Lakshmi Iyer
(Harvard Business School)
Impact and Policy Conference 2012
ECONOMIC CONDITIONS,
POLITICAL INSTITUTIONS
AND CONFLICT
Inter-state aka war
Intra-state/internal
Civil war
Terrorism
People-to-people
Organized crime
Targeting specific sections of society
Women
Religious minorities
CONFLICT: MANY TYPES
Inter-state aka war
Intra-state/internal
Civil war
Terrorism
People-to-people
Organized crime
Targeting specific sections of society
Women
Religious minorities
CONFLICT: MANY TYPES
0
10
20
30
40
50
60
Countries with external conflict Countries with internal conflict
INTERNAL CONFLICT IS BECOMING MORE
COMMON THAN INTER-STATE CONFLICT
Source: Author’s calculations from UCDP/PRIO Armed Conflict Dataset Version 4-2012 Conflict is defined as “A contested incompatibility that concerns government and/or territory where the use of armed
force between two parties, of which at least one is the government of a state, results in at least 25 battle-related
deaths.”
0
5
10
15
20
25
30
35
40
45
Conflicts with 25-999 deaths Conflicts with at least 1000 deaths
LOW-LEVEL CONFLICTS ARE BECOMING
MORE WIDESPREAD
Source: Author’s calculations from UCDP/PRIO Armed Conflict Dataset Version 4-
2012
Bangladesh
India
Nepal
Pakistan
Sri Lanka
-8-6
-4-2
02
Log(
conf
lict d
eath
s pe
r ca
pita
, 199
8-20
04)
6 7 8 9 10 11Log GDP per capita 1998 (PPP)
ECONOMIC ROOTS: POOR COUNTRIES
EXPERIENCE MORE INTERNAL CONFLICT
Source: Iyer (2011).
Conflict data is from Global Terrorism Database 2. GDP and population data from World Development
Indicators.
AchhamArghakhanchi
Baglung BaitadiBajhang
BajuraBanke
Bara
Bardiya
Bhaktapur
Bhojpur
Chitawan
Dadeldhura Dailekh
Dang
DarchulaDhading
DhankutaDhanusa
DolakhaDotiGorkha
Gulmi
HumlaIlam
Jajarkot
Jhapa
Jumla
Kailali
Kalikot
Kanchanpur
KapilbastuKaskiKathmandu
KavrepalanchokKhotang
Lalitpur
Lamjung
Mahottari
Makwanpur
Manang
Morang
Mugu
Myagdi
NawalparasiNuwakot
Okhaldhunga
PalpaPanchthar Parbat
ParsaPyuthan
Ramechhap
Rautahat
Rolpa
Rukum
Rupandehi
Salyan
Sankhuwasabha
SaptariSarlahi
SindhuliSindhupalchok
Siraha
Solukhumbu
Sunsari
Surkhet
SyangjaTanahu
TaplejungTerhathum
Udayapur
02
46
0 .2 .4 .6 .8 1Poverty rate 1995-96
Conflict deaths per 1000 population Fitted values
POOR REGIONS WITHIN COUNTRIES
EXPERIENCE MORE CONFLICT: NEPAL
Conflict deaths in Nepal’s civil war 1996-2006
Source: Do and Iyer (2010).
Notes: Conflict data collected at district level by the Informal Sector Service Center,
Kathmandu. Poverty rate from Nepal Living Standards Survey, 1995-96.
0
500
1000
1500
2000
2500
Poverty<20% 20%<Poverty <=40% Poverty>40%
# conflict deaths per 1000 population 1998-2007
POVERTY AND CONFLICT ACROSS INDIAN
DISTRICTS
Source: Iyer (2011). Notes: Conflict data is from the RAND-MIPT Terrorism Incident Database (1998-2007). Conflict is defined as “Violence
calculated to create an atmosphere of fear and alarm to coerce others into actions they would not otherwise
undertake, or refrain from actions they desired to take. Acts of terrorism are generally directed against civilian
targets.” District level poverty rates are Head Count Ratios (percentage of population below the state-specified poverty
line) computed from National Sample Surveys 1999-2000.
Mandated political representation of women increasingly
common
India’s Panchayati Raj: 1/3rd of all local (district/village) councils to
consist of women
50% in Afghanistan village development councils (Fotini et al, 2012)
Many other developed and developing countries have gender quotas
for elected officials.
Panchayati Raj elections in India held at dif ferent dates by
dif ferent states.
