This study was presented during the conference “Production and Carbon Dynamics in Sustainable Agricultural and Forest Systems in Africa” held in September, 2010.

1. The economics of sustainable land management
practices in the Ethiopian highlands
Menale Kassie, University of Gothenburg; Precious Zikhali, Centre for World Food Studies (SOW-VU); John
Pender, United States Department of Agriculture (USDA); Gunnar Köhlin, University of Gothenburg
ABSTRACT: This paper uses data from household and plot-level surveys conducted in the highlands of Ethiopia. We
examine the contribution of sustainable land management practices to net value of agricultural production in areas with
low versus high agricultural potential. A combination of parametric and non-parametric estimation techniques is used to
check result robustness. Both techniques consistently predict that minimum tillage is superior to commercial fertilisers,
as are farmers’ traditional practices without commercial fertilisers, in enhancing crop productivity in the low agricultural
potential areas. In the high agricultural potential areas, by contrast, use of commercial fertilisers is superior to both
minimum tillage and farmers’ traditional practices without commercial fertilisers. The results are found to be insensitive
to hidden bias. Our findings imply a need for careful agro-ecological targeting when developing, promoting, and scaling
up sustainable land management practices.
***
DISCUSSION AFTER THE PRESENTATION: The presentation was followed by a question regarding strategies on how
to use the study to fill the gap between research and policy processes. It was replied that the study had been presented
to the Ethiopian Ministry of Agriculture and that workshops have been organised over the three years at the regional
level to discuss the results together with local and international researchers and policy makers. Discussions have also
taken place with the World Bank on how to bring these kinds of studies together and synthesise the results in order to
develop a tool to guide the promotion of land management strategies in various agro-ecological areas.
There was also another comment suggesting that it is not only relevant to carry out research on which land management
strategies work where, but also to look at the approaches in order to promote local participation and farmers’ own
research and innovations. A final question concerned the Ethiopian extension system, which is highly politically driven.

2. Rationale
• Ethiopian economy highly dependent on agriculture
• Severe land degradation
• Low agricultural productivity
• High dependency on food aid
• Response from Government, NGOs and donors:
– massive programs of natural resource management to
reduce environmental degradation, reduce poverty, and
increase agricultural productivity and food security

3. However…
• …Success has been limited!
• Low adoption, dis-adoption or reduced use of
technologies
– e.g., 16 kg of nutrients per hectare (EEA/EEPRI
2006)
• Continued low productivity!

4. Why limited success?
• Blanket recommendation: Technology packages are not
site or household specific and are disseminated through
a ‘quota’ system, eg:
-Commercial fertilizer: 100 kg of Di-Ammonium
Phosphate (DAP) and 100 kg of urea per hectare is
promoted all over Ethiopia
-Uniform SWC technologies released and promoted
disregarding local agro-ecological and socio-economic
variations

5. Realize!
• Economic returns to different farm technologies vary
by agro-ecology:
– e.g. physical soil and water conservation investments (e.g. stone
terrace) impacts on productivity are greater in low moisture and low
agricultural potential areas than in high moisture and high agricultural
potential areas (Gebremedhin et al. 1999; Benin, 2006; Kassie et al.
2008)
• Need rigorous empirical research on where particular SLM
interventions are likely to be successful, to ensure sustainable
adoption of technologies and beneficial impacts on
productivity and other outcomes

6. Three comparisons:
Impacts on net value of production in high and low
rainfall areas:
1. Commercial Fertilizer (CF) versus Farmers’ Traditional
Practices (FTP) (i.e. traditional tillage without CF)
2. Minimum Tillage without commercial fertilizer
(MTWOCF) versus FTP and,
3. Minimum Tillage (MTWOCF) versus Commercial
Fertilizer (CF)

7. Data-1
• Household-and plot-level data conducted in 1998 and 2001 in the
highlands (above an altitude of 1,500 m.a.s.l) of the Tigray and
Amhara regions of Ethiopia.
• A stratified random sample of 99 Peasant Associations was
selected from highland areas of the two regions.
• The Tigray region is typically low moisture and generally low
agricultural potential region (Benin, 2006).
• The Amhara region has greater variation in agro-ecological zones
that have been classified in ”high potential” and ”low potential”
areas, primarily based on rainfall patterns.

