POLICE-SEARCH STATISTICS DON'T LIE, BUT STATISTICIANS CAN

All statistics are not created equal. Some shine a spotlight to
illuminate reality, but others spread a thick fog of confusion. An
example of the fog effect is a recent study that uses weak statistics
to back up claims that Texas police engage in widespread racial profiling.

The study was commissioned by a coalition of civil rights groups,
including the American Civil Liberties Union.

"Blacks and Hispanics stopped for traffic violations are more likely
to be searched than Anglos in six of every seven jurisdictions in
Texas, a statewide study of police records shows," The Dallas Morning
News recently reported.

The same story went on to say the study found that "minorities also
are pulled over and ticketed at a higher rate than white drivers by
three-fourths of law enforcement departments."

The report itself concluded that "the statistical racial disparities
in stop and search rates suggest a pattern of racial profiling by law
enforcement agencies across Texas."

It all sounds as simple as saying one plus one equals two. But the truth is,
we still don't know if racial profiling is a big problem in Texas. That's
because the study, titled "Racial Profiling: Texas Traffic Stops and
Searches," simply doesn't stand up to the tough standards that researchers
impose to determine if a statistical study is valid.

Before elected officials, police, advocacy groups, community leaders
and other Texans rush to change the ways that police operate, let's
step back and take a close look at three important flaws in the study.

Flaw 1: The racial profiling report concludes that blacks and Latinos
stopped by police were much more likely to be searched than white
drivers, even though the authors of the study don't know the reasons
for the police searches. Yet such reasons matter, because officers
have little discretion in many searches. For example, probationers and
parolees generally aren't afforded the same rights to privacy as
others. More than 500,000 Texans are on probation or parole, and more
than half of them are black or Latino.

Flaw 2: The location of a search also matters when analyzing racial
bias. Search rates can vary greatly across a city regardless of race,
with the highest rates in the areas with the most crime. In some
high-crime areas, police search black and white drivers at comparable
rates.

If white drivers tend to drive in neighborhoods where officers
generally are less likely to search people they stop, white drivers
are at less risk of being searched.

Because the Texas racial profiling report's analysis doesn't account
for locations, we can't tell whether the observed disparity in
searches is due to a racial bias or to where people drive.

Flaw 3: The report attempts to measure racial bias in vehicle stops
essentially by comparing the racial distribution of stopped drivers to
the census, Texas transportation data and U.S. transportation survey
data. The three sets of data provide a "benchmark" - a target race
distribution that, presumably, the stop data should reflect if police
weren't targeting black and Latino drivers.

But in this case, three bad benchmarks don't make for a strong
analysis. We can't discern whether the reported disparities are due to
racial profiling or to differences in driving behavior and exposure to
law enforcement.

Even if white and nonwhite drivers committed offenses at equal rates,
police will be stopping black drivers more frequently if black drivers
travel roads with twice the law enforcement presence as the roads that
white drivers travel.

Today, many law enforcement agencies in Texas are failing to produce
data that can allow analysts to reach a coherent and conclusive
assessment on the extent of racial bias in policing in the state.

The problem can be solved if police collect and release more detailed
statistical information on the vehicles they stop, the places where
the stops take place and the motorists who are searched.

Because such information now is lacking, the only valid conclusion
that can be drawn from the Texas racial profiling study is that it
failed to collect enough data to reach a valid conclusion.

Greg Ridgeway is a statistician and Jack Riley is director of RAND
Public Safety and Justice, a unit of the RAND Corporation, a nonprofit
research organization in Virginia. They are working on a study of
racial profiling.