Can assess the effect by comparing states before and after
women gain such representation.
POLITICAL EMPOWERMENT OF TARGETED
GROUPS
EFFECTS OF FEMALE POLITICAL
REPRESENTATION IN INDIA
Source: Iyer et al (2012). Crime data obtained from National Crime Records Bureau, New Delhi.
Crime variables are number of crimes per 1000 population (gender-specific population for gender-specific crimes).
Results control for demographic characteristics, real per capita state GDP, policy strength per capita, state and year fixed
effects and state-specific time trends.
-20%
-15%
-10%
-5%
0%
5%
10%
15%
20%
25%
30%
35%
% Change after Panchayati Raj Implementation
CRIMES AGAINST WOMEN OTHER CRIMES ARRESTS FOR CRIMES
AGAINST WOMEN
• Evidence from two newly created databases Religious identity of state level legislators (based on names)
Incidents of Hindu-Muslim violence based on news reports in the Times of India (extend the Varshney-Wilkinson database from 1995 to 2010).
Episodes of religious violence are not rare in India: Hindu-Muslim riots occurred in every year over the period 1980-2007, and in all but two states.
On average, 7-8% of state legislators; this is below the Muslim population share of 13%.
There is a decline in the occurrence of Hindu -Muslim riots after 1992 (except for the events of 2002 in Gujarat).
WHAT ABOUT POLITICAL REPRESENTATION
OF RELIGIOUS MINORITIES?
What is the impact of having a Muslim MLA on the
probability of occurrence of a Hindu -Muslim riot?
Confounding factors: places which elect Muslim MLAs might
be very dif ferent from places which do not.
We will therefore compare places where Muslim candidates
narrowly won elections to places where Muslim candidates
narrowly lost, as a means of controlling for such
unobservable dif ferences across dif ferent areas.
MUSLIM LEGISLATORS AND HINDU-
MUSLIM VIOLENCE
Sample All districts Districts with close elections between Muslims and
non-Muslims
Vote margin to define close elections 5% 3% 2%
Average probability of riot in sample 5.8% 9.5% 9.5% 8.4%
Effect of having a Muslim MLA in the district 0.3% -1.0% -4.5% -6.0%
MUSLIM LEGISLATORS AND HINDU-MUSLIM
VIOLENCE (PRELIMINARY EVIDENCE)
Source: Bhalotra, Clots-Figueras and Iyer, 2012. Effect of having a Muslim MLA is computed from regressions which control for year and district fixed effects.
Internal conflict/civil war has strong economic roots.
Political representation can be an effective means to protect
targeted communities.
Rebuilding economic growth and political institutions are
likely to be instrumental in preventing further conflict.
But these are big challenges in post -conflict environments.
Survey data from South Asian countries indicates the
following characteristics of job markets in conflict zones:
Preponderance of agricultural jobs and a lack of manufacturing and
construction jobs.
Education and skill deficits among the workforce.
Greater labor force participation, but a large incidence of unpaid and
casual jobs.
POLICY CONCLUSIONS
Source: Iyer and Santos (2012).
Bhalotra, Sonia, Irma Clots-Figueras and Lakshmi Iyer (2012),
“Politician Identity and Religious Violence,” Work in Progress.
Do, Quy Toan and Lakshmi Iyer (2010), “Geography, Poverty and
Conflict in Nepal.” Journal of Peace Research , 47 (6), 2010.
Iyer, Lakshmi (2011), “Managing Conflict.” In Ejaz Ghani, ed.,
Reshaping Tomorrow: Is South Asia Ready for the Big Leap? ,
Oxford University Press.
Iyer, Lakshmi, Anandi Mani, Prachi Mishra and Petia Topalova
(2012), “The Power of Political Voice: Women's Political
Representation and Crime in India." American Economic Journal:
Applied Economics (forthcoming).
Iyer, Lakshmi and Indhira Santos, 2012. “Creating Jobs in South
Asia’s Conflict Zones.” World Bank Policy Research Working
Paper WPS 6104.
PAPERS CITED
RECONSTRUCTING INSTITUTIONS AFTER VIOLENT CONFLICT
F O T I N I C H R I S T I A , M I T
Bangkok
2012
MOTIVATION
• Presumed causal link between institutional quality
and development.
• Post-conflict areas replete with new institutional
designs.
• Context renders assessment highly cumbersome.
