Xiaodong ChenKennedy School of Government
Harvard University
Agent-based Modeling of the Effects of Social Norms on Enrollment in
Payments for Ecosystem Services
Global Conservation Investments
Conservation investments are far below the requirements for conserving ecosystems worldwide
The efficiency of conservation investments may be improved through Payments for Ecosystem Services (PES)
Factors considered for efficient conservation investment Biological values, demographic conditions, economic conditions,
political conditions
Little is known about the effects of social norms
Social norms are shared understanding of how individual members should behave in a community
Social norms can be represented as actions of members in a community
Social norms may have substantial impacts on human decision-making
Changes in human decision-making due to social norms may in turn change social norms
Emergence and Evolution of Social Norms
Agent-based modeling is a bottom-up approach focusing on individual decision making
Agents are capable of interacting with other agents to obtain ‘knowledge’ about others’ actions and perceive social norms
Perceived social norms of agents may be imperfect and contain random noise
Agents improve their ‘knowledge’ about social norms through ‘learning’ from interactions with others
Agent-based Modeling of the Evolution of Social Norms
Main objective: increase forests and grassland to prevent soil erosion
Secondary objective: restore ecosystems and provide wildlife habitat
Payment: 3450 yuan/ha in southwest 2400 yuan/ha in
northwest (currently, 1 USD = 6.7
yuan)
Payments for Ecosystem Services in China-- Grain-to-Green Program (GTGP)
An example GTGP plot
Farmland in cabbage
Wolong Nature Reserve, ChinaWolong Nature Reserve, China
Habitat to over 6000 plant and animal species
Home to about 4500 people
Objectives
Develop agent-based model to simulate effects of social norms on land enrollment in PES programs
Assess effects of PES program design on patterns of social norms
Methods
Household Interviews
Sample: 304 of 1197 households
Response rate: 95%
Questions: demographic, socioeconomic,
land reconversion (22.6%)
Policy scenario questions for people who plan to reconvert GTGP plots when current payments end
Policy Scenarios
GTGP land that household planned to reconvert after the payment ends
Re-enrollment plans if a new policy was in place
Policy Attributes Payment: 1500, 3000, 4500 yuan/ha
Neighbors’ behavior: 25%, 50%, 75% would reconvert
Attribute combinations varied across respondents
Opportunity Cost Estimation
),0|(*)(
)(1)(
reconvertpayreenrollPreconvertP
reconvertPreenrollP
jj
jj
)( jreenrollP is the probability the jth GTGP plot is re-enrolled
)( jreconvertP is the probability the jth GTGP plot is reconverted to crop production after the payments end
),0|( reconvertpayreenrollP j is the probability of re-enrolling the jth GTGP plot under a new payment program, for plots that will be reconverted to crop production after initial payments end
Opportunity Cost Estimation
P(reenroll j) is estimated at different payments
The per hectare opportunity cost of a land plot is the payment level at which the land plot will be re-enrolled
Simulation Rules -- PES Program
PES program contracts last for one unit of time
All households make reenrollment decisions for all of their GTGP plots at each time point
Simulation Rules -- Agents
Each household was modeled as an agent
Agents were simulated for multiple units of time to allow for multiple opportunities for making reenrollment decisions
Agents interact with each other to perceive social norms as measured by reenrollment decisions of their neighboring agents at previous times
Agents would reenroll a GTGP plot if the payment is larger than the opportunity cost of the plot
Social norms in each community were measured as the proportion of households reconverting their GTGP plots
Agents cannot obtain all information on social norms through one round of interaction
Perceived social norms time = 1 = 0.5
Perceived social norms time = j =
knowledge time = j * neighbors’ action time = j-1 +
(1 – knowledge time = j) * random norm
Knowledge increases through agents’ ‘learning’ from additional interactions with other agents
Simulation Rules – Emergence and Evolution of Social Norms
Simulation Experiments
One-time reenrollment of GTGP land under different payments
Dynamics in land reenrollment due to social norms under different levels of payments, initial_knowledge, and learn
Results
Independent variables Parameters
Proportion of neighbors’ reconverting GTGP plots -
Payment level (yuan) +
Number of people in the household -
Cropland of the household (ha) +
Age of household head +
Chi-square ***
*** p < 0.001
Pooled Logit Estimation of Re-enrollment under A New Payment Program
Amount of GTGP Land Reenrolled at Different Payments
current payment
Dynamics in GTGP Land Reenrollment under Different Payments
6.4 ha
7.7 ha
6.2 ha
Dynamics in GTGP Land Reenrollment under Different Levels of Initial_knowledge about Social Norms
Dynamics in GTGP Land Reenrollment under Different Levels of Learn about Social Norms
Conclusions
Over 15% more GTGP land can be reenrolled at the same payment if social norms are leveraged through multiple rounds of interactions.
The effects of social norms were largest at intermediate payments.
Land enrollment may converge to different levels at different times due to different levels of ‘initial_knowledge’ and ‘learn’ about social norms
Acknowledgments
People F. Lupi, L. An, R. Sheely, A. Vina, J. Liu
Financial Support National Science Foundation National Institutes of Health National Aeronautics and Space Administration Michigan Agricultural Experimental Station MSU Environmental Research Initiative
Thank You!