Incentives and performance in public service
delivery
Oriana Bandiera, LSE
SITE Conference on Public Sector Effectiveness
Stockholm, December 2017
Selection and motivation in public service delivery
• Individuals sort into jobs according to their preferences,
skills, and the jobs’ own attributes (Roy 1951)
• Recent theory (Akerlof and Kranton 05, Besley and Ghatak
05) and survey evidence suggest that:
• individuals who sort into public service jobs have
stronger pro-social preferences
• these jobs tend to have low material incentives
Do incentives crowd out pro-sociality?
Pro-social preferences and material incentives both align
agents’ interests with the principal ⇒ negatively
correlated in equilibrium
Yet, the correlation is often taken as evidence that
offering more generous rewards would:
worsen the performance of existing workers by crowding
out their pro-social motivation
attract agents with weak pro-social preferences who perform
poorly on non-incentivised dimensions
This is often used to justify keeping public sector pay low
Yet, incentives improve performance
In Ashraf et al (15) we design an experiment to test
crowd-out directly in the context of public health:
We measure pro-social motivation with a lab game
We then test whether agents with higher motivation
perform worse if offered incentives
We find crowding in, that is, more pro-social agents
respond *more* to both financial or social rewards
RCT evidence in education and health in LICs show large and
sustained improvement in performance
Muralidharan&Sudaraman 11, Duflo et al 12, Miller et al 12,
Olken et al 12)
and actually crowd in prosocial motivation
Selection experiments are rarer
Dal Bo et al (15) randomise wage offers for a civil
service position in Mexico and find that higher wages
attract better qualified applicants with the same PSM
(Perry)
but maybe once on the field they will underperform?
Deserranno (17) shows that earnings can act as a signal
for the nature of the job when this is uncertain (in her
case hybrid sales+health promotion job)
but will this be relevant for standard civil service
positions?
We should look on the extensive margin
Context: rural Zambia, hard to fill positions in remote
locations, common to all SSA
• In 2010 MoH launched a new program: the CHA
• Meant to address shortages in rural areas by
formalising community health workers positions
• Career advancement (salary progression) within the civil
service can, but does not have to, be used as incentive
• “What will happen now that they see themselves as
civil servants? will they retain their connection to the
community?” (Mr Mwila, MOH HR director, 6/ 2010)
We run a nationwide nurse recruitment experiment
The experiment
• Test how career incentives affect (i) who applies to
public health jobs (ii) who is selected and (iii) their
performance through selection
• Key challenge: identify selection effects (frommotivation on the job)
Identification strategy
• New cadre, job attributes unknown to potential applicants
• Identification strategy:
• vary career incentives when recruiting agents
• [opens selection channel]
• provide the same incentives to all agents once hired
• [shuts down effort channel].
Experimental design
First phase of CHA program: recruit 2 CHAs from 165
communities in 48 (out of 58) districts
Experimentally vary the salience of career and social
benefits by means of different recruitment posters at
the district level
Experimental design
Randomized at district level across the country (48
districts), stratified by province & rate of high school
attainment
Yields balanced sample ⇒ CHA applicants in two
treatments are drawn from similar areas, to work in
similar areas
Same eligible population
– 4% eligible, of which 13% unemployed, 8%
housework, 30% self-employed (mostly farming), 34%
employees
Context map
Control
REPUBLIC OF ZAMBIA
MINISTRY OF HEALTH
ONE-YEAR COURSE IN COMMUNITY HEALTH
The Ministry of Health of the Republic of Zambia is launching a new national Community Health Worker (CHW) strategy and invites
applicants to participate in the inaugural training of community health workers.
The training will begin on 30th August 2010 and will be held at the Provincial level for selected applicants. All participation costs,
including transportation, meals and accommodation will be covered by the Ministry of Health.
BENEFITS:
Learn about the most important health issues in
your community
Gain the skills you need to prevent illness and
promote health for your family and neighbors
Work closely with your local health post and
health centre
Be a respected leader in your community
QUALIFICATIONS:
Zambian National
Grade 12 completed with two “O” levels
Age 18-45 years
Endorsed by Neighborhood Health Committee
within place of residence
Preference will be given to women and those
with previous experience as a CHW
APPLICATION METHOD:
Submit to the DESIGNATED HEALTH CENTRE
indicated above:
Completed application form with necessary
endorsements. If no blank forms are attached to
this notice, kindly obtain a blank one at the
nearest health centre.
Photocopy of school certificate documenting
completion of Grade 12 and two “O” levels.
Photocopy of Zambian national registration card.
For more information: Contact the designated health
centre indicated above.
CLOSING DATE: 30th JULY 2010. Only shortlisted candidates will be contacted for interview.
