DA is cherry picking science

"Follow the science" has become almost a ritual chant in public policy debates in recent years. But the strength of the evidence behind these claims varies widely. Some claims are indeed based in solidly established science. Some are supported by one side of a genuine scientific debate. Some have hardly any basis at all, with the bulk of the research pointing the other direction.

Immediately upon taking office last December, Los Angeles District Attorney George Gascón announced several abrupt and controversial changes in policy. Among these was a directive to all deputies in the office to stop seeking sentence enhancements in all cases. The shift was claimed to be based on research.

Sentence enhancements have been an essential part of California sentencing since the enactment of the Determinate Sentencing Law in 1976. They are part of the system for tailoring the sentence to the specifics of the crime and the defendant's criminal history, rather than lumping everyone who violates a particular statute into a single category.

Everyone who steals property by force is guilty of robbery, but people who rob with a gun create a greater danger than those who merely wrestle property away, and they deserve more severe punishment. Rape is a heinous crime under any circumstances, but a rapist who additionally inflicts great bodily injury on the victim deserves more severe punishment. Those who commit crimes repeatedly need to be dealt with more severely than first offenders, nearly everyone agrees. For these reasons, California law has sentence enhancements for gun use, infliction of great bodily injury, and prior conviction of serious felonies.

For a district attorney to effectively nullify these long-standing rules of law by announcing in advance that he will never seek enhancements is remarkable, to put it mildly. A very strong justification would be needed for such a step.

District Attorney Gascón claimed that research supports his policy. His special directive announcing it said, "While initial incarceration prevents crime through incapacitation, studies show that each additional sentence year causes a 4 to 7 percent increase in recidivism that eventually outweighs the incapacitation benefit." Despite the plural "studies," only one unpublished paper was cited to support this assertion.

Punishment may affect crime rates in three primary ways. The fact that crimes are punished may deter some would-be criminals from committing those crimes. This is general deterrence. Locking up a criminal in jail or prison will, in most cases, prevent him from committing any crimes on people outside the prison for the duration of his confinement. This is incapacitation. Finally, the experience in prison may increase or decrease the likelihood that a person commits further crimes after release through rehabilitation, specific deterrence, or a criminogenic effect.

Gascón's claim is that extending a prison sentence not only has a net criminogenic effect but that the effect is so strong that it outweighs the incapacitative effect. In a subsequent press release on his policies, he said, "We are doing all of this because the science and data tell us so." Yet the single, unpublished 2015 study he cites concedes that its findings are "in sharp contrast with prior research." To get a clearer picture of what the science and the data actually tell us, the Criminal Justice Legal Foundation undertook a review of the research literature.

For research through 2009, there already was a thorough review of the literature. The lead author is criminologist Daniel Nagin of Carnegie Mellon University, a well-known researcher in the field. This review notes that it is important to separate two different aspects of sentencing. For offenders at the lower end of the criminal scale, the choice is between imprisonment and some alternative such as probation. For those guilty of more serious crimes and those with more extensive criminal records, the question is not between prison and something else but rather the length of the prison sentence. The research results are different for these two groups and their different sentencing choices. Only the latter is relevant to the enhancement question. Enhancements, by definition, are only for those defendants who are going to prison for a base term.

The Nagin review refers to the "length" choice and its effect on recidivism rates as a "dose-response relationship." That review found "little convincing evidence on the dose-response relationship between time spent in confinement and reoffending rate." The results of the studies to that time were "quite varied." Some showed a decrease in recidivism rate with length of sentence, and some showed an increase, but none of the observed effects were strong. To the extent that the research at that time showed anything, it showed only that there was no effect strong enough to produce consistent study results.

CJLF's review of published, peer-reviewed literature from 2009 to the present found that the overall research picture has not changed much, and what change there is points in the opposite direction from Gascón's claim. There are several more studies, but they have variable results and most do not find strong effects. Even the same sets of data have yielded different results when different researchers apply different methods to them. The two largest studies on long sentences, the most relevant to the present controversy, were those by Benjamin Meade and others in 2012 and by the United States Sentencing Commission last year. Both found that for sentences greater than five years, longer sentences were associated with decreased recidivism rates after release.

Given the state of the published research, could anyone with even a minimal understanding of science honestly say that he is following "the science and the data" based on a single, unpublished, outlier study that is so sharply different from the body of published work in the field? The answer is quite obviously no.

One of the most important aspects of good empirical science is reproducibility. That is true even in those sciences where one can do controlled experiments. Many years ago, great hopes were raised by a scientist who claimed to have discovered "cold fusion," but those hopes were soon dashed when no one could duplicate his results. Reproducibility is even more important in a social science where one can only observe and not control the events producing the data. Making a major policy change on the basis of a single outlier study without confirmation by any other researchers is not evidence-based policy, but quite the contrary.

Following the science requires a careful evaluation of the research literature as a whole, with careful attention to the limitations of each study and the confirmation by other researchers or lack of it. Cherry-picking a single study merely because its bottom line fits one's predetermined agenda and then claiming to be following the science is a misuse and debasement of science and a fraud on the public. 