Adaptation to land constraints:
Is Africa different?
Derek Headey
International Food Policy Research Institute (IFPRI)
Thom Jayne
Michigan State University (MSU)
1
1. Introduction
 Some 215 years ago, Malthus argued that pop. growth
cyclically outstrips agricultural productivity
 In much of the world, economic history has not been
kind to Malthus, because of “induced innovations”:
1. Endogenous reductions in fertility (Becker)
2. Endogenous intensification of agriculture (Boserup)
3. Policy-induced, scientific intensification of agric. (GR)
4. Migration out of agriculture
 But what about Africa? Famines still not a thing of the
past, and still huge dependency on food aid
1. Introduction
 Structurally, there are 6 inter-related reasons to be
concerned about the specter of Malthus in Africa:
1. Very poor, and poverty still heavily rural
2. Low inherent agric. potential (incl. low irrigation)
3. Mixed success with agric. intensification
4. Climate change: secular changes, and more shocks
5. Rapid population growth (double by 2050) and small
and shrinking farm sizes
6. Very limited success with industrialization
1. Introduction
 This paper is mainly concerned with farm size
evolution: trends, patterns, and adaptive behaviors
 Framework based on decomposing growth in farm
income:
𝑂𝑂𝑂𝑂𝑂𝑂
∆𝑙𝑙
𝑃𝑃𝑃.
=
𝐿𝑎𝑎𝑎
∆𝑙𝑙
𝑃𝑃𝑃.
𝑂𝑂𝑂𝑂𝑂𝑂
+∆𝑙𝑙
𝐿𝐿𝐿𝐿
Growth in rural population is the sum of fertility & net migration:
=
𝑂𝑂𝑂𝑂𝑂𝑂
∆𝑙𝑙
𝐿𝐿𝐿𝐿
+ ∆ ln 𝐿𝐿𝐿𝐿 − ∆ ln 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 − ∆ ln 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠
1. Introduction
 Our overarching objective is to assess international
experience in these 4 adaptations to land pressures
 There is a large literature exploring Boserup’s
hypothesis, as well as policy-induced intensification
 There is much smaller literature on land expansion
 There is essentially no literature on farm sizes &
fertility rates
 And there is some indirect literature on farms sizes,
rural nonfarm activity and migration
 For each of these adaptations, we also ask whether
Africa is different, and why?
1. Introduction
 In terms of data and methods, we make use of:
1. FAOSTAT ag production and land data;
2. Census (FAO) and survey data on farm size
distributions
3. DHS data on rural fertility rates & occupations
4. Some WB data on remittances
 We combine these data in an unusually rich data set
on agricultural and rural development
 (though we also acknowledge that some of the
numbers are fairly speculative)
 On methods, our approach is necessarily exploratory
 Establishing causation is an under-recognized
problem with Boserup’s theory
 Problems of simultaneity, omitted variables, selection
biases, parameter heterogeneity. Some examples:
1.
2.
3.
4.
Agroecological (AE) factors & market access jointly determine
settlement patterns and intensification
Boserupian intensification depends on AE potential
Unsuccessful intensification encourages out-migration
Policies promote intensification, discourages out-migration
 IV not plausible with this data, but we aspire to
identification via control vars., FE & first differencing
2. Land expansion
 If farm sizes are shrinking, why not expand land use?
 Africa is typically thought of as land abundant, but
this neglects the heterogeneity within Africa
Region
Period
1970s
Africa - high densityb (n=5) 2000s
1970s
Africa - low densityb
(n=11)
2000s
South Asia
1970s
(n=5)
2000s
China & S.E. Asia
1970s
(n=4)
2000s
Hectares per agric. Hectares per holding
worker (FAO)
(censuses)
0.84
1.99
0.58
1.23
1.65
2.65
1.37
2.82
0.78
2.01
0.55
1.19
0.80
2.08
0.68
1.58
2. Land expansion
 Several important facts & mysteries emerge from
census, FAO and FAO-IIASA data:
1. Farm sizes are shrinking in high-density Africa.
2. Some high-density countries still have unused land,
but smallholders face major constraints to using that
land (e.g. Ethiopia, Madagascar).
3. Even in countries with unused land (e.g. Ethiopia),
there are major constraints to using new lands:
different agronomics, disease burdens, infrastructure
4. Farm sizes are unchanged (on average) in low
density Africa, but still very small on average
3. Agricultural intensification
 In the framework above, the most welfare-relevant
indicator of intensification is just output per hectare
 Boserup focused more on cropping intensity, and the
ag-econ profession & CGIAR looks a lot at yields
 But switching to high value crops is obviously also a
potentially important adaptation, especially in SSA.
