Demographic Pressure and Institutional Change:
Village-Level Response to
Rural Population Growth
in Burkina Faso
Margaret S. McMillan, William A. Masters and Harounan Kazianga
NBER African Successes Project Conference, Zanzibar
August 3-5, 2011
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Can local public services and land rights
be explained by population size?
•
•
•
•
Study design and preview of results
Identification strategy
Data and survey methods
OLS and 2SLS results
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Study design in one slide
• Burkina Faso has had unusual shocks to rural population
– Onchocerciasis control => migration into river valleys;
– Cote d’Ivoire (CDI) politics => forced repatriation
• Spatial variation in shocks across villages could permit
identification of demographic effects on public action
–
–
–
–
Low/rising densities could explain weak/stronger public sector
We have population census data for 1985, ‘96 and 2006
Use village distance to rivers and CDI to instrument exposure
Use village elders to recall changes in village amenities
• Causality is still unclear; goal is to describe correlation
between population shocks and village amenities
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Measuring amenities with time-varying indexes
• Public services – four distance metrics of proximity to:
(1) Public Services and Utilities: office used to register births, savings
& loan facility, fixed-line telephone, mobile phone reception;
(2) Public Infrastructure: road passable all year, seasonal road, bus
stop, primary school, secondary school, health center;
(3) Religious Services: any church, mosque or temple;
(4) Markets: open-air with storage facilities, livestock, or shop.
• Property rights – a principal-components index of:
(a) Use rights to cropland (undefined, communal, familial, individual)
(b) Transactions for cropland (parcels ever sold or rented)
(c) Authorities for resolving conflicts (informal, traditional or elected)
(d) Commons demarcation and regulation (pasture land, forest)
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Linking amenities to population shocks
• Migration adds variance to natural increase in population,
and migration is tied to river valleys and the CDI border,
• Interpretation of causality is limited, as usual.
– Migrants could differ from non-migrants, and
– Proximity to rivers and CDI could drive both via other channels.
• Novelty is in the data on rural amenities, and their
correlation with population shocks
– Here, we pre-committed to study design before data collection
so no selection bias in our methods or hypotheses
– Future papers can do much more with these data…
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Preview of results
• We find strong links between amenities and population
– Instrumented population effects are much larger than OLS
– Elasticities of distance to population are between 0.4 and 1.4
• Magnitude of variance and effect sizes are large
– The increase in proximity of public services associated with
moving from the 1st to 5th quintile of predicted village population
is larger than the 10 years of time trend from 1996 to 2006
• No correlation between population and property rights
– Instrumented estimate is larger but not significantly positive
– Significant time trend with 2006 values above 1985 and 1996
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Onchocerciasis Control in West Africa (1974-2002):
Aerial spraying over 11 countries for 14 years
to eliminate blackfly vectors
Estimated Onchocerciasis Prevalence in West Africa
Prior to control (1974)
After control (2002)
Source: WHO, Onchocerciasis Control Programme (www.who.int/apoc/onchocerciasis/ocp).
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Repatriation from Cote d’Ivoire (1999-2003):
As many as 1 million returnees in total,
out of 2.25 m. total Burkinabe in CDI in 1998,
contributing to 13 m. total population in Burkina in 2003
Foreign population in Cote d’Ivoire census of 1998
Source: IRIN News, “In-depth: Cote d’Ivoire crisis”, November, 2002 (www.irinnews.org).
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Villages’ distance to rivers and to Cote d’Ivoire
Mali
Niger
Ghana
Cote d’Ivoire
Togo
Benin
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Survey Method
• Sample is all 747 villages in the national farm survey
of the Office of Agricultural Statistics in Burkina Faso
• Respondents were a focus group of village leaders
convened by the enumerator
• Survey asks respondents to characterize:
–
–
–
–
the situation now and when it was established;
the previous situation and when it was established;
and the situation before that, and when it was established;
variables are either continuous (distance to services), or
categorical (land rights).