8. Descriptive statistics
Variables Amhara region Tigray region
Sampled household 396 357
Sampled villages 98 100
Sampled plots 1365 1113
Rainfall 1981 mm 641 mm
Population density 144 person/km2 141 person/km2
Minimum Tillage plots 15% 13%
Fertilized plots 30% 35%
Extension system Same Same
Rural credit service Same Same
Seed and fertilizer markets
and distribution systems
Same Same
Net value of production 2140 ETB 1730 ETB

9. Estimation methods
• Semi-parametric method:
– Propensity score matching (PSM) method:
construction of the counterfactual and reduce
problems arising from selection biases. Find a group
of non-adopters plots similar to the adopters
• Parametric method:
– Switching regression framework: to differentiate
each coefficient for adopters and non-adopters
• The parametric analysis is based on matched samples of
adopters and non-adopters obtained from the propensity
score matching (PSM) process.

10. PSM matching quality
Common support/overlap region for comparisons
Effect of CF compared to FTP in high potential
areas of Amhara region
Effect of CF compared to FTP in low
potential areas of Amhara region
Effect of MT compared to FTP in high potential
areas of Amhara region
Effect of MT compared to FTP in low
potential areas of Amhara region
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support

11. Common support/ overlap region
Effect of CF compared to FTP in Tigray region Effect of MT compared to FTP in Tigray
region
Effect of MT compared to CF in Amhara region Effect of MT compared to CF in Tigray
region
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support
0 .2 .4 .6 .8 1
Propensity Score
Untreated Treated: On support
Treated: Off support

12. Reminder: Three comparisons of net value
of agricultural production
Three comparisons undertaken to assess Minimum Tillage
(MT) and Commercial Fertilizer impacts on productivity.
1. Commercial Fertilizer (CF) versus Farmers’ Traditional
Practices (FTP) (i.e. traditional tillage without CF)
2. Minimum Tillage without commercial fertilizer
(MTWOCF) versus Traditional Practices and,
3. Minimum Tillage versus Commercial Fertilizer

13. 1. Commercial Fertilizer (CF) vs Farmers’ Traditional Practices
(FTP) (Average adoption effects - Semi-parametric method)
High potential areas Low potential areas
Amhara Amhara Tigray
NNM KBM NNM KBM NNM KBM
Average adoption
effect 1377A 1083A 118 279 56 142
Standard error 349 257 488 399 234 186
Number of observations within common support
Number of treated 313 313 46 45 356 356
Number of control 447 447 331 331 607 607
A significant at 1%; B significant at 5%.
Notes: NNM = nearest neighbor matching; KBM = kernel based matching;

14. 2. Minimum tillage (MTWOCF) vs FTP
(Average adoption effects (ATT)-Semi-parametric method)
High potential areas Low potential areas
Amhara Amhara Tigray
NNM KBM NNM KBM NNM KBM
Average adoption
effect
19 253 510B 277 715A 694A
Standard error 994 446 246 219 313 316
Number of observations within common support
Number of treated 19 21 131 131 109 109
Number of control 391 391 349 349 606 606
A significant at 1%; B significant at 5%.
Notes: NNM = nearest neighbor matching; KBM = kernel based matching;

15. 3. Minimum tillage (MTWOCF) vs Commercial Fertilizer
(Average adoption effects (ATT)-Semi-parametric method)
Amhara Tigray
NNM KBM NNM KBM
Average adoption
effect
-1240A -935A
949A 303
Standard error 519 412 372 465
Number of observations within common support
Number of treated 370 370 92 92
Number of control 112 112 357 357
A significant at 1%; B significant at 5%.
Notes: NNM = nearest neighbor matching; KBM = kernel based matching;

