Minimum Data Analysis of Technology Adoption and Impact

Report
Minimum-Data Analysis of
Technology Adoption and Impact Assessment
for Agriculture-Aquaculture Systems
John Antle
Oregon State University
Roberto Valdivia
Montana State University
AquaFish CRSP RD-IA Meeting, Seattle Oct 4-7 2010
Reminder: Our Goals
• Learn about data, model for IA
• Design IA for AquaFish investigations
• Plan data collection
–
–
–
–
What data do we need?
Does your project have data?
Do you need to collect more data?
Set up a work plan for data collection and
preparation.
• Plan analysis & presentations
Impact Assessment using the
TOA-MD model
• Technology Adoption leads to Impacts
– Define the relevant farm populations
– Describe existing system and new system(s)
based on improved technology (without vs
with, baseline vs counterfactual)
– Simulate adoption of new system(s) in the
farm populations
– Quantify impacts of adoption using Indicators
Logical structure of TOA-MD: Adoption analysis
























Farm population w/base
tech & base indicators
(poverty, sustainability)



Adoption





















Sub-populations:
non-adopters (base tech & indicators)
adopters (improved tech, indicators)
Result: r% adopters, (1-r)% non-adopters
Logical structure of TOA-MD: Impact analysis
Example: Poverty
System 1 income
distribution (Poverty = 65%)
System 2 income distribution
(Poverty = 25%)
1-r% non-adopters
r% adopters
System 1 & 2 income
distribution (Poverty = 45% )
TOA-MD Components
Design
Population (Strata)
System characterization
Impact indicator design
Data
Opportunity cost distribution
Outcome distributions
Simulation
Adoption rate
Indicators and
Tradeoffs
Design
• Populations and strata
– Spatial: agro-ecozones, political units,
watersheds, location
– Socio-economic: farm size, wealth, age, gender
• Systems
– Crop, livestock, aquaculture subsystems
– Sub-systems composed of activites
– Farm household
• Impact indicators
– Population mean
Design (cont.)
• Impact indicators
– Economic
• Mean farm or per capita income
• Poverty
– Environmental
• On-farm: soil quality
• Off-farm: water quality, GHG emissions
– Social
• Nutrition
• Health
• Intra-household distribution
Design (cont.)
• Indicator design
– Population means
– Probability of exceeding a threshold
• Headcount poverty: % below poverty line
• Nutritional threshold
• Environmental threshold
TOA-MD Components
Design
Population (Strata)
System characterization
Impact indicator design
Data
Opportunity cost distribution
Outcome distributions
Data
• Economic
– Farmed area in each stratum (ha)
– Population mean net returns for each system in each
stratum
• Yield, price, costs of production, and land allocation to
each activity in each system
• Or … direct observation of returns to the whole system
– Standard deviation of net returns by system and stratum
– Correlation between returns to each system
• Environmental and social
– Mean and CV for each outcome for each system
– Correlation with net returns
Data (cont.)
• Household: means and CVs for
– Farm (ha), herd (head or TLU) and pond (ha) sizes
– Family size (number of persons)
– Non-agricultural income
TOA-MD Components
Design
Population (Strata)
System characterization
Impact indicator design
Data
Opportunity cost distribution
Outcome distributions
Simulation
Adoption rate
Indicators and
Tradeoffs
Simulation
• Adoption
– Baseline: all farms using system 1
– System 2 becomes available, r% adopt, (1-r)%
continue to use system 1
• Impact Indicators
– Indicators calculated for each system and each
stratum
– Population indicators are a weighted average
of indicators for farmers using each system in
each stratum.
Implementing IA: Agricultural Systems Design
Farm Household
• Income
• Health & nutrition
• Distribution & assets
Crop system
• productivity
• costs (variable, fixed)
• input use (fertilizers, other)
• land allocation
• soil & water management
Environmental processes
• on-farm (soils)
• off-farm (runoff, erosion, GHG,
pond effluent)
Aquaculture system
• productivity
• costs
• feed
• water & waste management
• pond construction
and maintenance
Livestock system
• productivity
• costs
• herd health
• feed quantity & quality
• manure management
The Malawi Case Study: Integrated Agriculture-Aquaculture
• population: southern Malawi
• strata: 5 southern districts
• systems:
• system 1: maize, beans other subsistence crops
• system 2L: system 1 with small ponds, low integration
• system 2H: system 1 with larger ponds, irrigated vegetables,
high integration
• random survey of farms by strata and system
Subsistence
crops
Aquaculture
Irrigated
vegetables
The Malawi Case Study
• We will use the case study to show you the kinds of data we need
to do impact assessment of aquaculture systems
• At the same time, we will evaluate the data needed for your
projects and design a plan to acquire them and implement IA
Subsistence
crops
Aquaculture
Irrigated
vegetables
Exercise 1: Populations, Systems & Indicators
1. Define the populations and strata for your investigations
(you may want to draw a map)
2. Develop a schematic diagram for the existing or baseline
system (system 1).
3. Develop a schematic diagram for the system(s) with
improved technologies (system 2).
4. Define the relevant indicators for your investigations.
Modeling Adoption Rates
•
Farmers choose practices to maximize expected returns
•
Farmers expect to receive v1 ($/ha/season) for system 1
•
Farmers who adopt system 2 expect to earn v2 ($/ha/season)
•
The opportunity cost of changing from system 1 to system 2 is
defined as ω = v1 – v2
Note: ω < 0 means gain from adoption of system 2
ω > 0 means loss from adoption of system 2
•
Returns vary spatially, so opportunity cost varies spatially, and
is described with the distribution (ω)
(ω)
Construct spatial
distribution of
opportunity cost
ω>0
ω<0
(adopters)
(non-adopters)
Opportunity cost () and adoption rate (r)
System 1
System 2

