In the first two lectures, we began with the supply-and-demand diagram.

In Lecture 1, we asked what the diagram shows and what it hides. The diagram shows an equilibrium price and quantity, but it does not show the market process that could produce that outcome.

In Lecture 2, we looked more closely at one hidden assumption behind the diagram: reservation prices. We saw that demand and supply curves require buyers and sellers to have fixed cutoff prices. But real people often adapt to circumstances, alternatives, urgency, expectations, and interaction with others.

In this lecture, we step back and ask a deeper question:

What is a model?

This question matters because this course uses agent-based models, while conventional textbooks use a very different kind of model. Before we compare these two approaches, we must understand what models are for.

Are models mainly tools for prediction?
Are they tools for explanation?
Are they simplified maps of reality?
How do we decide whether one model is better than another?

These questions will help us understand why Agent Based Microeconomics offers a different way of studying economic reality.

1. Models Are Maps

A model is not reality itself. A model is a simplified representation of reality.

In this sense, models are like maps.

A road map is useful for driving. A subway map is useful for commuting. A hiking trail map is useful for walking through the mountains. Each map simplifies the same reality in a different way. Each map leaves out many details. But each map includes the details needed for a particular purpose.

A road map does not show every tree, building, or footpath. A subway map does not show the exact geography of the city. A hiking map may show slopes, trails, and rivers that a road map ignores.

None of these maps is the complete territory. But each one may be useful for its own purpose.

Economic models work the same way.

Every economic model simplifies. The question is not whether a model leaves things out. Every model must leave things out. The real question is:

Does the model include the details that matter for the purpose at hand?

If our purpose is to predict a price, one type of model may be useful. If our purpose is to understand how buyers and sellers interact, another type of model may be needed.

This distinction is central to the course.

2. Different Goals Require Different Models

A model designed for prediction may not be the same as a model designed for understanding.

Suppose a weather forecast tells us that rain is likely tomorrow. That prediction may be useful. But it is different from understanding the atmospheric process that produces rain.

Similarly, an economic model may predict that a price will rise. But that does not necessarily mean it explains how the price rises.

Did buyers bid against each other?
Did sellers raise prices after observing shortages?
Did information spread through the market?
Did contracts change?
Did agents imitate each other?
Did expectations shift?

A model that predicts the outcome may still fail to explain the process.

This is why the distinction between prediction and explanation is so important.

A crystal ball would be valuable if it predicted the future accurately. But it would not give us understanding. It would not tell us why the prediction is true.

In this course, we are not mainly looking for a crystal ball. We are looking for understanding.

We want models that reveal causal mechanisms. We want to see how outcomes arise from the behavior, information, interaction, and rules governing agents.

3. Prediction Versus Explanation

Prediction focuses on observable outcomes.

In the supply-and-demand model, the observable outcomes are market prices, quantities demanded, quantities supplied, and the equilibrium point where demand equals supply.

These are important. But they are not the whole market.

The textbook diagram does not explain how buyers and sellers actually meet. It does not explain what they know. It does not explain who makes the first offer. It does not explain how bargaining takes place. It does not explain whether contracts form before equilibrium. It does not explain whether prices change quickly or slowly.

Those are questions about process.

If our goal is only prediction, we may ignore these details. But if our goal is understanding, we cannot ignore them.

Explanation requires us to understand the process that produces the outcome.

This is where agent-based models become useful. Instead of beginning with the final equilibrium point, we begin with the agents in the market. We ask:

Who are the agents?
What do they know?
How do they meet?
How are offers made?
How are contracts formed?
How do prices change over time?

These are not minor details. If our purpose is understanding, these are exactly the questions we must ask.

A process model explains how outcomes arise. It does not simply say that supply equals demand at equilibrium. It asks how buyers and sellers interact under specific rules, and how their interaction generates the observed result.

The contrast is simple:

Prediction focuses on the outcome.
Explanation focuses on the process that produces the outcome.

4. The “As-If” Defense

Conventional textbook models are often defended in terms of prediction.

Economists have long known that many textbook assumptions do not describe how people actually behave. Firms do not always calculate marginal revenue and marginal cost in the textbook way. Consumers do not actually behave like utility-maximizing machines with complete information and stable preferences.

Milton Friedman’s famous “as-if” defense responded to this problem.

According to this defense, a model does not need realistic assumptions if it gives good predictions. Firms do not need to literally maximize profits. It is enough if they behave as if they maximize profits. Consumers do not need to literally maximize utility. It is enough if they behave as if they maximize utility.

