likely to be appropriate for subjects where there is no scope for further increasing its bursary first. 
• wider policy factors, such as higher rates for teachers in schools with higher levels of pupil disadvantage, which the evidence shows struggle more with recruitment and retention. 

4) More research is needed 

Any evidence base is always somewhat incomplete and could benefit from further research. However, we see several key policy areas relating to retention payments where further research can be particularly useful: 

1. Replication studies provide value. Our comparative analysis of the impact of MPRP reveals that estimated impacts on retention are sensitive to the particular approach taken to defining eligibility and estimating impact. Multiple studies by different research teams can therefore provide a more rounded and robust overall picture than one single study. Replication studies can often be seen as needless duplication, but are actually crucial for high-quality scientific enquiry.
2. Reversal effects could still be an issue. Our analysis explored whether retention rates worsen again after teachers become ineligible for payments; for example, by moving into their sixth year of teaching when payments are eligible for those in their first five years. Whether there are reversal effects matters enormously for assessments of long-term value for money. However, our sample size was too small to investigate this robustly, so more data from future years will be needed to answer this important question.
3. Recruitment impacts could provide an additional teacher supply boost. Assessing the impact of retention payments on recruitment to ITT was out of scope for our study and challenging to evaluate. Retention payments could have a positive recruitment impact if longer-term policy information was available to potential recruits about future eligibility, but this has not generally been the case in the schemes since 2018. Both recruitment and retention effects could also potentially be stronger if payments were integrated more formally into the pay framework, as current and future eligibility becomes more predictable to teachers.
4. New and on-going schemes can add to our knowledge further. Our evaluation was unable to look at the impact of LUP in its second year, the TRI payment scheme and the longer-term impacts of ECPs and MPB. Therefore, there remain future opportunities to learn about the retention impacts in future data.
5. Impacts may be different in further education (FE). The TRI scheme is an expansion of the LUP scheme, which includes expansion to some teachers in sixth form and further education colleges. The impacts may be different on FE given its unique context. We are currently working with CFE Research for DfE to evaluate the impact of the TRI scheme on the FE workforce.

Recommendations for Government 

Overall, our evaluation findings are partially supportive of the conclusion that retention payments improve teacher retention, but are far from conclusive. 

The implications of our findings for future policy are that retention payments may be a valuable additional tool for improving recruitment and retention when targeted at shortage subjects and may offer good value for money. However, the marginal cost appears to be considerably higher than for training bursaries, suggesting retention payments should only be considered as part of a wider strategy for teacher recruitment and retention for subjects with bursaries that are at a clear maximum. 

We recommend that the Government should maintain a policy of teacher retention payments focussed on shortage subjects, as the impact of eliminating retention payments could lead retention to worsen. Sustaining them long-term as an offer for future early career teachers could also prompt additional recruitment benefits. Alongside, the Government should retain a policy of raising bursaries for subjects experiencing teacher supply challenges where bursaries are low and maintain high bursaries for shortage subjects, raising them over time with the level of the teaching starting salary.