Tuesday, October 23, 2012

Competing risks

In this post, Mark Thoma casually points us to the recent Phillips Curve as it has recently revealed itself to us.

I may be going out on a limb here, but I think he is subtly suggesting we exploit that relationship to do some social good.

Certainly I would go out on that limb myself.

The standard answer is the expectation augmentation answer.

And that's fine, as far as it goes. But then that's an empirical question. How well anchored are inflation expectations now? Are they well anchored or poorly anchored? I would argue that they are fairly well anchored.

In my mind, the chance that I am wrong about that seems a lot lower than the chance that I am right. And if I am wrong I think we know what to do about high inflation. And also if I am wrong we should get a sense of it before its too late because we should be able to see whether inflation is accelerating or not. If there is really a terrifying new NAIRU it's going to show itself.

When I think about the competing risks here, I feel more and more like an old Keynesian on the Phillips Curve. Hell, it took those scoundrels almost two decades to cause any problems back in the third quarter of the twentieth century, and even then it was really exogenous oil shocks that did the heavy lifting, not the old Keynesians themselves.

But this strikes people as imprudent, and I don't understand why.

Did expectations augmentation become a greater risk after we understood it than it was before we understood it? That doesn't seem to make sense. The fact that old Keynesians who did not think in terms of expectations augmentation and managed a strong economy for two decades before any trouble happened suggests to me that we modern Keynesians, who do think in terms of expectation augmentation (and therefore know the warning signs) don't pose all that much of a stagflationary risk in the current depressionary environment.

The assessment of these risks seems so clear to me that it almost sounds silly to title this post "comepting risks".

7 comments:

  1. Out of curiosity, isn't the Phillips curve also based on the standard normal distribution? That's at least according to Dr. Michael Emmett Brady, but I think he could be wrong on this one. The Phillips curve's shifts in expectations doesn't necessarily have to be based on the standard normal distribution...then again, Dr. Michael Emmett Brady might be right on this one and have a point.

    ReplyDelete
    Replies
    1. I'm not sure I understand - could you explain?

      If you're actually estimating a Phillips Curve I imagine normal distributions are used. But I can't imagine how that would change things.

      You bring up this point a lot, but I feel like I don't always grasp exactly what you're meaning by it.

      Delete
    2. Let me put it this way: surely the tendency of the Phillips Curve not to stay put over the last several decades is not a result of a poorly specified error structure, right?

      I mean - we generally think that's an actual parameter shift or shift in actual expectations. The normal distributions seems like a triviality in the face of that sort of stuff.

      Delete
    3. Blue Aurora can correct me if I am wrong, but I think that if the Phillips Curve is a best fit using Gaussian error, then a normal distribution of error is implied.

      One problem with that which has gained some currency is the idea of fat tails.

      In general, I do not believe in Gaussian errors for economic data, so I have been using Chebyshev's method of curve fitting, which is based upon minimax.

      Delete
    4. Daniel Kuehn: "I mean - we generally think that's an actual parameter shift or shift in actual expectations."

      I took a quick look at the graphs Mark Thoma posted. The difference since 2008 looks like a statistical artifact, caused by truncation. In the current regime, unemployment is not a good predictor of inflation. That statistical fact causes the slope of the curve to flatten.

      I agree that inflationary expectations have changed, but not because of statistics. :)

      Delete
    5. Sorry, that should be since 2009, not since 2008.

      Delete
  2. "If there is really a terrifying new NAIRU it's going to show itself."

    How would it do that? Seriously, I am interested in the empirical underpinnings of NAIRU. When and where has it shown itself in the past? Thanks. :)

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.