Expert Witness
Explaining to a court what a regression does and does not prove — while the other side's expert, with the same data, testifies to the opposite.
What the job actually is
Courts routinely decide questions that turn on statistical evidence: whether a pay gap reflects discrimination, whether a merger raised prices, whether a drug caused a harm. Somebody has to translate an econometric result into a claim a judge can rely on, and be cross-examined on it.
The technical work is causal inference, and it is mostly about what is missing. A coefficient with three stars beside it is evidence of association in one sample. Whether it is evidence of causation depends on what was left out of the model, whether the sample selected itself, and whether the thing being explained also causes the thing explaining it.
The professional hazard is subtler than lying. Both experts are honest, both ran defensible models, and the specifications differ in ways that produce opposite conclusions. The job is knowing which choices in your own analysis are doing the work — before opposing counsel finds them for you.
The calls that define it
- Whether the evidence supports a causal claim or only an association
- Which omitted variable would overturn the result, and whether it plausibly exists
- How much of the finding survives a different but equally defensible specification
- What can be stated with confidence, and what must be conceded
Instructed by one side but formally owing a duty to the court, which is a genuine tension and the reason the role carries the professional risk it does.
What you'd need to know
The concepts this chair runs on — each one links to a plain-English definition.
Sit in the chair
Reading about a job is the trailer. This simulation makes you do it — free, in the browser, nothing to install and nothing locked.
Where you learn it
The course behind this chair — bite-size levels that teach exactly what the job is tested on.
🎯 Play Regression Range →