What is the TOA-MD Model? Basic Concepts and an Example
John Antle
Roberto ValdiviaAgricultural and Resource Economics
Oregon State University
www.tradeoffs.oregonstate.edu
What is the TOA-MD Model?• The TOA-MD Model is a unique simulation tool for multi-dimensional impact assessment that uses a statistical description of a heterogeneous farm population to simulate the adoption and impacts of a new technology or a change in environmental conditions. • TOA-MD is designed to simulate what would be observed if it were possible to conduct a controlled experiment. In this experiment, a population of farms is offered the choice of continuing to use the current or “base” production system (System 1), or choosing to adopt a new system (System 2).• In fact it is never possible to carry out such ideal experiments, so TOA-MD is designed to utilize the available data to attain the best possible approximation, given the available time and other resources available to conduct the analysis. • Additionally, TOA-MD is designed to facilitate analysis of the inevitable uncertainties associated with impact assessment.
• There are two components in the TOA-MD analysis: • First, the model simulates the proportion of farms that would adopt a new system (system 2), and the proportion that would continue to use the “base” system (system 1) • Second, based on the adoption rate of the system 2, the TOA-MD model simulates selected economic, environmental and social impact indicators for adopters, non-adopters and the entire population.
• How is the TOA-MD simulation approach related to the experimental and non-experimental statistical approach to impact assessment (estimation of “treatment effects”)?
• The underlying conceptual model is very similar: economic agents (farms in this case) self-select into “treatment” or choose to adopt a technology (or adapt to an exogenous change such as climate change), and related “outcomes” occur to those self-selected into “treatment” (adoption)• For “ex post” assessment the simulation approach can be used together with experimental methods to parameterize the model and simulate outcomes of interest, including both “mean” indicators and other “threshold” indicators such as poverty rates and indicators based on other quantifiable outcomes. It can also be used to simulate “policy relevant” effects, such as technology adoption induced by payments for ecosystem services. • For “ex ante” assessment, where statistical methods cannot be used, the simulation approach provides a “laboratory” to explore potential impacts, using various types of available data to construct the future “counterfactual,” e.g., the impacts of and adaptation to climate change.
TOA-MD approach: modeling systems used by heterogeneous populations
Systems are being used in
heterogeneous populations
A system is defined in terms of household, crop, livestock and
aquaculture sub-systems
(ω)
0
Map of a heterogeneous region
Opportunity cost, system choice and adoption
Opportunity cost = v1 – v2
follows distribution ()v1 = returns to system 1V2 = returns to system 2 System 1: > 0
(non-adopters) System 2: < 0
(adopters)
opportunity cost
• Outcome distributions are associated with system choice– Farms select themselves into “non-adopter” and “adopter” sub-
populations, generating corresponding outcome distributions for these sub-populations
• Impact indicators are based on system choice and outcome distributions– TOA-MD produces mean indicators and threshold-based
indicators• Analysis shows that impacts depend on the correlations
between adoption (opportunity cost) and outcomes– Many impact assessments ignore correlations– Yet these correlations are often important for accurate impact
assessment!
