Reader's Guide to "Econometrics as Observation"

1.  "The Lucas Critique":  In the 1950s it was noticed that there was an inverse relationship (called "The Phillips' Curve) between the rate of inflation and the level of unemployment.  With higher inflation, unemployment was lower; with lower inflation, unemployment was higher.  Since increasing the money supply arguably leads to more inflation and decreasing the money supply lowers inflation, it seemed to many economists that the Phillips Curve provided a recipe for controlling unemployment: to lower it, print more money. What Lucas argued is that the relationship described in the Phillips curve breaks down if policy-makers attempt to use it to control unemployment.  Suppose that the way the relationship works is that when prices increase, firms observe that the price for their products have increased and interpret this as an increase in demand.  They then employ more workers to meet this demand and unemployment drops.  If, however, firms know that the price increase is due to attempts by the government to increase employment, then they will not interpret the price increase as an increase in demand, and they won't employ additional workers and the relationship will break down.  The relationship is not "invariant to intervention."  This leads to the question: how can economists find relationships that are invariant to intervention?

2.  Causation: It is arguable that the distinction between X causing Y and a mere correlation between X and Y is whether the relationship is invariant to interventions that change the value of X.  Consider the relationship between barometer readings and the onset of storms. These are correlated, because they are effects of a common cause.  If I try to prevent a storm by moving the pointer on my barometer, I will not succeed: the correlation will break down.  On the other hand, if I had the power to change atmospheric pressure (the true cause of storms), then I could bring on or prevent storms.

3.  Identification:  Suppose that the quantities individuals demand of some commodity depend on its price and the quantities individual firms supply depends on the price as well.  Although individuals and firms take prices as given, prices then result from the bidding down and up of those demanding and supplying the commodity.  The underlying relations that determine prices are thus  D = f(p) and S = g(p).  The economist comes along and observes the price.  He or she wants to know the supply and demand functions -- f(p) and g(p).  But from just observing prices, there is no way to "identify" f(p) or g(p).  That's the identification problem.

Hoover is concerned with these problems and how to solve them.

A few specific questions to answer while reading Hoover:

1.  What's wrong with saying "A diamond is hard enough to scratch glass except when one attempts to use it to scratch glass"?

2.  What are the "autonomous" relations discussed by Haavelmo? Is the relationship between the depression of a car's gas pedal and the car's speed autonomous?

3.  What is "strong apriorism" and how does it respond to the Lucas Critique?

4.  What are "confluent relations" (p. 306) and how do they differ from autonomous relations?

5.  Hoover argues that econometrics is best thought of as a way of observing confluent relations.  What does he mean? How is this a response to the Lucas Critique? What do you make of his analogy between econometrics and observations using a telescope?