Transcript Introduction Minimum Data Analysis

Minimum-Data Analysis of Technology Adoption
and Adaptation to Climate Change
John M. Antle
Department of Ag Econ & Econ
Montana State University
Workshop on Adaptation to Climate Change
Nairobi Sept 2008
Motivation
•
Understanding technology adoption: what farms adopt
where, and why? What changes or incentives are
required to achieve a target adoption rate?
•
•
E.g., orange flesh sweet potato; practices that reduce
off-farm externalities
Adaptation to climate change: what are impacts of
climate change with present practices? What are benefits
of adaptation? Will new technologies facilitate
adaptation?
•
E.g., drought resistant varieties, more resilient crops
such as sweet potato
Methodological issues
•
Need timely, feasible analysis to inform policy decisions
•
Representative farm vs heterogeneous population
•
Technical vs economic potential
•
Data availability
Machakos: heterogeneity leads to different activities, adoption
rates, vulnerabilities
Modeling Adoption Rates in Heterogeneous
Populations
•
Farmers choose practices to max expected returns
v (p, s, z) ($/ha)
p = output & input prices, s = location, z = system 1, 2
•
Farmers earn v (p, s, 1) for current system
•
Farmers can adopt system 2 and earn
v (p, s, 2) – TC – A
where TC = transaction cost, A = other adoption costs
•
The farmer will choose system 2 if
v (p, s, 1) < v (p, s, 2) – TC – A
•
The opportunity cost of switching from 1 to 2 is
 = v (p, s, 1) – v (p, s, 2) + TC + A
 adopt system 2 if  < 0.
•
Suppose Government or NGO wants to encourage
adoption by providing incentive payment PAY (e.g., to
reduce negative externalities of syst 1, or encourage
positive externalities of syst 2)
 adopt system 2 if  < PAY.
•
Opportunity cost varies spatially, so at some sites farms
adopt system 1 and at other sites adopt system 2
Construct spatial
distribution of
opportunity cost
Derivation of adoption
rate from spatial
distribution of
opportunity cost
PAY
Case3
PAY0
Case 2
()
100
Case 1
Rate
Effect of the Changing the Variance of the Opportunity Cost: “representative
farm” is limiting case with zero variance in opportunity cost, adoption curve is a
step function
PAY
s=0
A
()
0
Rate
Analysis of Adaptation to Climate Change
•
Impacts of climate change: Productivity of vulnerable
crops declines more than resilient crops, e.g., maize vs
sorghum, tomato vs sweet potato
•
•
PAY is amount needed to compensate for loss
Adaptation is adoption of practices that are relatively less
vulnerable under the changed climate
•
Reduces loss due to climate change, or increases
gains
PES
Effect of climate
change: impact and
adaptation
Without
adaptation
()
100
With
adaptation
Rate
Minimum Data Methods to Simulate Adoption Rates
(Antle and Valdivia, AJARE 2006)
•
How to estimate the spatial distribution of opp cost of
changing practices?
•
Use “complete” data to estimate site-specific inprods
and simulate site-specific land management decisions
to construct spatial distribution of returns
•
MD approach: estimate mean, variance, covariance of
net returns distributions using available data
Need to know mean and variance of
 = v (p, s, 1) – v (p, s, 2) + TC + A
MD approach: use available data to estimate mean and
variance of 
Mean: E () = E (v1 ) – E (v2 ) + TC + A
Suppose system 1 has one activity, then:
•
E (v1 ) = p11 y11 – C11 is usually observed
•
E (v2 ) = p21 y21 – C21 is estimated using Inprods* and
cost data:
•
y21 = y11 {1+ (INP21 – INP11)/INP11}
* Inprod = inherent productivity = expected yield at a site with
“typical” management
•
•
C21 is estimated using C11 and other information on
changes in practices
TC and A are estimated using available data, if
relevant
Variance of returns:
•
Observation: cost of production c   y where  is a
constant and y is yield
Then v = py – c  (p - ) y and CV of v is equal to
CV of y
•
Recall:  = v (p, s, 1) – v (p, s, 2) + TC + A so we
know s2 = s12 + s22 - 2s12
•
Usually observe s12, can assume s12  s22
•
s12 difficult to observe. Can assume correlation is
positive and high in most cases. If s12  s22 = s2 then
s2  2s2 - 2s12  s2 = 2s2(1 – 12)
•
Most systems involve multiple activities (crops, livestock).
s12 and s22 depend on variances and covariances of
returns to each activity. In the MD model, we assume all
correlations between activities within system 1 are equal
(1), and make the same assumption for system 2 (2).
•
In general, incentive payments are calculated as
PAY = PES * ES
Where PES = $/unit of ES, ES = services / ha
•
For adoption analysis, set ES = 1, then
PAY = PES ($/ha)
Conclusion: to implement MD approach we need:
•
Mean yields for system 1
•
Either mean yields for system 2, or Inprods for each
activity in each system
•
Output prices and cost of production for each activity
•
Variances (or CVs) of returns (yields) for each system
•
Correlation of returns to activities within each system (1
and 2)
•
Correlation of returns between systems 1 and 2 (12)
Implementation: Data files and programs (Excel, SAS, R)
Inprods
Data for system 1
Data for system 2
TEMPL1
TEMPL2
and ES rate
REGIONS
Calculate mean and variance of opportunity cost
Tradeoff file
TRDMD
Sample distribution of opportunity cost
Adoption decision by field
Aggregate to get regional adoption curve or ES
supply curve
Validation: Compare Detailed Models to MD Model
•
3 carbon sequestration studies
-
Montana
-
Kenya
-
Senegal
Comparison of EP and MD Models: carbon supply for Montana
wheat system
100
90
80
70
$/MgC
60
50
40
30
20
10
0
0
50000
100000
150000
200000
250000
300000
350000
MgC/yr
EP Model
MD Rho=0.90
MD Rho=0.95
MD Rho=0.99
400000
450000
Comparison of EP and MD models: Carbon contract participation
in Machakos, Kenya Case Study (Full model = 700 parms, MD = 75)
16000
14000
Carbon Price (KS/MgC)
12000
10000
8000
6000
4000
2000
0
0.60
0.65
0.70
0.75
0.80
0.85
Participation Rate
Full Model
MD Model Rho=0.8
MD Model Rho=0.9
MD Model Rho=0.8 and V1=V2
0.90
Comparison of EP and MD models: Carbon Contract
Participation in Senegal Peanut Basin
100000
87500
Carbon Price
75000
62500
50000
37500
25000
12500
0
0
10
20
30
40
50
60
70
80
90
Participation Rate
rho=.6
rho=.7
rho=.8
rho=.9
rho=.95
F_EP
100
Where do I get minimum data?
• Regional (AEZ) data
– Soils & climate
– Observed yields, yield trials
– Budgets for costs of production, observed prices
• Farm survey data
– Yields & yield variability
– Mean costs of production & prices
MD Examples
• Adoption of new variety for an existing crop
– Inprods, CV, price, cost of production, correlation
• Adoption of new crop
– Yield in TEMPL1, weights
– Partial adoption of new variety: add new variety
as new crop, set weights
• Ecosystem services
– ES rate, Inprods
MD Examples (cont.)
• Climate change
– Modify inprods, adjust weights?
• Adaptation
– Modify inprods for base, for adapted system with
climate change
Climate impact & adaptation
• REGIONSCC: climate only
• REGIONSCM: climate with improved
maize
• REGIONSCP: climate with sweet potato
Country Team Work Plans
•
•
•
•
•
•
System & scenario identification
Data acquisition
Team composition
Collaboration plan
Follow-up workshop
Publication