TOWARDS A GENERAL APPROACH FOR COMPUTER-ASSISTED ANALYSIS OF LEGAL DECISIONS: IN DOMAIN NAME DISPUTES, DOES THE PANELIST MATTER?

Abstract

This paper introduces an automated methodology for descriptive legal analysis using data on 27,000 domain name disputes decided under the Uniform Domain Names Dispute Resolution Policy. Using the methodology, I examine whether characteristics of adjudicating panels affect who wins such disputes. Controls for legal factors are synthesised by applying text mining to pre-processed decision texts. L1-penalised logistic regression is then used for estimation. Significance tests are conducted using post-selection inference methods. I find that panellist identity and size does not significantly affect dispute outcomes. Further, the synthesised controls are effective proxies for legal factors, suggesting that the methodology can be applied to other legal fields.