In Search of a Spatial Equilibrium in the Developing World
Douglas Gollin
Martina Kirchberger
David Lagakos
Oxford
Trinity College Dublin
UCSD and NBER
April 5, 2017
0 / 64
Large Rural-Urban Income Gaps in Developing World
1 / 64
Large Rural-Urban Income Gaps in Developing World
• Value added per worker (Gollin, Lagakos and Waugh, 2014)
• Consumption measures (Young, 2014)
• Wages (Herrendorf and Schoellman, 2015)
1 / 64
Large Rural-Urban Income Gaps in Developing World
• Value added per worker (Gollin, Lagakos and Waugh, 2014)
• Consumption measures (Young, 2014)
• Wages (Herrendorf and Schoellman, 2015)
• Important role in accounting for income differences across countries
• Developing countries have low income largely b/c they employ most of
their workers in the least productive activities/areas • Restuccia, Yang, Zhu (2008); Vollrath (2009); McMillan & Rodrik (2011)
1 / 64
Appealing Explanation: Spatial Equilibrium
• Assumption: utility equalized across regions (Rosen, 1979; Roback, 1982)
• If not, people would have moved
• Implication: higher consumption but lower amenities in urban areas
• Almost every paper with a spatial equilibrium: amenities backed out of
structural model, not measured directly
2 / 64
Very Incomplete List of Papers Assuming a Spatial Equilibrium
• Baum-Snow and Pavan (2012) and Desmet and Rossi-Hansberg (2013) on
U.S. city wage and size distribution • Desmet and Rossi-Hansberg (2014) on dynamics of U.S. manufacturing
• Allen and Arkolakis (2014) on welfare impacts of transportation
infrastructure • Glaeser & Gottlieb (2009) on agglomeration economies in U.S.
• Bryan and Morten (2014) on spatial misallocation of labor in Indonesia
• Ahlfeldt, Redding, Sturm, Wolf (2015) on fall of Berlin wall
3 / 64
What We Do • Go searching for spatial equilibrium in 20 African countries • Link 2005 Demographic & Health Surveys (DHS) and other data to
population data from Gridded Population of the World (GPW) • Measure how consumption and amenities vary with population density
within each country • Measures of durable goods, housing quality, child health, indoor air
pollution from DHS • Outdoor air pollution using satellite-derived data • Crime data from Afrobarometer, Living Standards Measurement Surveys • Internal migration data from DHS 4 / 64
What We Find
• Almost all metrics, in almost all countries, constant or increasing in
population density • True even within people of similar education levels
• Migration is on net to cities and large in magnitude
• Hard to reconcile with spatial equilibrium
5 / 64
What We Find
• Almost all metrics, in almost all countries, constant or increasing in
population density • True even within people of similar education levels
• Migration is on net to cities and large in magnitude
• Hard to reconcile with spatial equilibrium
• Easier story: residents of developing world moving toward cities since offer
better living conditions on average • Of course, looking for something and not finding it doesn’t prove its not
there – we can’t provide proof of non-existance 5 / 64
Simple Model
6 / 64
Model Environment
• Economy divided into J regions • Identical households with preferences: U(c, h, a) I I I I
consumption, c housing, h amenities, a strictly increasing in all three
• Households mobile, choose where to locate • Three characteristics of each region: I I I
wage, wj housing price, pj amenities, aj
• Household budget constraint: wj = c + pj h
7 / 64
Spatial Equilibrium
• Assumption: utility equated across regions; common utility value ≡ u¯
8 / 64
Spatial Equilibrium
• Assumption: utility equated across regions; common utility value ≡ u¯
• Basic property I: For any regions j and k, if cj > ck , then either hj < hk or
aj < ak or both
8 / 64
Spatial Equilibrium
• Assumption: utility equated across regions; common utility value ≡ u¯
• Basic property I: For any regions j and k, if cj > ck , then either hj < hk or
aj < ak or both • Basic property II: For any regions j and k, households do not prefer to
migrate from j to k or vice versa
8 / 64
Spatial Equilibrium
• Assumption: utility equated across regions; common utility value ≡ u¯
• Basic property I: For any regions j and k, if cj > ck , then either hj < hk or
aj < ak or both • Basic property II: For any regions j and k, households do not prefer to
migrate from j to k or vice versa • Typical approach: go with these assumptions, back out amenities from
model • We will look at direct measures of cj , hj and aj
8 / 64
Spatial Equilibrium: Role of Housing Prices
• “Aren’t housing prices higher in cities? Couldn’t this mean aj = ak ?”