For some purposes, this defense may have value. Unrealistic models can sometimes produce useful predictions. A model may predict an outcome even if it does not describe the real causal process.

But this defense does not help us in this course.

Why not?

Because our goal is not merely to predict prices and quantities. Our goal is to understand market processes.

We want to know how agents actually behave, how they interact, and how market outcomes are produced. The “as-if” defense sets aside exactly the thing we want to study.

If we want explanation, we cannot be satisfied with saying that agents behave “as if” the model were true. We need to model the process itself.

5. The Methodological Divide

This gives us the basic methodological divide between conventional textbook economics and Agent Based Microeconomics.

Conventional textbook methodology often bypasses the actual behavior of firms and consumers. It does not always ask how firms really set prices or how consumers really make choices. Instead, it asks whether the model can generate useful predictions about observable outcomes.

Agent-based modeling takes a different path.

In ABM, we try to build a representation of the market process itself. We explicitly model agents, information, trading rules, search, bargaining, contracts, timing, and institutions.

The difference is not simply that one model is mathematical and the other is not. Nor is it that one model is complicated and the other is simple.

The deeper difference is the goal.

Textbook models often move quickly toward observable outcomes: prices, quantities, and equilibrium.

Agent-based models put the market process at the center.

This is why ABM is especially useful for understanding. It allows us to replace “as-if” models with more accurate representations of buyer and seller behavior.

6. Mistaking the Map for the Territory

There is an important mistake we must avoid.

A model is a map, not the territory. It may be useful for a purpose, but it should not be confused with reality itself.

The “as-if” defense says that unrealistic assumptions may be acceptable if the model predicts well. But textbooks often go further. They present the model as if it describes how markets actually work.

This creates confusion.

If a model is defended only as a prediction tool, then we cannot also claim that its assumptions describe real behavior. We cannot say: “The assumptions do not need to be realistic,” and then teach students as if those assumptions are true descriptions of real people.

For example, if the model assumes that consumers maximize utility, that does not prove that real consumers behave this way. If the model assumes that firms maximize profit, that does not prove that real firms actually make decisions in that way. If the model assumes fixed reservation prices, that does not prove that real buyers and sellers have fixed cutoff prices.

A map may be useful. But we should not mistake the map for the territory.

If we want to understand real markets, we must go back to the agents, the information, the rules, the contracts, and the actual process by which market outcomes arise.

7. Emergence

One of the central ideas in agent-based modeling is emergence.

In many textbook models, we begin with an observed outcome and then build a model that generates that outcome. The model is judged by how well its results match what we observe.

In agent-based modeling, the direction is different.

We begin by specifying the process. We define the agents, their behavior, their information, their rules of interaction, and the structure of the market. Then we let the outcome emerge from those interactions.

This is powerful because we are not forcing the model to generate a particular result from the start.

Instead, we ask:

If agents behave in this way, under these rules, with this information, what happens?

Then we can vary the process.

What happens if information is limited?
What happens if search is costly?
What happens if contracts are final?
What happens if contracts can be renegotiated?
What happens if agents bargain?
What happens if agents learn from experience?
What happens if agents cooperate instead of acting selfishly?

This is what makes ABM powerful for understanding. It shows us how outcomes arise from market processes.

8. What We Have Learned

This lecture gives us the methodological foundation for the rest of the course.

First, models are maps, not mirrors of reality. They simplify the world for a purpose.

Second, different maps are useful for different purposes. In the same way, different economic models are useful for different goals.

Third, prediction and explanation are different goals. A model that predicts an outcome may not explain the process that produces it.

Fourth, conventional textbook models are often defended in terms of prediction. Agent-based models, as we use them in this course, aim at explanation and understanding.

Finally, if we want to understand markets, we must not mistake the supply-and-demand map for the territory. We must look beneath the diagram to the agents, information, rules, contracts, and institutions that generate market outcomes.

Our goal is not merely to know where the curves cross.

Our goal is to understand the market process that could produce such an outcome.

Preparing for Lecture 4

Lecture 4 begins the next unit of the course.

In the first three lectures, we prepared the ground. We asked what the diagram hides, what assumptions are required to draw the curves, and what models are for.

In Lecture 4, we will begin building.

We will construct a simple agent-based model of a rental market. We will start with students, homeowners, rooms, information, offers, and contracts. Then we will ask:

What kind of market process can generate the textbook supply-and-demand outcome?