• As a result, lack of consistent rigorous empirical
evidence on institutional effect on governance,
broader stability and economic well being.
VARIATION IN INSTITUTIONAL QUESTIONS OF INTEREST
• Afghanistan (with Andrew Beath (WB) and Ruben Enikolopov (NES))
• Counterinsurgency Context
• Institution under Evaluation: CDD Program
• Bosnia-Herzegovina (with Marc Alexander (Stanford)
• Ethnic Reintegration Context
• Institution under Evaluation: First Integrated School
OPERATIONALIZATION
Country Units Partners Funding Years $
Afghanistan 500 villages;
~15,000
respondents
MRRD; WB; 7
local & intl.
NGOs
MRRD; WB;
USAID;
UNWFP;
CIDA; IGC
5 2 M
Bosnia and
Herzegovina
3 schools;
244 students
Mostar
Municipality;
OSCE
Russel Sage;
Harvard Univ.
1 15K
AFGHANISTAN: CDD PROGRAM
• The National Solidarity Program (NSP) is Afghanistan’s largest development Program.
• Present in over 29,000 of Afghanistan’s 38,000 villages.
• Sponsored by international donors and run by MRRD.
• Implemented by NGOs in two main stages:
• Election of Community Development Councils (CDCs) through secret-ballot election.
• Project Selection.
Similarity of Treatment
and Control Villages
Create
CDCs
Treatment
Villages
(NSP)
Select
Projects
Implement
Projects Projects
Finished
Control
Villages (Non-NSP)
May – Oct. 2009 May-Nov. 2011
Interim Estimates Final Estimates
Baseline
Survey
Aug. – Sep. 2007
1st
Follow-
Up
Survey
2nd Follow-
Up Survey
followed by
Wheat
Distribution
The evaluation estimates impacts by collecting data over four years in 500 villages:
250 NSP (treatment) & 250 non-NSP (control)
Structure of Evaluation and Data Collection
SUBTREATMENT INTERVENTIONS
• Comparing two ways
of electing councils
• Single Member Districts
• Multi-member Districts
• Comparing two ways of
selecting projects
• Referendum
• Community Consultation
Meeting
QUESTIONS FOR CDD EVALUATION
• Random assignment allowed us to examine range of
impact-related questions:
• Both substantive:
• Can development aid reduce insecurity?
• Can development aid improve women’s lives?
• And operational:
• Do electoral rules affect who gets elected?
• Does method of project selection affect capture of resources?
NSP’S INTERIM IMPACTS
• NSP reduces insecurity by “winning hearts and minds.”
• NSP improves women’s economic and social status in the community.
• Higher quality council members if elected in multi-member districts.
• Lower levels of elite capture if projects selected via referenda.
NSP BEHAVIORAL INTERVENTION
• Additional behavioral intervention to triangulate
attitudinal results.
• Instead of behavioral games, we measured behaviors
through real life event in rural Afghanistan.
• Compared outcomes in targeting and corruption
during a food aid distribution, in villages with
customary governance structures versus villages with
democratically elected councils.
NSP BEHAVIORAL INTERVENTION FINDINGS
• Existence of multiple institutional structures can lead
to institutional competition and underperformance.
• When in the lead, better aid targeting and higher
participation.
• When not in the lead, higher levels of embezzlement
and lower levels of participation.
OTHER RANDOMIZED CDD EVALUATIONS
• CDD increasingly popular: US$1.3 billion per year in lending in 2000-2008 by WB towards ~50 CDD projects.
• Fearon et al (2009) in Liberia find little impact on economic well being but some on governance.
• Barron et al (2009) in Indonesia and Casey et al (2011) in Sierra Leone identify some positive effects on economic outcomes but none on sociopolitical measures.
• Humphreys et al (2012) in DRC find no effect.
BOSNIA: SCHOOL INTEGRATION
• Does post-conflict institutional integration improve inter-ethnic cooperation?
• Partial integration of high schools leaving student allocation to segregated or integrated environment to chance.
• N-person public goods game to assess institutional effect on people’s willingness to contribute to a public good.
FINDINGS OF SCHOOL INTEGRATION ON PUBLIC GOODS PROVISION IN
BOSNIA • On average, integrated
institutions significantly increase contribution to public goods.
• Availability of sanctions in integrated institutions drives up contributions significantly but has no such effect in segregated institutions.
• Diversity of groups leads to lower public goods contribution only in the context of segregated institutions.
