5. Stimulus

这也可以？！
Carmen Reinhart and Ken Rogoff presented a research paper called “Growth in a Time of Debt” in 2010, at a time when politicians everywhere were fiercely arguing about the wisdom of getting into further debt in the hopes of kick-starting the economy. From a bunch of statistical correlations between countries’ growth rates and their debt/GDP ratios—which are a simple way to measure how much money a country’s government has borrowed, relative to how big the economy is—Reinhart and Rogoff presented what quickly became a famous result: If a country’s debt/GDP ratio rises about 90 percent, economic growth tends to be substantially slower.

Politicians who favored spending cuts jumped on this result, as you might expect. Paul Ryan, later the vice-presidential running mate of Mitt Romney, mentioned the 90 percent growth collapse while arguing the case for the budget proposals that made his name. Olli Rehn, the European Union’s top man on economics, also mentioned the 90 percent cutoff. Professors Reinhart and Rogoff were invited to address a group of U.S. senators. And their work was much mentioned by journalists. It was seen as relevant, of course, because efforts to stimulate the economy involved cutting taxes, increasing
government spending and borrowing more money in the short term—which for many countries meant approaching or exceeding that dangerous-sounding 90 percent debt/GDP ratio.

Now the story switches to the University of Massachusetts, Amherst, where a graduate student in economics, Thomas Herndon, was set a routine assignment: choose an interesting economics paper, get the data and try to repeat the analysis. This is called a replication exercise, and it’s good practice for young researchers. Herndon picked the Reinhart-Rogoff research, and quickly ran into trouble: he simply could not replicate the results from “Growth in a Time of Debt.” And of course his heart sank
because, well, he was just a student and Reinhart and Rogoff were Harvard professors.

Eventually, Thomas Herndon approached Reinhart and Rogoff directly, and they sent him not only the data that was publicly available from their website, but the actual spreadsheet they had used to crunch the numbers. And he found—after blinking, rubbing his eyes and asking his girlfriend to check —that Carmen Reinhart and Ken Rogoff had made a pretty basic error in Excel: they omitted some of the rows, and thus didn’t include the data for Australia, Austria, Belgium, Canada and Denmark.

Oops. Actually, Herndon raised other questions about the paper that ended up making a much bigger difference. He found that when more recently available data was included, the results changed substantially. He also picked a methodological fight with Reinhart and Rogoff; who wins that one is more a matter of opinion. Of course, pro-stimulus politicians and commentators milked the discovery of errors in the paper just as enthusiastically as pro-austerity politicians had milked the original paper.

This was overblown on both sides. An Excel spreadsheet full of correlations from wildly different countries in wildly different circumstances didn’t prove much in the first place, so discovering errors in that spreadsheet doesn’t disprove much, either. The bottom line is that lots of debt seems to be correlated with lower growth, as you would expect, but that eye-catchingly sharp cutoff at 90 percent is imaginary. And finding a correlation is no proof that debt causes slow growth: the idea that slow growth causes debt is at least as plausible.