The Art of Drawing Causal Inferences from Nonexperimental Data

Our next item of business is to try to figure out what causes recessions.

In chap. 12, we discovered that housing often declined before the recessions officially began. In chap. 13, we learned that an “inverted” yield curve and a drop in weekly hours in manufacturing often preceded recessions. Before discussing those important findings we issued the warning not to jump to causal interpretations merely from those temporal orderings. Movements in these early warning signs can predict the recessions, but that is not the same as saying they “cause” the recessions. We know that weather forecasts do not cause the weather, even though the forecasts regularly precede the weather. Nor does the flowering of plants cause Summer to arrive, even though Summer has always been preceded by a period of exceptional plant flowering that poets call the “Spring.”