The death penalty, they say, is dying. And in many important respects, this is true. Support for capital punishment has decreased dramatically. In 1998, there were nearly 300 new death sentences imposed in the United States. In 2018, there were just 43. In the past decade alone, six states have abolished the death penalty by legislation or court order, while four more have imposed moratoriums on executions. Abolition is increasingly embraced by conservatives, who have spearheaded repeal efforts in a number of states, most recently and energetically in Wyoming.

This trend is partly rooted in research that continues to expose enduring problems with the death penalty, particularly its racism, arbitrary application, and failure to deliver on claims of public safety. Such findings are significant, but they are also familiar. They mirror the very same evidence that led the U.S. Supreme Court to strike down the death penalty more than 47 years ago in its historic decision in Furman v. Georgia. Decided amid dampening support for the death penalty and a long pause on executions, Furman declared that the death penalty was arbitrarily and capriciously applied. There were clear signs of racial bias and no evidence that it worked as a deterrent. Nevertheless, the response to Furman was swift: States immediately began enacting new death penalty laws designed to pass constitutional muster. Just four years after Furman, in 1976, the court upheld a new set of statutes in Gregg v. Georgia, signaling the start of the “modern” death penalty era.

Although the past few decades have ushered in significant reforms — such as improvements in the quality of capital defense, a major factor in reducing new death sentences — states continue to carry out executions in the face of serious questions about the influence of racism, mental illness, or intellectual disability; biased, shoddy, or simply incomplete police investigations; prosecutorial misconduct and woefully deficient defense lawyering; and the persistent influence of junk forensic science.

The failure to grapple with these problems has high stakes. The case of Rodney Reed is only the most recent example of a state rushing toward an execution despite serious concerns over innocence. Reed’s execution date was one of a slew of executions Texas planned to carry out through early December — a group of cases that illustrate all too clearly the problems plaguing capital convictions. Scores of people have been exonerated from death row across the country, and research published by the National Academy of Sciences suggests that there are hundreds more innocent individuals facing execution. Given this landscape, there is vanishingly little doubt that innocent individuals have been put to death — including in Texas, which has executed more than 550 people since 1976, far more than any other state.

People hold hands while praying during a protest against the execution of Rodney Reed on Nov. 13, 2019, in Bastrop, Texas.

Collecting the Data

As the 40th anniversary of the U.S. Supreme Court’s decision in Gregg v. Georgia neared in 2016, it seemed clear to us that the death penalty remained as problematic today as it was in 1972. But we wondered what the numbers would show. We were curious not only about who was on death row in the country’s remaining death penalty states, but how many people had been removed from death row — a sizable but largely invisible population in the ongoing death penalty debate. As far as we could tell, this particular dataset did not exist. So we decided to compile one ourselves. We began collecting information on individuals sentenced to death starting on July 2, 1976, the day Gregg was announced. Such data would allow us to see who had ended up on death row, where they were from, and the outcome of each case. The goal was to create a high-level snapshot of four decades of “modern” capital punishment.

We set out to collect information about every individual sentenced to die by the federal government, the U.S. military, and states with an “active” death penalty — those jurisdictions that still hand out death sentences or carry out executions, even if they’re rare. We also included states that had ended capital punishment but where people remained on death row.

No sooner did we begin collecting data than states began to drop off our list. Delaware’s death penalty was struck down by the state Supreme Court in August 2016, with the last two condemned men resentenced last year. In New Mexico, which ended capital punishment in 2009, the last remaining death sentences were reduced to life this year. In October 2018, Washington effectively abolished the death penalty by court order.

For the most part, however, the scope of the project remained unchanged. Our first step was to contact prison officials with the federal government and in each of our target states to ask for a full roster of individuals sentenced to death row beginning in July 1976. We also asked for specific data points: dates of birth, race, gender, sentencing dates, and current status. This last item is where things got tricky. It’s relatively easy to find lists of people who are on death row or have been executed, but each of these categories represents a minority of the individuals sentenced to die.

