Behavioral Economics and Graph Theory: How They Connect

Charts & pivot tables · Updated November 22, 2026

Behavioral economics and graph theory connect when economists use networks to study how people’s choices are shaped by their relationships. Graph theory represents people or organizations as points linked by connections; behavioral economics explains how factors such as social preferences, limited attention, and biased beliefs influence decisions within those networks. Together, they help analyze peer effects, cooperation, information spread, and other social behavior.

What the Connection Means

Behavioral economics studies how actual human decisions differ from the predictions of simplified models of fully informed, consistently rational choice. Graph theory is a branch of mathematics for describing relationships and pathways. The two are complementary rather than interchangeable: graph theory supplies a way to represent who is connected to whom, while behavioral economics supplies assumptions or evidence about how those people make decisions.

This combination is often called behavioral economics on networks, network economics, or the study of social and economic networks. Those labels cover a broad set of work, not one standardized theory. A researcher might use a graph to represent friendship ties and then examine whether a person’s saving, exercise, or adoption of a new technology is associated with the behavior of their friends. The important question is not simply whether connections exist, but how the structure and meaning of those connections affect choices.

For instance, an employee may take part in a retirement plan partly because coworkers discuss it, partly because they learn practical information, and partly because they want to fit in. A network model can distinguish the employee’s links to coworkers from the behavioral explanations for responding to them. Different explanations imply different predictions and different interventions.

How a Social Graph Represents Relationships

A graph consists of nodes, also called vertices, and edges connecting pairs of nodes. In a social or economic application, nodes might be people, households, firms, or communities. Edges might represent friendship, advice, trade, communication, or another defined relationship. The choice of edge matters: being friends is not the same as exchanging money or receiving health advice.

Graphs can encode more than a simple yes-or-no tie. A directed edge can show who gives information to whom; a weighted edge can represent how frequently two people interact. A network can also change over time, or contain several types of relationships at once. These distinctions are useful when the question depends on whether a person influences another, merely observes them, or interacts with them repeatedly.

Several structural measures help summarize a graph:

  • Degree: the number of a node’s direct connections. In a directed graph, incoming and outgoing links can be counted separately.
  • Path length: the number of links along a route between nodes. Short paths can make information reachable across a network.
  • Clustering: the extent to which a person’s contacts are also connected to one another.
  • Centrality: a family of measures describing a node’s position, such as its number of links or role in connecting otherwise separated groups.

These measures describe structure; they do not, on their own, establish that a well-connected person persuades others. A central individual may be influential, may simply be exposed to more people, or may share traits with those people.

Behavioral Mechanisms in Networks

People can respond to their connections through several distinct mechanisms. Social learning occurs when someone uses others’ choices as evidence about an uncertain option. Seeing neighbors adopt a product might suggest that it works, even if the observer does not know why they adopted it. Learning can be useful, but it can also produce information cascades: people imitate earlier choices while overlooking their own private information.

Social preferences mean that a person’s satisfaction may depend partly on other people’s outcomes. They may care about fairness, reciprocity, or helping a friend. Social norms create another channel: someone may follow a behavior because they believe it is expected in their group, or because deviation risks disapproval. These motives are not identical. A person can imitate others to learn, comply to avoid social costs, or act generously because they value a recipient’s welfare.

Networks can also affect beliefs and attention. People may notice information from close contacts more than from distant sources, trust familiar messengers, or give disproportionate weight to vivid examples. Behavioral models can represent such patterns, but the details need evidence. A graph alone cannot tell a researcher whether a person copied a friend, learned from that friend, or independently made the same choice.

Network Games and Choice Models

One way to combine the fields is to build a network game. Each person chooses an action, and the payoff from that action depends on their own circumstances and sometimes on the actions of linked people. In a coordination setting, such as adopting a communication platform, a person may benefit when many contacts use it. The behavior of neighbors can act as a strategic complement: more adoption nearby makes adoption more attractive.

Other settings involve strategic substitutes. For example, if several people compete for a limited opportunity, one person’s effort may reduce the payoff available to others. Network games describe these interdependencies, while behavioral assumptions can modify how choices are made. Instead of assuming perfect calculation, a model might allow for noisy decisions, limited information, present-focused choices, or different beliefs about what neighbors will do.

Researchers also use simpler models. A diffusion model can ask whether an innovation spreads when a person adopts after enough contacts have adopted. A threshold might represent practical usefulness, perceived safety, or a social norm. Such a model is an analytical tool, not a claim that people literally count their neighbors. The appropriate model depends on the behavior and evidence being studied.

Concrete Examples and Applications

Consider a public-health program encouraging vaccination. A network can represent family, friendship, or community ties. People may learn about clinic access through contacts, respond to trusted recommendations, or hesitate when misinformation circulates in their group. Studying only the average person may miss whether information reaches isolated groups or whether a few trusted messengers connect different communities. The graph helps locate those patterns; behavioral research is needed to understand trust, risk perceptions, and responses to messages.

