January 19, 2017
A revised empirical look at outcomes achieved by federal public defenders and court-appointed attorneys
In this post from this past summer I noted an intriguing empirical paper posted on SSRN by Yotam Shem-Tov in which the authored, after taking a deep dive into "data from all multiple defendant cases in federal courts between 2001-2014," reached the finding that "that defendants assigned a public defender in co-defendant cases had slightly worse outcomes." A few federal public defenders let me know that the author was checking his data after receiving feedback, and the revised paper now, available here, carries a slightly different title and a significantly different key finding in its abstract:
"An Investigation of Indigent Defense Systems: Public Defenders vs. Court-Appointed Attorneys" by Yotam Shem-Tov
Abstract: To provide essential, constitutionally mandated legal services for defendants without financial means, US courts employ indigent defense systems composed of private court-appointed attorneys and public defenders’ organizations. I investigate the public defender’s causal effect on defendant sentencing outcomes relative to private court-appointed attorneys using a new “twins design” identification strategy. I argue and show empirically that in multiple defendant cases the decision of who is assigned to the public defender organization in jurisdiction X, a large urban locality, can be treated as close to a randomized experiment, which can be utilized to measure the effectiveness of court-appointed private attorneys relative to public defenders. I find that public defenders out-perform court-appointed attorneys in a range of sentencing outcomes. Employing a similar identification strategy in federal courts finds that public defenders perform at least as well if not better then court-appointed attorneys in multiple defendant cases. I provide strong evidence of selection in the assignment of attorney types to defendants in both jurisdiction X and federal courts, which makes a naıve comparison invalid and misleading.
My understanding is that the new empirical analysis now more properly accounts for the fact that public defenders typically will represent the lead defendant (and thus the one usually most culpable defendant) in multiple defendant cases, and thus a proper analysis needs to account for this critical variable.
January 19, 2017 at 10:11 AM | Permalink
A placebo condition (no active treatment) should have been added, in the form of pro se representation. If no significant difference could be felt at the gut level, it would show, neither representation adds any value to the defense. A difference smaller than 30% is difficult to experience at the gut level.
The differences claimed in the abstract are around 5%. Any statistical significance is caused by the gigantic sample sizes. They did not adjust the statistical testing for gigantic sample size, nor for repeated measures.
If one believes that competence comes from doing a task 10,000 hours, the differences should have been greater. Imagine the difference in outcomes in other fields between beginners and seasoned experts.
This study is worrisome. Spend a lot of tax money, get little of value back.
Posted by: David Behar | Jan 21, 2017 12:11:43 AM