A family member who has a history of opposition to efforts to reduce racial discrimination emailed me an article by Paul Sperry entitled, “Eric Holder believes all cops are racists…” Sperry claims that Holder’s efforts to reduce unconscious racial bias have caused Seattle police officers, “to ignore signs of criminal behavior and threat indicators they’ve gleaned from years of street experience.” Sperry says that attempts to reduce discrimination by police officers, “puts their own lives in danger — and risks the safety of residents.” Sperry claims that anti-discrimination efforts among the Seattle police have been a disaster:
Since Holder stepped in, crime is up 13% overall in Seattle. But it’s not just minor infractions. It’s the biggies — aggravated assaults up 14%, car theft up a whopping 44% and murders up 21%.
Those statistics immediately piqued my interest because it is really hard to get crime rates to swing that much, so I checked them. Sperry provided lots of links, but oddly, he didn’t link to any sources for his statistics, so we will probably never know where he got his data. But I searched Google and the first article I found at the top of my Google search was a column from the Seattle Times last week that points out that crime statistics fluctuate by these sorts of magnitude (13%-44%) from month to month all the time. Sperry attributes the large (13%-44%) increases in crime to the changes that Holder started making in 2012, but it is hard to make annual crime statistics fluctuate by 13% or more. And a 21% increase in annual murders would be big-time national news. I’m guessing that Sperry carefully selected a month for his article where he found big random changes in the statistics because I looked at the last year of data and found pretty much what I would expect. Monthly stats randomly swing by large amounts, but the annual trend in the latest crime data for the past year is very small. The annual change is much smaller than the variance (monthly swings) which means that there probably isn’t any statistically significant trend over the past year at all.
Sperry’s article is a good example of how statistics make smart people smarter, and stupid people stupider. It is harder to lie with statistics than to lie with stories because in the information age it is easier to check statistics than to check stories. Paul Sperry is either stupid or lying to his readers. Either way, his use of statistics makes it really easy to discredit him with a quick Google search and that makes us all smarter. Now we know that we should not trust him nor his publisher. His publisher is the New York Post which is literally the Fox News of newspapers. Both are run by Rupert Murdoch. My sister also said that there is a graph of Seattle crime data in the article, but it wasn’t there a few hours later when I looked. ¡¿It is weird that the New York Post took data out of their article without explanation, huh? A reputable newspaper would post a correction when they make a substantive change to an article. Otherwise, readers who email relatives about the article will look stupid when their source alters the story and hides part of the original work.[Update: My sister emails to say that she probably saw the graph somewhere else and wrote the source wrong.]
Mark Twain complained about “lies, damned lies, and statistics”, but that was a long time ago when statistics were nearly impossible to check. Today I might have found Sperry’s story to be plausible if not for his stupid use of easy-to-check statistics. Without his dramatic statistics, his claims fall apart, so maybe Seattle’s efforts to reduce discrimination are worthwhile after all.
Similarly, Darrell Huff who authored the bestselling book, How To Lie With Statistics, was paid by the tobacco companies to lie to Congress twice to try to deny the growing evidence that tobacco causes cancer. Although Huff undoubtedly helped Big Tobacco’s battle temporarily, they still lost the war and the truth of statistics won. Huff made a little money, but he permanently tarnished his legacy and lost credibility.
Medianism 