 So I’m going to show you a series of graphs, and then
some more formal econometric tests.
 We decompose ag output into cereal and non-cereals.
 Cereals output can be also decomposed into yields &
cereal cropping intensity
3. Agricultural intensification
Agricultural output per hecatre (2005 int. dollars)
0
2000
4000
6000
EGY
CRI
Ag output per hectare
JOR
CHL
LBN
COL
CHN
VNM
ECU
URY
JAM
ARM
UZB
PHL
VEN DOM
BRA
MYS
PER
MKD
TKM
PAK
THA
GTM
MMR
GEO
ALB
SOM
TJK
SWZ
IRN
PAN
KGZ
TUR
HND
IDN
SLV
MEX
LAO
IND
ARG
PRK
BLR
SRB
PRY
SYR
BTN
MRT
KEN
LTU AZE
ROM
MDG
BWA
MNE
FJI
BIH
BDI
NGA
MNG
BOL
CIV
HTI
KHM MWI
LVA
GHA
MAR
ZAF
GUY
TUN
LBR BEN
BGR
UGA
COG
UKR
NIC
TZA
CMR
AGO
SDN
MDA
NAM
DZA
LBY
GIN
TMP
GAB
ETH
ZAR
IRQ
COM
MLI
GNBSLE
CAF
ERI
RUS
AFG
LSO
MOZ
ZMB
ZWE
KAZ
SENBFA
TCD
TGO
GMB
NER
0
BGD
NPL
LKA
RWA
200
400
600
Agricultural population density (person per sq km)
800
1500
3. Agricultural intensification
EGY
Cereal output per hectare ($/ha)
1000
500
Cereal output per hectare
VNM
DOM
GUY
IDN
CHN
MMR
MYS
CRI
CHL
COL
URY
PER
ECU FJI
UZB
BGD
LKA
LAO
THA
PHL
KHM
NPL
RWA
0
ARGMDG
PRK
VEN
BRA
IND
ALB
PAN
PAK
MEX
TKM
BIHSRB
LTU
LVA
ZAF
IRN
BTN
AZE LBRSLV
MKD
TJK
BLR
ARM
KGZ
GIN
MWI
PRY
GTM
TUR
SLE
COM
NIC
CIV
ZMB
LBN
UKR
AFGGNB
BGR
BOL
GEO
RUS
TMP
KEN
HND
TZA
CMR
UGA
BDI
SYR
GAB
NGA
ROM
ETH
GHA
IRQ
TGO
HTI
JAM
BEN
BFA
MDACAF SEN
COG
ZAR GMB
MOZ
JOR
ZWE
AGO
SWZ
MAR
LSO
0
200
400
600
Agricultural population density (person per sq km)
800
150
3. Agricultural intensification
Cereals cropping intensity (%)
50
100
Cropping intensity in nonAfrica sample is heavily
explained by irrigation:
R-sq = 0.56
NPL
BGD
VNM
EGY
ERI
LAO
MMR
0
CHN
KHM
PHL
KAZ
BFATHA
MRT
TCD
BTN
NER
PAK
MLI
IND
ETH
GMB
SYR
MAR
IRQ
LTU
SRB
IRN
ROM
TKM LSO
TUR
GIN
NGA
URY
ZWE
SDN
BGR
MDG
TMP
SLE
MWI
PRK
SOMTJK
LVA
BLR
KGZ
MDA
AGO
TZA
IDN
UKR
HTI
MKD
AFG
SLV
MEX
BEN
MOZ
ECU
SEN
KEN
AZE
GEO
DZA
ARM
VEN
UZB
RUS
PRY
TGO
ARG
NAM
COLJORHND
BRA
CHL
BIH
LBR
PER
BOL
TUN
SWZ GNB
GUY
GTMZAR
LBN
ZMB
ZAF
PAN
NIC
UGA
BWA
CMR
GHA ALB
LBY
BDI
MNG DOM
CAF
COM
CRI
CIV
MYS
COG
GAB
MNE
FJI JAM
0
LKA
RWA
200
400
600
Agricultural population density (person per sq km)
800
CRI
EGY
JOR
Non-cereal output per hectare
LBN
CHL
COL
CHN
ECU
JAM
URY
ARM
UZB
VEN DOM
BRA MYS
MKD
PER
GTM
PHL
TKM
GEO
SWZ
ALB
PAK
TJK
HND
PAN
KGZIRN
TUR
SLV THA
MEX
BLR
PRY
ARG
SYR
KEN
PRK
AZE
SRB
IND
BTN
IDN
FJI
LTU
ROM
BDI
MMR HTI
CIV
BIH
BOL
GHA
MAR
NGA
MWI
ZAF
COG
LBR
UGA
MDG
BEN
LVA
BGR
AGO
CMR
LAO
NIC
GAB
MDA
TZA
UKR
ZAR
COM
GUY
CAF
TMP
IRQ
GNB
ETH
LSO
GIN
SLE
MOZ
ZWE
RUSZMB
SENBFA KHM
AFG TGO
GMB
VNM
NPL
RWA
LKA
BGD
0
Non-staples output (% total crop output)
1000
2000
3000
4000
5000
3. Agricultural intensification
0
200
400
600
Agricultural population density (person per sq km)
800
Table 4. Log-log estimates of agricultural value per hectare
and its three components
Regression No.