The greatest number of people sentenced to death since the summer of 1976 are no longer on death row, not because they were executed but for a host of other reasons: They’ve died or killed themselves; they have been exonerated or had their sentence commuted; or they’ve been resentenced to a lesser term. Many have been released from prison. This population of people no longer on death row makes up a stunning 43 percent of The Intercept’s dataset.

The lack of care in collecting such important data was fairly shocking — especially when you consider that pro-death penalty prosecutors and politicians tell us these cases represent the “worst of the worst.”

As we began to receive records, it became clear that much of what we were getting from the government was flawed. States like Florida and Texas — both extremely active death penalty states with robust sunshine laws — had the data on hand and were quick to provide it. Texas publishes most of its data online, but many other states simply do not track the information we were trying to find, or do so in a haphazard way that leads to incomplete and inaccurate data.

What we learned was that the states’ record-keeping is abysmal, on balance. Oklahoma told us that the information probably existed in hardcopy form but would be difficult to locate. The corrections department hadn’t had a computer upgrade since the early ’90s, and the documents were likely stored somewhere in a vast warehouse. Louisiana, where death row is housed at the Supermax prison known as Angola, gave us a copy of the so-called Angola Death Book, a binder full of details on a seemingly random assortment of cases. We were told that the Death Book had never before been released. After we questioned the accuracy of Nebraska’s public information regarding people no longer on death row, officials eliminated it from their website.

The lack of attention and care in collecting and storing such important data was fairly shocking — especially when you consider that pro-death penalty prosecutors and politicians always tell us that these cases represent the “worst of the worst,” people so dangerous it is necessary to murder them on our collective behalf. But perhaps even more vexing is the fact that the Bureau of Justice Statistics, an agency within the Department of Justice, collects detailed demographic and other data about states’ death-row populations, yet Congress has blocked its public disclosure. Although BJS has made some of its findings available in annual reports, the U.S. Code specifically exempts that research from public release, despite the fact that the majority of the collected information is a matter of public record. We were denied access.

Although we initially believed that the data collection would be a relatively straightforward task, it soon became clear that we would have to find alternate sources of information. Much of this data was acquired via a network of attorneys across the country, including lawyers in public defender offices at the state or federal level. Other sources were connected to advocacy groups, and some had individually been tracking this data in their respective states.

Another important resource was a series of quarterly reports published since the mid-1970s by the NAACP’s Legal Defense Fund, titled “Death Row USA,” which offers a running, state-by-state list of individuals sentenced to death. The reports include racial information for each individual and details about those who have been executed, including who abandoned their appeals and “volunteered” to die. The reports were particularly useful in trying to put together a list of people sentenced to death in Oklahoma, where the state could not provide any records concerning individuals removed from death row for a reason other than execution. We were able to use the NAACP documents going back to 1976 to create a list of 160 people who had been removed from death row there.

Finally, we undertook extensive online research to fill in various gaps, using the Death Penalty Information Center to keep our data current, and archival news reports, court opinions, and other sources to find information on older cases.

Once the dataset was largely complete, our next step was to provide it to a small group of lawyers and academics for review. They helped find additional errors and encouraged us to move forward. Though a wealth of research and data on capital punishment has been collected by others — notably the Death Penalty Information Center and law professor Brandon Garrett, whose research has long informed our reporting —  it was clear that our dataset could build upon such resources and reveal even more about the country’s capital punishment system.

The Intercept’s dataset contains individuals sentenced to die since July 2, 1976, in active death penalty jurisdictions, including 29 states and the federal government.

A Failed Policy

The end result finally came together in the fall of 2019. It includes 7,335 individual entries from 29 states and the federal government.

Some of the numbers from our dataset may appear to conflict with statistics that are widely cited when referring to the death penalty in the U.S. For example, in June, the country marked a grim milestone: the 1,500th execution since the Gregg decision. In contrast, our dataset contains 1,448 executions. This is because our dataset is limited to sentences handed down since Gregg in death penalty jurisdictions that remain active.

Nonetheless, the dataset provides a damning portrait of capital punishment in the U.S. Of the 7,335 entries, 1,448 people have been executed, 20 percent of the total. Even those who insist that executions are effective in combating crime must concede that this indicates a failed public policy.

