A school system offers teachers a bonus if student results improve. The logic sounds straightforward: reward strong performance and people will focus more intensely on what produces it.
But a performance-pay scheme does not reward “good teaching” directly. It rewards a measurement rule.
That distinction changes everything.
If the bonus depends on test scores, teachers may focus more attention on tested content. If it depends on pupil growth, the model must decide whose growth belongs to which teacher. If a threshold determines payment, effort may concentrate around pupils near the cutoff. If teachers compete for individual rewards, collaboration can weaken. If the metric is noisy, a teacher can work exceptionally well and still lose the bonus—or receive one because of cohort variation rather than better teaching.
The Education Endowment Foundation’s current Teaching and Learning Toolkit estimates only a small average positive impact for performance pay—around one month of additional progress—and rates the evidence security very low. EEF also warns that incentives can narrow attention toward tested outcomes or pupils near performance thresholds. These concerns do not prove every incentive scheme is harmful. They show that incentive design changes behaviour through the metric chosen, sometimes before it changes instructional quality.
This article owns one narrow canonical job on eduKateSG: teacher bonuses or salary progression linked to measured performance, including metric design, effort allocation, threshold effects, attribution noise, collaboration and unintended consequences. It does not own recruitment or retention premiums, ordinary salary scales, professional appraisal, teacher development or school accountability more broadly.
The useful question is not, “Should excellent teachers earn more?” It is: what behaviour does this particular pay rule make rational, and does that behaviour actually improve learning?
The 50-second answer
Teacher performance pay works through incentives. The scheme selects a measured outcome, attaches money to it and changes the costs and benefits of teacher choices.
- Define the educational outcome precisely.
- Use measures teachers can influence but not manipulate easily.
- Reduce noise through multiple evidence sources where appropriate.
- Watch for threshold effects and curriculum narrowing.
- Protect collaboration.
- Avoid rewarding outcomes outside a teacher’s reasonable control.
- Make rules transparent.
- Keep professional development and working conditions in the system.
- Evaluate distributional effects, not only average scores.
- Compare performance pay with alternative uses of the same money.
The shortest principle is: performance pay does not incentivise teaching in the abstract; it incentivises whatever the payment formula counts.
1. Incentives change attention
Teachers already have professional motives, relationships and responsibilities. A financial incentive adds another signal about what the organisation values.
That signal can increase effort toward measured outcomes.
It can also shift attention away from unmeasured outcomes.
2. Measurement design is the centre of the intervention
A bonus based on examination pass rates creates a different incentive from one based on growth, classroom observation, attendance or team outcomes.
Each metric has strengths and vulnerabilities.
The pay scheme cannot be understood separately from its measurement rule.
3. Test-score incentives can narrow curriculum
If only Mathematics and English scores count, teachers and leaders may rationally protect those subjects at the expense of others.
Within a subject, tested formats can receive more attention than broader knowledge or transfer.
Curriculum narrowing is an incentive response, not necessarily misconduct.
4. Thresholds create discontinuities
Suppose a bonus depends on the percentage of pupils reaching a benchmark.
Moving a pupil from far below to moderately below may not affect payment. Moving a pupil just below the benchmark to just above it may.
Teachers can rationally concentrate effort near the threshold unless design counteracts the effect.
5. Growth measures try to reward progress rather than level
A teacher with a low-attaining intake should not be judged only on final attainment.
Growth measures compare outcomes with a baseline.
They can be fairer conceptually and statistically noisy in small classes.
6. Value-added models depend on assumptions
Statistical models attempt to estimate the contribution of teachers while accounting for prior achievement and other variables.
Results can vary with model specification, cohort size and missing data.
Do not treat a single estimate as a perfect measure of teacher quality.
7. Small classes create high statistical volatility
One or two unusual pupils can shift average results materially.
A teacher should not receive large pay changes because of random cohort composition.
Multi-year measures can reduce noise and make incentives slower.
8. Team teaching complicates attribution
Students may learn from subject teachers, specialists, tutors and teaching assistants.
Which adult “owns” the outcome?
Individual bonuses can misrepresent collaborative production of learning.
9. Team incentives solve one problem and create another
A school-wide bonus can reward collective success and protect collaboration.
An individual teacher may then feel their effort barely changes the probability of payment.
Incentive strength and collective responsibility trade off.
10. Subject comparability is difficult
Standardised results may be available in some subjects and not others.
A system can end up rewarding teachers in tested subjects under one rule and everyone else under another.
Perceived fairness matters for motivation and retention.
11. Observation-based pay adds professional judgement
Classroom observations can capture questioning, explanation, routines and curriculum implementation.