• Responses permit construction of 3-step time series
– we use only the situation in 1985, 1996 and the present
– some villages failed to report some data, but all n>2,000
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Questionnaire design: distance to services
N°
V.1
V.1.1
V.1.2
V.1.3
V.2
V.2.1
V.2.2
V.2.3
Questions
Réponse
Distance (en km)
Année d’établissement
Distance entre le village et l’administration centrale (pour les registres des naissances)
La situation actuelle
|____|____|____|
|____|____|____|____|
La situation précédente
|____|____|____|
|____|____|____|____|
La situation antécédente
|____|____|____|
|____|____|____|____|
Distance entre le village et la route praticable par car ou camion toute l’année
La situation actuelle
|____|____|____|
|____|____|____|____|
La situation précédente
|____|____|____|
|____|____|____|____|
La situation antécédente
|____|____|____|
|____|____|____|____|
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Questionnaire design: land rights
N°
Questions
Réponse
VIII.1
Type de droit appliquée pour les terres de culture
(si la réponse est non, mettre des croix à année de début d’application)
Type de droit appliquée
(1=Oui ; 0=Non)
Année de début
d’application
VIII.1.1 Propriété individuelle
|____|
|____|____|____|____|
VIII.1.2 Propriété collective-familiale
|____|
|____|____|____|____|
VIII.1.3 Propriété collective-communautaire
|____|
|____|____|____|____|
VIII.2
Location, vente et prêts de terres de culture
(si la réponse est non, mettre des croix à année de début d’application)
Possibilité de transaction Année de début
(1=Oui ; 0=Non)
d’application
VIII.2.1 Est-ce que la terre peut-être louée ?
|____|
|____|____|____|____|
VIII.2.2 Est-ce que la terre peut-être vendue ?
|____|
|____|____|____|____|
VIII.2.3 Est-ce que la terre peut-être prêtée ?
|____|
|____|____|____|____|
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Construction of indexes for distance to services
• Distance responses were aggregated into four types:
– Public Services: any office used to register births, savings & loan
facility, fixed-line telephone, and/or mobile phone reception;
– Public Infrastructure: any road passable all year, seasonal road,
bus stop, primary school, secondary school, health center;
– Religious Service: any church, mosque or temple;
– Marketplaces: any market with storage, livestock, or shops.
• Distance indexes constructed in three ways:
• Proximity of the farthest service in the basket
• Proximity of the nearest service in the basket
• Average proximity of all goods in the basket
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Mean value of distance indexes for census years
All types of public services are being provided
increasingly close to surveyed villages
Proximity to all sources (km)
(Average distance to all services)
Public Public Religious MarketYear Services Infrast. Services places
1985
Proximity to closest source (km)
(Distance to closest site in each set)
Public Public Religious MarketServices Infrast. Services places
26.968 14.798
6.476
8.603
18.349
3.628
3.551
4.828
1996 25.142 11.629
4.740
7.788
15.155
1.856
2.308
3.734
2010
3.043
6.116
8.648
0.502
1.175
1.948
19.724
8.164
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Construction of indexes for land rights
• Institutions governing land rights were aggregated using
principal components analysis, into a single index whether:
– cropland rights are undefined, communal, familial or individual;
– transactions for cropland include sales or rentals;
– authorities to resolve conflicts are informal, traditional or elected;
– pasture and forest is demarcated and restricted via quotas/taxes
• Index is defined with higher values for greater rule of law
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Mean value of land rights index
Land rights improved sharply from 1996 to 2006
Year
1985
1996
2006
Principalcomponents
index of land
rights1
-0.096
-0.035
0.140
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Regression specification
• Main regression is:
i jt     Pjt   jt   jt
– Where:
i is infrastructure or institutional service to the village;
P is population of the village;
 are fixed effects for 45 provinces and 3 years;
 is error term;
j indexes villages (≈16 per province);
t indexes time (1985, 1996, 2006)
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
OLS estimates of main regression
Population
Y=1996
Y=2006
Constant
Observ.
