Police routinely changed queries used for drunk driving stats, producing unreliable statistics, auditor says.
Auditor’s report found many of the same issues first identified by 2016 Statesman investigation.
Police data can be essential for police and policymakers to figure out how to enforce laws, protect safety.
The Austin Police Department has changed how it tallies drunken driving wrecks, now using the Texas Department of Transportation’s database, according to a new report by the city auditor’s office.
That change alone increased the average annual number of drunken driving wrecks in Austin by 52 percent — more than 600 collisions — when compared with numbers pulled from the Police Department’s own database, the auditor’s report found.
In the past, the department routinely changed how it crunches drunken driving statistics — key data used to help shape law enforcement decisions made by city and police officials, according to the report.
The American-Statesman spotlighted the Police Department’s statistics problem in May 2016 — during the closing days of the $10 million fight over Austin’s ride-hailing regulations — when the department provided contradictory counts on drunken driving crashes in the city.
A subsequent analysis by the Statesman in August found significant deficiencies in how authorities have analyzed cases of driving while intoxicated — findings echoed in the city auditor’s report released this week.
“APD has routinely changed the techniques used for analysis of DWI data,” the audit report said. “These changes … have resulted in fluctuations in both the overall DWI arrest and crash statistics used for research and/or reported to stakeholders during the scope period October 2013 through March 2017.”
RELATED: Police revise drunken driving crash stats key to Prop 1 campaign
The report also plainly laid out the importance of the data: “These changes affect the accuracy of DWI incident data used by the department for decision-making and the consistency of reporting on DWI incidents.”
The department didn’t immediately respond Wednesday to questions about what — if any — changes it made in how it produces DWI counts, in the wake of the Statesman’s investigation and the auditor’s review.
However, the auditor’s report says the department made a key change the month after the Statesman published its investigation into the department’s DWI statistics.
The drunken driving data became a critical part of a political campaign seeking to overturn Austin’s ride-hailing regulations during the spring of 2016.
At the time, police data showed a 23 percent drop in drunken driving wrecks from 2013 to 2014, when ride-hailing giants Uber and Lyft began to officially serve the city — a figure they used to support Proposition 1, which would have repealed City Hall regulations that required drivers be fingerprinted.
But, in the closing days of the campaign, the police provided the Statesman with a second set of figures that showed the drop was just 12 percent.
Uber and Lyft went on to lose by 12 percentage points.
Then, just days after the election, the police sent the Statesman a third set of numbers that showed the drop was 17 percent.
“The data that was given to us in the heat of the battle turned out to be wrong,” said District 7 Council Member Leslie Pool, who requested the review. “To the extent that we can control for some of this, we should be controlling for some of this.”
Pool said she will raise the issue of the Police Department hiring a full-time statistician during upcoming budget talks this summer as one way to reduce such errors in the future.
The Statesman investigation that followed identified key deficiencies in the Police Department’s statistics unit and how it crunches the drunken driving data. Like this year’s audit, the Statesman found that the department failed to standardize the criteria it used to pull and crunch the statistics from its database.
The investigation also found the department’s statistics unit had no formal system for storing these queries or the results they produce, depriving it of a key way to check its work, and that its staff has little formal training in computerized mapping, databases or statistics when they were hired.
“A tip of the hat to (the Statesman),” Pool said, for uncovering the shortcomings.