They also contain observer error and can encourage performative lessons when high stakes.
Training, multiple observations and clear rubrics reduce but do not eliminate these problems.
12. Student surveys can add perspective and create popularity pressure
Students experience clarity, respect and classroom climate directly.
Survey data can be useful and influenced by grading, subject difficulty, demographics and expectations.
Use cautiously in pay decisions.
13. Multiple measures can reduce single-metric gaming
A balanced formula may combine achievement, observation and professional contribution.
Complexity can become opaque.
Teachers should be able to understand how decisions are made.
14. Complexity weakens the behavioural signal
If the formula contains twenty weighted indicators, teachers may not know what actions affect the reward.
That can reduce incentive strength while increasing administrative burden.
Design needs enough breadth for validity and enough simplicity for intelligibility.
15. Targets can be absolute or relative
An absolute target rewards anyone who reaches a standard. A tournament rewards the top performers relative to peers.
Relative competition can discourage collaboration and create uncertainty about what is enough.
Absolute standards can become too easy or impossible depending on context.
16. Incentives can change effort rather than skill
A teacher may work longer hours, run extra classes or provide more feedback.
If weak subject knowledge or pedagogy is the bottleneck, additional effort may have limited effect.
Pay cannot substitute for capability-building.
17. Professional development changes the production function
Training can improve what teachers know how to do.
Incentives can change how intensely they do it.
The two mechanisms can complement one another but should not be confused.
18. Resources constrain response to incentives
A teacher may want to improve laboratory instruction and lack equipment. A teacher may want to provide more feedback and face an extreme marking load.
Financial incentives do not remove structural constraints automatically.
Working conditions belong in the analysis.
19. Intrinsic motivation need not disappear when money appears
Teachers can care deeply about students and respond to incentives.
The simple story that money always “crowds out” professionalism is too strong.
But poorly designed incentives can communicate mistrust or distort priorities.
20. Perceived fairness influences response
Teachers are more likely to accept a scheme when measures are transparent, achievable and linked to work they can influence.
Opaque or apparently arbitrary payouts can damage morale.
Procedural justice is part of implementation.
21. High-stakes metrics invite gaming pressure
When substantial money depends on one measure, schools may alter student entry, testing conditions, classification or instructional emphasis.
Not all strategic response is fraud. Some is predictable optimisation around the rule.
Design should anticipate behaviour at the margin.
22. Excluding difficult students is an unacceptable incentive pathway
If results improve when low-performing pupils are removed from the denominator, the scheme contains a dangerous signal.
Eligibility and attribution rules must protect students from strategic exclusion.
Equity needs to be designed into the formula.
23. Incentives around attendance require causal caution
A teacher can build relationships and routines that affect attendance and cannot control illness, transport or family circumstances.
Rewarding raw attendance can penalise teachers serving higher-need communities.
Context and reasonable influence matter.
24. Pupil thresholds can shift effort distribution
EEF explicitly warns about attention concentrating on pupils near test thresholds.
Growth-based designs can reduce this incentive, though no formula is perfect.
Audit who receives additional support under the scheme.
25. High achievers can be neglected under minimum-threshold incentives
If the reward depends on pass rates, a student already safely above the threshold contributes little marginal value.
Stretch and enrichment can receive less attention.
Every incentive formula contains an implied distribution of effort.
26. Very low achievers can also be neglected
When reaching the threshold seems unlikely, teachers may rationally focus elsewhere.
This is precisely the group that often needs intensive support.
Equity auditing is essential.
27. Individual bonuses can weaken sharing
Teachers may become less willing to share materials or strategies if peers are competitors for a fixed reward.
Most schemes do not intend this outcome.
Competition design can produce it anyway.
28. Collaborative cultures produce shared learning
Curriculum planning, moderation and professional learning depend on trust.
Performance pay should be evaluated for its effect on staff networks, not only pupil scores.
Team incentives or non-competitive bonuses may protect collaboration differently.
29. Recruitment and retention premiums are different
Paying more to attract scarce teachers to hard-to-staff subjects or schools changes labour supply.
It does not reward measured performance after teaching.
EEF distinguishes these mechanisms, and this article does too.
30. Base salary still matters
A small bonus on top of an uncompetitive salary may have little motivational or retention effect.
Performance pay cannot repair an unsustainable compensation system.
Compare incentives with broader workforce policy.
31. Bonus size affects salience
A tiny payment may not influence behaviour. A very large payment increases stakes, gaming pressure and perceived risk.
The relationship is not automatically linear.