OTHER STUDIES
• RCTs on peer effects, teacher pupil ratios, teacher
incentives, teacher performance, report cards,
uniforms, textbooks, school meals, primary school
deworming, educational incentives for parents and
children. But not in conflict contexts.
• Exception: Burde and Linden (2012)who find that in
Afghanistan village-based schools in villages that
lack public schools, significantly increase enrollment
and test scores and eliminate gender disparity.
ISSUES AND RECOMMENDATIONS
NO ONE-SIZE-FITS-ALL INSTITUTIONAL INTERVENTION
• Fetishization of new institutional frameworks post
conflict, with emphasis on representation.
• New institutions not necessarily a panacea as
indicated by variant results of CDD impact.
• Focus should be placed on specific contextual
needs and interventions that also look to existing
institutional frameworks and interactions between
old and new.
PARTNERSHIP WITH ACADEMICS
• Organizations faced with post-conflict challenges often find it too cumbersome to pursue rigorous evaluation for their projects.
• Opportunities for organizations to link up with academics who can design optimal evaluation relevant to the context taking advantage of existing discontinuities and exogenous variation.
• Academics can also raise considerable funds, allowing program to afford better assessment than its budget would allow.
BETTER COORDINATION FOR REPLICATION
• Hard to generalize or even compare findings of
studies that look at post-conflict institutions as
different:
• Context
• Intervention
• Measures
• Replication remains very important and requires
higher levels of coordination in interventions studied
as well as in indicators and measures employed.
MORE SHARING
• Notable progress in sharing research designs.
• Posting of pre-analysis plans also important in protecting from data mining and reporting bias.
• But too many people in the field still facing same logistical and operational problems because not enough discussion on realities of implementation in post-conflict context.
• EGAP, JPAL and other venues should encourage more public sharing on field challenges pertaining to better access, more appropriate enumeration, better human subjects protection.
Reintegrating Combatants into Civilian Life
Cyrus Samii, New York University
Rigorous evaluations of reintegration programs
• Humphreys & Weinstein
(2007, Sierra Leone)
• Annan, Blattman, Carlson &
Mazurana (2007, Uganda)
• Barron, Humphreys, Paler &
Weinstein (2009, Aceh)
• Gilligan, Mvukiyehe & Samii
(2012, Burundi)
• Blattman & Annan (2012,
Liberia)
• Studies involved university,
IGO, and NGO partnerships.
• Study costs varied from tens
of thousands to hundreds of
thousands.
Typical program concept: first generation
economic integration
inputs
political integration
Typical program concept: critiques
economic integration
inputs
political integration
context:
social
economic
psychological
Typical program concept: second generation
economic integration
inputs
political integration
context:
social
economic
psychological
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Studies in Burundi, Liberia, and Sierra Leone.
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Positive impact with poverty reduction, quality of livelihood
(Burundi, Liberia).
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Positive impact with poverty reduction, quality of livelihood
(Burundi, Liberia).
• No “downstream” effect in Burundi, and Sierra Leone,
economic & political outcomes uncorrelated. But in Liberia,
some evidence that jobs deters participate in violence.
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Positive impact with poverty reduction, quality of livelihood
(Burundi, Liberia).
• No “downstream” effect in Burundi, and Sierra Leone,
economic & political outcomes uncorrelated. But in Liberia,
some evidence that jobs deters participate in violence.
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Studies in Sierra Leone and Uganda.
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Abusiveness of faction predicted social acceptance (Sierra
Leone), but not economic or political outcomes.
• Social uncorrelated with economic or political (Sierra Leone).
• Emphasis on psych. factors seems misplaced (Uganda).
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Studies in Aceh and auxiliary evidence from Burundi.
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
• Community perceptions of ex-combatants were actually
worsened by CDD programming.
• In Burundi, coordination of community-based and individual
based programming failed (major time gaps).
Evaluating the program concept
economic integration
inputs
political integration
context:
social
economic
psychological
Implications
• For policy:
– We are pretty good at livelihood enhancement, but it’s not clear that
how much this contributes to political integration. We need more
study of the latter.
– We have failed to understand or find ways to contribute to producing
social contexts more conducive to integration.
• In building knowledge capacity:
– Despite the sensitive contexts, rigorous research can be done to
study program concepts for combatant reintegration.
– The results so far indicate that much more study is indeed needed to
ensure limited programming resources are applied efficiently.