Jim Tankersley has been doing the Lord’s work by following up on questionable arguments people have made about our current economic weakness being something other than a demand crisis. First, he asked Alberto Alesina about how all that expansionary austerity is working out from the vantage point of this year. Now he looks at the Economic Policy Uncertainty (EPU) index (Baker, Bloom, Davis) as it stands halfway into 2013.
And it has collapsed. The EPU index has been falling at rapid speeds, hitting 2008 levels. Yet the recovery doesn’t seem to be speeding up at all. Wasn’t that supposed to happen?
I’ve been meaning to revisit this index from when I looked at it last fall, and this is a good time to do so. It’s worth unpacking what actually drove the increase in EPU during the past five years, and understanding why there was little reason to believe it reflected uncertainty causing a weak economy. If anything, the relationship is clearly the other way around.
Let’s make sure we understand the uncertainty argument: the increase in EPU “slowed the recovery from the recession by leading businesses and households to postpone investment, hiring and consumption expenditure.” (To give you a sense, in 2011 the authors argued in editorials that this index showed that the NLRB, Obamacare and "harmful rhetorical attacks on business and millionaires" were the cause of prolongued economic weakness.)
As commenters pointed out, it would be easy to construct an index that gets the causation to be spurious or even go the other way. If weak growth could cause the Economic Policy Uncertainty index to skyrocket, then it’s not clear the narrative holds up as well. “There’s uncertainty over whether or not Congress and the Federal Reserve will aggressively fight the downturn” isn’t what the index is trying to measure, but that’s what it seems to be doing.
Let’s take a look at the graph of EPU. When most people discuss this, they argue that the peaks tell them the index is onto something, as it peaks during periods of major confusion (9/11, Lehman bankruptcy, debt ceiling showdown).
But what is worth noting, and what drives the results in a practical way, is the increase in the level during this time period. And that happens immediately in January 2009:
How does economic policy uncertainty jump the first day in 2009? The index has three parts. The first is a newspaper search of people using the phrase “economic policy uncertainty.” I discussed that last fall, arguing that it was mostly capturing Republican talking points and the discipline of the GOP machine rather than actual analysis.
The second is relevant here, and that’s the number of tax provisions set to expire in the near future. (In the first version of the paper this was total number of tax provisions, while in the current version it’s total dollar amount of those provisions.) It’s heavily discounted, so tax cuts that are expiring in a year or two are weighted at a much higher level than those that are further in the future.
What does this look like over the past few years?
So what happened starting in early 2009? The stimulus, of course. And the stimulus was in large part tax provisions that were set to expire in two years. This mechanically increased economic policy uncertainty, even though it was a policy response designed to boost automatic stabilizers. Also, the Bush tax cuts were approaching their endgame, and the algorithm gave a disproportionate weight to them as they entered their last two years.
Then, in late 2010, the Bush tax cuts and some tax provisions from the stimulus were extended to provide additional stimulus to the economy while it was still weak.
Here’s how the creators of the index describe this move: “Congress often decides whether to extend them at the last minute, undermining stability of and certainty about the future path of the tax code... Similarly, the 2010 Payroll Tax Cut was a large tax decrease initially set to expire in 1 year but was twice extended just weeks before its expiration.”
But this decision was not orthogonal to the state of the economy. A major reason the administration waited and then extended the Bush Tax Cuts and the payroll tax cut was the fact that the economy was still weak, and they wanted to boost demand. The only policy uncertainty here was how aggressive and successful the administration would be in securing additional stimulus, which itself was a function of the weakness of the economy. To retroactively argue that the government’s actions in securing additional demand were creating the crisis they are trying to fight requires an additional level of argument not present.
The third part of their index has the same issue. They draw on a literature (e.g. here) that uses disagreements (dispersion of predictions) among professional forecasters as a proxy for uncertainty -- disagreements about the predicted growth in inflation, and predictions of both state and federal spending, one year in advance.
The problem comes from trying to push their definition of EPU onto these disagreements. Debates over how much the federal government will spend through stimulus, how rough the austerity will be at the state level, or how well Bernanke will be able to hit his inflation target, which drives this index, are really debates about the reaction to the crisis. The dispersion will increase if people can’t figure out how aggressively the state will respond to a major collapse in spending. But this is a function of a collapsing economy and how well the government responds to it, not the other way around.
This is why we should ultimately be careful with studies that take this index and plop it into, say, a Beveridge Curve analysis. As Tankersley notes, the government decided to fight a major downturn with stimulus, and the subsequent move away from stimulus before full employment hasn’t helped the economy. In other breaking news, if you carry an umbrella because it is raining, and then toss the umbrella, it doesn’t make it stop raining.