Policy Lessons from India`s Total Sanitation Campaign

Policy Lessons from India’s
Total Sanitation Campaign
18 July 2012
open defecation is common.
open defecation is
important.
three starting points
1. open defecation harms early
life health
2. early life health matters for
life-long human capital and
productivity
3. open defecation has negative
externalities, which require
government responsibility
importance of safe excreta disposal
why are children in India shorter, on
average, than children in African countries
that are poorer, on average?
(Deaton, PNAS 2007)
1
-2
-1
0
importance of safe excreta disposal
-3
0
20
40
60
percent of households without toilet or latrine
80
-2.5
-2
-1.5
-1
similar trend among Indian states
0
.2
.4
.6
.8
fraction of population openly defecating
95% CI
linear best fit
1
lasting effects of early life health
• disease in early life has enduring
consequences for human capital (Almond and
Currie, 2011)
• poor health and inadequate nutrition in early
life cause persistent deficits in cognitive
development and ability (e.g. Case and Paxson
2010).
• health promotes growing tall, smart, and
productive
negative externalities of open
defecation
one household’s open defecation can
hurt everybody else, even if
everybody else disposes
of their feces safely
negative externalities of open
defecation
in economic theory, negative
externalities are an important
indicator of government
responsibility
sanitation and the TSC
evidence from India’s
Total Sanitation Campaign
• partially subsidized pit latrine construction, with an
incentive to local leaders to socially motivate use
– over 10 years from 2001 to 2011, about one household
latrine per 10 people in rural India
• real, full-scale implementation by the Indian
government
– external validity (Ravallion 2012, and others)
– estimates average over administrative losses
– large scale  detect effects on mortality
India’s Total Sanitation Campaign
• announced in 1999, began spending money to
build latrines in 2001
 $1.5 billion over 10 years
• focused on low-cost pit latrines; incomplete
subsidy
• emphasis on outcome: becoming open
defecation free
two features may have helped:
1. ex post incentive
for desired
outcome
(Holmstrom &
Milgrom, 1991)
2. made use of
existing social
structure
two types of evidence
randomized experiment
• three districts in
Maharashtra
– … but in fact in only one was
the experiment actually
implemented
• villages randomly selected
for TSC-type intervention in
February 2004
• are children taller in August
of 2005?
non-experimental
• three identification
strategies  IMR & height
– year-to-year variation in
latrine construction by district
– long difference in IMR from
2001 to 2011 census data
– discontinuity in the incentive
to local leaders
• can study actual, large-scale
implementation
experimental results
in Nanded and Nandurbar there was no experiment
mean difference
between
treatment and
control villages,
height-for-age z
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
-0.05
-0.1
before
after
Ahmadnagar
Nanded & Nandurbar
Source: Chaudhury, Moulik, Hammer, Pokharel, & Spears (in progress)
non-experimental results I
are infants born in districts and years
in which many TSC latrines have been
constructed more likely to survive
their first year of life, relative to other
infants born in different years in that
district or in different districts?
data: district level household survey 3
• conducted mainly in 2008, birth history since
January 1st, 2004
• from this, construct repeated cross section of
198,287 infants born alive
• dependent variable: survived: 0; died: 1,000
• independent variable: TSC latrines built by
first year of life from administrative records
effects of TSC on IMR, DLHS-3
evidence of a causal effect
• similar with district-specific linear trends
• no pre-program correlation (“parallel trends”)
• effect on post-neonatal mortality, not on
neonatal mortality
• interactions indicate plausible mechanisms
– bigger effect on children given non-breastmilk food
earlier in first year
– larger effect where population density is greater
• Granger causality: no effect “back in time”
non-experimental results II
census organization’s
Annual Health Survey
census
2001
TSC latrines built
2010-2011
did districts in which more TSC latrines
were built between 2001 and 2010 see
a greater decline in rural IMR than
other comparable districts?
long difference in differences
• similar result: decline in IMR of about 4 deaths
per 1,000 babies born alive
– able to control for other district-level factors
• no evidence against parallel trends
– no “effect” on change in IMR in 1990s or 1980s
• falsification tests
– no “effect” on urban IMR
– no “effect” of institutional delivery program
– no “effect” of public works program with similar
data
effects of early life health on
human capital
cognitive
children’s achievement
height
lower IMR
do children who live in districts that
had more TSC latrines in their first year
of life subsequently grow taller,
relative to other children born in
different years or different districts?
