A Fascinating Mistake.

I'm going to enjoy my quantitative economics course (EC475). Our first handout described the odd results of Henry J Moore in 1914 when he tried to measure the slope of the demand curve for pig iron. When he regressed the price of iron on the quantity in the market, he discovered that it was in fact positive -- that the more iron there was, the higher the price was. This led him to claim that he had "discovered a new type of demand curve with a positive slope".

His problem was one we studied extensively in our theory classes, but never discussed in the context of a concrete example, called the endogeneity of the regressors. Forget the fancy name -- the idea is simple. Think of the price as the intersection of a downward-sloping demand curve and an upward-sloping supply curve. If the system has no noise, you will only ever observe a single quantity and price -- the equilibrium. However, we can imagine that we observe variation because shocks move the demand or supply curve around. Now, suppose the demand stays fixed, and all the shocks move the supply curve. Then you will observe a series of points that describe the demand curve, and a regression will measure its (negative) slope. However, if the shocks only move the demand curve, you observe the supply curve's (positive) slope. In real life, you need to argue which of these cases -- or something in between -- your data describes if you want to do a simple regression of price on quantity.

(This is called endogeneity of the regressors because the error terms might affect the regressors -- in this case, your unobserved shocks may shift the quantity in a systematic way as well as the price by shifting the demand as well as the supply curve.)

One interesting thing is how this error can now be neatly and impeccably described in a page and a half of very clear statistical theory, but at the time it took thirteen years (until Working's 1927 paper) to discover the mistake. At the time, the idea of using statistics to measure and test economic theory was relatively young, but now our understanding has been refined to the point that an error made by one of the great founders of the field can be expressed in terms that an undergraduate could immediately grasp. It kind of makes you wonder what sort of boneheaded things we're all doing now that will seem trivial ninety years from now.