In consumer behavior, a firm or public agency might study how a new service spreads between households. Adoption could result from direct recommendations, observation, or the practical benefit of joining a platform that friends already use. Those mechanisms suggest different strategies: making reliable information available, improving the service’s usefulness, or reducing barriers to trying it. Simply targeting the person with the highest degree may not work if that person is not trusted or does not connect to the relevant audience.

In workplaces and schools, network analysis can help examine mentoring, knowledge sharing, cooperation, and peer effects. In finance, it can represent links among institutions or investors, while behavioral assumptions can help explain panic, imitation, or changing risk perceptions. These applications require care: a pattern found in one network or population may not carry over to another.

How Researchers Choose a Method

The method should follow the question. Researchers first define the relevant decision and the relationship that might affect it. Then they decide whether the goal is to describe network structure, predict behavior, or estimate a causal effect. Those are different tasks: a useful prediction does not automatically reveal why a behavior occurred.

  1. Specify the population and ties. Decide who counts as a node and what an edge means, including whether ties are directed, weighted, or time-dependent.
  2. Identify plausible mechanisms. Separate learning, social pressure, shared incentives, and preferences for others’ outcomes where possible.
  3. Choose data and a design. Surveys, administrative records, direct observation, and randomized interventions each capture different information and have different limitations.
  4. Match the model to the evidence. Use a descriptive network measure for a structural question and a design capable of identifying causal effects for a causal question.

For example, an experiment could randomly provide selected participants with accurate information and then observe whether it reaches their contacts. Researchers would still need to account for how participants were selected, whether they shared the information, and whether the intervention changed other behavior. Network research is most informative when its measures and design fit the mechanism being tested.

Common Problems and Limits

A major challenge is the reflection problem: people in the same group may behave alike because they influence one another, because they face the same conditions, or because similar people formed ties in the first place. If friends exercise at similar rates, that resemblance alone cannot show that one friend caused the other to exercise. Shared neighborhoods, schedules, resources, and preferences may explain some or all of the association.

Network data can also be incomplete or inaccurate. People may forget contacts, report ties differently, or change relationships during a study. Defining a boundary is difficult: a survey of close friends may omit online contacts or family members who live elsewhere. If relevant connections are missing, measures of centrality and exposure can be misleading.

Another limitation is that network structure is not a measure of influence. High degree does not necessarily mean high trust or persuasive power. Likewise, a model that predicts diffusion does not prove the proposed psychological mechanism. Stronger conclusions typically require careful measurement, a credible comparison or experimental design, and attention to alternative explanations.

Finally, network data can be sensitive. A map of relationships may reveal identities, health concerns, political ties, or workplace interactions. Researchers and organizations should consider consent, privacy, data minimization, and the risk that an intervention could stigmatize or disadvantage a group. The fact that an influential contact can be identified does not automatically make that person an appropriate target for persuasion.

How It Differs from Related Approaches

Traditional behavioral economics often studies individual decisions in settings such as laboratory experiments, surveys, or consumer choices. Network analysis adds a formal account of relationships and interdependence. It does not replace behavioral experiments: experiments can help test whether a particular message, incentive, or social cue changes decisions, while network models help explain how the effect might travel between connected people.

Graph theory is also not the same as social network analysis. Graph theory provides mathematical concepts and tools; social network analysis applies network ideas to social structures and questions. Nor is every network model behavioral. A model can use graphs while assuming people optimize perfectly. The behavioral element comes from the account of human choice, motivation, belief, or attention—not merely from drawing a network.

Frequently Asked Questions

Is graph theory part of behavioral economics?
Not exactly. Graph theory is a separate mathematical field that behavioral economists can use to represent relationships. It becomes part of a behavioral-economics analysis when the model or research question connects network structure to human decisions and behavior.

What is a simple example of behavioral economics on a network?
Imagine a person deciding whether to try a new service after hearing about it from two friends. A network describes the person’s links to those friends; behavioral economics can examine whether the person is learning from their experience, responding to social approval, or valuing the same features independently.

Does a person with more connections have more influence?
Not necessarily. More connections can mean greater exposure or reach, but influence also depends on trust, the type of relationship, the message, and the setting. A connection count is a structural measure, not proof of persuasion.

Can graph theory prove that friends cause similar behavior?
No. Similar behavior among connected people may result from influence, shared circumstances, or selection into friendships. Establishing causation requires an appropriate research design and evidence that addresses these alternatives.

What is the difference between a network effect and a peer effect?
A network effect usually means that a product or action becomes more valuable as more people use it, as with a communication service. A peer effect means one person’s behavior or outcome is affected by other people’s behavior or characteristics. The ideas can overlap, but they are not synonymous.

Why the Combination Is Useful

Graph theory makes relationships explicit, and behavioral economics gives researchers ways to describe how people interpret and respond to those relationships. Together, the fields can clarify why information spreads unevenly, why group choices reinforce one another, or why an intervention works in one community but not another. Their strongest contribution is not a universal formula for social influence; it is a framework for asking more precise questions about connections, decisions, and evidence.

Related

Try it: the Percentage Formula Generator: percent of, percent change, increase and decrease, with the formula.

More in Charts & pivot tables

Sources: Function behaviour and menu paths: Microsoft Support documentation for Excel. Method & sources

Spotted an error in this guide? Let us know.