R1
R2
Agric. output Cereal output
Dep. var.
per ha
per ha
Population density
0.33***
0.18***
Density*Africa
-0.11**
-0.23***
Road density
0.14***
0.09**
Number of ports
0.14***
0.21***
Urban agglom (%)
0.29***
-0.09
Regional fixed effects?
Yes
Yes
Sign of SSA dummies? + in E.Africa
Zero
AE controls
Yes
Yes
No. Obs
243
243
R-square
0.8
0.74
R3
R4
Cereal crop
intensity
0.20***
-0.01
-0.03
0.03
0.31***
Yes
Neg.
Yes
243
0.67
Non-cereal
output per ha
0.28***
-0.01
0.19***
0.15***
0.31***
Yes
+ in E.Africa
Yes
243
0.79
Table 5. Log-log estimates of specific agricultural inputs
Regression No.
R1
Fertilizers
Dep. var.
per hectare
Population density
0.76***
Density*Africa
-0.32**
Road density
-0.08
Number of ports
0.50***
Urban agglom (%)
0.38
Regional fixed effects
Yes
Sign of SSA dummies?
Zero
AE controls
Yes
No. Obs
0.73
R-square
0.69
R2
R3
R4
Cattle/oxen
per hectare
0.42***
0.15*
0.31***
0.07
0.03
Yes
Neg.
Yes
0.77
0.74
Irrigation per
hectare
0.59***
-0.47***
0.04
0.24***
0.24**
Yes
Zero
Yes
0.92
0.91
Capital per
hectare
0.24***
-0.10***
0.07**
0.12***
-0.03
Yes
Zero
Yes
0.77
0.73
Noncereals
Low productivity of cereals sector
Table 7. Potential explanations of Africa’s agricultural
intensification trajectory
Stylized facts
Low fertilizer
application
Low adoption
of improved
varieties
Potential explanations
Agronomic constraints (e.g. low soil fertility) Poor
management practices, low human capital High transport
costs (see regression 1 in Table 4); Low rates of subsidization
(structural adjustment)
More varied agroecological conditions and crop mix
Lower returns because of limited use of other inputs (e.g.
irrigation); Lower investment in R&D
Low use of
Tsetse fly in humid tropics Feed/land constraints in some
plough/ tractors densely populated areas
Low rates of
irrigation
High non-cereal
output per
hectare
Hydrological constraints; High costs of implementation and
maintenance; Poor access to markets limits benefits
Agroecological suitability; Colonial introduction of cash crops;
Non-perishable cash crops (cotton, coffee, cocoa, tea,
tobacco) not limited by poor infrastructure and isolation
3. Reducing rural fertility rates
Rural fertility rates (# children)
2
4
6
8
Figure 3. Rural fertility rates and rural population density
African gradient
0
Non-Africa gradient
0
500
1000
Rural population density (person per sq km)
1500
10
Figure 4. Desired rural fertility & population density
Desired fertility (# children)
2
4
6
8
NER
TCD
NER
NER
TCD
SEN
CMR
MLI
MRT
MLI NGA
ZAR
SEN
CAF
MLI
NGA
CMR
MLI
LBR
NGA
ERI
CMR
ZMB
NGA
ERISEN
TZA
SDN
MOZ
BFABEN
CIV
SEN
GIN
GIN
BFA
MDG
SEN
TGO
ZMB
TZA
ETH
NAM
MDG
COG
GAB
MOZ
CIV
COM
GHA
TZA
TZA
BEN
ZWE
SLEBEN TMP
BDI
ZMB
LBR
ZMB
TZA
MWI
MDGTGO
ETH
BWA
GHA
MDG
RWA
ETH
KEN
ZWE
GHA
JOR
JOR
RWA
BDI
PRYZWE
JOR
PAK
RWA
GTM MWI
MWI
PAK
JOR
ZWEKEN
MWI
JOR SLV
GTM
MAR
ZWE
KEN
KEN
KGZ
MAR
KEN
TUN
