Can Watson Do Economics?

Congratulations to the silicon-based genius for beating Jeopardy's two best carbon-based geniuses. But can Watson do economics?

Watson could certainly win on plenty of economics challenges:

Answer: This Scot wrote that people seeking their own self-interest would help others.

Question: Who is Adam Smith


Answer: The U.S. GDP in 1987.

Question: What is $4.7364 trillion?


Answer: Technology triggered this sector's reduced volatility, contributing to the Great Moderation.

Question: What are inventories?

Well, I'm not certain Watson would have gotten that last question, or perhaps he wouldn't have hit the buzzer as fast as I would have.

What about forecasting? There are some good reasons for doubting Watson's ability there. Begin with the matter of foreign exchange rates.

Answer: the direction of the dollar due to Middle East tensions.

Question: what is . . . up?????   Or what is down??????

Why is this issue so difficult? Let's begin with multiple causation. Exchange rates, like most everything else, are affected by everything else. So today (February 16, 2011) there was news of Middle East tension, but also further strength in the U.S. economy. Further, the dollar has strenthened the past two weaks while the Middle East situation deteriorated, so maybe the tensions were already built into the price.


That's a key point about forecasting the results of auction markets, such as currency exchange rates. When people see news that should affect the exchange rate in the coming year, they react immediately. Meaning that if you measure the change in the dollar from an instant after the news hit until some time in the future, you may get odd results–the change had already happened.

Can Bill Conerly do better than Watson? I've tracked my foreign exchange rate forecasts, back when I was doing them regulary. This is the worst forecast category for me, as well as for all other professional forecasters.

What about the stock market?

Answer: The strategy that would have made money during the 2010 "flash crash."

Question: What is "Trust the computer."

As I wrote last May, at least some computers doing high-frequency trading were telling the humans to buy when the market was abnormally low, a strategy that could have earned millions or billions of dollars. So the computers can do a good job.

Stock market and interest rates are both hard to forecast because whatever change is dictated by the news may already have happened by the time I read the morning's Wall Street Journal. That holds true for Watson as well.

There are plenty of interesting things to forecast that are not auction-market results, such as GDP or the CPI. My fellow economists and I do a much better job forecasting these variables than exchange rates. Our challenge is that they are based on an unfathomable factor: human behavior. That's not to say that we don't have any idea what people will do, just that they surprise us a lot.

Forecasters affect these outcomes, just as they affect current exchange rates and interest rates. When the outlook is rosy, according to us forecasters, then consumers and businesses spend money, accentuating the rosy forecast. Then others who had initially been skeptical are won over to the rosy scenario. The boom may become unsustainable. Does any of this sound familiar?

Going forward, there is a lot of room for improved understanding of the economy, and the massive data exploitation that Watson handles points in an interesting direction. At times I've used very large econometric models of the economy. That experience contributed to my skepticism about the models. But more data is usually preferable to less data, so I watch all the numbers as they float across the news.

Answer: Yes

Question: Can supercomputers help economic understanding.

One element that Watson may be able to help with is relative magnitudes. We humans tend to give more weight to the headlines, and to the things we understand. Sometimes small sectors take on disproportionate significance simply because we hear about them and we understand them. Other sectors are not so well understood, so we give them little weight. Watson could certainly remember the relative size of the health care sector and the manufacturing sector.


Answer: The distant future

Question: When will Watson be able to accurately forecast the economy?