Pilot and evaluate.
32. Payment timing matters
A bonus delivered a year after the measured behaviour may be weakly connected psychologically.
Faster payment can strengthen the link and may require waiting for validated data.
Accuracy and immediacy trade off.
33. Teachers should know the rules before the performance period
Retroactive criteria undermine trust.
Publish metrics, weights, eligibility and dispute procedures.
An incentive cannot guide behaviour if the rule is unknown.
34. Appeals processes matter because metrics can be wrong
Data errors, student attribution mistakes and missing scores occur.
Teachers need a credible route to correct factual errors.
Due process supports legitimacy.
35. Performance pay can increase administrative work
Observation, score processing, target setting, verification and appeals consume time.
Include these costs in evaluation.
A scheme should not create more bureaucracy than educational value.
36. School leaders can become metric managers
When bonuses depend on data, leaders may spend substantial attention validating evidence and negotiating targets.
That can improve clarity and displace instructional leadership.
Opportunity cost applies to management time too.
37. Students can perceive incentive pressure
Teachers may communicate higher stakes around tests when pay depends on results.
This can increase focus and anxiety.
Monitor student experience.
38. Parents may misinterpret bonuses as objective quality rankings
A teacher receiving no bonus may still be excellent under a noisy formula.
Do not publish simplistic league tables from compensation metrics.
Performance measures are decision tools, not complete identities.
39. Short-run gains may not represent durable learning
Intensive test preparation can raise measured performance without broad transfer.
Use assessments that reflect the curriculum and examine later outcomes where possible.
Teaching to worthwhile constructs is different from teaching to item quirks.
40. Narrow metrics can miss relational and developmental work
Teachers support belonging, behaviour, language and long-term intellectual habits.
Not every valuable outcome is easily monetised in a formula.
Measurement systems should acknowledge limits.
41. No metric system can fully observe teaching quality
Performance pay therefore operates under incomplete information.
The design problem is to choose measures sufficiently valid for incentives without pretending they capture everything.
Humility is part of governance.
42. Pilot schemes should pre-specify unintended outcomes
Do not measure only the target score.
Track curriculum breadth, teacher collaboration, subgroup support, retention, workload and perceived fairness.
Incentives can improve one metric while damaging another.
43. Compare against alternative spending
The same money might fund coaching, smaller classes, curriculum resources, retention premiums or additional planning time.
EEF’s average impact estimate is small and evidence security very low.
Opportunity cost is central.
44. A bonus can reward existing advantage
Teachers serving high-performing or more advantaged students may find targets easier depending on formula design.
Contextual growth models attempt to address this and introduce statistical complexity.
Distributional analysis is essential.
45. A bonus can reward improvement in difficult settings if designed carefully
Growth, team outcomes or context-aware targets can recognise progress from lower baselines.
Check whether the model really does so rather than assuming fairness from technical terminology.
Simulate payouts on historical data before launch.
46. Incentives should not replace professional standards
Teachers still need clear expectations, appraisal, support and accountability independent of bonus eligibility.
A payment scheme is one policy instrument.
It should not become the entire theory of teaching quality.
47. Performance pay should have a stop rule
If evaluation shows negligible gains, high gaming, poor fairness or collaboration damage, the scheme should be redesigned or ended.
Policies should not persist merely because launch costs were high.
Incentive systems need feedback too.
48. The endpoint is better teaching, not better metric performance
The central test is whether the incentive changes teacher behaviour in ways that produce broad, durable learning.
If the metric rises while the curriculum narrows or inequity grows, the rule has succeeded administratively and failed educationally.
Worked case 1 — The threshold problem
A bonus depends on the proportion of pupils achieving a pass grade. Teachers begin running extra sessions mainly for pupils just below the cutoff.
Results improve, while pupils far below and far above receive less additional attention.
The district redesigns the formula around growth bands and audits support distribution.
The original behaviour was rational under the original metric.
Worked case 2 — The noisy small class
A teacher of twelve advanced students loses a bonus after two pupils have unusual exam-day circumstances. Her observed teaching remains strong.
The school moves from one-year raw results to multi-year evidence plus moderated observations.
The measure becomes less volatile, slower and more complex.
Worked case 3 — Competition weakens sharing
A school awards only the top twenty percent of teachers. Shared-resource meetings become less open and teachers stop swapping revision materials.
Leaders had wanted excellence and created a tournament.
The scheme changes to a non-rank-based school-and-individual model with common performance standards.
Worked case 4 — The tool that could not fix capability
A system introduces generous bonuses for Mathematics growth in schools where teachers report weak subject-specific professional development.