9 / 64
Spatial Equilibrium: Role of Housing Prices
• “Aren’t housing prices higher in cities? Couldn’t this mean aj = ak ?”
• Say j is the city, and wj > wk but pj > pk
9 / 64
Spatial Equilibrium: Role of Housing Prices
• “Aren’t housing prices higher in cities? Couldn’t this mean aj = ak ?”
• Say j is the city, and wj > wk but pj > pk • In a spatial equilibrium, two possibilities
9 / 64
Spatial Equilibrium: Role of Housing Prices
• “Aren’t housing prices higher in cities? Couldn’t this mean aj = ak ?”
• Say j is the city, and wj > wk but pj > pk • In a spatial equilibrium, two possibilities 1
Could be: cj > ck but hj < hk ; household gets worse housing
9 / 64
Spatial Equilibrium: Role of Housing Prices
• “Aren’t housing prices higher in cities? Couldn’t this mean aj = ak ?”
• Say j is the city, and wj > wk but pj > pk • In a spatial equilibrium, two possibilities 1
Could be: cj > ck but hj < hk ; household gets worse housing
2
Could be: cj < ck but hj > hk ; household gets worse consumption
• ⇒ higher housing prices in cities not enough to conclude that there must
be a spatial equilibrium; need to look at quantities, which is what we do
9 / 64
0
15
16
17
Utility
18
Consumption/Amenities 5 10
19
15
20
Consumption, Amenities and Utility in Spatial Equilibrium
0
2 Consumption
4 6 Population Density Amenities
8
10 Utility
10 / 64
Data
11 / 64
Data • Household-level data from Demographic and Health Surveys (DHS) I
Nationally representative surveys, consistent methodology
I
Use all 2005 DHS surveys with available GPS coordinates for survey clusters
I
276,000 households from 20 countries
• Satellite-derived measures of air pollution data • Crime data from Afrobarometer surveys and Living Standards
Measurement Surveys (LSMS) • Data on population density measures from Gridded Population of the
World (GPW) I
Resolution of ∼ 1km at equator
I
Based on census data; minimal amount of modeling
I
Restrict attention to countries with sufficiently high spatial detail
12 / 64
Countries Studied, DHS Samples and Populations Country
Households in Sample
Country Population
Benin Burkina Faso Cameroon Dem. Republic of Congo Ethiopia
17,332 13,617 14,189 16,344 16,037
10,050,702 16,460,141 21,699,631 65,705,093 91,728,849
Ghana Ivory Coast Kenya Liberia Madagascar
11,574 9,394 9,033 9,333 17,578
25,366,462 19,839,750 43,178,141 4,190,435 22,293,914
Malawi Mali Mozambique Nigeria Senegal
24,210 10,105 13,899 38,170 7,780
15,906,483 14,853,572 25,203,395 168,800,000 13,726,021
Sierra Leone Tanzania Uganda Zambia Zimbabwe
12,629 9,282 8,939 7,164 9,442
5,978,727 47,783,107 36,345,860 14,075,099 13,724,317
Total
276,051
769,082,846 13 / 64
Linking DHS Data to Density
• DHS clusters are assigned an approximate GPS location (randomly
displaced by small amounts to provide anonymity to respondents) I
Urban clusters are randomly displaced in the data by 0-2 km
I
Rural clusters are randomly displaced in the data by 0-5 km, with 1% of clusters randomly selected to be displaced by 10 km
• We take 5 km buffers around both urban and rural clusters and compute
average population density across the buffer. I
Insures that we capture the actual cluster but also nearby areas.
