In the early hours of June 5, 2015, Gary Bravo was leaving Sammy T’s in downtown Huntsville, Alabama. The club was hosting a gay night because the last of the city’s few gay bars had closed and some downtown bars were picking up the slack.

As Bravo walked out with two co-workers, they encountered a group of young men. One grabbed Bravo’s friend, and Bravo intervened. The next thing he remembers, someone spun him around, and he was on the ground being punched and kicked while his attackers shouted homophobic slurs. Faggot. Cocksucker.

“A couple more hits and I would have ended up being brain dead,” he recalled.

Bravo suffered extensive injuries from the attack. His right eye was bloodied and swollen, and he couldn’t see from it for weeks. His eye socket had to be reconstructed.

Despite his attackers’ words during the beating, police did not investigate it as a hate crime, or report it to state or federal authorities as one.

Bravo’s case is just one of thousands lost each year to a deeply flawed system for collecting hate crime data, one that has left the U.S. with unreliable, incomplete official counts and little handle on the true scope of bias-motivated violence.

Under a federal law passed in 1990, the FBI is required to track and tabulate crimes in which there was “manifest evidence of prejudice” against a host of protected groups, including homosexuals, regardless of differences in how state laws define who’s protected. The FBI, in turn, relies on local law enforcement agencies to collect and submit this data, but can’t compel them to do so.

The evidence suggests that many police agencies across the country are not working very hard to count hate crimes. Thousands of them opt not to participate in the FBI’s hate crime program at all. Among the 15,000 that do, some 88 percent reported they had no hate crimes. According to federal records, the Huntsville Police Department has never reported a hate crime.

Local law enforcement agencies reported a total of 6,121 hate crimes in 2016 to the FBI, but estimates from the National Crime Victimization Survey, conducted by the federal government, pin the number of potential hate crimes at almost 250,000 a year — one indication of the inadequacy of the FBI’s data.

“The current statistics are a complete and utter joke,” said Roy Austin, former deputy assistant attorney general in the Department of Justice’s civil rights division. Austin also worked at the White House on data and civil rights and helped develop an open data plan for police data.

It’s true that many hate crime cases fall away before they start because about half the victims never report them to authorities.

But to understand why so many cases that are reported to authorities still fall through the cracks, ProPublica requested incident reports or aggregate data from more than 350 law enforcement agencies in 48 states, including the 50 largest agencies nationwide, on the bias-motivated crimes they had investigated since 2010.

More than 280 agencies responded, but in many cases only to say they hadn’t investigated any such incidents, or had no records, or that their records were bad. When we followed up with agency public information officers, they acknowledged that investigators frequently did not mark down incidents as motivated by bias, even if there was evidence suggesting this was so — a spray-painted swastika, for example, might be classified simply as vandalism and not also as a hate crime.

The FBI and some larger agencies champion a two-tiered process in which incidents classified by the first responding officer as potentially bias-motivated are re-evaluated by a second investigator, who determines if the incident should be counted officially as a hate crime. Few of the agencies that responded to our records request appeared to follow this procedure.

In most states, local law enforcement agencies send their hate crime data to the state, which is then supposed to submit it to the FBI, but we found several instances in which this chain broke down. The Orlando Police Department, for example, told us it had reported five hate crimes for 2015 to the Florida Department of Law Enforcement, but the FBI data shows no hate crimes investigated by Orlando that year. Orlando police say FDLE acknowledged the city’s data hadn’t been sent to the FBI “due to a system error,” but an FDLE spokesperson told ProPublica the Orlando police hadn’t submitted the information on time.

Variations in how states define hate crimes also cause confusion that contributes to the undercount. That appears to be one factor in what happened with the Huntsville case.

Alabama has a hate-crime statute, but it only applies to acts of bias based arising from “race, color, religion, national origin, ethnicity, or physical or mental disability,” not sexual orientation. A prosecutor on Bravo’s case acknowledged Bravo’s sexual orientation clearly played a role in his attack, but a spokesman for the Huntsville police said the assault couldn’t be reported as a hate crime because gay people aren’t protected under the state’s law.

The 1990 federal law, however, includes crimes against gay people among those it tracks and asks agencies to report such cases even if no hate crime charge was prosecuted.

Jack McDevitt, a longtime hate crimes researcher and the director of Northeastern’s Institute on Race and Justice, has reviewed police incident reports to check if investigators missed indicators that assaults or vandalism were actually hate crimes.