Adopt
Example 1
Farm 1
100
80
20
r = Σ Adopt /N = 0/1 = 0
0
Example 2
Farm 1
75
85
-10
r = Σ Adopt /N = 1/1 = 1
1
Example 3
Farm 1
Farm 2
100
75
80
20
85
-10
r = Σ Adopt /N = 1/2 = 0.5
0
1
Example 4
Farm 1
Farm 2
Farm 3
100
75
90
80
20
85
-10
120
-30
r = Σ Adopt /N = 2/3 = 0.67
0
1
1
Opportunity cost and adoption rate: 1 farm

ω =20, r = 0
()
0
r (%)
100
Opportunity cost and adoption rate: 1 farm

()
0
100
ω =-10, r = 1
r (%)
Opportunity cost and adoption rate: 2 farms

()
Rate
0
100
Opportunity cost and adoption rate: 3 farms

()
Rate
0
100
Opportunity cost and adoption rate: Many farms

()
Rate
0
100
Derivation of adoption rate
from spatial distribution of
opportunity cost

r(a) 
d
0<ω
r
()
100
ω<0
r (0)
Varying the Adoption Rate
•
•
We often want to use the model to see the effects of a range
of different adoption rates
•
E.g., low adoption rates may occur because farmers cannot
get access to the technology or lack resources to invest
•
E.g., high adoption rates may occur because farmers are
required by government regulations to use a technology
that reduces pollution
We assume farmers choose system 2 if it is more profitable
than system 1
•
As we “force” the model to lower or higher adoption rates,
average farm income in the population will be lower than
at the profit-maximizing adoption rate.
Varying the Adoption Rate (cont.)
•
To simulate the effects of farmers using different adoption
rates, the model can simulate the effect of giving all farmers an
incentive payment or penalty (PAY) for using system 2. This
payment is used to calculate the adoption rate, but it is not
included in the calculation of farm income.
•
With an incentive payment, famers use system 2 if:
v1 < v2 + PAY
•
or if
v1 – v2 < PAY
or if
ω < PAY
So when PAY = 0 we get the adoption rate that occurs without
any incentive payment or penalty. This is the adoption rate that
gives the highest average farm income in the population.
Adoption rate with incentive
payment (Pay)
, Pay
ω > Pay0
0 < ω < Pay0
()
ω<0
100
Adoption rate with Pay0 > 0
Adoption
Rate
PAY0
PAY > 0
()
PAY < 0
Adoption rates
with Pay <0
100
Adoption
Rate
Using PAY to simulate
effects of different
adoption rates
, Pay
Adoption rates with Pay >0
Parameters of the Distribution of Opportunity Cost
Farms are heterogeneous so  = v1 – v2 varies across farms.
We use the mean and variance of  in the MD model.
Mean: E () = E (v1 ) – E (v2 ) ($/ha)
Suppose system 1 has one activity, then:
E (v1 ) = P11 Y11 – C11 – f11
P11 = price of output for activity ($/Y/time)
Y11 = yield of activity 1 (Y/ha/time)
C11 = variable cost of activity 1 ($/ha/time)
f11 = fixed cost of activity 1 ($/ha/time)
Note: subscripts are system, activity
System 2 with 1 activity:
E (v2 ) = P21 Y21 – C21 – FC21
P21 = price of product of activity 1, system 2
Y21 = output of activity 1, system 2
C21 = variable cost of activity 1, system 2
f21 = fixed cost of activity 1, system 2
Note: subscripts are system, activity
Variance of :
i  standard deviation of returns, system i
ik  covariance of returns, system i and k
ρ12  12/12
2  E( - E())2 = E({v1 - E(v1)} – {v2 - E(v2)})2
= 12 + 22 – 212
= 12 + 22 – 212ρ12
Multiple Activities in a System
Suppose system 1 has 2 crop activities
Each activity k uses a share W1k of the land
E.g., activity 1 = maize, activity 2 = beans
