I thought the story would be simple: Using publicly available data, we would count up how many criminal homicides were reported to the Los Angeles Police Department in 2018, compare it to years past and write a story.

This would be a straightforward 2018 data article for us at Crosstown, a data journalism news organization based at the Annenberg School for Communication and Journalism.

Of course, assuming anything in journalism is a mistake.

When we first looked at the LAPD data in early January 2019, it told us there were 251 reported criminal homicides in 2018.

But on Jan. 15, Chief of Police Michel Moore told the police commission that there 258 criminal homicides in 2018. So, what happened?

Answering that question proved to be a lesson in how slippery data are, even those that count something seemingly concrete such as murders. Answering that question has been a lesson for us in making sure we understand how the data are entered into the system, processed and updated. We’re still learning.

Following the Chief’s report back in January, we called the LAPD. We emailed them the spreadsheet we had constructed from their data, including salient details about each case, to ask about the discrepancy between the Chief’s report and the publicly available data. They suggested we submit a public records request.

On Jan. 16, 2019, we did just that. But the next day, Jan. 17, we received an email that our request was closed because the data were publicly available, sending us to the exact site where I had started.

When we looked at the data again on Jan. 24, the count had jumped. There were now 266 criminal homicides listed for 2018.

What was going on? No one at the department seemed to have an answer. So, at the advice of an LAPD communications officer, we asked, publicly, at a press conference.

Eventually, we sat down with the head of their CompStat team, the group within LAPD that pores over crime data to look for trends. He agreed to look at our list and compare it against his own.

He told us that some homicides, upon investigation, had been reclassified as cases of legitimate self defense, one was reclassified as a suicide, another happened in 2018, but wasn’t officially deemed a homicide until January 2019, so it won’t count for last year.

Another criminal homicide listed was the Trader Joe’s incident, where a manager, Melyda Corado, was killed during a shootout exchange between police and the suspect, who was holding people hostage inside the Silver Lake grocery store. Corado’s death is not recognized by the FBI as a homicide, the LAPD told us, though the LAPD is still charging the suspect with a homicide. And finally, the department counted one case of negligent manslaughter as a homicide.

So why weren’t these changes in the public database? It appears that when LAPD investigations change the status of a crime, the cops do not always update the publicly available data. Or at least not right away. There also seems to be some communication or bureaucratic breakdown between the data maintained privately on LAPD servers and data maintained publicly on the Open Data Portal.

We wrote three pieces on the 2018 criminal homicide numbers:

• Where reported criminal homicides were committed
• How racial disparities persist among victims
• How guns were used far more in criminal homicides than other weapons

The deeper we dig into crime data, the more idiosyncrasies and discrepancies we discover.

Two criminal homicides this year raised red flags for us.

The first was the murder of Victor McElhaney, a 21-year-old music student at the University of Southern California, who was killed not far from campus on March 10. His murder was coded as “gang-related,” yet no news article we found about the murder mentions gangs. Additionally, while the police said the murder happened after an attempted robbery, there was no mention of that in the LAPD codes. When we asked the department about these details, they declined to comment, citing the ongoing investigation. We wrote about Victor’s murder here.

The second was the high-profile murder of rapper and community activities Nipsey Hussle, 33, who was shot at 3:20 p.m. on March 31 in the parking lot outside of his store, The Marathon Clothing.

In the publicly available data on Nipsey Hussle’s murder, we noticed two tags in the data the police initially released. One was “0395: Murder/Suicide,” which was inconsistent with other descriptions of the crime. The second was “1822: Stranger.” That also didn’t match other information, as LAPD Chief Michel Moore intimated that Nipsey knew the alleged shooter, Eric Holder, 29, whom law enforcement ultimately arrested for the murder. Holder pleaded not guilty. The LAPD has since removed those codes from the publicly available data.

We are seeing data discrepancies in other major crimes, such as those coded as hate crimes. These may appear in the publicly available data one week, to then have the code for hate crime removed a week, a month or more later.

As we learn more about how the data are collected, recorded and updated, we will write more about our process here and on our site: xtown.la.