“We’ve found quite a bit of misclassification or missed cases that could have been investigated,” he said. “You’ve got police officers who aren’t trained to ask the right questions.”

It might not seem like being left out of the official hate crime count would rank high among Bravo’s concerns, but he said it did. Not being counted made him feel as though he and victims like him are being erased — and that there’s no accountability for the bias that motivates their attackers.

“If it happens to someone else, they’re going to be treated the same exact way, and it’s going to be unfair,” he said.

---

When we asked local law enforcement agencies across the country for their hate-crime reports and data, it quickly became clear that this was an area of substantial uncertainty and discomfort for many of them.

In Omaha, Nebraska, Deputy City Attorney Bernard J. in den Bosch said he couldn’t release older hate crime records in response to ProPublica’s request because the police department suspected that former employees had classified such crimes incorrectly in reporting them to the FBI.

“The Omaha Police Department does not feel comfortable providing the numbers since they are concerned about their accuracy,” in den Bosch said, adding that it was difficult to identify hate crimes properly and that the employees entrusted to do so had done a job the department was now “leery of.”

A spokesman for the Madison, Wisconsin, police department acknowledged its officers weren’t trained in identifying bias-motivated incidents and had not been submitting reports properly as a result.

For one thing, officers weren’t sure if an incident had to be prosecuted as a hate crime to be counted (it doesn’t). After receiving ProPublica’s information request, department officials said they altered their policies on reporting such crimes to the FBI and reclassified dozens of incidents.

In several cases, the information agencies sent ProPublica about hate crime investigations since 2010 was at odds with what they had submitted to the federal government.

According to FBI data, the Anne Arundel County Police in Maryland, which polices half a million people between Baltimore and Washington, D.C., reported no hate crimes between 2012 and 2015. But the department sent us 68 pages of reports for the past 5 years, showing that county police responded to well over 100 incidents, including many in which there was evidence of crimes motivated by bias.

In one case, a woman brought her new boyfriend, a black man, to the home of her ex-boyfriend. He allegedly swung a hammer at the visitor, saying “I don’t want any niggers in my house, get this nigger out of my house.” According to another report, someone scratched the same word into the side of a black man’s car and stole from it.

Officials with the Maryland State Police said many of these cases weren’t reported to the FBI because the Anne Arundel County Police — unsure if they were hate crimes — had flagged them as “inconclusive” in paperwork. Lt. Ryan Frashure, a spokesman for the county police, said the agency has started marking more of its reports as conclusive if they have the appearance of a bias motivation. Indeed, in the 2016 FBI hate crime report, the most recent available, the Anne Arundel County Police reported 16 hate crimes.

Officials at some police departments acknowledged that their officers were simply not in the habit of classifying incidents as possible hate crimes, even when there was evidence that this was the case.

The Miami-Dade Police Department, Florida’s largest law enforcement agency, reported just two hate crimes to the state since 2010. Records officials told us they could not find any reports from that period in which investigating officers noted that a crime had a bias motivation.

“I was on the road for 9 years, but when I wrote a report, I don’t ever remember titling it a hate crime,” said Detective Argemis Colome, a spokesman for the department. “If they would have done graffiti on a wall, it would have been titled a vandalism.”

After speaking with ProPublica, the department’s director ordered an audit to see whether the department had failed to identify hate crimes properly and, thus, to report them to the state and the FBI.

In general, police departments nationwide have embraced data in the last 20 years, using an ever-more sophisticated array of it to track where crimes are happening, to allocate resources, and to hold commanders accountable for results. The New York Police Department’s CompStat initiative is among the most prominent of such efforts and has been credited with playing a role in broad, lasting reductions in violent crime in what was once considered an ungovernable city.

Yet, for many law enforcement agencies, collecting and using data on hate crimes has remained a stubborn exception. According to Brian Levin, a former New York City police officer who studies hate crimes, some law enforcement agencies lack a commitment from the top to properly investigate such incidents and collect data on them.

“You have to have a combination of training, executive leadership, and some kind of infrastructure that is sustained and continuing,” Levin said.

Myesha Braden, a former civil rights prosecutor in the Department of Justice, said police departments often miss opportunities early on to communicate to officers that hate crimes are a priority. “If an officer knows coming into the police academy that hate crime is one of the important crimes they’ll be investigating, at least the seed is planted,” she said.

