Selective statistical analysis may promote longer federal sentences

In the 1987 case McCleskey v. Kemp, the U.S. Supreme Court was confronted with damning statistics demonstrating clear and unmistakable racial bias in the application of the death penalty in Georgia. But rather than accept the reality of that bias -- and the wide-ranging implications for the administration of justice in America -- a five-justice majority rejected Warren McCleskey's use of statistics to compare similar cases and those show the prevalence of racial bias. In a fiery dissent, Justice William Brennan accused the majority of fearing "too much justice."

Fast forward nearly 40 years, and the criminal justice system continues to grapple with the use of statistics and comparisons between cases in order to achieve standardization in sentencing. To that end, the United States Sentencing Commission recently developed a website known as Judiciary Sentencing Information, or "JSIN," which contains a database of sentences imposed in federal criminal cases. This database, "specifically developed with the needs of judges in mind" (in the words of the Sentencing Commission), reflects a sharp departure from the federal judiciary's historic reluctance to rely upon statistics to determine the appropriateness of sentences imposed upon "similarly-situated defendants" (again, the words used by the Sentencing Commission).

Using selected data points, JSIN allows judges to compare sentences for other individuals defendants sentenced throughout the country. In creating the database, however, JSIN's architects consciously elected to omit relevant, readily available data which excludes lenient sentences and promotes existing disparities. By excluding these data points, JSIN makes it appear that judges hand down harsher sentences than those actually imposed. Although the goal of encouraging consistency -- and reducing disparity -- is a laudable one, JSIN may well achieve the opposite -- while making federal criminal sentences more punitive at the same time.

JSIN (available here) allows judges to input the applicable guideline section, the adjusted offense level, and drug type and quantity, to compare a particular case against other cases with similar data points sentenced in the previous five years. These broader trends include: (1) the number of individuals meeting the same criteria; (2) the average length of imprisonment; (3) the percentage of sentences involving imprisonment; (4) the percentage of sentences receiving a fine or probation only; (5) the percentage of sentences imposed within, above and below the guideline range; and (6) the percentage of sentences involving a reduction for cooperating defendants (known as a "5K departure").

While it may appear that JSIN is designed to provide many data points for comparison, the reality is that the decision to exclude other data points creates misleading -- or even false -- picture of sentencing decisions. This is because JSIN:

• Excludes noncustodial or fine-only sentences from its calculation of average and median length of sentences. Under JSIN, these sentences, which are imposed in a sizable number of cases, are treated in this context as if they do not exist. The number of cases excluded here is hardly insignificant. In FY2020, these sentences were imposed on more than 9% (one in 11) defendants. Excluding these relatively lenient sentences skews the results to make it appear that judges impose more punitive sentences than reality reflects.

• Excludes sentences with 5K departures from various metrics, including sentence type and average- and median-length sentences of imprisonment. In FY2020, 5K departures were awarded to more than 8% of defendants sentenced that year, ranging from 23% of defendants sentenced in the Eastern District of Pennsylvania to 1.2% in the District of New Mexico. Excluding cases with 5K departures again results in an upwards skew.

• Does not provide results by circuit or district; JSIN only provides national results. The Sentencing Commission's decision to refuse to provide numbers more applicable to particular districts (say, the Eastern District of Pennsylvania), provides an incomplete and misleading snapshot of the sentences imposed in a particular district.

• Does not distinguish time-served sentences from other custodial sentences. Time-served sentences, which can range from a number of hours to a number of years, are counted as sentences of imprisonment, even where they are de facto probationary sentences. These sentences, which often have more to do with pretrial detention orders and court's calendars, are given no different weight than non-time served sentences -- additionally skewing the results.

• Fails to account for the impact of mandatory-minimum and career-offender sentences, both of which serve to drive sentences upward irrespective of the guidelines. In addition, both are disproportionately imposed on individuals of color. In FY2020, more than 75% of career-offender sentences and nearly 70% of mandatory-minimum sentences were imposed against African-American and Latino individuals.

• May include unknown data points that are not publicly available. According to the Sentencing Commission's website, the reported sentencing information is "dynamic and will change over time." What this caveat portends for any objective assessment of the data used by the Sentencing Commission is anyone's guess.

• Excludes demographic data such as race, gender, age, and similar characteristics. Although justice should be colorblind, recent experience, such as the distinction between crack and powder cocaine, and the prevalence of orchestrated stash house robbery prosecutions involving people of color, demonstrates that it is not. JSIN ignores these data points, despite the fact that they are present in every presentence report and are used by the Sentencing Commission in other contexts (such as breaking down charging decisions). Their exclusion here -- where they might actually demonstrate the sort of disparities that Warren McCleskey set forth in his attack in the 1980s -- is striking.

The decision to include -- or exclude -- particular data points to compare "similarly-situated defendants" reflects an effort on the part of the Sentencing Commission to distill the complexities of federal sentencing into seemingly easily digestible morsels for comparison purposes. In doing so it has made conscious decisions regarding how to characterize sentences contained in the JSIN database. Those decisions resulted in a resource that mischaracterizes sentences actually imposed, disregards the reasons underlying those sentences, and ignores pertinent demographic information. In its current incarnation, JSIN will exacerbate disparities and result in higher sentences. Instead of ignoring or discounting statistics which demonstrate disparity and inconsistency in sentencing -- as the Supreme Court did in McCleskey -- the Sentencing Commission has instead chosen to bake them into the sentencing calculus. Although recent years have seen many reforms designed to make the criminal justice more just, JSIN represents a step in the opposite direction. 