• The density distribution that we achieve is quite similar to what we would
get from census data. I
We may miss the extreme tails of the density distribution by averaging.
14 / 64
Example: DHS Clusters in Dar Es Salaam, Tanzania
15 / 64
.4 0
0
.1
Frequency of census EAs .2 .3
Frequency of DHS Households .1 .2 .3
.4
Distribution of Densities in Tanzania Census and DHS
0
4 8 Log of population density
0 12
4 8 Log of population density
12
→ Correlation between GPW density estimates and census estimates of population density is 0.93 with a p-value of 0.000.
16 / 64
Kernel density .2 .4 .6 0
0
Kernel density .2 .4 .6
Distribution of Population Density in Select Countries
0
2
4 6 8 Log of population density
2
4 6 8 Log of population density Mali
Kernel density .2 .4 .6
Malawi
10
0
0
2 4 6 8 Log of population density Mozambique
10
0
Nigeria
2
4 6 8 Log of population density Senegal
10
Sierra Leone
.6
0
.6
0
Madagascar
Kernel density .2 .4 .6
Liberia
10
17 / 64
DHS Urban/Rural Classifications – Often Not Reliable
BurkinaFaso
Cameroon
DRC
Ethiopia
Ghana
IvoryCoast
Kenya
Liberia
Madagascar
Malawi
Mali
Mozambique
Nigeria
Senegal
SierraLeone
Tanzania
Uganda
Zambia
Zimbabwe
.4 .2 0 .6 .4 .2 0 .6 .4 .2 0
Kernel density
.6
0
.2
.4
.6
Benin
0
2
4
6
8
10
0
2
4
6
8
10
0
2
4
6
8
10
0
2
4
6
8
10
0
2
4
6
8
10
Log of population density Rural
Urban 18 / 64
Urban-Rural Density Distributions
• Urban and rural densities are not as distinct as one might expect. • In a number of countries, there is substantial overlap when administrative
definitions are applied. • Frequently used “urban-rural” dichotomy may be flawed for some
purposes: I
Not necessarily very precise if we are thinking about urbanization in terms of population density, market thickness, agglomeration externalities.
I
Still useful, however, for thinking about some characteristics that are linked to administrative classifications; e.g., public service provision?
19 / 64
Durables Ownership
20 / 64
Telephone Ownership by Population Density
.2
.4
Phone .6
.8
1
Phone ownership in Ethiopia, Nigeria, Senegal and Tanzania
2
4 Ethiopia
6 Log of population density Nigeria
Senegal
8
10 Tanzania
21 / 64
Telephone Ownership by Population Density
0
.2
.4
Phone
.6
.8
1
Phone ownership in entire sample
2
4
6 Log of population density
8
10
22 / 64
Durables Ownership Differences by Density Quartiles
Differences From Q1
Telephone
Television
Automobile
Motorcycle
Q2
Q3
Q4
Regional Std. Dev
0.07
0.20
0.43
0.47
14
18
20
0.03
0.16
0.46
15
18
20
0.00
0.02
0.09
6
15
20
-0.01
-0.01
0.00
8
13
14
0.41
0.18
0.26
23 / 64
Durables-Density Gradients
• Durable ownership clearly increasing in population density • This is clearly counted in consumption and income, so consistent with
previous studies • Not a contribution per se, but frame of reference for other metrics,
confirmation of previous conclusions in literature about income
24 / 64
Housing Quality
25 / 64
1
Percent with Finished Roof and Finished Walls
Lowest Density Quartile .4 .6 .8
BEN
CMR GHA NGA ZWE CIV SLE
MWI CMR
GHA ZMB KEN BEN ETH BFA MDG MDG
.2
BFA
SLE
MWI COD MOZ ETH