A recent ProPublica review of training academy standards found that only 12 states have laws requiring police to learn how to identify and investigate hate crimes at that stage; few agencies provide such training once officers leave the academy.

Officers at several police departments told ProPublica they thought it was up to prosecutors to decide if an incident was a hate crime, or that they needed a suspect in custody to categorize an offense as a hate crime in their reports.

Such misconceptions stop officers from classifying hate crimes as they should, said Michael Lieberman, the Anti-Defamation League’s counsel in Washington, D.C.

“To say that there’s not enough to prosecute it, that’s not the question. The question is what happened at the scene of the crime,” he said. It wouldn’t take much to encourage better reporting, he added — for starters, more police departments could add a field to incident report forms allowing officers to mark that a crime had a bias motivation.

“What’s on the form is what’s important,” Lieberman said. “You have to have a dropdown box that says ‘hate crime.’”

The Boston Police Department is frequently held up as a model because it has a dedicated civil rights unit that handles hate-crime investigations. Officers in the unit train BPD recruits to identify bias motivations when they fill out crime reports, and the unit serves as secondary investigators on every potential hate crime, Sgt. Det. Carmen Curry, a supervisor in the unit, said in an email. On top of that, officers in the civil rights unit search through police reports daily for terms like “gay,” so that potential hate crimes don’t get missed. 

BPD has reported 996 hate crimes to the FBI since 2010, averaging about 142 a year.   “It’s a very difficult thing to maintain trust in law enforcement right now,” Curry said. “The fact that the BPD goes out front and has a unit to respond to these types of incidents sends an important message.”

---

Momentum to pass a federal law to count hate crimes started to build in the late-1980s, spurred in part by a notorious case in New York.

In 1986, Michael Griffith, a 23-year-old black man, was killed by a car in Howard Beach, Queens, while running in terror from four white men who hurled epithets as they chased him from their neighborhood. Ed Koch, then the mayor of New York, called it a “modern-day lynching.”

When the House debated a bill in 1989, Democratic Rep. Barbara Kennelly called Griffith’s death an example of why crimes motivated by bias needed to be distinguished from other crimes, tracked and counted separately.

“These crimes of prejudice must be recorded as such if we are to come up with informed ways to prevent them in the future,” she said. “The basic idea behind hate crime legislation is simple: The more we know about these crimes of hate, the better chance we have to prevent them. Accurate data on when, where and how often these crimes occur will help.”

Legislation to collect hate crime statistics encountered stiff resistance from the Justice Department, which said requiring police to divine criminals’ intentions would “discredit” the agency’s overall data efforts, and from the White House, which said it would “impose unnecessary burdens” on police. There were also those who opposed including crimes against gay people in the count.

“In my opinion, our society should not enshrine homosexuality on a pedestal alongside race and religion as the primary focus of our civil rights laws,” Rep. William Dannemeyer, R-Calif., said before a debate on the bill. “Do not let the sexual revolutionaries hijack the freedom train.”

Ultimately, the Hate Crime Statistics Act that passed in 1990 directed the attorney general to keep track of crimes motivated not only by race, color, religion or national origin — the groups protected by the federal hate crime statute at the time — but also those based on sexual orientation.

The Justice Department assigned the task to the FBI, which asked local law enforcement agencies to submit detailed information — including the bias motivation, offender’s race, type of offense and date of each possible hate crime — as an addendum to the information they submitted to the Uniform Crime Reporting Program on a variety of crimes, from burglary to homicide.

To help police understand what to do, the FBI organized training sessions and put together two large gray books packed with information on investigating hate crimes. By 1996, the FBI had held 61 regional conferences, training more than 3,600 law enforcement officers from nearly 1,200 agencies. The conferences covered “investigation, identification, reporting and appropriate handling” of hate crimes, according to congressional records from that period.

The first hate crime data released by the FBI in 1991 had information from 2,771 law enforcement agencies in 32 states. By the next year, that jumped to 6,181 agencies in 41 states and the District of Columbia. By 1999, the number had nearly doubled to 12,122 agencies. In the latest release, 2016, the FBI reported that 15,254 agencies participated. Still, participation just means agencies submitted information of any kind — even if it’s only the form that says they had no hate crimes. The percentage of agencies reporting zero hate crimes has grown from 73 percent in 1991 to 88 percent in 2016.

Today, the government does little to help local law enforcement to collect accurate and complete hate crime statistics.

The Federal Law Enforcement Training Centers, a branch of the Department of Homeland Security, once sent hate crimes specialists around the country to work with state and local cops, but no longer does so.

