Winnertakeall prizes versus proportional payments to reduce child malnutrition in India Will Masters Friedman School of Nutrition amp Department of Economics Tufts University International Conference of Agricultural Economists ID: 716184
Download Presentation The PPT/PDF document "Performance bonuses for public services:" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Performance bonuses for public services:
Winner-take-all prizes versus proportional payments to reduce child malnutrition in India
Will MastersFriedman School of Nutrition & Department of Economics, Tufts University
International Conference of Agricultural Economists
Vancouver, BC – 2 August 2018
Prakarsh Singh
Department of Economics, Amherst CollegeSlide2
A test of performance bonuses for ICDS workers in ChandigarhSlide3
Performance pay is difficult to use and evaluateMeasurement of performance is costly, affected by noise, time lags and confoundersRewards may crowd out other motivations, and reduce effort on other tasksRewards may drive selection into participation and allocation of effortEfforts to improve child nutrition are difficult to monitor and improve
Inputs (dietary intake and disease exposure) are usually not observedOutcomes (body size, disease state) are difficult to measure and compareLinks between inputs and outcomes are unknownIndia’s ICDS program offers a large-scale opportunity to interveneOver 1 million centers each serving over 30 preschool children, with salaried Anganwadi worker providing mid-day meal, advice to mothers, some teaching
Government aims to improve performance for both nutrition and educationObjectives include reduced weight-for-age malnutrition, which is still widespreadLow weight-for-age, defined as WAZ < -3 or -2 standard deviations below median of a healthy population, can be due to either inadequate diet or disease burden
Performance bonuses for child nutritionmotivation | trial design | outcomes | mechanismsCan performance bonuses help salaried workers improve public services?Slide4
Our contribution: design of incentivesPrevious work in other settings test many kinds of performance payIn health, reviewed for developing countries by Miller &
Babiarz 2014In education, e.g. Behrman et al. 2015 for secondary schools in MexicoPrevious work with ICDS workers to reduce child malnutrition has found:Information plays a central role (Singh 2015 in JEBO, from earlier Chandigarh trials)Goal cards for workers, recipe books for mothersPrincipal mechanism is targeting communication with mothers about at-risk children
Magnitude of incentive matters (Singh & Mitra 2017 in EER, from Kolkata trials)Significant impact for 200 but not 100 Rs per childPayments affect intrinsic motivation (Singh & Masters 2017 in JHE, from Chandigarh)Fixed bonuses of 200
Rs elicit almost as much gain as 200 Rs per child who improvesNow, focus on contests with known payout of equal expected valueContests are widely used in many settings (McKinsey & Co 2009)Fixed timeline and known payout facilitates budgetingRank order tournaments create high-powered incentives (Lazear and Rosen 1981) …but also discouragement (Brown 2011, Cason, Masters & Sheremeta 2010, 2017)Divisible prizes can be paid in proportion to success, incentivizing all improvements
Performance bonuses for child nutrition
motivation
| trial design | outcomeSlide5
Summary of resultsTrial compares two incentive schemes using identical information, offering performance bonuses of identical expected value (<5% of salary) For 85 workers serving about 2,000 children in government-run ICDS day-care centers in urban slums of Chandigarh, India
Mechanism checks address heterogeneity, workers’ job satisfaction and efforts reported by mothers, with check for parallel trends before interventionWe randomize individuals workers within neighborhood clustersCompare standard winner-take-all (WTA) prize paid to highest performer to a proportional reward payment (PRP) divided among all successful caregivers Criterion is number of children whose weight-for-age status improves
Contest is within clusters, offering up to 600 Rs per worker in that clusterGains from proportional rewards exceed gains from winner-take-all prizeespecially after contest ends, with after 6 mo. (5.9 pct pts) vs 3-mo. contest (4.3
pct pts) mechanism is greater impacts from poorer performers, who also report higher job satisfaction after the contest ends Performance bonuses for child nutritionmotivation | trial design | outcomesSlide6
Context for this trial
Previous paper:
fixed bonus vs piece rate
This study:Incentive design
In urban slums of Chandigarh -- Planned city in far north India
-- Capital of both Punjab and Haryana
-- Population size < 2 million
Design and implementation done with ICDS management
-- Geographically separated blocks
-- Previous paper compares block 1 as
controls for trends and seasonality to block 2 for fixed bonus vs. piece rate
-- This study’s data collected in 4 rounds
at 3 month intervals,
Oct 2014 - July 2015,
with surveys of workers,
children and their mothers
Performance bonuses for child nutrition
motivation
| trial design | outcomesSlide7
Design of this trialPerformance bonuses for child nutrition
motivation | trial design | outcomes
Randomization occurred within clusters
Number of Anganwadi workers (centers)
Total
WTA
PRP
Adarsh
3
3
*
Badheri
4
2
2
Buterla
4
2
2
Dhanas
38
19
19
Ramdarbar
29
14
15
Sector 52
7
3
4
85
43
42
Results of individual randomization within neighborhood clusters
Number of children (at first
endline
)
Total
WTAPRPAdarsh121121 *Badheri1326270Buterla935637Dhanas973505468Ramdarbar777374403Sector 522441071372,3401,2251,115
* Adarsh had only 3 centers, so could be allocated only one type of contest, which was randomly picked to be WTASlide8
• Baseline I (October 2014)
• Baseline II (January 2015)
• Assigned to treatment in early February 2015
• Endline I (April 2015)
• Payouts given in early May 2015, no further incentive offered
• Endline II (July 2015)
Timeline of the trial
Pre-trends?
