The Florida killing of Trayvon Martin, an unarmed black teen earlier this year, has brought national attention to the laws that allow people to use lethal force to defend themselves.

At least 20 states have laws with provisions that don’t require civilians to flee from an intruder before fighting back, according to the National Conference of State Legislatures. Of those, eight states, all of them in the south, specifically use the phrasing, “Stand Your Ground.” That includes Florida.

Since Martin’s killer, George Zimmerman, invoked the stand-your-ground defense, these laws have been defended by gun rights groups for empowering civilians. They’ve also been criticized by civil rights groups for encouraging violence and being racially biased.

A recent study suggests that laws may lead to more deaths. According to a June study [pdf] by researchers at Texas A&M University, the rates of murder and non-negligent manslaughter increased by 8 percent in states with Stand Your Ground laws. That’s an additional 600 homicides per year in the states that have enacted such laws.

The study, which analyzed FBI crime data nationwide from 2000-2009, says it could mean either that more people are using lethal force in self-defense, or that situations are more likely to escalate to the use of violence in states with the laws. “Regardless, the study said, “the results indicate that a primary consequence of strengthening self-defense law is increased homicide.”

But how do you measure racial bias? John Roman, a senior fellow at the Urban Institute’s Justice Policy Center, recently conducted a study examining racial disparity using FBI data on 43,500 homicides from 2005 to 2009, the most recent years for which data was available. He sifted out only the killings in which there was a single shooter and a single victim, both of whom were strangers to each other — narrowing the pool to about 5,000 homicides.

And Roman looked specifically at “justified” homicides, which as defined by the FBI, are when police determine a private citizen has killed someone who is committing a felony, such as attempted murder, rape or armed robbery.

Roman found that the killings of black people by whites were more likely to be considered justified than the killings of white people by blacks.

But the analysis didn’t compare Stand Your Ground states to those without the laws.

At FRONTLINE’s request, Roman analyzed the pool of 43,500 homicides by race in states with Stand Your Ground laws* and those without them. Because he wanted to control for multiple variables — the races of the victim and the shooter, whether they were strangers, whether they involved a firearm and whether the murders were in Stand Your Ground states — Roman used a technique known as regression analysis, which is a statistical tool to analyze the relationship between different pieces of data.

Using this analysis, Roman found that a greater number of homicides were found justified in Stand Your Ground states in all racial combinations, a result he believes is because those states yielded more killings overall.

Roman also found that Stand Your Ground laws tend to track the existing racial disparities in homicide convictions across the U.S. — with one significant exception: Whites who kill blacks in Stand Your Ground states are far more likely to be found justified in their killings. In non-Stand Your Ground states, whites are 250 percent more likely to be found justified in killing a black person than a white person who kills another white person; in Stand Your Ground states, that number jumps to 354 percent.

You can see the breakdown of the killings in the chart below. The figures represent the percentage likelihood that the deaths will be found justifiable compared to white-on-white killings, which was the baseline Roman used for comparison:

So the disparity is clear. But the figures don’t yet prove bias. As Roman points out, the data doesn’t show the circumstances behind the killings, for example whether the people who were shot were involved in home invasions or in a confrontation on the street.

Additionally, there are far fewer white-on-black shootings in the FBI data — only 25 total in both the Stand Your Ground and non-Stand Your Ground states. In fact, the small sample size is one of the reasons Roman conducted a regression analysis, which determines the statistical likelihood of whether the killings will be found justifiable.

And lastly, whether a homicide is ruled justifiable only tells part of the story. Stand Your Ground laws can be applied at multiple points during an investigation.

In Florida, for example, if a shooter invokes the Stand Your Ground law, police can determine whether to make an arrest when they arrive on the scene. If they do arrest him, the suspect then appears before a judge who determines whether Stand Your Ground applies to the case. If it does, the prosecutor then decides whether to go to court.

The system offers substantial discretion to authorities at every level, which is much more difficult to monitor and evaluate — and much more vulnerable to creeping bias.

That’s why the U.S. Commission on Civil Rights announced last month that it would investigate concerns about racial bias in the law’s application.

Michael Yaki, the commissioner leading the probe, said in an email that he wasn’t surprised by Roman’s findings. “It reinforces even more the need for a comprehensive investigation,” he said.

He added: “What the [commission] will do is more complete, more thorough, and ultimately aimed at determining whether SYG statutes by their nature, enforcement, or application, create opportunities for racial bias to enter into the system.”

Yaki said he planned to examine a handful of states that have enacted Stand Your Ground laws, most likely including Florida. His study should take at least a year.

*Roman considered any state with a broad self-defense law to be a “Stand Your Ground” state, although states have passed different variations of the law and some states didn’t enforce the laws until the end of 2009.

Sarah Childress, Former Series Senior Editor, FRONTLINE