Crime report statistics give method to madness
By Evan Brandt
An old trope on the subject of statistics say, “There are three kinds of lies: Lies, damned lies and statistics.”
While statistics can certainly help illustrate the crime problem in Pottstown, the numbers should not be mistaken for the only way to look at the problem, or even the final word.
“The job of a statistician is all about uncertainty,” said Marc Scott, associate professor of applied statistics at New York University. “With statistics, it’s always about what is being compared to what.”
(Full disclosure: Scott is a childhood friend of Evan Brandt.)
For example, The Mercury compared crime statistics from 2014 to an average of the last 10 years of crime statistics, including 2014.
While that showed that in most crime categories, 2014 crimes were below the average, that conclusion can be uncertain if the numbers are very low, Scott warned.
“When the number of events, and thus the rate, is low, comparisons with other rates may be misleading,” Scott wrote in an e-mail.
Pottstown’s murder “rate” is a good example.
With no more than three murders a year in the past 10 years, we not only get a ridiculously impossible average of 1.5 murders per year, but it is misleading to say Pottstown’s murder rate in 2014 was 33 percent higher than the 10-year average because there were two instead of 1.5 murders that year.
“You have two data points and 20,000 people,” said Scott. “When you have very low numbers, it can be dangerous to draw too many conclusions.”
“You may see an increase in the numerical sense that isn’t necessarily an increase in the statistical sense,” said Scott.
“One way we might approach this question is, looking at this data, to ask how many murders might you expect in a year and how many more would you need in a given year to say that something has changed,” said Scott. “In this case, I would say not until you reach six murders.”
Also, it’s important to recognize, Scott said, that the murderer, or victim, may not be from Pottstown, just in Pottstown.
With one murder among 22,546 people in 2012, readers should remember that “It’s not like there are one out of 20,000 murderers in Pottstown,” he said. “But when it’s framed as a ‘hit rate’ (to be blunt) then it makes sense. Just like a disease, a certain percentage of the population will get hit with it. In other words, 20,000 people are exposed to the potential to be murdered but only one was.”
Further, said Scott, statisticians would adjust that “exposure” for other factors.
“Looking at those numbers, you might conclude that everyone in Pottstown that year had a one-in-20,000 chance of being murdered. “
“But if the targets of the murders are drug dealers, or users, your chances of being a murder victim are much lower, ” unless you are a drug dealer, Scott said.