Treatment effects within contest period
Treatment effects after contest ends
Performance bonuses for child nutrition
motivation
|
trial design
| outcomesSlide9
TreatmentsScalable design: Bonus pool in each cluster total Rs. 600 per worker Equal expected value and timing of payment in each treatment arm
In WTA treatment, the highest performer wins the entire bonus poolIn PRP treatment, each successful performer wins their share of the bonus poolSimple implementation: Goal cards with gains needed for each child
Status improvements can be from severe (WAZ<-3) to moderate (WAZ<-2) or to noneStatus improvements exclude any cases of overweight relative to height (WHZ>+1)Bonuses are net of any declines in status into moderate or severe malnutritionBonuses have lower bound of zero
Performance bonuses for child nutritionmotivation | trial design | outcomesSlide10
Baseline descriptive statisticsPerformance bonuses for child nutrition
motivation | trial design | outcomes
WTA
PRP
WTA
PRP
Demographic characteristics
Child diet (foods consumed at least 2x/week)
Child age
4.6
4.6
Milk
1.00
0.99
Child is male
0.50
0.51
Dal
0.96
0.95
Mother age
27.2
27.2
Green Veg
0.60
0.56
Number of children
2.35
2.29
Fruit
0.48
0.49
Household income
8,639
8,380 Sweets0.130.23Mother illiterate0.380.36Egg0.250.27Worker age37.7038.95Dalia0.890.88Worker college0.430.45Chicken0.110.13Roti0.920.94Child weight outcomesChips0.430.38Weight for Age (WAZ)-1.54-1.48Malnourished (WAZ<-2)0.270.28Worker effort Weight (kg)13.7313.81Home visits7.616.41
Talks about child
10.6211.26Worker nutrition knowledgeGroup AW meetings4.293.73Quiz 1 (18 questions)10.7211.10Other mother mtgs8.329.35Quiz 2 (25 questions)15.2115.76Quiz 3 (7 questions)4.484.66Slide11
Expected and actual payouts* In the WTA arm, 3 of 6 clusters had no net improvement, hence no payout
Performance bonuses for child nutritionmotivation |
trial design | outcomes
WTA
PRP
Total number of workers
43
42
Total number of slum clusters
6
5
Expected payout per worker
600
600
Average payout per worker
279*
600
Average payout per payee
4000
1575
Number of workers receiving payouts
3*
16
Number of clusters without payouts
3*
0
Ex-post outcomes
Ex-ante designSlide12
Weight-for-age outcomesPerformance bonuses for child nutrition
motivation | trial design | outcomesSlide13
Results: ATE for child nutrition outcomesDependent variables are Weight (kg), Wfa z (WAZ),
Wfa mal (status indicator)Significance levels shown are * 10%, ** 5%, *** 1%.Performance bonuses for child nutrition
motivation | trial design |
outcomesImpacts of PRP (vs. WTA)
Weight
Wfa z
Wfa mal
Weight
Wfa z
Wfa mal
Weight
Wfa z
Wfa mal
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Within the contest period (after 3 months)
Proportional
0.0764
0.0705*
-0.0432*
0.138
0.0638
-0.0381
0.138
0.0642
-0.0369
Child and mother-level controlsXXXXXXWorker ControlsXXXN234823422342166516651665166516651665
Over the longer term (after 6 months)
Proportional0.2020.0951*-0.0589**0.209**0.0875*-0.0517*0.200*0.0824*-0.0496*Child and mother-level controlsXXXXXXWorker ControlsXXXN232522722272193519341934193519341934Slide14
Results: Heterogeneity among workersDifference to mean payout coefficients are in thousands of rupees
Dependent variables are Weight (kg), Wfa z (WAZ), Wfa mal (status indicator)Significance levels shown are * 10%, ** 5%, *** 1%.