TRAINING OPPORTUNITY
DESIGNATED HEALTH CENTRE: FOR POSTING AT:
Form RP-C
Treatment
REPUBLIC OF ZAMBIA
MINISTRY OF HEALTH
ONE-YEAR COURSE IN COMMUNITY HEALTH
The Ministry of Health of the Republic of Zambia is launching a new national Community Health Worker (CHW) strategy and invites
applicants to participate in the inaugural training of community health workers.
The training will begin on 30th August 2010 and will be held at the Provincial level for selected applicants. All participation costs,
including transportation, meals and accommodation will be covered by the Ministry of Health.
BENEFITS:
Become a highly trained member of Zambia’s
health care system
Interact with experts in medical fields
Access future career opportunities including: o Clinical Officer o Nurse o Environmental Health Technologist
QUALIFICATIONS:
Zambian National
Grade 12 completed with two “O” levels
Age 18-45 years
Endorsed by Neighborhood Health Committee
within place of residence
Preference will be given to women and those
with previous experience as a CHW
APPLICATION METHOD:
Submit to the DESIGNATED HEALTH CENTRE
indicated above:
Completed application form with necessary
endorsements. If no blank forms are attached to
this notice, kindly obtain a blank one at the
nearest health centre.
Photocopy of school certificate documenting
completion of Grade 12 and two “O” levels.
Photocopy of Zambian national registration
card.
For more information: Contact the designated
health centre indicated above.
CLOSING DATE: 30th JULY 2010. Only shortlisted candidates will be contacted for interview.
TRAINING OPPORTUNITY
DESIGNATED HEALTH CENTRE: FOR POSTING AT:
Form RP-T
Career advancement = salary increase (CHA=$290, Nurse= $530, EHT=$615, doctor=$1625)
Part I: Sorting and Selection
(the prequel)
If hired an applicant will get utility U=M+sH(a), where
M=material benefits
H= public health output
a= ability (cognitive & non-) H’>0
s= prosociality
Thus U is increasing in a,s
Assume cov(a,s)=0 in the population
his outside option in the private sector is V(a), V’>U’>0
V’>U’ (reward to talent higher in the private sector)
Is there a trade-off between talent and pro-sociality?
A candidate will apply if U>V
a
V(a)
U(a)
a’
U, V
agents with a<a’ will apply
The threshold is increasing in s
a
U(s1)
U(s2)
a(s1) a(s2)
U, V
V(a)
even though (a,s) are not correlated in the population
higher ability applicants are more prosocial
The application frontier
a(s) st U(a(s))=V(a(s))
a
M/b
The figure is drawn under the assumption that U=sW+M and V=ba
U>V, apply
U>V, do not apply
s
The effect of increasing M
increasing M
s
a
common support linea(s=1, M=MC)
The effect on prosociality depends on ability
increasing M
s
a
common support linea(s=1, M=MC)
prosociality is lower for inframarginal applicants
prosociality is higher for marginal applicants
Crowd- out?
o for all levels of ability in the common support, increasing M attracts less pro-social applicants
o but increasing M attracts more talented applicants
o thus the marginal applicant will be more pro-social
o effect on average applicant is ambiguous
o effect on hired applicant depends on the selection mechanism
o we analyse this next
CHA inputsfacility
utilisation
health
outcomes
treatment
applicants
CHAs
treatment
applicants
CHAs
new data on the universe of applicants
1585 applicants: skills, pro-sociality, social connections
161 selection panels: composition, ranking of every applicant
Measuring social preferences
o “Adapted Inclusion of Others in Self (IOS) scale”
(Overlapping circles): measures the extent to
which individuals perceive community and self-
interest as overlapping (Aaron et al 1992)
o Plans to remain in the community
IOS Scale
grade 12 scores
the number of courses taken in biology and other natural
sciences
career motivation
next TE on the applicant pool
Measuring skills
TE on the pro-sociality of the applicants is negative only
at low ability levels
prosociality pink line is treatment effect by skill rank
low ability applicants are less pro-social in T
high ability applicants are equally pro-social in T
TE on applicants skills is positive or nil
skill
pink line is treatment effect by skill rank
=1 if
selected p-value
=1 if
selected p-value
=1 if top 3 in skills X treatment 0.121*** 0.158***
(0.0287) (0.0351)
=1 if top 3 in skills X control 0.122*** 0.128***
(0.0374) (0.0391)
=1 if top 3 pro-sociality X treatment 0.0952** 0.0810*
(0.0386) (0.0408)
=1 if top 3 pro-sociality X control 0.0576* 0.0560
(0.0291) (0.0341)
=1 if aims to higher rank X treatment 0.0973** 0.0933**
(0.0410) (0.0378)
