We’ve come a long way. Not long ago, an economist entering into a consulting contract with the federal government might have to explain to the human resource department that economic “model building” didn’t involve balsa wood airplanes or toy boats.
Today, as a recent newspaper commentary headline pointed out, “Models Will Run the World.” It is a recognition that both industrial producers and marketing firms like Amazon owe their success to mathematical modeling of their operations. And the growth of models is just getting started.
Many of these models are quite complex, reflecting the world we live in. The economic model developed and used by the New York Federal Reserve Bank to support monetary policy decisions, for example, combines so many variables that wrangling them into a coherent picture would be unthinkable without today’s powerful computer assistance.
The Federal Reserve’s model is called “DSGE” — for Dynamic Stochastic General Equilibrium — and each word represents a key characteristic of how it reflects the movements of the U.S. economy.
The model works very well, but sometimes it misses changes in the economy that gum up the forecasts. Researchers are hard at work to diagnose and fix the cause of the problem, but a good guess is that it might involve a mix of expectations and anecdotal evidence.
The use of anecdotal evidence is of particular significance in monetary policy decision-making. Alan Greenspan served as chairman of the Federal Reserve Board from 1987 to 2006. As his term progressed he because increasingly concerned about the data he relied on to asses the economy. It took so long to develop that it often arrived too late to take action in the financial markets.
Eventually, he admitted that because of its timing he depended more on anecdotal information he gathered himself than on the official data or the economic models they fed. He said that he spoke by telephone with a substantial number of key people in various sectors of the economy; people who were not only decision-makers themselves but were observant of market shifts. From these, Greenspan formed his picture of the economy’s speed and direction and from these formed the basis for his monetary policy views.
Economic models, in fact mathematical models of any sort, are not very good at integrating anecdotal evidence into their process. The problem is that the math deals with orderly, numerical quantities that can be assembled into a logical pattern. Anecdotal evidence is anything but orderly, and instead typically is a jumble of observations, impressions, opinions, and other dissociated elements.
A recent example of anecdotal evidence that is useful in understanding the economy — but extremely difficult to integrate into a mathematical model — is the observation of more un-mowed lawns.
With the usual logic of a mathematical model, we might think that in a robust economy just the opposite would happen; households with more money to spend would hire a landscaping company to keep their lawns tidy. It turns out, though, that because of the economy’s job growth, and other structural obstacles, landscape companies are having trouble finding enough workers to do the jobs they have, let alone new ones.
Integrating anecdotal evidence into mathematical models isn’t easy, but in some cases, something called “expert systems” can be used to impose an artificial order on anecdotal evidence. The largest-scale effort in this area was something called the “Hamlet Evaluation System” developed to assess the security situation in rural areas during the Vietnam War. It was essentially a matrix of the anecdotal observations and descriptions provided by experienced officers (experts) of hamlets from the peaceful to the ones virtually under the enemy’s control.
The HES system acquired an underserved reputation for useless paperwork but despite its flaws it was, especially by Vietnam War standards, remarkably accurate and useful.
Despite its usefulness HES demonstrated some of the limitations of expert systems. One of those restricts the type of problem that an expert system can solve: the selection of experts. In our diverse economy, for example, deciding who the experts are would be a major problem.
The limitations of expert system have led many systems developers to look instead to artificial intelligence to integrate diverse information into decision-making support logic. Amazon, for example, is reportedly using AI in its warehouse operations model to position its robots.
In economics, AI might allow a systematic way to add the kind of inherently disorderly evidence of the type that Alan Greenspan used to sort out and weigh in his head.
That would be a genuine achievement by itself. But while an AI-bolstered economic model would boost our capability, forecasting the American economy is often a humbling experience. Even Alan Greenspan admitted that he missed the clues leading up to the 2007 crash.
James McCusker is a Bothell economist, educator and consultant.
Talk to us
> Give us your news tips.
> Send us a letter to the editor.
> More Herald contact information.
