91³Ô¹ÏÍø

Methodology

The non-representation project team coded data from a total of 2,370 cases across four counties. Data were coded from a variety of primary documents and electronic data including court case files, county attorney files, the files of magistrate judges, and jail databases. Project staff traveled to all four counties in person and obtained copies of original documents and data directly from electronic and paper filing systems in each location. While on site, project staff also met with and interviewed local justice system officials and staff. Documents, field notes, and electronic data were transported back to the Deason Center for coding and analysis by the research team. Project Coordinator Sirivore developed a coding system for cases that could capture each defendant’s legal representation status at three points in their cases: their magistration, their arraignment, and their case disposition. The coding scheme also required coders to review and record details of any applications for counsel the defendant submitted, as well as many other details about the cases including dates of court appearances, the nature of case dispositions and sentences, the defendant’s demographics, their criminal history, and any time spent in jail pretrial. All research assistants underwent several days of training in use of the coding instrument, and their work was subject to continual review for quality during their time on the project. In coordination with county officials, the project team sought to identify samples of approximately 600 closed cases. Because the team intends to perform comparative analyses between cases where defendants were represented by retained, appointed, or no counsel, attempts were made to assure a relatively even balance across these three categories among the cases sampled. Because, of course, these three groups do not in fact appear evenly in the data, over-sampling and re-weighting were used to compensate. Magistration records in Espejo were unavailable for cases magistrated before 2019. In Cactus, a subset of cases with magistration dates in the early 2000s were administratively closed during our data collection period, and these also lacked any surviving magistration records. These cases were dropped from all analyses on this page because the lack of records made it impossible to chart defendant counsel status at these early stages of their cases. In Hinkley County, unclear data on counsel type sometimes prevented the project team knowing whether counsel was acting under court-appointed or privately-retained auspices. Those cases were retained in the data, but limit their usefulness, as reflected in the caveats included with the analyses above. The disparity findings reported are based on logistic regression models run in each county. The analyses examined differences in rates of three types of representation – appointed, retained, and non-representation – at each of three stages of defendants’ cases – magistration, arraignment, and disposition. Reported findings are accordingly based on results of nine models run for each county, except Hinkley where data were insufficient to compare retained and appointed counsel rates. The regression analyses held constant the original charge against the defendant (11 categories: assault (family violence), assault (other), marijuana, drugs (other), driving while intoxicated (first degree), driving while intoxicated (second degree), driving with license suspended, theft, theft by check, traffic offenses, and ‘other’); whether the defendant had a co-occurring case; and the defendant’s race and Hispanic ethnicity. Co-occurring cases were recorded when a defendant was simultaneously accused in a separate case in the county. In two counties – Bluebonnet and Cactus – this was inferred from a search of the county’s court case database. In Espejo and Hinkley, it was inferred from the timelines of all cases in which the defendant appeared in the project dataset. Where a defendant was either accused in two cases arising from the same incident, or already had a prior unresolved case at the time they were arrested, they were considered to have a ‘co-occurring case’. Regression results verified that co-occurring cases were significantly (and positively) related to representation by counsel for defendants in several counties. Differences for Hispanic defendants in rates of representation were regarded as significant at the p<0.05 level. Regression models were run with several specifications, including collapsing race categories in different ways, none of which altered the substantive conclusions reported on this page. Full results can be obtained from the authors on request.