GTM
UZB DOM
MEX
GTM
NAM
HTI
KHM
NIC
PHL RWA
TKM
PRY
ECU
DOM
PHL
HTI
KHM
RWA
KAZMAR
NAM
NIC
DOM
HND
DOM
HTI
LSO
IDN
PHL
BRA
KHM
DOM
IDN
KAZ
IND
PHL
PER
COL
IDN
LKA
GUY
COL
IDNIDN
IND NPL
COL
ECU
THA IDN
ARM
NPL
PER
BOL
ARM
ARM
ALB
PER
BOL
TUR
COL
SWZ
AZE
PER
IND
BRA
COL
NPL VNM
BOL
UKR
NPL
African sample gradient
VNM
EGY EGY
EGY
EGY EGY
EGY
BGD BGD
BGD BGD
BGD
Full sample gradient
0
SLV
0
500
1000
Rural population density (person per sq km)
1500
40
F5. Unmet contraception needs (%) and rural population density in Africa
RWA
Unmet contraception needs (% women)
20
25
30
35
RWA
GHA
LBR
KEN
GHA
SEN
GHA
GHA
MDG
MWI
BFA
MLI
ERI
CIV
ERI
ZMB
ZMB
MDG
GIN
NAM
NAM
RWA
COM
ETH
BEN
GAB
MWI
TGO
SEN
LSO
MLI
ZMB
SLE
TZA
BFA
KEN
TZA KEN
MWI
BEN
BEN
ZAR
MOZ
TCD
GIN
TZA
TZA
CMR
NGANGA
CMR
CMR
NER
MOZ
15
NER
COG
NER
0
Sources
ETH
NGA
NGA
100
200
300
Rural population density (person per sq km)
400
Table 8. Elasticities between rural fertility indicators
& rural population density
Regression number
1
Dependent variable
Actual fertility Actual fertility Desired
fertility
Linear
Log-log
Linear
Desired
fertility
Log-log
b/se
b/se
b/se
b/se
Pop density (per 100 m2)
-0.14***
-0.09***
-0.11***
0.00
Density*Africa
0.05
0.09***
-0.34***
-0.07***
Female sec. education (%)
-0.02***
-0.05***
-0.01**
-0.08***
Ag. output per worker, log
-0.58***
-0.13***
0.01
0.06***
Africa dummy
1.25***
-0.15
2.13***
0.67***
Number of observations
165
165
164
164
R-square
0.75
0.76
0.77
0.81
Model
2
3
4
4. Nonfarm diversification
 Much neglected in 1980s literature on Boserup
 Subsequent literature on both RNFE and migration &
remittances shows that RNF income is big
 But not much specific literature looking at pop density
 On RNF activity, often suggested there is a U-shaped
relationship between farm size and RNFE: landless
poor are pushed into RNFE, rich are pulled in
 Very difficult to look at rural-urban migration
 Int. remittances have boomed in last 10 years,
particularly in densely population South Asia – now
22% of rural income in Bangladesh
Table 9. Speculative estimates of rural nonfarm
employment shares for men and women in the 2000s
High density Africa
Country
Low density Africa
W
M
Benin
50.4
23.7
Congo (DRC)
14.0
Ethiopia
Country
Other LDCs
W
M
Country
W
M
Burkina Faso
12.9
8.1
BGD
53.4
44.5
23.5
Chad
13.7
9.6
Bolivia
71.4
25.9
34.3
9.7
Cote d'Ivoire
31.7
22.1
Cambodia
36.0
Kenya
47.1
37.3
Ghana
50.1
26.6
Egypt
69.4
Madagascar
17.8
15.3
Mali
44.6
16.0
Guatemala
79.1
Malawi
41.5
36.0
Mozambique
5.2
23.0
Haiti
24.0
Nigeria
65.5
37.0
Niger
60.2
35.8
India
22.4
Rwanda
Sierra Leone
Uganda
7.3
25.2
15.5
14.2
20.1
20.3
Senegal
Tanzania
Zambia
63.7
7.2
30.1
37.1
10.5
19.5
Indonesia
Nepal
Philippines
59.2
90.5
16.2
19.0
39.5
34.2
42.6
Table 11. Elasticities between RNF employment indicators
and rural population density for women and men
Regression No.
