Modeling the U.S. Economy

Will the U.S. economy grow 4% in 2028? What would happen to interest rates if Americans began to save more? These are the some of the questions economists ponder when they seek to understand the U.S. economy.

Miguel Faria-e-Castro

To answer these kinds of questions, economists build economic models. What exactly are economic models? They are simplified representations of the economy or parts of the economy. These models rely on a series of economic assumptions and use logic or mathematics to analyze the relationship between different variables, such as interest rates and consumption.

“Mathematical models are very useful tools to capture and ‘simulate’ that type of complexity, allowing us to obtain quantitative answers to some of the questions we economists ask,” said Miguel Faria-e-Castro, a senior economic policy advisor and an economist who runs the macroeconomic model at the St. Louis Fed.

Economic Models’ Long History

French physician François Quesnay is credited with creating the first example of an economic model in 1758.

Image via Wikimedia Commons

Economic models certainly aren’t new. Many consider the 1758 “Tableau Économique” the first example of an economic model. François Quesnay, a physician in King Louis XV’s court, created a theoretical model of flows of money and goods in a society, with the intent of fixing France’s ailing mercantilist economy. Quesnay would become one of the intellectual leaders of the Physiocrats, members of a school of economic thought that would influence Adam Smith.

In the centuries that followed, models have become the building block of economic analysis. One example of a simple yet elegant economic model is supply and demand, in which a higher price simultaneously elicits an increased supply of products by producers and a decline in demand for those products by consumers.

Building a Model for the Whole Economy

The problem, however, becomes more complex when you’re looking at the entire economy. It is shaped not only by a multitude of markets, all interacting with each other, but also by the decisions of many different economic agents, such as consumers and business owners.

And the interactions are further complicated by feedback loops, Faria-e-Castro pointed out.

“Aggregate consumption depends on variables like gross domestic product, real interest rates and inflation,” he explained. “But then inflation depends on consumption, GDP, investment and government spending.”

A macroeconomic model—one that simulates the entire economy—requires an economist to identify the key economic variables, use mathematics to define the relationship between these variables, write the computer code and then input the data—a lot of data.

“At the end of the day, the model is just a huge system of equations,” Faria-e-Castro said. “And each equation describes the relationship between different macroeconomic variables, like consumption, and how they interact with each other.”

Testing Economic Models

Building the model is just the first step. Validating the model requires testing to see how well it explains past events.

“If a model’s results don’t match the historical data, then it wouldn’t be considered a good model,” Faria-e-Castro said. “The validity of the model is also tested in real time, since we constantly produce forecasts and validate them against the data that actually comes in.”

Constant testing is important. Models can break down when the economy is hit by powerful economic shocks that disrupt the way markets interact and economic agents behave.

“The 2008 financial crisis was a big challenge for economic models, since nothing similar had happened since the Great Depression,” he said. “Similarly, the COVID-19 pandemic posed a lot of challenges for macroeconomic models because it caused many variables to exhibit unprecedented amounts of volatility.”

For example, stay-at-home orders in response to the pandemic caused a surge in U.S. unemployment in 2020, as can be seen in the graph below from online database FRED. Within a single month, the jobless rate soared to its highest level since World War II.

With the advent of the generative artificial intelligence (AI) revolution, it’s also an exciting time to build economic models.

“AI can potentially help us solve and estimate more complex models, which may be more realistic and be able to capture patterns in the data that current models are not really suited to capturing,” Faria-e-Castro said.

What Is the Value of Economic Models?

So why would an economist use an economic model? Reasons include:

• Understanding the cause and effect between different variables. For example, economists can better understand how inflation affects real economic growth, and vice versa.
• Creating counterfactuals. An economist can create “what if” scenarios, such as what could happen to investment if there was a sudden, sharp rise in inflation.
• Forecasting. By understanding how the pieces of the economy fit together, economists can project the possible paths that the economy might take in the future.

Key Model for Understanding a National Economy

While economists use different types of economic models, the dominant paradigm for macroeconomic policy analysis is the dynamic stochastic general equilibrium (DSGE) model, Faria-e-Castro noted. Here’s how each of those elements contribute to the model’s structure:

• Dynamic: It accounts for people’s behavior today based on their expectation of the future. For example, if people believe the economy will have strong growth in future years, they’ll spend and invest differently than they would if they believed the future economy will grow much slower.
• Stochastic: It allows for random disturbances to hit the economy and cause economic fluctuations.
• General equilibrium: It estimates the equilibrium prices that will clear all the model’s markets simultaneously, so that supply meets demand. For example, a sudden decline in the oil supply would result not only in higher pump prices but also reduced consumer spending on nonpetroleum goods and services.

Of course, these models aren’t built to be a precise replica of the economy.

“Analyzing all the markets in the economy is much more difficult,” Faria-e-Castro said. “DSGE models do this in a very simplified way, considering a few different abstract markets at a time, such as goods and services, labor and financial assets.”

The St. Louis Fed DSGE Model

Seminal works by Lawrence Christiano, Martin Eichenbaum and Charles L. Evans in 2005 and Frank Smets and Rafael Wouters in 2007 laid down the foundations for the standard model used by the Federal Reserve and other central banks, according to Faria-e-Castro.

Even this standard model has variations. The St. Louis Fed DSGE model, which Faria-e-Castro manages, is different from models at other regional Federal Reserve banks, such as Chicago, Cleveland, New York and Philadelphia.

“What sets our model apart from others is that ours allows for a more detailed analysis of fiscal policy,” he said.

Other models use the household as the basic economic agent in the model, but the St. Louis Fed model breaks up households into two types: workers and owners of capital. In this manner, the model can better gauge the impact of changing tax policies and transfer payments, which will affect these two groups differently, Faria-e-Castro said.

“This is valuable if you want to study the possible effect of stimulus checks,” he said, adding that it also allows a deeper analysis of fiscal policy counterfactuals (e.g., what could happen if income taxes were to increase by a certain amount) and monetary-fiscal interactions (e.g., whether monetary policy might need to change if an increase in transfer payments were to cause higher inflation).

A Conceptual Description of the St. Louis Fed DSGE Macroeconomic Model

 

The dark blue boxes represent the main agents in the model: workers, capital owners, goods producers, government and the central bank. The flows show the movement of goods and money between these agents. Even in a simplified form, this illustration shows the complexity of the agents’ relationship in the economy; each agent’s decision will affect the others. Transfers include direct government payments like Social Security.

Using the St. Louis DSGE Model for Research

The results from the St. Louis Fed model give economists insights into what might be around the corner for the U.S. economy. For economists looking at narrower aspects of the U.S. economy, such as housing, they can input the model’s macro variables, like GDP growth, into their own models.

Recently, Faria-e-Castro and St. Louis Fed Senior Economic Policy Advisor Serdar Ozkan used the St. Louis Fed model to study how expectations of future productivity gains from generative AI might affect output, investment and inflation, and how these effects could require different policy decisions by the Federal Reserve.

Again, Faria-e-Castro and Ozkan’s work looked at what could happen, not what will happen. Macroeconomic models are not crystal balls. They are simplified representations of the economy, stripping away extraneous elements and providing a logical and disciplined way to study the world. Because of their simplicity, however, models can be a powerful tool for testing economic hypotheses and theories.