Understanding Delegation Through Machine Learning: A Method and Application to the European Union

Delegation of powers represents a grant of authority by politicians to one or more agents whose powers are determined by the conditions in enabling statutes. Extant empirical studies of this problem have relied on labor-intensive content analysis that ultimately restricts our knowledge of how delega...

Full description

Saved in:
Bibliographic Details
Published in:The American political science review 2020-02, Vol.114 (1), p.291-301
Main Authors: ANASTASOPOULOS, L. JASON, BERTELLI, ANTHONY M.
Format: Article
Language:English
Subjects:
Algorithms
Artificial intelligence
Authority
Bias
Coding
Cognitive style
Content analysis
Delegation
Delegation of authority
Estimation
Institutional change
Language Processing
Law
Legislation
Machine learning
Methods
Political science
Politicians
Politics
Ratios
Regulation
Regulation of financial institutions
Selection Criteria
Sentence Structure
Statutes
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Delegation of powers represents a grant of authority by politicians to one or more agents whose powers are determined by the conditions in enabling statutes. Extant empirical studies of this problem have relied on labor-intensive content analysis that ultimately restricts our knowledge of how delegation has responded to politics and institutional change in recent years. We present a machine learning approach to the empirical estimation of authority and constraint in European Union (EU) legislation, and demonstrate its ability to accurately generate the same discretionary measures used in an original study directly using all EU directives and regulations enacted between 1958–2017. We assess validity by training our classifier on a random sample of only 10% of hand-coded provisions and replicating an important substantive finding. While our principal interest lies in delegation, our method is extensible to any context in which human coding has been profitably produced.
ISSN:0003-0554
1537-5943
DOI:10.1017/S0003055419000522