Taking a range of risk factors into consideration, Florida, Georgia, Texas and Alabama are among the most likely states to make serious mistakes in capital cases while Connecticut and Colorado are low risk, according to Columbia Law School Professor James Liebman, the study's lead researcher.
While race, politics and an overburdened legal system play a strong role, Liebman said areas that relied heavily on the death penalty as punishment, even in weaker cases, were most likely to impose a flawed capital sentence.
"What our study shows is that aggressive death sentencing is a magnet for serious error," Liebman told Reuters.
The central premise of Liebman's study - that an overturned sentence means that a serious error happened in the trial - is challenged by the Criminal Justice Legal Foundation.
It says it's not new or surprising that a large percentage of capital verdicts are overturned. According to the foundation, the issue is whether that happens because of mistakes, as death-penalty opponents contend, or because of unreasonable obstructions placed in the way of such sentences, as advocates of capital punishment assert.
The study, which looks at why mistakes occur in capital cases, follows a report by Liebman in 2000 which found that 68 percent of all death verdicts reviewed from 1973-1995 were reversed by courts due to serious error.
Of those reversals, 82 percent ended in less harsh sentences, and 9 percent of those people were found not guilty and eventually freed.
Since the death penalty was reintroduced in America in 1973, 99 death row inmates have been exonerated, raising questions about its validity and leading to mounting pressure among opponents for it to be scrapped or at least suspended.
"If you have a scattershot death penalty policy, you are going to miss most of the time," said Liebman, a strong opponent of the death penalty.
The researchers found broad differences from one area to another within the same state. For example, in Lexington County in South Carolina, the death sentence was imposed in 93 per 1,000 homicides. In Richland County, just a couple of miles away, the rate was 9 per 1,000 homicides.
What the study showed said Liebman, was that the death penalty should be reserved only for the "very worst of the worst" cases.
The study estimated when death sentences increased from a quarter of the national average to the highest rate, the predicted increase in reversal rates was sixfold to about 80 percent.
The more aggravating circumstances found in a case - such as multiple victims, a defendant who has a long history of prior violent behavior or physical torture - the less likely a mistake would be made.
"As you add those aggravating circumstances, the likelihood of reversl goes way down," said Liebman.
Looking at particular cases, researchers identified three key errors that often led to reversals: incompetent legal counsel, police or prosecutors who suppressed evidence and judges who gave jurors the wrong instructions.
High capital reversal rates were also more likely in densely populated states and in areas where the risk of homicide was higher for whites and in those areas with a weak record of catching and imprisoning serious criminals.
Another trend was that the more often state trial judges were subject to election and the more partisan those elections, the higher the error rate.
The report suggested 10 reforms, including proof beyond any doubt that a defendant committed a capital crime, barring the death penalty for defendants with obvious extenuating circumstances such as for juveniles and the mentally ill.
Other suggestions included making life imprisonment without parole an alternative to death, making all police and prosecution evidence available to the jury, insulating sentencing judges from political pressures and the appointment of competent defense counsel.
For the most part, Liebman said little if any compensation was given to people wrongly sentenced to death.
"In all of these cases, it's a case of wasted lives, time and money," he said.
© MMII, CBS Worldwide Inc. All Rights Reserved. This material may not be published, broadcast, rewritten, or redistributed. Reuters Limited and The Associated Press contributed to this report