I’m having a bit of a debate
over at Zubin’s economics blog, and rather than continue it in the comments
there, I thought I might as well hoist it up here, into its own blog entry.
Zubin’s found a chap called Matthew Hanson who claims
that he might have found a "Giffen good" – something which becomes
more popular the more expensive it gets. The title of Hanson’s paper: "Are
Improvised Explosive Devices a Giffen Good?".
Unfortunately, it seems quite clear to me that no, IEDs are not a
Giffen good, and that Hanson uses some rather shoddy logic to come to his conclusion.
There are quite a few places where Hanson’s assumptions can be challenged,
but let’s leave them to one side, and concentrate on the logical leap that Hanson
makes at the beginning of his paper:
We can treat the percentage of IED attacks that are effective as an inversely
correlated proxy for the price of an IED attack, since a decrease in the percentage
of IED attacks that are effective increases the resources necessary to conduct
an effective IED attack.
Let’s say that Iraqi insurgents, in some given time period, make A
attacks, of which E are effective. Then "the percentage of IED
attacks that are effective" is just E/A. Hanson says
that this number is "an inversely correlated proxy for the price of an
IED attack" – in other words, as E/A goes down,
the price of an IED attack – let’s call it a – goes up.
If the cost of an IED attack is a, then le’ts say the cost of an effective
IED attack is e. Since only E in every A attacks
are effective, we can say that e=aA/E.
Now let’s take another look at Hanson’s reasoning: he says that "a decrease
in the percentage of IED attacks that are effective increases the resources
necessary to conduct an effective IED attack". In other words, as E/A
goes down, e goes up.
Well, yes. Hanson’s saying here that E/A is inversely correlated
to aA/E – which is all but tautological. Most simply,
you can just hold a constant, and come to the conclusion that E/A
is, quite literally, the inverse of A/E.
What Hanson has signally failed to show is that there’s any correlation
at all, inverse or otherwise, between E/A and a –
the proposition which he seems to think that he’s demonstrated.
So it seems to me that Hanson hasn’t come close to showing that IEDs, or the
cost of an IED attack, are a Giffen good. As I said in the comments to Zubin’s
original blog entry, a Giffen good isn’t just any old thing which you buy more
of and therefore spend more money on: it’s something which actually becomes
more expensive on a per-unit basis even as it becomes more popular. And for
all Hanson’s lovely charts of attack effectiveness against total IED incidents,
I don’t think he’s found any product which is behaving in a Giffen-like manner.
Update: Hanson responds, in the comments. If I understand
him correctly, he’s now saying not that IEDs are Giffen goods, but rather that
"the resources expended to cause a unit of damage" – something
which fits into "the standard consumer theory interpretation of price"
– are (or might be) Giffen goods. In which case Hanson needs to change
more than his sentence, he needs to change his title. Intuitively speaking,
something as abstract as "the resources expended to cause a unit of damage"
isn’t a good at all – a point that dsquared makes, also in the comments.