ausblenden:
Schlagwörter:
normative claims, frequentist statistics, significance, power, structural equation model, finite mixture, Bayesian statistics, prediction, machine learning
JEL:
A12 - Relation of Economics to Other Disciplines
JEL:
C01 - Econometrics
JEL:
C11 - Bayesian Analysis: General
JEL:
C12 - Hypothesis Testing: General
JEL:
C18 - Methodological Issues: General
JEL:
C81 - Methodology for Collecting, Estimating, and Organizing Microeconomic Data; Data Access
JEL:
H41 - Public Goods
JEL:
K00 - General
JEL:
K41 - Litigation Process
Zusammenfassung:
To their credit, empirical legal scholars try to live up to the highest methodological standards from the social sciences. But these standards do not always match the legal research question. This paper focuses on normative legal argument based on empirical evidence. Whether there is a normative problem, and whether legal intervention promises to mitigate the problem, requires a decision. If uncertainty cannot be completely removed, the legal decision-maker must weigh the risk of false positives against the risk of false negatives. This may call for an adjustment of the significance level. The fact that all legal choice is historically contingent, that legal problems tend to be ill-defined, and that strategic actors have an incentive to bias the generation of evidence defy frequentist statistics. Yet the law can capitalize on the adversarial principle. Competition among interested parties helps contain the strategic element and spurs the creative search for better evidence. This leads to suggestive, but institutionally contained empirical evidence.
