Teachers Incentives: Theory and Practice

1
Teachers Incentives Theory and Practice

• Victor Lavy
• The Hebrew University of Jerusalem, CEPR and NBER
• OECD Conference, Amsterdam
• November 19, 2004

2
Background and Policy Context
? Performance pay for teachers being introduced
in many countries, amidst opposition from
unions ? Rational teachers may be motivated
by incentive pay ? However, there is little
evidence on the effect of teachers
incentives at schools.
3
Examples of recent incentives programs
- In USA Denver, Dallas, Cincinnati,
Phila Iowa, Arizona - The UK, 2002
agreement with teachers unions - New Zealand,
2001, a system wide program - Israel,
2000-2002, a system wide program ?
4
Four key questions

• ? What are the principles that should guide the
  design of optimal teacher incentives?
• ? How much is performance affected by incentives
  for current teachers?
• How much is performance affected by changing the
  pool of teacher applicants?
• Do teacher preferences align with those of their
  students and of society in general?
• What The policy implications and ways to remedy
  existing distortions.
• ?

5
Payment For output

• ? Teachers compensation
• Payment on output
• Payment for input
• Payment on output has two advantages incentives
  and sorting
• Incentives
• - Align teacher incentives with those of
  students and society
• - Align agenda of teachers and students
• - Align preferences of teachers and students
• But two major problems with incentives
• - Presumes that we can agree on what is
  desirable
• - Assumes that we can measure it accurately
• - Potential distortions
• - Measuring output
• Continuous or discrit standard
• Team Incentives

6
Payment on output disadvantages

• Payment on output has also disadvantages
• - Earning potential is relevant goal of
  education
• - But it is observed only later in life
• - Instead proxies are used
• - Test scores or other performance measures
• - Potential distortions teach to the test,
  cheating
• - Measuring output
• Continuous or discrete standard
• Team Incentives
• ?

7
Summary of Design Challenges
? Paying for output versus paying for inputs?
? How should performance be measured? ?
Absolute/Relative P-measures? ? Continuous or
discrete standard? ? Individual teachers
versus team contributions? ? How should the
rewards be structured? ? How generous should
rewards be? ? Distortions teaching to the
test, cheating ? Spillover effects /
externalities due to incentives?
8
Paying For Performance, and Teachers Effort,
Productivity and Grading EthicsVictor
LavyNBER Working paper 10622June 2004
9
The Experiment

• An individual performance bonus paid to teachers
  on the basis of their own students achievements
• - English, Hebrew/Arabic, Mathematics teachers
• - Teaching in advance of matriculation exams
• - Structure rank order tournaments
• - Multiple tournament entry, ranked each time
• - Ranking criterions passing rate and mean
  score
• - Ranking base actual outcome - predicted
  outcome
• - All teachers who had a positive residual in
  both

10
The Experiment
Outcomes were divided into four ranking groups
A/score P/rate
T/points First 16 20 36 Second 12 15 27 T
hird 8 10 18 Fourth 4 5 9
11
The Experiment
Prizes
3036 points 7,500 2129 points5,750 1020
points3,500 9 points1,750
12
The Experiment
Participants 629 teachers English - 207 (97
awarded) Math 237 (124 awarded) Languish 185
(80 awarded)
13
The Experimental Design
- 100 eligible schools - 50 chosen for
treatment - Assignment rules of eligible to
treatment matriculation rate lt 45
14
Individual Incentives Effects on Math Outcomes
15
Individual Incentives Effects on English Outcomes
16
Individual Incentives Effects Grading Ethics
17
Additional Results
Who are the successful teachers? Teachers
ranking in tournament is not correlated with
teachers characteristics gender, age,
experience, education
Teachers behavioural changes observed?

• Effort significant additional after school
  instruction
• more intensive after school preparation before
  exam
• Teaching methods more individualized
  instruction,
• more tracking by
  ability,
• adapting
  t-methods to students ability
• Grading Ethics no change, due to monitoring and
  penalties

18
The Effect of Teachers Group Performance
Incentives on Students AchievementsVictor
LavyJournal of Political EconomyDecember 2002
19
The Experiment

• A team performance bonus paid to all teachers on
  the basis of the school average performance
• - All teachers in school share the bonus
• - Ranking criterions relative level and
  relative gain
• - Relative to other school with same SES
  background
• - Rank order tournament
• - Performance measures dropout, credits up to
  minimum required for matriculation, and mean
  matriculation score
• - The top third schools win awards

20
Group Incentives Effects On High School Outcomes
21
Targeted Individualized Instruction to Under
Performing StudentsLavy Victor and Analia
SchluserNBER Working Paper 10575May 2004
22
The Estimated Effect of Targeted Individualized
Instruction on the Matriculation Rate of Treated
Schools
23
The Estimated Effect of Targeted Individualized
Instruction on the Matriculation Rate of Treated
Students
24
The Estimated Effect of Targeted Individualized
Instruction on the Matriculation Rate by Ability
Quartiles
25
Comparison of Alternative ProgramsCosts and
Benefits
26
Cost Benefit Comparison

• The teacher incentive program (Lavy, 2004)
• Cost 170 per student - a 4 increase MR, from
  41 to 45.1
• Student bonuses (Angrist and Lavy 2004)
• Cost 300 per student - a 68 increase in MR,
  from 19 to 26

27
Conclusions

• Teachers may be expected to respond to
  incentives inherent in compensation structures
• General theories of compensation should apply to
  teaching
• Those theories suggest that output-based pay is
  best used when output is well defined and easily
  measured
• Input-based pay is best when jobs are inherently
  risky and when output is not easily observed.
• The experimental evidence shown suggest that
  teachers incentive are effective in schools
• Individual more than team incentives
• The distortions created by incentives can be
  controlled
• More experiments are needed