Adoption, Outcome Distributions and Impact Indicators
Adoption and outcome distributions
Entire Population with adoption: 55% >
r(1,a)% non-adopters
System 1: 20% >
System 1 before adoption: 25% > threshold
System 2: 90% >
Outcome z
(z|1)
r(2,a)% adopters (z|1,a) (z|2,a)
(z|a)
Components of TOA-MD Analysis
System characterization
Adoption rate
Population (Strata)
Impact indicator design
Opportunity cost distribution
Outcome distributions
Indicators and Tradeoffs
Design
Data
Simulation
An Example: Integrated Agriculture-Aquaculture• Based on Dey et al (2010) Agricultural Economics: economic analysis of IAA• stratified survey of farms, without and with IAA
• Design of TOA-MD analysis• population: farm households in southern Malawi where aquaculture is
feasible• strata: 5 southern districts• systems:
• Subsistence crops• Crops + aquaculture, low or high integration
Subsistence crops
Aquaculture
Irrigated vegetables
-1000
-800
-600
-400
-200
0
200
400
600
800
0 10 20 30 40 50 60 70 80 90 100
Opp
ortu
nity
Cos
t
Adoption Rate (%)
Zomba Mwanza Mulanje Thyolo Mangochi
Adoption Rate and Opportunity Cost of Adopting IAA in Southern Malawi – Predicted Adoption Rate is Point Where Curves Cross the
Horizontal Axis
Poverty Rate and Adoption Rate of IAA, Southern Malawi
60
65
70
75
80
85
90
95
100
0 10 20 30 40 50 60 70 80 90 100
Pove
rty
Rate
(%)
Adoption Rate (%)
Zomba Mwanza Mulanje Thyolo Mangochi
Mean Monthly Protein Consumption and Adoption of IAA, Southern Malawi
0
0.5
1
1.5
2
2.5
0 10 20 30 40 50 60 70 80 90 100
Mea
n M
onth
ly P
rote
in C
onsu
mpti
on p
er p
erso
n (k
g)
Adoption Rate (%)
Zomba Mwanza Mulanje Thyolo Mangochi
Note most improvement occurs for those districts with lowest
protein consumption
Relationship between adoption and Protein Consumption, Non-adopters and Adopters of IAA, Mulanje Dist., Malawi
Non-adopter sub-population and entire
population are equal at 0% adoption
Adopter sub-population and entire population are equal at 100% adoption
Slope of relationship between indicator and adoption rate has
same sign as the correlation between opp cost and the
outcome variable (negative in this case)
Relationship between adoption and Mean Returns per Farm, Non-Adopters and Adopters of IAA, Mulanje District, Malawi
Economic outcomes that are positively related to net returns have a maximum in the entire population at the
predicted adoption rate (41% in this example)
-1000
-800
-600
-400
-200
0
200
400
600
800
0 10 20 30 40 50 60 70 80 90 100
Op
po
rtu
nit
y C
ost
Adoption Rate (%)
Zomba Mwanza Mulanje Thyolo Mangochi
The various “treatment effects” discussed in the experimental and non-experimental statistical literatures are either equivalent to or closely related to the mean indicators generated by the TOA-MD model:
• AveTreatment Effect = mean population indicator• Marginal Treatment Effect = change in mean indicator• Treatment on Treated = related to mean indicator of adopters• Treatment on Untreated = related to mean indicator of non-
adopters• Policy relevant treatment effect = derived from mean
indicators
Ave. farm income ($/year) Poverty rate (%)
Mean Monthly Protein Consumption (kg/person)
Strata Adoption rate (%)
base (no adoption)
% Change on
population
% Change
on adopters
base (no adoption)
% Change on
population
% Change
on adopters
base (no adoption)
% Change on
population
% Change
on adopters
ZOMBA 49.22 112.47 54.60% 135.62% 87.50 -15.81% -42.48% 1.41 12.86% 38.95%
MWANZA 49.40 89.01 50.77% 137.61% 99.16 -16.01% -51.88% 1.94 0.30% 10.64%
MULANJE 40.81 81.01 54.46% 179.51% 84.30 -11.38% -44.26% 0.65 53.10% 191.35%
THYOLO 41.92 170.85 41.85% 116.92% 95.93 -16.48% -56.11% 1.75 -0.49% 28.63%
MANGOCHI 37.95 188.62 30.77% 116.63% 72.24 -10.53% -53.66% 0.77 56.42% 178.33%
REGION 44.49 123.90 45.23% 132.70% 87.11 -11.25% -30.45% 1.29 15.32% 59.00% 1
Summary: Impacts of IAA Adoption on Farm Population and IAA Adopters
Conclusions• TOA-MD is a unique simulation tool for multi-dimensional impact assessment of agricultural systems• The Malawi case study illustrates how it can be used with available data to simulate:
• the adoption rate of a new technology• the economic, environmental or social impacts of the new technology
• The model can also be used for analysis of ecosystem services, and impacts of climate change and other environmental change• Training in use of the model, and the model software are available from the TOA Team.
• More info is available at : http://tradeoffs.oregonstate.edu