Some teachers increase hours but struggle to change explanations and diagnostic practice.
Leaders add sustained instructional coaching. Incentives and capability-building are evaluated separately.
Practical route for policymakers and leaders
Begin with the educational behaviour you hope to change, then ask whether the metric validly captures it. Simulate likely incentives, especially thresholds and subgroup effects. Protect curriculum breadth and collaboration. Publish rules, data-quality procedures and appeals.
Pilot, measure unintended consequences and compare performance pay with other uses of the budget.
Practical route for teachers
Understand the formula but do not let it replace professional judgement. Document data errors, watch whether the metric distorts who receives attention and continue collaborating where possible.
If the scheme exposes an instructional weakness, seek development rather than simply adding hours.
Practical route for parents and students
Do not interpret bonuses as complete rankings of teachers. Ask how teaching quality is evaluated and whether incentives affect curriculum or assessment pressure.
A compensation metric is one imperfect signal inside a much larger system.
Common failure modes
- Rewarding a metric as if it were teaching quality itself.
- Using raw attainment without accounting for context.
- Using noisy growth estimates for tiny classes.
- Creating threshold incentives that redirect effort.
- Neglecting pupils far above or below the cutoff.
- Narrowing curriculum to tested material.
- Creating competitive tournaments that weaken collaboration.
- Using opaque formulas teachers cannot understand.
- Ignoring data errors and attribution problems.
- Assuming more effort fixes weak pedagogy.
- Failing to provide professional development.
- Rewarding outcomes teachers cannot reasonably influence.
- Penalising teachers in higher-need contexts through poor modelling.
- Ignoring subjects without standardised tests.
- Using high-stakes observations that encourage performance theatre.
- Ignoring student anxiety and experience.
- Failing to cost administration.
- Calling recruitment premiums performance pay.
- Never evaluating unintended effects.
- Keeping a scheme because it is politically visible rather than educationally effective.
Frequently asked questions
What is teacher performance pay?
It is compensation—usually bonuses or salary progression—linked to measured performance such as student outcomes, observations or other indicators.
Does performance pay improve attainment?
EEF currently estimates a small average effect of around one month of progress and rates the evidence security very low. Effects vary with scheme design.
Why can incentives narrow teaching?
People rationally allocate effort toward what is measured and rewarded. If the metric samples only part of the curriculum, other outcomes may receive less attention.
What are threshold effects?
If payment depends on crossing a benchmark, effort can concentrate on pupils or outcomes near that benchmark because moving them has the largest effect on payout.
Are growth measures fairer?
They can better reflect progress from different starting points, but statistical estimates can be noisy and depend on modelling choices.
Should bonuses be individual or team based?
Each has trade-offs. Individual incentives can be stronger and risk competition; team incentives support collaboration and can weaken the link between individual effort and reward.
Is performance pay the same as paying more in hard-to-staff schools?
No. Recruitment or retention premiums change labour supply and placement. Performance pay links compensation to measured outcomes after or during teaching.
Can observations be used?
Yes, but observations need trained raters, multiple samples and transparent criteria. High stakes can change teacher behaviour during observed lessons.
What should leaders monitor besides scores?
Curriculum breadth, collaboration, subgroup attention, teacher workload, retention, perceived fairness, gaming and student experience.
What is the strongest design principle?
Model the behaviour the payment formula makes rational before launching it.
Evidence boundary and caveats
Performance-pay studies vary in bonus size, metrics, labour-market context, teacher autonomy and baseline compensation. Some schemes combine incentives with coaching or accountability, making components difficult to isolate. Student outcomes are also affected by many factors outside one teacher’s control.
EEF’s very-low-security rating is therefore important. A small average effect does not justify assuming any formula will work. Incentive design is the intervention.
Sources and further reading
- Education Endowment Foundation, Performance pay, Teaching and Learning Toolkit, updated October 2025: https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/performance-pay
- Education Endowment Foundation, teacher professional development and implementation resources: https://educationendowmentfoundation.org.uk/
Continue exploring on eduKateSG
- How X Works Hub
- How Rapid Learning Cycles Work in Education
- How Instructional Coaching Works
- How Class Size Changes Learning
- How Academic Learning Time Works
- How Education Works | Teacher Professional Development
The final idea
Performance pay is an experiment in organisational attention.
The formula tells teachers which outcomes have financial weight. Good design can focus effort on valuable learning. Bad design can make the measurable become more important than the important.
The question is not whether money motivates people. It often does. The educational question is whether the metric points motivation toward teaching that remains broad, fair and durable after the bonus is paid.