While the FBI still does technical training on how to send data to the UCR program, “the UCR trainers do not provide training regarding the investigation of hate crimes,” according to Stephen Fischer, a spokesperson for the division that used to run hate-crime-specific training. Fischer said the agency is working with local police to modernize their computer systems, which is designed to facilitate all reporting.

Cynthia Deitle, the chief of the FBI’s Civil Rights Unit from 2008 to 2011, is now program director for the Matthew Shepard Foundation, which has sponsored a handful of hate crime conferences for police similar to those she used to organize at the FBI. In October, she held one at the National Civil Rights Museum in Memphis that was attended by about 50 people, some in law enforcement, others from community groups like OUTMemphis, a local LGBT group.

Six officers came from the Shelby County Sheriff’s Office, in part because the state noticed the agency’s hate crime statistics were wrong. According to Chief Deputy Floyd Bonner, the sheriff’s office had wrongly classified several incidents as hate crimes, but further review revealed they were not. For example, he said, when two African-American kids got into a fight at school, the reporting officer checked off the box for a hate crime “for some reason.” Those incorrectly categorized crimes amounted to more than half of the hate crimes the agency reported for the year, a problem Bonner attributed to a lack of training.

“Time and time again reporting issues come up. So it’s about teaching officers what to recognize,” he said. “We all think that hate crime is on the rise because of the news. But it’s still very important because we should know exactly.”

Dennis and Judy Shepard gave the keynote address at the conference, telling attendees about their son, Matthew, who died in 1998 after being beaten and tortured by two men because he was gay, and how they had worked for a decade to push for legislation that expanded the federal hate crime law to cover crimes motivated by a person’s sexual orientation or gender identity. The Shepard/Byrd Act passed in 2009.

After their speech, they spoke with me about their frustration that attacks like the one that took their son’s life continue not to be reported as, or labeled, hate crimes, even since the new law took effect. When communities don’t know the full toll that bigotry exacts, they have no way to defend themselves, they said. They said the 1990 hate crime statistics statute should have made it mandatory for law enforcement agencies to report to the FBI, not voluntary.

“The one real flaw of the law is not requiring reporting,” Dennis Shepard said. “If they don’t mark it, then they can’t prosecute it.”

---

Four months after Gary Bravo’s June 2015 assault, the Huntsville Police Department quietly made a single arrest in the case, charging a southern Alabama man with second-degree assault. They said their investigation had led to only one assailant. Since the man was 18 at the time of the attack and had no prior felonies, he cut a plea deal under Alabama’s youthful offender statute, receiving up to three years of probation. His criminal record remains sealed.

Publicly, the outcome of the case drew almost no attention. The police and prosecutors had not labeled it a hate crime and dispensed with it as they would any routine street fight.

The incident did prompt Marianne Landers to open Deja Vieux, a new gay bar — and now the oldest operating one in the city — in a cavernous, windowless space just across a highway from the Huntsville Police Department. Landers said she felt compelled to open a place for gay people after she read about Bravo’s beating and filed paperwork to form a company just 24 days after the attack. She said she thought of her nephew, who doesn’t live in Huntsville, but is also gay.

Still, the bar’s operator, Cecil Gordon, insisted Bravo’s attack couldn’t have been a hate crime. He said local law enforcement agencies were always good to leaders of the gay community and surely would have given Bravo’s case more attention and been tougher on the perpetrators if they knew Bravo was attacked because he was gay.

“You can’t fall back on the gay card,” Gordon said of Bravo’s story.

On an October visit to Deja Vieux, the talk turned to the attack on Bravo. Several of the men who had come for drinks said they did not know if Bravo had been the victim of a hate crime or agreed with Gordon that he wasn’t.

This is what happens when bias crimes aren’t counted and publicly called out as acts of hate, said James Robinson, the CEO of Free2Be, a statewide LGBT advocacy group that does counseling in Huntsville.

Robinson tried to publicize the attack on Bravo and helped raise funds for his recovery. He understood the assault couldn’t be prosecuted as a hate crime under Alabama law, but nonetheless thought the police should acknowledge that anti-gay bias had motivated the attack. Otherwise, how would anyone know the violence was aimed not only at Bravo, but at the larger community to which he belonged?

“Words have power in two kinds of ways: They have power when you speak them and power when you don’t,” Robinson said. “The words need to be said. This community needs to be recognized.”