This study examines the adoption of improved seeds and fertilizer by farmers in Ethiopia between 1999 and 2009. The researchers find that a farmer's decision to adopt new technologies is strongly influenced by the adoption decisions of their neighbors. A one standard deviation increase in the percentage of neighbors adopting a technology increases a farmer's likelihood of adoption by 11-12% for seeds and 19% for fertilizer. While learning from others helps drive diffusion, it is a gradual process limited by current local adoption levels.

1. The Influence of Neighbours
P. Krishnan & M. Patnam
Cambridge University, UK
Adoption of Seed and Fertiliser in
Ethiopia

2. Slow Adoption
• Adoption of fertiliser and improved seeds key
to increased land productivity
• However, the adoption and diffusion of such
technologies has been slow
• Many potential reasons

3. Reasons for low adoption
• Constraints on supply/distribution
• Credit constraints (Dercon)
• Behaviorial (overestimating patience) (Duflo
et al)
• Heterogenous and uncertain returns (Suri)
• Imperfect information (Udry & Conley)
• We concentrate on the last

4. ERHS 1999-2009
1999 2009
Adopt new seed % 0.18 0.23
Use fertiliser % 0.62 0.64
Neighbours adopting new seed % 0.17 (0.25) 0.21 (0.25)
Neighbours using fertiliser % 0.59 (0.34) 0.63 (0.34)
Correlation: own and neighbour seed adoption 0.51 0.29
Correlation: Own and neighbour fertiliser adoption 0.31 0.17

5. Seed and Fertiliser adoption
Adoption of new seed across the decade
did not adopt seed adopted new seed
2009 2009
did not adopt seed 1999 663 (69%) 157 (16%)
adopted new seed 1999 82 (8%) 70 (7%)
Adoption of fertiliser across the decade
did not adopt fertiliser adopted fertiliser
2009 2009
did not adopt fertiliser 1999 258 (27%) 177 (18%)
adopted fertiliser 1999 92 (9%) 445 (46%)

6. Differences between adopters and
non-adopters
• Seed: More educated, less oxen, no real
differences in wealth
• Fertiliser: Wealthier, more and better land.
better educated
• Both: More extension visits in 99
• Key difference: adopters are more likely to
have neighbours who are adopters too

7. Identifying the effects of neighbours
• Spatial neighbours based on a distance of 1
km from the household.
• Instrument for the average neighbour's
decision to adopt : the non-overlapping sets
of neighbours - or neighbours of neighbours
• Affect the decisions of spatial neighbours
directly - but not the household's own
decision.

8. Impacts of neighbours on seed
adoption
Adoption 1999 Adoption 2009
Probit IV Probit Probit IV Probit
Neighbours adopt (sd.23/.24) -0.15 (0.07)** 0.46 (0.20)** -0.11(0.07) 0.47 (0.25)**
Cragg-Donald F 136.47 57.94
Sample Size 972
Changes over time in seed adoption
OLS IV
Neighbours adopt -0.17 (0.08)** 0.49 (0.28)**
Cragg-Donald F 72.46
Sample Size 972
*Significant at 10% ** Significant at 5% ***Significant at 1%

9. The importance of neighbours
• An increase of one standard deviation in the
average neighbours' adoption raises the
probability of own adoption by about 11% in
1999 and 12% in 2009.
• Average adoption rates range from 0.18-0.23,
so this is large - more than double current
levels.

10. Neighbours influence on fertiliser
adoption
Adoption fertiliser 1999 Adoption fertiliser 2009
Probit Probit IV Probit Probit IV
Neighbours adopt 0.18 (0.09)* 0.53 (0.21)** 0.10 (0.11) 0.53 (0.36)
Cragg-Donald F 22.98 57.94
Sample Size 972
Change over time in fertiliser adoption
OLS IV
Neighbours adopt 0.09 (0.07) 0.41(0.22)*
Cragg-Donald F 120.35
Sample Size 972

11. Impacts of neighbours again
• A one standard deviation increase in the
average fertiliser adoption of neighbours
raises own probabilities of adoption of
fertiliser by 19%.
• The effects are similar in both years
• A substantial effect given that adoption is
already about 62% in the survey areas.

12. Impact on own adoption, given levels
of adoption by neighbours

13. Summary
• Seed: The speed of diffusion through learning
from others increases until local diffusion
levels of 70 percent have been reached
• Fertiliser: these benefits from learning appear
to tail off at about 30 percent diffusion levels.
• In both cases: an increase by 10 percent in
diffusion in the neighbourhood increases the
probability of adopting by about 5 percent at
current levels of diffusion

14. Is this really learning? So what?
• We examine improved seed that is bought
• Potentially, supply constraints might mean
sharing of seed more important
• Yes, likely to be learning but..
• Learning from others is a powerful tool, but is
not amenable to rapid change through policy,
as it reflects steady but careful learning from
the experiences of others.