Some 2,752 remain on death row, roughly 38 percent of the total. The single largest group of people are no longer on death row. Out of 3,135 entries, more than 2,000 have been resentenced and hundreds have been released. Hundreds more have died while awaiting execution; this group makes up 8 percent of our dataset.

Of the 7,335 entries, 1,448 people have been executed, 20 percent of the total. Even those who insist that executions are effective in combating crime must concede that this indicates a failed public policy.

In 32 percent of the cases represented in the dataset, individuals have ultimately been resentenced, meaning that courts have intervened to correct errors that occurred at trial. In other words, roughly a third of all death penalty prosecutions were flawed in some way that required readjudication. We call this the basic failure rate of the capital punishment system.

Fifteen states had above-average failure rates — some far above it. In Colorado, the failure rate for capital cases is at least 63 percent; in South Carolina, it is 60 percent. Idaho, Indiana, Kentucky, and Pennsylvania have failure rates of 50 percent or more.

In these cases, involving 2,373 individuals, the vast majority were ultimately resentenced to a lesser term in prison, most often to life (with or without parole). The remaining individuals were resentenced to death row, to a discrete term of years, or to time served and released. In fact, within the dataset are at least 333 people who have been released from prison after being on death row; at least 132 have been exonerated.

When it came to race, our findings were startling, if not entirely shocking. The disproportionate punishment of defendants of color — and black people specifically — is one of the death penalty’s defining historical characteristics. But our dataset shows that rather than becoming more equitable over time as new death sentences become rarer across the country, the death penalty appears to be more racially biased than ever. In state after state, our data showed that the percentage of people of color sentenced to death in recent years is larger than in the decade immediately following Gregg.

The numbers are especially stark in states that are home to the country’s largest death-row populations. In the first full decade after Gregg, people of color made up 51 percent of those sentenced to death in Texas. This percentage has grown to 75 percent in the past 10 years. In California, people of color made up 52 percent of those sentenced to death after Gregg, compared to 78 percent in the last 10 years. In Florida, the proportion jumped from 40 percent to 52 percent over the same two periods. And in Oklahoma people of color leapt from 28 percent of those sentenced to die in the decade after Gregg to 80 percent in the last 10 years.

Notes on Record-Keeping

Despite years of work, the dataset remains incomplete. One problem we found was a slew of missing names across states from the late 1970s and early 1980s. This is because, although Gregg may broadly mark the beginning of the “modern” death penalty era, many states had to repeatedly revise their new statutes before they could withstand judicial review. Although people condemned to die under such statutes were generally resentenced, the process varied from state to state — and information about these early cases can be hard to come by. Then there is the sizable set of names missing from our data altogether: the men and women sentenced to die in the chaotic era after Furman and before Gregg. Hundreds were pulled into the system during this time, and these invisible cases are an emblem of the false distinction between the so-called modern death penalty era and everything that came before.

As we prepared to make this data public, we wanted to be transparent about the ways in which we sacrificed nuance and detail in favor of simplicity and cohesion.

Race: As our reviewing experts pointed out, some of the most important evidence showing bias in our death penalty system is revealed by the race of the victim in a given case. Numerous studies have shown that a defendant’s chance of receiving a death sentence is far higher if the victim is white. In fact, a report published this past August revealed that in Georgia — whose death penalty law provided the model for other states back in 1976 — defendants convicted of killing a white victim are 17 times more likely to be executed. While some lawyers and academics seek to track such data in some jurisdictions, there was no way to obtain it for all the states and cases represented in our dataset.

Another limitation is in the categories we use to classify race. In order to standardize the information we received — and in an effort to reconcile the myriad ways that different states categorize death-row populations according to race — we settled on five broad categories: Asian, black, Hispanic or Latino, Native American, and white. Where we had no racial information or were unable to reliably classify an individual, we used the category “other or unknown.” Such labels are, of course, woefully inadequate in capturing the true racial and ethnic makeup of death rows across the country. Unfortunately, so were the materials we drew upon to form the dataset to begin with. A more accurate accounting of race would be a project unto itself and not one we could responsibly undertake based on the information upon which we were relying.

