exogeneity
The condition where a variable's variation is determined outside the system and not driven by the outcome or by confounders, enabling valid causal inference.
As defined in Noah Smith on Whether Economics is a Science
Where it comes up · 1
Most important claims · 5 of 5Search all claims for “exogeneity” →
- There is always a potential omitted confounding variable—the 'dreaded third thing'—in any empirical study, even rolling balls down ramps; the question is not whether one exists but how plausible and worrisome it is, which is why econometrics ultimately comes down to grumpy seminar-room debates over whether the omitted variable concerns are credible.Noah Smith on Whether Economics is a Science
- Judging which empirical study has better 'identification' always requires a guess and a leap of faith that all reasonable confounders were controlled for; this is not unique to economics—every natural-science experiment also requires the assumption that the experimenter achieved good controls, so the difference between economics and physics is one of degree, not kind.Noah Smith on Whether Economics is a Science
- Natural experiments are extremely hard to find in macroeconomics because macro phenomena are global rather than bounded—there are no clean borders where a policy works in some places and not others, and even cross-country comparisons are contaminated by linkages—so macro is almost by definition the domain where data is worst, which is precisely why priors and ideology play a larger role in macro debates.Noah Smith on Whether Economics is a Science
- The $800 billion stimulus was not a natural experiment because it was not randomly decided—Obama enacted it for a reason responding to economic conditions, and the state-to-state variations are also non-random and depend on assumptions about national-level effects—so its inconclusive aftermath cannot prove anything, unlike a randomized lottery study which ranks much higher on the scale of convincing identification.Noah Smith on Whether Economics is a Science
- Roberts cannot think of any economic study so decisively executed that opponents had to throw up their hands and change their views, which suggests economics has serious problems claiming to be a science—and the recent $800 billion stimulus is a prime example, since brilliant credentialed economists with Nobel Prizes still cannot agree whether it succeeded.Noah Smith on Whether Economics is a Science