null hypothesis in political economics

Date view Thread view Subject view Author view

From: Dave Long (
Date: Fri Oct 20 2000 - 18:24:22 PDT

> point. As I said, we're all dead Keynesians now. Drucker has pointed
> out that Carter did everything a good believer in government influence
> would do, and none of it worked, or rather it had exactly the wrong
> effect. Reagan did just the opposite, and it did.

So how do we know that a hecatomb of 100 oxen, sacrificed on prime
time TV, wouldn't have had an equal effect (in either direction) on
the economy?

Looking in the Ibbotson SBBI 2000 yearbook, I see that the spread
of forecasted value (according to their models) in large company
stocks went from 82.42 at the 90th percentile to 5.30 at the 10th,
for a unit invested over the period 1976-2000. That's a 15-fold
spread, and given a decent ensemble of countries, all following
the same economic policy, we might expect to find a similar range
of outcomes.

How can we be sure we aren't guilty of data mining when examining
the historical record? Are the observed differences much larger
than that? Even if long term goverment bonds are a better proxy
for economies than large company stocks, there is still a 3-fold
range of outcomes for an ensemble of identical economies.

I'd imagine it would be difficult to assemble the data to do any
significance testing on economic policies, but surely someone's
tried it; anyone have pointers to the papers?


I've mentioned Jacob's book _Systems of Survival_ before,
<>, perhaps
there's a simple rule for making tradeoffs:
 Any economic areas where theory is demonstrated significant
 at a reasonable confidence level should be administered by
 socialist meritocracies (guardian moral syndrome), but all
 remaining areas should be left to markets (commercial moral
 syndrome). (what about the converse?)

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Fri Oct 20 2000 - 18:13:09 PDT