Date of birth: Fourteen states provided date-of-birth information; eight provided none. The remaining seven states provided some combination of information — some provided birthdates only for people currently on death row, while others provided the data only for people formerly on death row. Where full birthdates were not provided or we could not find them, we used published reports citing a person’s age to determine the most likely year of birth.

Suicides and “volunteers”: In all, our data contains 103 individuals who killed themselves while on death row. We feel certain that this is an undercount, in part because the results of death investigations aren’t always reported or made public. The data also probably undercounts the number of people who have given up their appeals and “volunteered” for execution, which is tricky to confirm absent an acknowledgement from the individual defendant or their lawyer. Our data contains 132 volunteers.

On death row: This field does not adequately capture the current status of all individuals in this category, particularly those in the middle of the appellate process. At any given time, a number of individuals will be caught in limbo: They’ve had their conviction overturned, but that decision is pending on appeal; they may have had their conviction or sentence overturned but are awaiting retrial or resentencing.

Exonerated: There are two main sources for tracking wrongful convictions: the National Registry of Exonerations and the Death Penalty Information Center. To determine who in our data had been exonerated, we adopted the registry’s definition, which does not rely solely on a court’s declaration of innocence. Applying this definition to individuals sentenced to death since July 2, 1976, across 29 states and the federal system, we identified 132 people exonerated after being sentenced to death.

Current sentence: Among the data are individuals who have been retried and resentenced multiple times after their first death penalty conviction, including individuals resentenced to death row more than once. Our data reflects each defendant’s most recent sentence.

Sentencing date: This can either be the date a jury handed down or recommended a death sentence or the day the defendant was later formally sentenced by the judge. Some states — most notably Texas — provided the date a person was received by the prison system, rather than the sentencing date. Typically, the incarceration date will follow fairly close to sentencing, though there are exceptions. In some cases, we could only narrow the sentencing date down to a particular year. Finally, although we have aimed to provide the original sentencing dates for all individuals, several of our entries are likely to reflect resentencing dates. Reversals in capital cases were common in the aftermath of Gregg. Some states, like Ohio, saw their new death penalty laws overturned, leading to mass resentencings. But others, like California, were repeatedly ordered to provide new trials or resentencing hearings in individual cases. Although some were resentenced to life, many others were sent back to death row. We have found that state records frequently fail to account for original sentencing dates in such cases. It is likely that the number of people we have designated as “resentenced” is lower than the actual total.

Commutations: People often conflate resentencing and commutation. In our dataset, a commutation denotes an executive branch action that reduced an individual’s sentence. Our data contains 68 commutations.

Duplicate entries: Within the data, we found a handful of people who had been sentenced to death in multiple states. We have counted at least eight such cases. In other words, our dataset contains 7,335 unique entries, not individuals.

After years of work to complete The Intercept’s death-row dataset, we nonetheless continue to find holes and inaccuracies. But we are sharing it publicly now because we believe that it is a unique and useful resource for anyone seeking to research capital punishment in today’s death penalty states.

We also believe that it provides a disturbing look at capital punishment as a whole. It strongly demonstrates across a number of metrics that the death penalty is a routinely fallible and ultimately failed policy. Nearly half of everyone sentenced to death since 1976 in the country’s active death penalty jurisdictions are no longer on death row, but not because they were executed. Hundreds have been released from prison; thousands are serving a lesser prison term. More than 130 have been exonerated. The thousands who have been resentenced demonstrate that capital prosecutions are flawed in at least a third of all death penalty convictions.

Far from delivering on the Supreme Court’s promise of a fairer and more equitable system in Gregg, the death penalty is as arbitrary as ever. People of color not only remain overrepresented as a whole, but these discrepancies appear to be getting worse — even as death sentences decline year after year. Nationally and in the country’s leading death penalty states, the percentage of people of color sentenced to die in recent years is becoming larger than the percentage sentenced four decades ago.

We hope that our findings will lead to further research. The Intercept’s death-row dataset is a living document, available on Github, that we will continue to update as new death sentences are meted out, old convictions are overturned, and executions continue. We welcome input. If you have information relevant to our dataset, please email us at deathrow@theintercept.com.