The Promise and Pitfalls of Text-Scaling Techniques for the Analysis of Judicial. Dyevre, A. Technical Report ID 2452863, Social Science Research Network, Rochester, NY, 2014.
The Promise and Pitfalls of Text-Scaling Techniques for the Analysis of Judicial [link]Paper  abstract   bibtex   
Among the most promising techniques that have emerged from the text-as-data movement, methods for estimating latent traits from textual data have found application across a broad range of topics. Yet attempts to apply these computer-based approaches to the analysis of judicial opinions have been both rare and unconvincing. The present paper considers the application of text-scaling methods to the analysis of a well documented body of case-law: the 16 opinions rendered by the German Federal Constitutional Court on European integration. Unsupervised text-scaling (Word⬚sh) appears to generate reliable results, especially for the ⬚rest 14 opinions. Judicial position estimates on what is interpreted as a pro/anti-integration dimension are broadly in line with the doctrinal accounts encountered in EU law textbooks and law review articles. They are also shown to be consistent with estimates from supervised text-scaling (Wordscores). These results demonstrate that the basic assumption underpinning the application of text-scaling techniques to party manifestos and legislative speeches that word choice is driven by rhetorical considerations also holds for judicial discourse.
@techreport{ dyevre_promise_2014,
  address = {Rochester, NY},
  type = {{SSRN} {Scholarly} {Paper}},
  title = {The {Promise} and {Pitfalls} of {Text}-{Scaling} {Techniques} for the {Analysis} of {Judicial}},
  url = {http://papers.ssrn.com/abstract=2452863},
  abstract = {Among the most promising techniques that have emerged from the text-as-data movement, methods for estimating latent traits from textual data have found application across a broad range of topics. Yet attempts to apply these computer-based approaches to the analysis of judicial opinions have been both rare and unconvincing. The present paper considers the application of text-scaling methods to the analysis of a well documented body of case-law: the 16 opinions rendered by the German Federal Constitutional Court on European integration. Unsupervised text-scaling (Word⬚sh) appears to generate reliable results, especially for the ⬚rest 14 opinions. Judicial position estimates on what is interpreted as a pro/anti-integration dimension are broadly in line with the doctrinal accounts encountered in EU law textbooks and law review articles. They are also shown to be consistent with estimates from supervised text-scaling (Wordscores). These results demonstrate that the basic assumption underpinning the application of text-scaling techniques to party manifestos and legislative speeches that word choice is driven by rhetorical considerations also holds for judicial discourse.},
  number = {ID 2452863},
  urldate = {2014-08-12TZ},
  institution = {Social Science Research Network},
  author = {Dyevre, Arthur},
  year = {2014}
}
Downloads: 0