The Economics Of Unusual Events Such As Pandemics, Bank Failures And AI

When the Covid-19 hit in 2020, many people said that economists could not possibly predict the economic impacts of lockdowns. It’s safe to say that forecasting would be difficult, but what is a business or household to do? Decisions must be made about spending, saving, and investments, and decisions depend on a vision of the future.

ChatGPT’s widespread use brings another business forecasting challenge: How will artificial intelligence change the decisions we must make about employment, investment, business processes, work, spending, and pretty much everything else?

And recently, the Federal Reserve had to assess the impacts on the economy of possible credit tightening after the Silicon Valley Bank failure, along with the other bank failures. Would bank problems be like another round of Fed tightening, or neglible in magnitude? Although we have seen large bank failures in the past, we have not seen a bank this large fail outside of a recession. The Fed had little historical experience to use in their assessment.

Some techniques can help both specialists and generalists make better decisions in new and unusual circumstances.

Postpone Large Decisions

Postponing large decisions does make sense. That billion dollar project that’s been in the planning phase for years? Can the decision be delayed a month or two? That’s not always possible, but it’s not a bad choice if feasible.

However, the threshold for postponing a major decision should be high. Too often business executives dither, using small issues to delay making decisions. Postponing the start of a new factory because of the Covid lockdowns made sense, but there were plenty of small changes in recent years that would not have justified postponing a major decision. (Incidentally, postponing a new factory’s construction in the early days of Covid made sense but was the wrong decision in hindsight. Wrong, because it turned out we needed more goods made. But sensible because of the great uncertainty the lockdowns caused.)

Create a Base Rate

Before jumping into details, the big picture should be understood. Before delving into the details of the current change, focus on the usual or normal pattern. For the overall economy, think about long-term economic growth rates. Then adjust up or down from there. For those bank failures, I looked at the long-run history of bank credit standards over past business cycles, then looked for deviations from normal.

This practice was emphasized in the excellent book Superforecasting by Philip E. Tetlock and Dan Gardner. It’s an excellent tool for improving predictions, and this tip is especially valuable in unusual circumstances.

Once a base rate is known, then changes in the current environment are used to move the forecast above or below the base rate. For example, the productivity of some workers will be boosted by improvements in artificial intelligence. How fast will the overall economy grow? We should start with a base rate of labor productivity growth, then adjust from there.

Examine Past Examples

A new thing will not be identical to past new things, but that’s a start to better understanding. For Covid-19, the average business leader didn’t know too much about past plagues, except maybe a little about the Spanish Flu. But a couple of minutes of web search or AI queries brings up a list of plagues with links to information about economic impacts.

Right now we economists studying the effect of artificial intelligence on the economy review the industrial revolution, electrification of manufacturing and the computer revolution. AI’s effects will not exactly follow past patterns, but many of us have reached some conclusions based on past technological changes. First, economic impacts lag inventions. Electrification of industry is a great example. The idea was proved two decades before widespread usage of the technology.

Second, the industrial revolution illustrates that new technology often makes consumer goods cheaper, though it also causes some people to lose their jobs.

With these lessons from past changes, we explore similar critical issues related to AI: speed of implementation, and who will gain and who will lose.

An academic researcher can wait for more information before assessing changes in business and the economy, but people running companies, non-profits or government agencies face decisions every day. They don’t have the luxury of simply waiting. The best results will come from postponing very large decisions, determining base rates and considering past examples.