CIV SEN MLI LBR SEN NGA ZWE
MOZ COD MLI LBR
ZMB
0
KEN
0
.2
.4 .6 Highest Density Quartile Finished roof
.8
1
Finished Wall
26 / 64
Lowest Density Quartile .4 .6 .8
1
Percent with Flush Toilet and Electricity
CMR GHA NGA SENCIV
.2
BEN ETH
CMR MOZ
NGA SEN MDG GHA ZMB CIV UGA COD TZA TZA ZMB LBR BFA KEN SLE
ZWE
MLI MOZ KEN
ZWE
0
MLI MWI ETH LBR BEN MDG COD SLE MWI BFA UGA
0
.2
.4 .6 Highest Density Quartile Flush toilet
.8
1
Electricity
27 / 64
Housing Quality Differences by Density Quartile Differences From Q1 Q2
Q3
Q4
Regional Std. Dev
0.03
0.19
0.51
0.41
16
17
20
0.02
0.21
0.51
12
18
20
0.04
0.19
0.48
13
17
20
Flush toilet
0.01
0.09
0.28
14
17
20
Water collection (min)
-4.1
-7.8
-16.4
11
16
18
0.09
0.26
0.48
16
18
17
Electricity
Tap water
Constructed floor
Finished roof
0.44
0.46
0.28
35.01
0.44
28 / 64
Housing Quality
• Housing quality much higher in urban areas
• True for just about every country, metric
• Put together with consumption data, must now be that amenities
increasing in density for spatial equilibrium to hold
29 / 64
Child Health
30 / 64
Child Health
• Child health measures functions of consumption (food) and also amenities
(health facilities, public health provision more generally) • Percent anemic
• Percent consuming below minimum acceptable diet
• Stunted: low height for age
• Wasted: low weight for height
31 / 64
Percent Anemic and Below Minimum Acceptable Diet
1
MDG ETH BFA SLE CMR UGA MLICIV ZWE SEN LBR NGA COD TZA BFA
Lowest Density Quartile .6 .8
GHA KEN MWI MLI SLE GHA ZMB CIV
BEN
MOZ
SEN
MWI CMR MOZ COD BEN
MDG
ZWE
TZA
ETH
.4
UGA
.4
.6 .8 Highest Density Quartile Anemic
1
Below minimum acceptable diet
32 / 64
.6
Child Health: Percent Stunted and Wasted
BEN MWI
Lowest Density Quartile .2 .4
MLI
CMR
BFA NGA CIV
TZA MOZCOD SLE KEN
ZMB
ETH MDG
UGA SEN GHA
BEN
MDG BFA
NGA
MLI
0
ETH KEN SEN GHA COD CIVSLE MOZ CMR ZMB LBR TZA ZWEMWIUGA
ZWE LBR
0
.2 .4 Highest Density Quartile Stunted
.6
Wasted
33 / 64
Child Health Differences by Density
Differences From Q1
Anemic
Stunted (low height for age)
Wasted (low height for weight)
Below minimum diet
Q2
Q3
Q4
Regional Std. Dev
-0.003
-0.041
-0.080
0.453
1
6
11
0.010
-0.017
-0.106
4
7
15
-0.005
-0.015
-0.022
4
7
8
-0.013
-0.027
-0.053
3
3
9
0.479
0.282
0.276
34 / 64
Child Health – Summary
• Child health (depressingly) poor on average
• But at least as good if not better on average in cities
• Probably reflects better consumption (diet) and better access to public
health facilities and programs • No attempt here to distinguish effects of diet and public health; very
important, not necessary for our purposes
35 / 64
Pollution
36 / 64
Pollution Variables
• Satellite data on outdoor air pollution at ground level I
Van Donkelaar et al (2015, 2016): Fine particulate matter (PM2.5)
I
Geddes et al (2015): Nitrogen Dioxide (NO2)
• DHS evidence on indoor air pollution I
Uses solid fuel source (wood, coal), and
I
Primarily cooks indoors
• Clear negative links of long-term exposure to health – literature enormous
(Pope & Dockery, 2006: meta analysis of meta analyses)
37 / 64
PM2.5 Concentrations in Nigeria (µg /m3 )
