Subject
Econometrics
Estimating relationships from data that was observed rather than assigned. A least-squares coefficient reports how the outcome differs across units that differ in the regressor; whether that difference was caused by it depends on what else differs alongside, and the model's own diagnostics cannot see the answer. The subject is about identification: naming what would have to hold for a coefficient to be an effect, quantifying the part of the gap that can be quantified, and judging the arguments offered to close the rest.
Learning paths
Econometrics
What a regression coefficient identifies when the data were observed rather than assigned, and what would have to be true for it to be read as an effect.
What this subject develops
Say what a coefficient identifies before reporting an effect
Diagnose why a least-squares coefficient estimated from observational data may differ from the causal effect, quantify the part of that difference the omitted-variable formula accounts for, and state the conditions an instrument must satisfy.
See detailed outcomes
Say what a coefficient identifies before reporting an effect
- The learner can identify the sources of endogeneity that make a least-squares coefficient differ from the causal effect, compute the direction and size of omitted-variable bias from the bias formula, explain why goodness of fit carries no information about that bias, and state what an instrumental variable would have to satisfy to repair it.