=1 if aims to higher rank X control 0.0698** 0.0654*
(0.0311) (0.0335)
=1 connected to village leader X treatment 0.00983
(0.0389)
=1 if connected to village leader X control 0.0283
(0.0267)
=1 connected to health centre staff X treatment -0.0383
(0.0676)
=1 if connected to health centre staff X control -0.0009
(0.0395)
0.62
0.68
0.98
0.4
0.57
0.54
0.63
0.59
Panels face different pools but value the same traits:
choose most talented, and hence most prosocial
TE on applicants: lower pro-sociality, higher skills
treatment control
Cognitive skills
random draw: median 24.9 > 23
Pro-sociality
random draw: median 2.45 < 2.54
TE on hired candidates: same pro-sociality,
higher skills
treatment control
Cognitive skills
panel selection 27.2 > 25.6
random draw: median 24.9 > 23
Pro-sociality
panel selection 2.55 = 2.55
random draw: median 2.45 < 2.54
Higher (lower) sensitivity of U to prosociality (ability)
implies that the marginal applicant will have the highest
ability AND highest pro-sociality
Treatment thus attract two groups of agents who would
have not applied otherwise
high ability/high prosociality
low ability/ low prosociality
In line with common concern: treatment attracts “the
wrong”/low prosociality applicants
but the concern is moot as they will not be selected
Taking stock on the elusive trade-off
CHA inputsfacility
utilisation
health
outcomes
treatment
better applicants
better
CHAs
Part II: Treatment Effect on
Performance
CHA inputsfacility
utilisation
health
outcomes
data
HH visits
Community meetings
HP records
Births
ANC
Vaccinations
Health practices
Anthropometrics
electronic platform + admin
MOH admin Own HH survey
Household visits: Context
CHAs work in remote, low-density area; median 78
square km area has 200 households
CHAs either walk or use bikes
Roads are bad: 10km take between 1.5 hours and
4.5 hours to traverse on bike
Step 1: Treatment agents do 31% more visits
40
Do career CHAs underperform on other tasks or other
dimensions?
Fewer meetings?
Fewer patients at HPs?
More likely to drop out?
Shorter visits?
Focus on easy-to-reach households?
Worse targeting within households?
Less responsive to emergencies?
Work shorter hours/devote less time to “informal” tasks?
41
Do career CHAs underperform on other tasks or other
dimensions?
Fewer meetings? actually twice as many
Fewer patients at HPs? no
More likely to drop out? no
Shorter visits? no
Focus on easy-to-reach households? no
Worse targeting within households? no
Less responsive to emergencies? no
Work shorter hours/devote less time to “informal” tasks?
no
42
Step 2: facility utilisation
CHAs' main remit is mother and child health, encourage:
institutional deliveries & regular checks for children and
immunisations
Using admin data we show that career CHAs:
⇑ Institutional deliveries by 30%,
⇑ children visits by 24%
⇑ children weighted by 22%
⇑ polio vaccination by 20%
43
Step 3: health outcomes
Information
Dependent variable
% of correct
answers in
medical
knowledge test
=1 if child
under 2 yr
old is
breastfed
=1 if child's
stool are
safely
disposed
number of
deworming
treatments
=1 if child
exposed to
CHA is on
track with
immunization
schedule
Career incentives 0.002 0.051** 0.121*** 0.225* 0.047**
(0.010) (0.023) (0.039) (0.129) (0.020)
household controls yes yes yes yes yes
child controls no yes yes yes yes
Mean of dep var in control .740 .641 .595 1.44 .058
Adjusted R-squared 0.057 0.561 0.161 0.263 0.024
N 738 613 736 659 462
Health practices
44
Bottom line: malnourishment falls by 25%
Dependent variable
=1 if weight for
age z score <2
SD (moderately
or severely
undernourished)
=1 if weight for
age z score <3
SD (severely
undernourished)
=1 if
MUAC<12.5
(moderately
or severely
wasted)
=1 if
MUAC<11.
5 (severely
wasted)
Career incentives -0.053* -0.028* -0.023 -0.014
(0.030) (0.015) (0.015) (0.014)
household controls yes yes yes yes
child controls yes yes yes yes
Mean of dep var in control .210 .051 .036 .014
Adjusted R-squared -0.006 0.003 0.018 0.017
N 582 582 581 581
Anthropometrics
45
Plan
more
inputs
more
instdeliveries
25% lower
malnutrition
treatment
better
applicants
better CHAs
Incentives affect selection both by changing the applicant
pool and the chosen agents
In line with common wisdom, offering material benefits
attract low prosociality agents but this is moot because
these have low ability and are not hired
Holds as long as (i) return to ability is higher in the private
sector and (ii) return to prosociality higher in public
More to gain where returns to ability are larger
there’s no tradeoff between talent and prosociality
Lessons for export