16. Results from switching regressions
(Average adoption effect (ATT)-parametric method)
AMHARA REGION TIGRAY REGION
High potential
areas
Low potential
areas
Entire
sample
Entire
sample
Entire
sample
CF
vs.
FTP
MTWOCF
vs.
FTP
CF
vs.
FTP
MTWOCF
vs.
FTP
MTWOCF
vs.
CF
Average adoption effect 1051A 293B 173 650B 785A
Standard error 229 149 145 245 302
Number of matched observations
Number of treated 313 131 356 109 92
Number of control 127 74 115 73 58
A significant at 1%; B significant at 5%.
Notes: CF = commercial fertilizer; FTP = farmers’ traditional practices; MTWOCF = minimum tillage without
commercial fertilizer.
Source: Own calculation

17. Conclusions-1
• Minimum tillage gives higher productivity gains compared to
commercial fertilizer in the low agricultural potential areas
• Commercial fertilizer gives higher productivity gains compared
to minimum tillage in high agricultural potential areas
• A one-size-fits-all approach in developing and promoting
technologies not recommended: different strategies are
needed for different environments

18. Conclusions-2
• Relying on external inputs (such as chemical fertilizers) in low-
potential areas, which has been the strategy in the past, is not
likely to be beneficial unless moisture availability issues are
addressed.
• Future research should investigate the combined effects of
minimum tillage or other moisture conservation practices and
commercial fertilizer.

19. Thank you!

20. Extra stuff

21. Results-3
• Covariate balancing indictors before and after
matching.
Amhara region Tigray region
High
potential
Low
potential
High
potential
Low
potential
Entire
sample
Entire
sample
Entire
sample
Entire
sample
CF
vs
FTP
CF
vs.
FTP
MTWOCF
vs.
FTP
MTWOCF
vs.
FTP
MT
VS
CF
MTWOCF
vs.
FTP
CF
vs
FTP
MT
Vs.
CF
Before matching
Mean standardized
difference
19.37 20.47 23.05 22.46 37.96 16.40 14.33 23.89
Pseudo 2
R 0.295 0.374 0.285 0.287 0.580 0.249 0.122 0.358
P-value of LR 2
χ 0.000 0.000 0.031 0.000 0.000 0.000 0.000 0.000
After matching
Mean standardized
difference
6.03 11.68 12.80 9.79 11.94 7.67 3.83 10.13
Pseudo 2
R 0.055 0.029 0.112 0.090 0.139 0.086 0.015 0.106
P-value of LR 2
χ 0.111 0.815 1.000 0.650 0.208 0.973 0.998 0.997

22. Results-6
• Rosenbaum bounds sensitivity test to hidden bias
Critical value
of hidden bias
( )Γ
CF
vs.
FTP
MTWO
CF
vs.
FTP
MT
Vs.
FPT
MTWOC
F
vs.
FTP
MT
VS
CF
High
potential
areas
Low
potential
areas
Entire
sample
Entire
sample
Entire
sample
1 0001.0< 001.0< 001.0< 001.0< 001.0<
1.10 001.0< 001.0< 001.0< 001.0< 001.0<
1.20 001.0< 001.0 001.0< 001.0< 001.0
1.30 001.0< 004.0 001.0< 001.0< 003.0
1.40 001.0< 026.0 001.0< 001.0 007.0
1.50 001.0< 026.0 001.0< 002.0 014.0
1.60 001.0< 050.0 001.0< 005.0 025.0
1.70 001.0< 085.0 001.0< 012.0 042.0
1.80 001.0< 135.0 001.0< 021.0 065.0
1.90 002.0 196.0 001.0< 034.0 096.0
2.00 006.0 267.0 001.0< 053.0 132.0