leveraging social networks to enhance agricultural extension: lessons from an rct study by paul...

Leveraging Social Networks to Enhance

Agricultural Extension Collaboration between Lori Beaman,

Ariel BenYishay, Paul Fatch, Jeremy

Magruder and Mushfiq Mobarak

MOTIVATION

Improving food security, raising farm incomes, and reducing environmental damage depend on smallholder adoption of new

technologies

Technologies that would minimize adverse environmental effects and increase long-term yields exist, but have yet to be adopted on a wide

scale

Low productivity in agriculture and environmentally unsustainable farming challenges are pressing development challenges for Malawi

2

Impact evaluation: Central Questions

What are the most effective ways to convey

information about new technologies to farmers?

What can MoAIWD do to increase rates of adoption

of technologies that will increase long-run

productivity and ensure sustainable use of natural

resources?

3

Motivation for peer farmers

When making decisions, people may be influenced by friends and neighbors

Will allow AEDOs to take advantage of existing channels of social networks, which may increase their ability to convey information

Finding the right partner farmers could be a low-cost way for the Ministry to boost adoption rates

4

Selected Districts

Conservation Agriculture

Districts

Mwanza

Machinga

Nkhotakota

5

Technologies

Conservation Agriculture

• Pit Planting • Use basins

instead of ridges • Very low

adoption at start of project

Crop Residue Management

• Composting

6

Role of the seed farmer

Implement the new technology

on their own farm

Talk to their friends and neighbors about what they are doing

Try to convince people in their social groups to

adopt the new technology

7

Ridges vs Pit Planting

Ridges Pit Planting

8

Evaluation Strategy

• Selection of partners to maximize adoption using theoretical diffusion model and detailed social network data

• 100 villages

Network Partners

• Use geography as a proxy to full social network mapping along with diffusion model

• Policy relevant alternative to Network Partners treatment, since low cost and scalable

• 50 villages

Geo Partners

• Business as usual: extension agent chooses partners

• 50 villages Benchmark

9

An Example Network 10

Data and Timeline 11

0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

0.400

Figure 1: Training Partner Farmers on Pit planting Increases Adoption

Trained Not trained

12

Adoption of PP Increases over Benchmark

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0.160

Year 1 Year 2 Year 3

Figure 2: Adoption Rates across Network, Geo and Benchmark partner villages

Network partners Geo partners Benchmark partners

13

Expect Strongest Effects in Places that didn’t know about technology before the project

-0.020

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0.160

0.180

Year 1 Year 2 Year 3

Figure 3: Adoption Rates in Villages with low pit planting use at Baseline

Network Partners Geo Partners Benchmark Partners

14

Does the choice of partner farmers matter? Yes

0.000

0.200

0.400

0.600

0.800

1.000

1.200

Year 1 Year 2 Year 3

Figure 3: Any adoption in the village (excluding trained partners)

Network partners Geo partners Benchmark partners

15

Who we train mattered

Gradual adoption over time in all villages (from 0 to 8% over 3 years), but

Social network treatments increased adoption over benchmark

Remember, this is above extension workers choosing carefully – not obvious that network data would beat this

16

Conclusions

• First: There are important and valuable technologies for which information is the only constraint to adoption

• Farmers, like the rest of us, are not perfectly informed

• Sometimes, even with these technologies, adoption is slow and difficult (still fairly low and increasing in year 3)

• Previous studies: social learning is important for tech adoption

17

Conclusions

• This study: We can identify partners that increase the speed of diffusion through social networks

• Looks like best is to treat the densest part of the network intensively rather than going for broad-based exposure.

• These partners are, though, hard to identify

• Avenue for future research (& collaboration to bring this to scale!): how to make networks work for policy in a more cost effective way

18