R-squared
Proximity of farthest source (km)
Proximity to all sources (km)
Services Infrastr.
Religion Markets Services Infrastr.
Religion
Markets
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
0.045*
0.003
0.153*** 0.182*** 0.041*
0.049** 0.145*** 0.187***
-0.059 0.174*** 0.127**
-0.030
0.025
0.193*** 0.167***
0.034
0.027 0.485*** 0.493***
0.072
0.255*** 0.538*** 0.515*** 0.222***
-3.29*** -3.18*** -2.44*** -3.20*** -2.97*** -2.71*** -2.14*** -3.03***
2,132
2,161
2,040
2,136
2,132
2,161
2,040
2,136
0.097
0.147
0.233
0.220
0.116
0.220
0.265
0.225
Population
Y=1996
Y=2006
Constant
Observ.
R-squared
Proximity to closest source (km)
Services Infrastr.
Religion Markets
(9)
(10)
(11)
(12)
0.108*** 0.116*** 0.125*** 0.228***
0.191*** 0.305*** 0.225*** 0.193***
1.109*** 0.611*** 0.490*** 0.628***
-2.69*** -1.40*** -1.50*** -2.64***
2,132
2,161
2,040
2,136
0.230
0.234
0.232
0.235
Land
rights
(13)
-0.001
0.063
0.230***
0.02
2,170
0.217
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
First stage results for IV estimates
Dependent variable: Village population
(1)
Distance to closest river
-0.130***
Distance to Cote d’Ivoire
Year==1996
0.083*
Year==2006
-0.212***
Constant
7.209***
Observations
R-squared
F-Stat Inst
2,130
0.176
(2)
-0.276***
0.081*
-0.213***
8.595***
2,130
0.167
(3)
-0.120***
-0.142
0.083*
-0.212***
8.101***
2,130
0.177
15.14
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
2SLS estimates of main regression
Proximity of farthest source (km)
Services Infrastr.
Religion
Markets
(1)
(2)
(3)
(4)
Pop. shock 0.417** 0.658*** 1.377***
0.555**
Y=1996
-0.091 0.122**
0.045
-0.060
Y=2006
0.103 0.623*** 0.794***
0.154*
-5.82*** -7.63*** -10.8***
-5.73***
Constant
Observ.
2,093
2,121
2,000
2,096
Pop. shock
Y=1996
Y=2006
Constant
Observ.
Proximity to closest source (km)
Services
Infrastr.
Religion
Markets
(9)
(10)
(11)
(12)
1.169***
0.471** 0.655*** 1.121***
0.104 0.271*** 0.181***
0.122*
1.331*** 0.677*** 0.620*** 0.835***
-9.89*** -3.81***
-5.12** -8.71***
2,093
2,121
2,000
2,096
Proximity to all sources (km)
Services Infrastr.
Religion
Markets
(5)
(6)
(7)
(8)
0.558*** 0.968*** 1.147***
0.707***
-0.019 0.119**
0.098
-0.008
0.360*** 0.734*** 0.761***
0.339***
-6.48*** -8.95***
-8.99***
-6.57***
2,093
2,121
2,000
2,096
Land
rights
(13)
0.367
0.037
0.308***
-2.48
2,130
Demographic Pressure and Institutional Change
Study Design | Identification | Data | Results
Conclusion
• Can local public services and property rights be explained
by size of local population?
– For public services, a strong correlation with population shocks:
– link to shift from 1st to 5th quintile of predicted population is
larger than the common time trend from 1996 to 2006
– For property rights, spatial variance is less significant:
– variance in this dataset is linked only to the national trend
• Looking forward, we’ve shown that:
– Villagers’ recall data can be used to study public sector changes;
– Low population density could help explain weak public services;
– Burkina Faso is responding to higher density with more services.