TSC  taller children
• IHDS 2005 data: individual-level identification
strategy identical to DLHS infant mortality
• at mean TSC intensity, children are 0.2 heightfor-age standard deviations taller
• robust to controls, including for height of
older sibling (who was not exposed to
program)
• effect only seen on rural children, not urban
are children who live in districts that
had more TSC latrines in their first year
of life better able to recognize letters
and numbers when they are six years
old, relative to other children born in
different years or different districts?
Source: Spears and S. Lamba (2012)
0.77
0.014
0.76
0.012
0.75
0.01
0.74
0.008
0.73
0.006
0.72
0.004
0.71
0.002
0.7
0.69
0
2001
never TSC
2002
some TSC
Source: Spears and S. Lamba (2012)
2003
latrines per capita
TSC latrines per rural capita
fraction recognizing numbers
TSC  ASER tests
NGP
the clean village prize
clean village prize
• a reward for open defecation free villages
• village chairman receives prize and monetary
incentive at a prestigious ceremony
• interesting to economists:
1. incentivizing the output, ex post
2. discontinuity in incentive
• “once the award was started, the numbers
increased like anything”
an incentive with discontinuities
a step function of village population
• a lot of money for rural India
• … but not enough to move IMR just by
making people richer
village chairmen’s motivation
incentive(population) > cost(population)
• incentive largely to village chairman
• in otherwise similar villages, the chairman will
implement the TSC with more intensity in
villages with populations just above cut-point
discontinuity-based causal identification
• IMR after the program should be lower in
villages with populations just above the cutpoints than in villages just below them
• … and similarly for districts with many villages
just above the cut-points
• discontinuity only for this program
• 2001 population set before program
• none of this uses official TSC data
effects of the NGP
• districts with a greater average prize per
capita among villages:
– built more latrines per capita
– experienced lower 2010-2011 infant mortality
• districts with more villages just above the
discontinuity experienced lower IMR; districts
with more just below had greater IMR
• instrumenting for 2011 census latrine
coverage replicates individual level IMR: -89
so, should the prize be increased?
only with an investment in better
monitoring and prize evaluation
• increasing the prize amount would increase the
incentive for undeserving applications
• this would further burden evaluation resources
• a resulting drop in the quality of monitoring
could further encourage bogus applications
• … and the NGP incentive unravels
Source: R. Lamba and Spears (2012)
TSC to NBA
policy opportunities and risks
cost per average infant death averted
2010 U.S. dollars; J-PAL method (Dhaliwal, et al.)
10
8
6
4
2
0
density
which level of decision-making?
0
.2
.4
.6
fraction of village openly defecating
Source: Kishore and Spears (in progress), NFHS-3
.8
1
village-level information is necessary
year in which TSC data was last updated, by village
2012
2008
2009
2011
2010
accessed 2/2012
policy lessons from the TSC
Improving sanitation – meaning safe
excreta disposal – must be a top priority
for India.
Because open defecation has negative
externalities, it is everybody’s problem,
and requires government action.
policy lessons from the TSC
By promoting and incentivizing latrine
use, the TSC has had positive initial
impacts on children’s health, human
capital, and cognitive achievement.
The TSC and clean village prize together
are a comparatively very inexpensive
way to save babies’ lives.
policy lessons from the TSC
Villages are a critical level of governance
for promoting sanitation and latrine use.
Incentives to local leaders for outcomes are
useful and should be strengthened by both
increasing the monetary incentive and
devoting resources to ensure accurate
evaluation and adjudication.
policy lessons from the TSC
Achieving total sanitation coverage will
require safeguarding the quality of
administrative data, by providing resources
for data sources that bypass political,
bureaucratic, and financial interests.
captured
whose
Interest?
policy
financing
emphasis
CSRSP
TSC
NBA
subsidy
ex post
incentives
?
latrine
construction
latrine
use
?
contractors
a few
pradhans
?