PM2.5
PM2.5 5 - 16 17 - 20 21 - 26 27 - 32 33 - 35 36 - 38 39 - 41 42 - 45 46 - 54 55 - 110
38 / 64
PM2.5 Concentrations in Nigeria – Excluding Dust
PM2.5 (Dust and Sea Salt removed)
PM2.5 0.5 - 5.3 5.4 - 6.0 6.1 - 6.6 6.7 - 7.2 7.3 - 7.9 8.0 - 8.5 8.6 - 9.3 9.4 - 10.1 10.2 - 11.2 11.3 - 20.5
39 / 64
Population Density in Nigeria
Population Density
12 - 32 33 - 47 48 - 60 61 - 73 74 - 91 92 - 121 122 - 151 152 - 216 217 - 349 350 - 19841
40 / 64
10
PM2.5 (micrograms per cubic meter) 20 30
40
PM2.5-Density Gradient in Nigeria
3
4 PM2.5
5 Log of population density
6
7
PM2.5 (dust and sea salt removed)
41 / 64
50
PM2.5 vs Density in all Countries
Lowest Density Quartile 20 30 40
NER
BEN
NGABFA
MLI SEN
GHA CIV TGO
SLE LBR CMR
0
10
GHA NGA TGO ZMB ZMB UGA BEN CIV CMR MWI UGA MWI BFA ZWE TZA KEN ZWE TZA LBR SEN SLE MOZ MOZ KEN NER MLI MDG MDG
0
10
20 30 Highest Density Quartile
PM2.5 Dust and sea salt removed
40
50
PM2.5
42 / 64
6
6.5 7 PM25 (micrograms/cubic meter) U.S.
7.5
PM25 (micrograms/cubic meter) India and China 0 20 40 60 80
PM2.5 vs Density in China, India and United States
2 China
4 Log of population density India
6 United States
43 / 64
NO2 vs Density in all Countries .5
TGO
Lowest Density Quartile .2 .3 .4
GHA
BEN SEN
NGA
BFA
CIV SLE ZWE UGA ZMB MWI MOZ TZA MDG
CMR KEN
MLI
.1
LBR
0
NER
0
.1
.2 .3 Highest Density Quartile
.4
.5
NO2
44 / 64
1
1.5 2 2.5 NO2 (micrograms/cubic meter) U.S.
3
NO2 (micrograms/cubic meter) India and China 0 2 4 6 8 10
NO2 vs Density in China, India and United States
2 China
4 Log of population density India
6 United States
45 / 64
1
Indoor Air Pollution vs Density
MLI MDG
ZWE
Lowest Density Quartile .4 .6 .8
SLE MDG LBR MLI ETH MWI MOZ COD BEN TZA UGA BFA
CIV ZMB GHA
KEN SEN NGA KEN CMR
ZWE
SEN NGA CMR GHA CIV
ETH
LBR MOZ SLE
BEN
0
.2
BFA
ZMB
UGA COD
MWI
0
.2
.4 .6 Highest Density Quartile
.8
1
Solid source of cooking fuel Solid source of cooking fuel x cooking indoors
46 / 64
Pollution Differences by Density Quartile
Differences From Q1
Nitrogen Dioxide
PM2.5
Cook inside with solid fuel
Q2
Q3
Q4
Regional Std. Dev
0.006
-0.007
0.002
0.085
11
13
12
-0.57
-1.30
-1.09
10
11
11
0.01
-0.05
-0.24
13
16
19
6.63
0.47
47 / 64
Pollution Variables – Summary
• Outdoor air pollution largely related to proximity to Sahara
• ...not population density
• Commentary on Africa’s under-developed manufacturing sector
• Indoor air pollution improves on average with density
• Caveat: Africa clearly different from India and China
• Still, biggest rural-urban income gaps in Africa, so simple theory of
pollution offsetting higher income in developing world hard to tell
48 / 64
Crime
49 / 64
Crime Variables • Main crime variables from Afrobarometer
• Matching clusters to density data more complicated – see paper
50 / 64
Crime Variables • Main crime variables from Afrobarometer
• Matching clusters to density data more complicated – see paper
• Property crime: “Over the past year, how often (if ever) have you or
anyone in your family had something stolen from your house?” • Violent crime: “... been physically attacked?”
