United States District Judges Edmond E. Chang, Sara L. Ellis, and Virginia M. Kendall comprised the fourth and final panel of the Northwestern University Law Review’s October 20, 2017 symposium, “‘A Fear of Too Much Justice’?: Equal Protection and the Social Sciences 30 Years after McCleskey v. Kemp,“ engaging questions of evidence, epistemology, and expertise on the contemporary bench. Professor Destiny Peery (Northwestern Law) facilitated the panel.

In McCleskey v. Kemp (1987), the Supreme Court was presented with an extensive and rigorous statistical study demonstrating that in Georgia courts, black defendants who had killed white victims were sentenced to death at far higher rates than any other race of defendant who had killed any other race of victim. The Court ultimately held that this evidence was insufficient to support an inference that decisionmakers who had sentenced Warren McCleskey, a black defendant who had killed a white victim, to death had acted with discriminatory purpose. In the years since McCleskey, scholars and courts have grappled with the role of social science in equal protection cases. Advocates seeking to establish equal protection violations in the wake of McCleskey have often been frustrated by the seeming impossibility of bringing any type of social science evidence—by nature aggregate and probabilistic—to bear on specific and particularized fact patterns.

Thirty years after this landmark case, Judges Chang, Ellis, and Kendall expressed a new, if cautious, openness of the bench to social science evidence. All three judges emphasized the importance of applying best practices of fact record development to the use of social science evidence. That is, advocates must show how the evidence is relevant to a particular element or claim, and must introduce it under the appropriate Federal Rule of Evidence. Judge Chang emphasized that lawyers should not cherry-pick quotes from studies that seem to support their argument without having a holistic understanding of the studies and confirming that their methods and conclusions truly support the point they are trying to make. Judge Chang drew a laugh from the audience when he expressed suspicion of briefs that, when using social science evidence, quote only from the first few pages of a study.

The judges also compared social science expertise to other kinds of expert information that are used in litigation. For example, they apply the Daubert standard to social science evidence coming into a case under Federal Rule of Evidence 702, evaluating such evidence with scrutiny comparable to any other expert or technical evidence that parties may seek to introduce. “Hard” social science may be easier for attorneys to introduce than “soft” social science, in part because the Daubert standard itself has been defined with reference to scientific methods more analogous to quantitative than qualitative methodologies. Additionally, judges often have a higher “comfort level,” as Judge Kendall put it, with quantitative methods. Judge Ellis, however, stated that she does not differentiate among social science disciplines in evaluating methodology, and expressed openness to various methodologies so long as they are rigorous and clear enough that she can have confidence in the results.

Evidence of all types is scrutinized more closely when the stakes are higher, Judge Chang noted. The judge hypothesized that this may account for courts’ historical reluctance to engage with social science evidence in, for example, civil cases with high dollar amounts at stake. Judge Kendall pointed out, however, that social science evidence has routinely been used for many years in sentencing hearings, which are among the highest-stakes proceedings in our legal system. In the end, the bench’s willingness to rely on social science evidence is context-dependent. However, advocates can take advantage of the contexts in which it is welcome, and, perhaps create new contexts by meticulously making social science evidence legible to the courts through established practices of developing the fact record.