Performance bonuses for child nutritionmotivation |
trial design | outcomesImpacts of PRP (vs. WTA)
Wfa z
Weight
Wfa z
Weight
Wfa z
Weight
(1)
(2)
(3)
(4)
(5)
(6)
Within the contest period (after 3 months)
Proportional (PRP)
0.0703*
0.0703*
0.076
0.0638
0.138
0.0642
0.138
Difference to Mean Payout
0.0111
0.0111
0.0256
0.0181
0.0342
0.018
0.0341
Difference to Mean Payout*PRP
-0.0101
-0.0101
-0.0376
-0.0158-0.0317-0.0163-0.0316 Child and mother-level controlsXXXX Worker ControlsXX N2342234223481665166516651665Over the longer term (after 6 months) Proportional (PRP)0.0947*0.203*0.0880*
0.210**0.0836*
0.202** Difference to Mean Payout0.0351**0.04030.0307**0.0667**0.0278**0.0610** Difference to Mean Payout*PRP-0.0508***-0.0935**-0.0447***-0.0920***-0.0377***-0.0783*** Child and mother-level controlsXXXX Worker ControlsXX N227223251934193519341935Largest gains were among workers with lower payoutsSlide15
Mechanism: Satisfaction and discouragementDifference to mean payout coefficients are in thousands of rupees
Dependent variables are Weight (kg), Wfa z (WAZ), Wfa mal (status indicator)Significance levels shown are * 10%, ** 5%, *** 1%.
Performance bonuses for child nutritionmotivation |
trial design | outcomesImpacts of PRP (vs. WTA) on workers’ self-reported satisfaction with their own ability, their work at ICDS, and their life in general
Ability
Work
Life
Ability
Work
Life
Ability
Work
Life
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Proportional (PRP)
-0.0256
-0.447
0.358
-0.0435
-0.464*
0.197
-0.0613
-0.461*
0.142
Diff. to Mean Payout
-0.0178
-0.170**
0.282***
-0.0188-0.183**0.275***-0.0240-0.186**0.253***Diff. to Mean Payout*PRP0.288***0.0910-0.432***0.295***0.108-0.416***0.314***0.111-0.354***Child and mother-level controlsXXXXXXWorker ControlsXX
X
N419241324242193419121957193419121957Workers who won larger payouts gained confidence in their ability from PRP…but PRP led to higher life satisfaction among those with lower payouts Slide16
ConclusionsContests with known payouts and fixed timelines can improve outcomes even (or especially) after the contest endsBonus of ~5% of salary lowered malnutrition rates by 6% after 6 months, 4% after 3 mo.
Larger gains at 6 months than 3 months could be due to Slow impact of workers’ actions on child weight gainSlow impact of contest design on workers’ actionProportional rewards encourage all workers, not just top performersLarger gains among those with below-average initial outcomesHigher life satisfaction among those below-average workersResults consistent and build on previous work in lab and field
Avoids discouragement effect found in laboratory trials (Cason, Masters & Sheremeta 2010, 2017) and athletic events (Brown 2011)Gains from gift-exchange mechanism (Akerlof 1982) as more workers win somethingReveals which actions benefit which children (Singh and Masters 2017)
Performance bonuses for child nutritionmotivation | trial design |
outcomesSlide17
Some states may want new incentivesPerformance bonuses for child nutrition
motivation | trial design | outcomes
Addressing the media after the budget, she said that her government has focused on delivering essentials….
…also announced 17.06 per cent higher allocation for Women and Child Department.
This includes Rs 250-500 per month incentive for Anganwadi workers based on performance…
Written by
Hamza Khan
| Jaipur | March 9, 2017
Could this be used in practice?Slide18
Thank you!