• Fear of crime: “... feared crime in your own home?”
• Feel unsafe in neighborhood: “... felt unsafe walking in your
neighborhood?”
50 / 64
Crime Variables • Main crime variables from Afrobarometer
• Matching clusters to density data more complicated – see paper
• Property crime: “Over the past year, how often (if ever) have you or
anyone in your family had something stolen from your house?” • Violent crime: “... been physically attacked?”
• Fear of crime: “... feared crime in your own home?”
• Feel unsafe in neighborhood: “... felt unsafe walking in your
neighborhood?” • Use also non-standardized crime variables from LSMS 50 / 64
Lowest Density Quartile .2 .4
.6
Property Crime and Violent Crime
CMR
LBR
UGA
MWI NGA TZA TGO GHA MLI NGAUGA MDG ZMB ZWE SLE SEN BFA MOZ MWI MDG TZA GHA BEN MLI TGO CIV
ZMB MOZ KEN
ZWE SEN
SLE
BFA
BEN
KEN NER CMR
CIV LBR
0
NER
0
.2 .4 Highest Density Quartile Property crime
.6
Violent crime
51 / 64
Lowest Density Quartile .2 .4 .6
.8
Fear of Crime
MDG KEN UGA MWI
KEN
CMR
SEN
BFA ZMB MLI TZA ZWE MWI MDG ZMBMOZ UGA ZWE MOZCMR CIV NGA BFA NGA TGOCIV LBR MLI SLE SEN TGO GHA GHA BEN SLE BEN NER TZA
LBR
0
NER
0
.2
.4 Highest Density Quartile
Feel safe in neighborhood
.6
.8
Fear of crime
52 / 64
Crime by Density Quartile
Population Density Quartile
Property Crime
Violent Crime
Fear of Crime
Feel unsafe
Q1
Q2
Q3
Q4
0.29
0.31
0.31
0.33
-
6
4
6
0.10
0.09
0.10
0.12
-
2
2
5
0.32
0.33
0.34
0.37
-
4
5
8
0.37
0.38
0.38
0.44
-
4
5
8
53 / 64
Crime Data in LSMS Surveys – Similar Picture, Less Comparable Data Malawi Anything stolen, personal attack
0
.4 .42 .44 .46 .48
.01 .02 .03
Ethiopia Theft, robbery, other violence
2
4 6 8 Log of population density
10
2
4 6 8 Log of population density
Tanzania Anything stolen/personal assault
0
.05
.1
.15
.2
.02 .04 .06 .08
Nigeria Theft/kidnapping/hijacking/robbery/assault
10
2
4 6 8 Log of population density
10
2
4 6 8 Log of population density
10
Uganda .02 .04 .06 .08 .1 .12
Theft/violence/conflict
2
4 6 8 Log of population density
10
54 / 64
Crime Variables – Summary
• Best evidence so far of negative amenities of cities
• Still, magnitudes small relative to size of urban-rural gap
• Bishop, Murphy (2011): San Francisco residents willing to pay $472 to
avoid 10% increase in violent crime; $472 / $57,276 = 0.8% of income • Cohen et al (2001): US residents in 2000: willing to pay $120 to reduce
chance of armed robbery by 10%; $120 / $34,432 = 0.4% of income • Ludwig and Cook (1998): US households in 1998 willing to pay $240 per
year to reduce chance of gunshot injury by 30%; $240 / $51,939 = 0.5% of income
