At a conference last week, Peter Praet, a member of the executive board of the European Central Bank, said something so breathtakingly obvious that, after a second, you begin to wonder what had been going on that he felt he had to state it: “… individual households are heterogeneous in many respects,” he said. Then, “… it is important to measure and analyse this heterogeneity because it can have important implications for aggregate figures.”
Families differ from each other, he meant, and we need to know how, because it could affect the way we measure the economy. He needed to say it because the idea that different families respond differently to the same event—that households are heterogeneous—is a relatively new way of looking at the economy.
There is a basic tension in economics between observation and assumption. An economist can’t possibly observe the full majesty of the universe, and data would be hard to manage if it were possible. So economists make assumptions about things they cannot observe. Over the 20th century, the profession found its purpose in predicting the future, all the stuff everyone would really, really like to observe, but can’t. To get there, economists bundled a bunch of assumptions together into economic models.
Models indicate what might happen if, say, we lower the tax rate. Or raise government spending. When you hear a politician say, “We can’t raise taxes because …,” he is referring, ultimately, to the conclusions of an economic model. A model relies on thoughts about what happens in reality, but it is necessarily not reality, because it is telling you about a thing that has not happened yet. And to make calculations even possible, most models do a funny thing: They rely on a “representative agent.” They assume that everyone is the same person—one agent in the economy, repeated 300 million times, each repetition with the same motivations and responses.
The models that make this assumption tend to differ, depending on what kind of person they think this representative agent is. A group of models known under the broad term “RBC” assumes a person who plans for the long term. This person knows that government spending will be matched, some day, by government taxing, and saves accordingly to pay for future taxes. IS-LM models (I could tell you what these acronyms stand for, but if you don’t already know, the names will not help) assumes its agent will make spending decisions based on what he has in his pockets right now. If the government spends, he spends.
Correspondingly, when economists study what actually happened in the past, they tend to rely on “aggregate data,” averaged over an entire population. Aggregated data have the advantage of being real. But if the entire U.S., averaged, has a certain response to an economic event, it’s hard to tell who did what and why. America is one nation under God, but not one nation under response to macroeconomic shocks. We are not homogeneous.
In his talk, Praet referred to a paper, Heterogeneous Consumers and Fiscal Policy Shocks, by Emily Anderson of Duke University, Atsushi Inoue of North Carolina State University, and Barbara Rossi of the Barcelona Graduate School of Economics. The paper does not describe a model but looks at actual, measured responses of households to two kinds of “policy shocks”—increases in in government spending and increases in taxation.
The measurements come from the Bureau of Labor and Statistics’ Consumer Expenditure Survey. It doesn’t follow everyone in America, but it does conduct interviews with representative households on spending and income. Anderson and her co-authors took the survey and divided it into groups by age, education, and income. The closer you get to reality, the more bewildering the math becomes. But the paper, at the very least, looks at aggregated groups, rather than aggregating all of America.
The paper found that when the government increases spending unexpectedly, the poorest and oldest quintiles increased spending the most. The richest quintile actually decreased spending. And after an increase in the tax rate, the richest quintile actually spent a little more money, while all other quintiles spent dramatically less. In both cases, the poorest and richest groups behaved dramatically different from the aggregate.
It turns out that one group behaves like the representative agents in the RBC model: rich people. And one group behaves like the representative agents in the IS-LM model: yep, poor people. The authors consider the possibility of access to credit. The richer you are, the more likely you are to have credit, which allows you to plan over time. The poorer you are, the more likely you are to make plans from your cash on hand. Both models—both assumptions—have merit in the real world. It just depends on whose real world you mean.
The rich behave differently from the poor. Households are heterogenous. Most surprising is that investigating this is still a relatively new field.