1. Learning About a New Technology: Pineapple in Ghana
Wenbo Huang

2. Introduction Technology adoption is fundamental to development
Characteristics of technology usually not transparent to new user – an investment in learning required
Multiple adopters in similar circumstances (often case with innovation in agriculture):
• process of learning may be social • learn of characteristics from each other
This paper investigates role of social learning in diffusion of new agricultural technology in Ghana

3. Over last decade in Akwapim South district of Ghana
Traditional established system of maize and cassava intercropping for sale to urban consumers
Recently transformed into intensive production of pineapple for export to European markets
Transformation involves adoption of new technologies
• intensive use of fertilizer and other agricultural chemicals
Measuring extent of social learning is difficult:
set of neighbours from whom can learn is difficult to define
distinguishing learning from other phenomena that may give similar observed outcomes
In absence of learning - act like neighbors: • interdependent preferences, technologies • subject to related unobservable shocks

4. Learning Model yi,t+1=wi,t f(xi,t)+εi,t+1
Simple model to motivate empirical specification Farmers trying to learn about responsiveness of output y to input x
yi,t+1=wi,t f(xi,t)+εi,t+1
Farmers do not know function f ---- object of learning
wi,t = positive exogenous growing conditions
(correlated across farmers and across time – observable to farmer but not to econometrician)

5. yi,t+1= output
xi,t = input (fertilizer)
εi,t+1= productivity shock
yi,t+1=wi,t f(xi,t)+εi,t+1

6. Choose input x to maximize subjective expectation of time t+1 profits for this input choice:
πi,t+1(x,wi,t)≡(wi,t f (x)+εi,t)− px
where yi,t+1=wi,tf(xi,t)+εi,t+1 Choosing inputs such that:
E π (x*,w )≥E π (x,w ) i,t i,t+1 i,t i,t i,t i,t+1 i,t
all x∈sup(xi,t)

8. Two main reasons for farmers' of x*i,t to be interrelated
4/24/2017
Two main reasons for farmers' of x*i,t to be interrelated
First {Spatial Effect}
Spatial or serial correlation in growing conditions induce correlation in farmers' fertilizer choices
-farmers face similar realizations of growing conditions
-make similar decisions about optimal amount of fertilizer use
Second{Learning Effect}
Farmers with related subjective information sets have similar subjective expectations of growing-conditions-adjusted output, g.
Similar input choices when face similar prices and growing conditions

9. 4/24/2017
Implications
Farmers tend to adjust input use toward surprisingly successful input levels, and higher than expected profits at the currently utilized input level will make farmers less likely to change from that level.
Farmers tend to adjust input use away from an input level that was less profitable than expected.
An oberservation of profit sufficiently above expecations will induce a farmer to switch to that level of input use.
The probability of changing input levels in response to a given piece of information is decreasing in a farmers's experience.

10. Implications for Actions
4/24/2017
Implications for Actions
Suppose farmer faces constant growing conditions.
New info allows him to locally learn about f at input level X(k,t)
With constant growing conditions--only expected profits associated with input level X(k,t) shift
-good news(higher than expected profits)--upward shift
-bad news(lower than expected profits)--downward shift
Responses to changes in expected profits different for farmers who used X(k,t) in previous planting vs. those who used alternative input level

11. Consider farmers who used X(k,t)
4/24/2017
Consider farmers who used X(k,t)
Incareases in expected profits, at input level X(k,t)
-reinforce farmers beliefs that X(k,t) is optimal
-continue to use it
Sufficiently larger decrease in expected profits at input level X(k,t)
-may switch to alternative
-bad news about X(k,t) provides no info about which alternative may choose

12. Consider farmers who previously used alternative to X(k,t)
4/24/2017
Consider farmers who previously used alternative to X(k,t)
Bad news： do not change , already thought X(k,t) was inferior option
Good news: may switch to X(k,t), if increase sufficientlt large.

13. Data

14. Two enumerators lived in or near each village and interviewed respondents in 15 rounds at intervals of six weeks.

15. 4/24/2017
We begin with information on pineapple being grown on 406 plots by 132 farmers. Of these plots, 288 were planted during oursurvey. Plot input data are missing on 3 of these plots, leaving 285; 77 of these were planted too late in our survey for fertilizer application to be completed before the end of fieldwork, leaving 208 plantings. We are missing data for some rounds on 8 of these, leaving 200 plantings; 87 ofthese are the first planting in our survey by particular farmers, leaving 113 observed changes in ertilizer use. GIS information is missing on 6 of these plots, leaving information on 107 changes in fertilizer use by 47 farmers.

17. The Figure above shows the patttern of adoption of pineapples in our sample villages: from less than 10 percent in 1990, pineapples spread very rapidly until more than 46 percent of farmers were cultivating pineapples in 1997.
For some of the analysis divide sample into 2 groups:
experienced (adopted before 1994) and inexperienced (adopted more recently) Pineapple cultivators are richer, male, more likely to be in each others information neighbourhood

18. Pineapple production in Ghana
4/24/2017
Pineapple production in Ghana

19. Pineapple production in Ghana
4/24/2017
Pineapple production in Ghana
Most salient departure from traditional techniques is use of new chemical input.
Pineapple cultivation sequence( period t is 6 weeks):
-plot planted at time t receives crucial fertilizer inputs during t+1 to t+4
-chemically forced flowering occurs approximately at t+5
-pineapple harvest is complete by t+9

20. Pineapple production in Ghana
4/24/2017
Pineapple production in Ghana
Know potential profitability of plot at time t+5, outcome is fully known at time t+9
Plot receives fertilizer t+1 to t+4---fertilizer inputs can be influenced by neighbours experiences.
X(i,t)=fertilizer applied during periods t+1 to t+4
can be influenced during this time.

21. Communication and Knowledge
4/24/2017
Communication and Knowledge

22. 4/24/2017
Pineapple farmers especially veteran pineapple farmers are more likely to be in each other's information neighborhood than would be expected by chance.
Table 2 provides a summary of our baseline information link distribution by experience. Over 20 percent of veteran pineapple farmers (within each village) have approached each other for advice about farming, while only 2 percent of nonpineapple farmers are in each others' information neighborhood.
A similar pattern is observed using our other information metrics. It may be the case that these information connections were important determinants of the adoption process.

23. 4/24/2017
Conclusion
Evidence that social learning is important in diffusion of knowledge regarding pineapple cultivation in Ghana.
Take advantage of data to identify learning effects in economy undergoing rapid technological change
Farmers more likely to change input levels of fertilizer use on receipt of bad about profitability of their previous level of fertilizer use
Less likely to change when observe bad news about profitability of alternative levels of fertilizer

24. Conclusion Magnitudes of innovatuions in fertilizer use:
4/24/2017
Conclusion
Magnitudes of innovatuions in fertilizer use:
Farmer incareses(decreases) his use after someone with whom he shares information achieves higher than expected profits when useing more (less) fertilizer than he did