55 / 64
Density Gradients by Education Groups
56 / 64
0
.2
Electricity .4 .6
.8
1
Electricity in Nigeria by Education Group
4
5
6 7 Log of population density
8
9
.1 .05 0
Fraction
.15
Histogram of hh heads with complete secondary or more schooling
Log of population density
.1 .05 0
Fraction
.15
Histogram of hh heads with complete primary and incomplete secondary schooling
Log of population density
.15 .1 .05 0
Fraction
.2
Histogram of hh heads with no or incomplete primary schooling
Log of population density
57 / 64
Slope coefficients 0.00 0.05 0.10 0.15 0.20
Slopes by Education Group
(a) Gradients for hh heads with no or incomplete primary schooling
electricity
floor
phone
roof
toilet
tv
wall
water
group
Slope coefficients 0.00 0.05 0.10 0.15 0.20
(b) Gradients for hh heads with at least primary schooling
electricity
floor
phone
roof
toilet
tv
wall
water
group
58 / 64
Internal Migration
59 / 64
Rural-Urban and Urban-Rural Migrants as Percent of Adults
Rural-to-Urban
Urban-to-Rural
Difference
Percent of Adults Dem. Republic of Congo (2007)
5.72
0.41
5.31∗∗∗
Ethiopia (2005)
5.19
0.46
4.73∗∗∗
Ghana (2008)
9.08
2.46
6.62∗∗∗
Kenya (2008-09)
11.03
1.72
9.31∗∗∗
Liberia (2007)
6.28
4.90
1.38∗∗∗
Madagascar (2008-09)
4.72
0.62
4.10∗∗∗
Malawi (2010)
5.08
1.41
3.68∗∗∗
Mali (2006)
5.52
1.60
3.92∗∗∗
Nigeria (2008)
7.31
1.94
5.37∗∗∗
Senegal (2005)
6.68
2.41
4.27∗∗∗
Sierra Leone (2008)
8.76
1.75
7.01∗∗∗
Zambia (2007)
14.00
2.03
11.96∗∗∗
60 / 64
Conclusion • We searched for a spatial equilibrium in 20 African countries
• Consumption and most amenities increasing in population density, at least
among “prime suspects” • Net migration is to the city in every country
• Hard to reconcile with a spatial equilibrium
• Easier explanation: households vote with feet, move to cities, which offer
higher consumption and at least as good amenities if not better • Dangerous to infer amenities assuming static spatial equilibrium on
cross-sectional data from developing world
61 / 64
Extra Slides
62 / 64
.8
Lack of Medicine and Food
TGO
SEN
Lowest Density Quartile .2 .4 .6
SLE
MDG ZWE CIV
ZWE SLE
SEN
UGA BFA BEN LBR MDG LBR BFA CMR ZMB MWI KEN MWI TZA KEN UGA MLI MLI NGA ZMB TZAMOZ MOZ NGA
NER
NER
TGOCIV
CMR
BEN
GHA
0
GHA
0
.2
.4 Highest Density Quartile
Lack of medicines
.6
.8
Lack of food
63 / 64
Trust Anyone, Trust Relatives
1
NER SEN CIV TGO TZA MLI BFAMWI ZWE MOZ GHA CMR UGA SLE MDG ZMBKEN
Lowest Density Quartile .4 .6 .8
BEN NGA LBR
NER
BEN BFA
MDG
SEN MLI
SLE
0
.2
ZWE LBR NGA UGATGO CMR MOZ GHA MWI CIV TZA ZMB KEN
0
.2
.4 .6 Highest Density Quartile Anyone
.8
1
Relatives
64 / 64