R1
R2
R3
R4
R5
R6
Women
Women
Women
Men
Men
Men
0.47
0.09
0.15
-0.33
-0.32
-0.31
Density*Africa
-0.19**
-0.22**
-0.15*
0.03
-0.02
-0.02
Africa dummy
-0.25
0.1
0.04
-0.43
0.09
0.09
Sec. educ. by gender
0.03
0.11
0.35***
0.35***
Road density
0.14*
0.15**
0.17*
0.17*
Electricity
0.20**
-0.07
0.09
0.09
Sample
Population density
Ag. Output/worker, log
0.46***
0.01
No. Obs.
162
122
95
74
74
74
R-square
0.2
0.53
0.24
0.55
0.55
0.55
25
Figure 6. National remittances earnings (% GDP) and
rural population density
LBN
Remittance earnings (% GDP)
5
10
15
20
HND
JOR
HTI
SLV
NPL
NIC
GTM
TGO
SEN
PHL
BGD
MAR
BOL LBR DOM
VNM
LKA
0
ECU
KEN
NGA
BEN
MLI
TUN
KHM
PRY
SDN
UGA
GIN
SYR
MEX
SLE
CRI
COL
PER
NER
ETH
DZA
IDN
MOZ
BFACHN
CIVCMR
PAN
MYS
THA
ZMB
GHA
URY
IRN
BRA
ZAF
ARG
COG
LAO
VEN
TZA
BDI
LBY
CHL
IRQ
0
EGY
PAK
IND
RWA
500
1000
Rural population density (person per sq km)
1500
Table 11. Estimating elasticities between national
remittance earnings (% GDP) and population density
Estimator
Structure
Density variable
Population density
Population density*Africa
Total population
Lagged remittances
Lagged population density
West Africa dummy
Central Africa dummy
East Africa dummy
Southern Africa dummy
1977-87 dummy
1987-97 dummy
1997-2007 dummy
Number of observations
R-square
OLS
Levels (logs)
Agricultural
Robust
First difference
Agricultural
OLS
Levels (logs)
Rural
Robust
First difference
Rural
0.25***
0.05
-0.24***
0.97**
-0.94
-1.31**
-0.21***
0.06
0.31***
0.04
-0.23***
1.17***
-1.22**
-0.82
-0.24***
0.06
-0.67*
-1.55***
-0.90**
0.14
0.15
0.33*
0.79***
231
0.39
-0.09
0.19
-0.49
-1.40***
-0.74*
0.24
0.12
0.28*
0.72***
-0.06
0.24*
147
147
231
0.4
159
0.22
5. Conclusions
 Land pressures are severe in much of Africa, esp. high
density SSA, where small farms are getting smaller,
and will continue to get smaller as pop. grows
 Yet history shows that rural people are generally adept
at adapting to mounting land pressures.
 Ag intensification is only part of the adaptation
 The question we posed is whether Africa is different
 In many ways, the answer is yes . . .
5. Conclusions
 Adaptation 1 – Agricultural Intensification
 Africa has intensified agriculture, but largely
through high value non-perishable crops (HVCs)
 Much less historical success with cereals, and much
less potential given limited potential for irrigation
 Should we shift emphasis of research and development
strategies from cereals to HVCs?
 CGIAR, for example, barely looks at cash crops like
coffee, tea, cotton, cocoa, tobacco (even though cash
buys food!)
5. Conclusions
 Adaptation 2 – Reducing fertility rates
 Higher densities (smaller farms) apepar to lead to a
desired reduction in fertility in Africa
 But desired reductions are not met by access to
contraceptive technologies
 High-density East Africa now shows mixed policies
 Ethiopia & Rwanda are investing in family planning
(*), but Museveni (Uganda) has resisted family
planning (population growth is “a great resource”)
Asian experience suggests FP yields high returns
Fertility: children per women
8.0
7.0
F1. Fertility trends in Pakistan and Bangladesh
6.0
Countries split
5.0
4.0
Pakistan
3.5
3.0
2.0
Bangladesh
2.3
1.0
0.0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
5. Conclusions
 Adaptation 3 – Nonfarm diversification
 Weak evidence, but evidence that is there suggests
that nonfarm sector doesn’t just grow without
engines like education, infrastructure, agriculture
(also true for African cities?)
 Boom in overseas migration and remittances is new,
and unexpected.
 20 years ago, BGD and Pakistan were regarded as too
big to benefit from remittances. Not true now.
 Why isn’t Africa getting more remittances?
5. Conclusions
 Finally, we ask whether the results we find warrant a
re-think in the way high density countries pursue
rural development
 Are SSA countries thinking through the implications
of rural pop. growth for farm sizes and rural welfare?
 Do SSA countries need rural development strategies
that are more integrated with respect to smallholder
intensification, commercial farms, family planning,
migration and rural nonfarm development?
 What are the costs of not doing so?