Loyal readers will remember this humble article where we introduced Tony Lawson, a big cheese over at Cambridge University. We said he was in the very thick of the mainstream. Turns out they label him as Post Keynesian [explained here], which, if true, means he’s at the extreme left wing of the mainstream, a lover of govt intervention and money printing, because that’s what a Post Keynesian is. And they have written all kinds of papers to refute him, all of which miss the point.
My good buddy Lio linked, in the comments to the previous article, to three papers that defend econometrics. The first is this, by some Ida Wolden Bache, who defends econometrics by writing [and all her words are in italics, with my snarky comments in normal font]:
The discovery of sustainable empirical relationships
[=meaning equations that are more than mere fairy tales, but actually represent reality]
is related to the discovery of structure.
Recognizing that structure does not have a unique meaning in econometrics, Hendry (1995a,
p. 33) defines it as “the set of basic, permanent features of the economic mechanism”. Necessary,
but not sufficient, conditions for a set of parameters to define a structure are that they are
invariant with respect to extensions of (a) the sample period, (b) the information set and (c)
[In other words, your brilliant equations have to be true tomorrow, too, not just yesterday.
And they have to be true not just for the people you surveyed, but for everyone else.
And they have to be true whether Obama is our Fearless leader, or some other politician.
So how on earth can you know that?
Do you have a crystal ball to see into the future?
Are you going to survey every last man on Earth?
Do you know for sure people will act the same under a stupid corrupt Fascist president, ignorant of the basics of economics, as under a normal person?
Ida Wolden Bache answers with a resounding yes to all three.].
The necessary conditions for structure are, in a limited sense at least, open to empirical evaluation.
[Well, not that resounding. Actually pretty unresounding.
The real problem is that she doesn’t know how unresounding it really is.
Let’s look at why she thinks there’s a “limited sense” we can test our equations]:
Tests of parameter constancy can be used to assess whether parameters are invariant
to extensions of the sample period. Omitted-variables tests can reveal whether the parameters
are invariant to extensions of the information set. Finally, tests of parameter invariance help
establish whether a parameter vector is invariant to regime shifts.
[Oh, goody. You aren’t sure if your number crunching has any significance? Crunch a few more.
Ida Wolden Bache: But Dave, we are using new numbers this time, in some of the tests. And even when we use the same old ones, we are crunching them differently.
Smiling Dave: Let me explain in a manner even a feminist will understand. Let’s talk about hemlines.
IWB: Hemlines? As in the length of women’s skirts?
SD: Exactly. For years, hemline lengths crunched perfectly with economic recessions. The longer the skirt, the worse things were, said the equations. You should have seen the formulas. They passed all the tests. It was wonderful.
IWB: And your point is?
SD: It worked fine, until it didn’t. One day the two things just stopped being correlated. But you should have seen the literature defending the thesis while it still worked. Tests of parameter constancy, omitted-variables tests, and tests of parameter invariance were trotted out to prove those hemlines will always and forever correlate to recessions, today and in the future, for all women on earth, no matter who is running the country.
IWB: They should have realized a woman has the right to change her mind.
IWB: So you’re saying it not just women, but everyone, who might change their mind? That all our equations will work until they don’t? That they are all meaningless techno-babble? That Tony Lawson and Austrian Economics is right?
SD: From the mouth of babes.
IWB: That’s sexist.]
Whether invariant features of the economy do in fact exist is an unresolved issue.
[So much for defending econometrics. You just admitted they have no leg to stand on. Unresolved issue is jargon for “no evidence exists whatsoever that there are invariants”.]
To be continued, if energy and inclination permit.