omitted variable bias
Distortion in an estimated effect that arises when an unmeasured variable influences both the predictor and the outcome, here unmeasured parental characteristics correlated with income.
As defined in Susan Mayer on What Money Can't Buy
Where it comes up · 4
- Daniel Carpenter2 claims
- Daron Acemoglu2 claims
- Jim Manzi on Knowledge, Policy, and Uncontrolled2 claims
- Susan Mayer on What Money Can't Buy2 claims
Most important claims · 8 of 8Search all claims for “omitted variable bias” →
- Comparing income measured before versus after an outcome (e.g., a child's reading scores or a teenage pregnancy) isolates bias: only income before an outcome should be able to affect it, yet income measured after the outcome also appeared to have a substantial effect, proving that stable parental characteristics—not the dollars—were driving the apparent income effect.Susan Mayer on What Money Can't Buy
- Patterns where larger, older firms enjoy quicker or more favorable regulatory decisions than smaller, newer firms can arise even when the regulator has zero implicit or explicit preference over firm type—because of uncertainty avoidance rather than capture or even risk avoidance.Daniel Carpenter
- 'Capture' is too seductive a term—thrown around whenever a pundit dislikes a regulatory decision—but it is orthogonal to whether an agency does a good or efficient job, and it is genuinely hard to prove because you need a counterfactual of an uncaptured agency ceteris paribus, while existing empirical studies suffer from omitted-variable bias.Daniel Carpenter
- Places that were more prosperous in 1500 (densely populated, urbanized) became relatively less prosperous today, while empty places without urbanization (US, Canada, Chile, Argentina) became richer—a reversal of fortune that is prima facie evidence against a purely geographic explanation of development.Daron Acemoglu
- The Mankiw, Romer, and Weil model implies returns to human capital four or five times larger than Mincer regressions show (which estimate roughly 6-10% earnings gain per year of schooling); the model suffers from omitted variable bias, and increasing schooling 3-4 years would only raise GDP per capita 25-30%, far too small to explain observed cross-country gaps.Daron Acemoglu
- Standard OLS regressions of child outcomes on income are inadequate for isolating the effect of income because parental income is correlated with a host of other characteristics that also influence children's outcomes, and it is essentially impossible to control for all of them since we don't even know the full list.Susan Mayer on What Money Can't Buy
- Omitted variable bias is not a minor nitpick but a massive problem in human systems: leaving out any causally important variable destabilizes all parameter estimates including the one of interest, and since data on all potential causes can never be gathered for complex social phenomena, all such regression models suffer significant omitted variable bias and cannot be relied upon to determine reliable, useful, nonobvious effects of interventions.Jim Manzi on Knowledge, Policy, and Uncontrolled
- There may be zero cases where a sophisticated multivariate regression in a high-causal-density situation with important policy stakes has produced consensus or made an opponent concede error; instead opponents always identify omitted variables (correctly) and redo the analysis, so the tools function as window dressing for pre-existing ideological biases rather than as instruments of certainty.Jim Manzi on Knowledge, Policy, and Uncontrolled