Data on AI's Risk of Bias

Employers' recent turn to AI to screen job applications raises concerns about algorithmic bias. This concern, in turn, has fueled efforts to "blind" algorithms to various applicant characteristics, such as, e.g., gender. Whether efforts to "blind" algorithms actually work endures as an empirical question.

Data on this question are analyzed in a recent paper, Blinding Gender: An Algorithmic Bias Experiment. In it, Charlotte Alexander (Georgia Tech--Business) et al. lever 9,000 recommendation letters submitted by applicants to medical residency programs in the service of testing "blinding" efforts.

The paper's core finding emphasizes that "even after blinding, our best performing model successfully predicts applicant gender 63% of the time, indicating that recommendation letters retain implicit linguistic patterns that continue to reveal identity." The authors conclude that "blinding strategies fail as a mechanical matter, in that they do not actually blind AI models to applicants’ gender" as the recommendation letters retain linguistic patterns that reveal applicants' identities (or here, gender). To be sure, as a naive ex ante prediction baseline should yield approximately 50%, whether the 63% rate found in the paper is "enough" to safely scaffold a conclusion that "blinding" efforts "fail" remains debatable. Regardless, and any quibbles notwithstanding, an excerpted abstract follows.

Tax Rates, Salary Caps, and NHL Team Performance

As I come across comparatively fewer empirical tax papers than I do for other law-related fields, I was delighted to stumble across Tax Differentials and Team Performance Under the NHL Salary Cap. In it, Martin Higgins (Concordia Univ. - Edmonton) and Mackenzie Frost Common (Alberta) exploit the NHL's adoption of a salary cap after the 2004-05 NHL lockout to assess the competitive implications of tax rate variation. This is not an abstract question as prior research on labor markets and institutional rules in professional sports implies that compensation constraints (e.g., salary caps) may influence competitive outcomes.

Graphing Summary Stats in Stata

As Stata graphics commands are filled with many deep and overly-complicated "rabbit holes," it's nice to come across developments that make nice, if small, improvements. A new user-created updated command, statplot, does just that. Specifically, the statplot command presents summary statistics for a variable and allows for labeling categories along the axis rather than in a legend. One example comparing output from the standard graph bar command (A) and what statplot can generate (B) follows. (Additional examples, with coding suggestions, are discussed here.)

Data on Shareholder Oppression

Shareholder oppression laws set out to protect minority shareholders in closely held corporations. As the scope and contours of these protections are notoriously opaque, however, related litigation is typically time-consuming, fact-sensitive, and protracted.

In a recent paper, An Empirical Analysis of Shareholder Oppression Law, the authors, Benjamin Means (S. Carolina) et al., set out to bring data to the shareholder oppression doctrine. To this end, the paper levers data from 850 shareholder oppression opinions decided nationwide from January 2010 through July 2021.

Data on Asset Forfeitures

Government extractions of assets through forfeiture remain an increasingly contested topic. Asset forfeitures are often significant for people entangled with the criminal justice system. Helping complicate motivations for forfeitures include the financial implications for federal, state, and local governments that generate revenue from forfeitures.

While governmental forfeiture continues to receive scholarly attention, much of the attention lacks data. However, a recent paper, Asset Forfeiture and Inequality, by Stephanie Holmes Didwania (Northwestern), explores forfeitures with the benefit of Department of Justice data on roughly 1.2 million federal asset forfeitures between 1998 and 2019 which are matched to county-level population data over that period. The paper's main take-away involves the claim that governments use asset forfeiture in ways that exacerbate inequality. An excerpted abstract follows.

More on the RE v. FE Specification Debate

In the xtivreg world, for example, while Sargan-Hansen testing helps assess instrumental variable efficacy, researchers also typically confront the question about whether a FE or RE specification is appropriate. Such a question was posed recently on the StataList (here) and an initial response echoes prior suggestions that focus on the Mundlak approach. Mundlak advances an alternative approach to the Hausman test insofar as, unlike the latter, the Mundlak approach may be used when the errors are heteroskedastic or have intragroup correlation. (An earlier discussion goes into greater detail.)

Moving SPSS files to Stata

For the underlying datasets themselves, movement across various software platforms are most easily achieved via third-party software (I've used StatTransfer for decades). Migrating command (or setup or do) files across platforms, however, typically takes a bit more work. A recent discussion on the StataList (click here) walks through coding suggestions for a migration from SPSS to Stata.

Data on Malingering in Insanity Cases Over Time

A steady decline in insanity claims by defendants in criminal cases over time juxtaposes with a concurrent increase in claims that defendants are "malingering" or "feigning" insanity or mental illness in the service of an acquittal or a reduced sentence. Neuroscientific evidence, on both sides, features prominently in these trends. And in this setting, conventional wisdom implies that state evidence introduced relating to a claim of defendant malingering may exert out-sized (negative) influence on a defendant's insanity claim.

CELS 2026 -- Submission Deadline Extension

Call for Papers -- Deadline Extension

Conference on Empirical Legal Studies

October 2-3, 2026

 

The 2026 Conference on Empirical Legal Studies (CELS) is now accepting submissions. CELS will consider empirical papers spanning all areas of empirical legal studies. Authors are encouraged to submit works-in-progress; however, submissions should be completed drafts that include principal results. Submitted papers must be unpublished and expected to be unpublished at the time of the conference.

The paper submission deadline has been extended to Monday, June 22, 2026 (11:59 PM Central Daylight Time). The deadline will not be extended further. There is no charge for submissions.

To submit a paper, please click here.

If you have any questions about the submissions process, please contact the conference organizers at: celsnorthwestern2026@gmail.com. 

Data on Judicial Recognition of Multiparent Families

Many topics in family law, for both obvious and perhaps less than obvious reasons, garner a bit more "heat than light." As well, family law topics are comparatively under-explored empirically. A recent paper by Douglas NeJamie (Yale) and Courtney Joslin (UC Davis), The Legal Recognition of Multiparent Families: Findings from an Empirical Study, breaks from tradition as it brings data to questions surrounding the legal recognition of multiparent families.

The paper levers two datasets, including all electronically available judicial decisions (through 2022) engaging with multiparent statutes as well as "judicial decisions applying a functional parent doctrine over four decades." When it comes to the judicial decisions themselves, Figure 6 (below) makes clear that judicial activity, while growing, is an emerging trend.