farm size = 2 ha, maize = 1.5 ha, beans = 0.5 ha
Then W11 = 1.5/2 = 0.75, W12 = 0.5/2 = 0.25
Now vik = Pik Yik – Cik – fik , for i,k = 1,2
E (v1 ) = W11 v11 + W12 v12
E (v2 ) = W21 v21 + W22 v22
Remember: first subscript = system, second subscript = activity
Data: Malawi case study
• Surveys
– Random sample of farms without IAA and with IAA in each stratum
(district)
• Areas, farm size, household size, income, health, nutrition…
• Other economic data from secondary sources
– Crop yields, prices, costs of production
– Yield variances
– Land use
• Environmental and social data, as needed:
– Pond effluent & water quantity, quality data
– Soil, climate data for crop system
• Estimate crop yields using crop simulation models?
• Estimate change in soil nutrients or soil carbon?
– Nutrition: e.g., protein consumption
Exercise 2: Data Preparation
1.
2.
3.
4.
Open Database\malawi_tables.doc. Review the data tables for
Malawi, note the way the data are organized by strata.
Open Database\MW_DATA_L.xls. This is the data template for the
model. This file is set up with data to analyze adoption of System 2L,
the aquaculture system with low integration.
Look at the Variable Description sheet in the data template, and
then the other sheets with highlighted variables. Identify how data
from the tables have been put into the sheets in the template.
As you review the data template, start preparing a list of data
needed for the systems you identified in Exercise 1 for your
investigations. Identify which variables are available, which ones
may need to be obtained from a survey or from secondary sources.
Exercise 3: Run the Model for Adoption of the LowIntegration System (System 2L)
1. Open the model Econmod\TOA_MD6_AF.xls.
2. Following the directions given, load the data
Database\MW_DATA_L.xls and run the model.
3. Examine the model output in the sheets and graphs.
4. Interpret the results, using the variable descriptions in
the data template.
5. What is the predicted adoption rate for System 2L?
6. What is the poverty rate with zero adoption? How much
is poverty reduced by adoption of System 2L?
7. Can you explain what happens to net returns per farm as
the adoption rate varies?
Exercise 4: Set up a model for adoption of
System 2H by farms using System 1
1. Find the data in the data tables for system 2H.
2. Open the data template for system 2L. Put the data for
system 2H into the sheets where system 2 data belong.
3. Run the model.
4. Compare the results for System 2L and 2H. Can you
explain the differences you see in adoption rates, poverty
and nutrition outcomes?
Exercise 5: Sensitivity analysis to the parameter RHO12
•
•
•
•
•
Explain what RHO12 represents in the model.
What happens when RHO12 = 1? (Hint: review the
formula for the variance of ω)
Run the model for adoption of System 2H with different
values of RHO12
Can you explain why the adoption rate changes as you
change RHO12?
Would you say the model is “sensitive” to this
parameter?
Exercise 6: Set up a model for adoption of
System 2H by farms using System 2L
Hint: you need data from the two data templates you have
already used.
Compare the results from the three models you have now
run: can you explain the differences?
Variance of opportunity cost and adoption rates

=0
r=0
()
0
r (%)
100
Note effect of variance on
adoption rate

similar documents