Regression: Economists' Favourite Toy

I just ran four million (growth) regressions.
Xavier Sala-i-Martin

Correlations are common in economics. Yet, it doesn’t mean that identifying whether the correlation between two or more variables represents a causal relationship is always an easy task. Countries who invest more in infrastructure would also have higher income levels or growth. Does it mean infrastructure raises income levels or is it the case that richer countries can simply invest more? To answer this sort of questions, economists employ regression analysis to quantify the relationship between one variable and the other variables that are thought to explain it. Regression analysis can also examine how close and well determined the relationship is. Nowadays, running thousands of regressions has become commonplace and easy. In fact, any empirical economic study appears to have a regression in it.

Despite its benefit, regression analysis is also subject to drawbacks and abuse. People like to use it, misuse it, or deliberately mislead other people with regression. Exhaustive, if not mortal, difficulties in performing regression analysis include the following: ommitted variables, reverse causality, mismeasurement, and too limited a focus. These difficulties make regression analysis suffer of inherent wounds, namely: First, It’s inductive. You can’t be 100% sure; Second, It suffers of some kinds of black box in explaining the mechanics of the observed relationships between variables.

Regarding growth regressions, Pritchett (2006) identifies other wounds commonly found in growth regressions:
1. Growth regressions never satisfactorily resolved the "symptoms versus syndromes" problem.
2. Growth regressions were widely seen as producing estimates of gains from policy reform that were orders of magnitude larger than the microeconomic estimates of those gains—without any particularly convincing economic explanation.
3. Growth regressions have a tough time dealing with the huge differences in country experiences.
4. Growth regressions cannot predict turning points— either accelerations or decelerations—and we know that developing country growth rates have these turning points.
5. Growth regressions did not help policymakers anticipate either the disappointments or the surprises of the experiences of the 1990s.

Should we abandon regression? It's indeed a tool. Like a pencil among many other tools in a carpenter's tool box. An amateur carpenter, someone who thinks that he can saw straigth without a pencil line, may not need a pencil. A professional carpenter, someone who thinks that he never be that good, will always find a pencil helpful in one way or another. Many things a pencil can't do, but a pencil surely can do something.