New research published in PLOS One sheds light on why the geographic distribution of death sentences in the United States is clustered in just a few jurisdictions. The findings indicate that a county’s legal and racial history plays a more important role than the homicide rate.
“I’m a quantitative social scientist and have been working in the area of the death penalty since publishing a book on how arguments about the death penalty are framed, and the impact of the ‘innocence frame’ on public opinion,” said study author Frank R. Baumgartner, the Richard J. Richardson Distinguished Professor of Political Science at UNC Chapel Hill.
His book, “The Decline of the Death Penalty and the Discovery of Innocence,” was published in 2008.
“Since then, I’ve been continuing with various studies of the death penalty, and I constructed a database of everyone executed in the United States, including the county from which the case derived,” Baumgartner explained. “A shocking pattern is an extreme concentration of cases in just a few counties. That was apparent in my data, but is also well known to those who study the matter, so it was not a new discovery.”
For their latest study, Baumgartner and his colleagues examined factors such as population size, number of homicides, violent and property crime rates, poverty rates, the relative share of nonwhite population, and history of lynchings. But they found a surprisingly weak relationship between the number of homicides in a county and the number of death sentences.
Instead, the better predictor of death sentences was the number of death sentences that a county had previously imposed. In other words, each death sentence made another death sentence more likely.
“The data are consistent with a ‘learning’ hypothesis where each time an event occurs, the expected time delay before the next event is shortened,” Baumgartner told PsyPost. “The most common use of such models is in analyzing such things as heart attacks or brain seizures. Given an underlying set of risk factors, having one heart attack makes a second one more likely. We used those statistical techniques and showed that a similar process occurred for U.S. counties.”
The findings are in line with previous research conducted by Baumgartner, which analyzed executions rather than death sentences.
“There have been about 9,000 death sentences in the United States since 1972, and about 1,500 executions,” Baumgartner said. “The findings are very strong with regards to death sentences, even stronger than in the earlier paper. So our point was not to identify something new; the fact of high concentration of use of the death penalty in just a few jurisdictions was already known. Rather, we explored where it comes from: Learning. Some counties get very good at it. Some never do.”
The study provides evidence that the “death penalty is not applied in a consistent manner,” Baumgartner told PsyPost.
“Even within the same state, there are stark differences in the odds that a crime with certain characteristics will lead to a death sentence,” he said. “A key factor is randomness. Just by flukes of their own histories, some counties get good at death sentencing, and others do not. This is not consistent with the U.S. Constitution, which guarantees ‘equal protection of the law.’”
The researchers also found evidence that racial dynamics played a role.
“To the extent that we could discern a pattern to the data beyond the learning aspect, we found two main patterns, one of less interest but one very troubling. The first is population size; larger counties have more death sentences. That’s not that interesting but is an important thing to keep in mind and to have as a statistical control. Once you control for population, homicides have no effect; that’s pretty interesting,” Baumgartner explained.
“The troubling thing is that racial dynamics have an effect. We found that a county’s history with Jim Crow-era lynchings still has a discernable statistical impact on the odds of a crime leading to a death sentence. So: ugly racial history, population size, and a lot of randomness and learning from one’s own history rather than having the penalty applied proportionately across jurisdictions in accordance, for example, with homicides.”
But the study — like all research — includes some caveats.
“I think we did a good job of explaining the concentration of cases in a small number of jurisdictions,” Baumgartner said. “At the same time, there are complicated trends toward and away from use of the death penalty over time, and these affect the entire nation. Death sentences became more common in the period from 1976 through about 1996, and have declined since then.”
“Similarly, progressive District Attorneys have now taken office through election in some of the death penalty ‘hotspots’: Houston, Philadelphia, Atlanta, and in those three counties they have completely suspended the use of the death penalty. So the trends we describe are not written in stone, and there are other factors that matter as well.”
“The legal profession is not typically very concerned with sophisticated statistical models,” Baumgartner added. “For those who are well versed in numbers, I think our repeated events model will be very convincing and show that the numbers cannot be squared with equal protection of the law. Whether the U.S. Supreme Court will understand that is another matter!”
The study, “Learning to kill: Why a small handful of counties generates the bulk of US death sentences“, was authored by Frank R. Baumgartner ,Janet M. Box-Steffensmeier, Benjamin W. Campbell, Christian Caron, and Hailey Sherman.