Agent (economics)

• 1. Introduction

  ›

  Agent (economics) is one of the most quietly consequential concepts in all of modern economic thought. Every time an economist builds a model of a market, a pension system, or a national economy, they must first ask a deceptively simple question: who is making decisions in this world?

  The answer is always some version of an agent. An agent, in economic modeling, is a decision maker. That is the whole definition, and yet from that spare starting point, economists have built some of the most complex and ambitious frameworks in social science. The agent might stand in for every consumer in a country, or it might represent a single household with a particular age, wealth, and set of choices to make. It might not even be a person at all.

  What does it mean to reduce a human life to an optimization problem? How do the choices economists make about agents shape the conclusions their models reach? And what happens when the agents in a model start interacting with each other, not as abstract types, but as computational objects following rules across space and time?

• 2. The Decision-Making Core

  ›

  Buyers and sellers are the classic example. In a partial equilibrium model of a single market, consumers on one side and producers on the other represent the two foundational agent types. Each is assumed to be solving some kind of optimization or choice problem, whether that problem is well-defined or ill-defined.

  The idea of optimization is central here. An agent is not just any participant; it is specifically a participant who is deciding something. That framing imposes a particular logic on how economists think about behavior. When you model someone as an agent solving a problem, you are committed to the view that their choices follow some internal structure, even if that structure is murky or hard to specify.

  Macroeconomic models take this further, often grouping agents into three broad categories: households, firms, and governments or central banks. Dynamic stochastic general equilibrium models, which are built explicitly on these microfoundations, treat each category as a distinct actor with its own objectives. Some models go even further, distinguishing workers from shoppers, or separating out commercial banks as a type of agent in their own right.

• 3. One Agent or Many

  ›

  A representative agent model assumes that all agents of a given type are exactly identical. Every consumer behaves the same way; every firm follows the same logic. This simplification lets economists describe the economy in the most stripped-down terms possible, which is often exactly what they want when the question they are asking does not hinge on differences between people.

  Heterogeneous agent models take the opposite approach, recognizing that agents within the same category are not identical. The choice between these two modeling strategies is not arbitrary. It is driven by what the research question demands. A model studying the economic effects of pensions almost certainly needs to treat agents differently by age, because age is directly relevant to how pensions work. A model studying precautionary saving or redistributive taxation needs to account for differences in wealth across households.

  The decision to use one type of model over the other is itself a theoretical commitment. It shapes which questions a model can even ask, and which conclusions it can reach. Economists may be obliged, as the source puts it, to use heterogeneous agent models when the differences among agents are directly relevant for the question at hand.

• 4. Principals, Agents, and Delegation

  ›

  Principal-agent models give the term a different and more specific meaning. Here, an agent is not just any decision maker in an economy. It is someone who has been delegated to act on behalf of another party, the principal.

  This framing captures a real and pervasive feature of economic life. Employees act on behalf of employers. Lawyers act on behalf of clients. Fund managers act on behalf of investors. In each case, the person doing the acting is the agent, and the question that motivates the model is whether that agent's interests align with the principal's, and if not, what mechanisms might bring them into alignment.

  The vocabulary of principal and agent has become one of the most widely used analytical tools in economics, finance, and organizational theory, precisely because the delegation problem it describes appears so often in practice.

• 5. Agents Without Bodies

  ›

  Agent-based computational economics pushes the concept of the agent into territory that has no obvious connection to real human decision makers. In these models, agents are described as computational objects modeled as interacting according to rules over space and time.

  These are not real people. They are code. The rules they follow are formulated to model behavior and social interactions based on stipulated incentives and information. The researcher decides what the agents want, what they know, and how they respond to each other, then watches what happens when they interact.

  The concept of an agent in this computational setting may be broadly interpreted to include any persistent individual, social, biological, or physical entity that interacts with other such entities in a dynamic multi-agent economic system. That is a deliberately wide definition. It leaves room for modeling not just consumers and firms but institutions, ecosystems, or any other entity whose behavior can be captured in rules and whose interactions with others produce outcomes worth studying.