Sessions' Sloppy Stats

Today I’m taking a break from the usual science reporting to critically discuss some other reporting on statistics. I watched a news clip this morning with great frustration as a pair of local newscasters reported on an argument regarding statistics but gave no conclusion. The attorney general, Jeff Sessions, reported that particularly violent crimes were way up in Minneapolis. The police of Minneapolis disagreed.

“Yeah, but it’s interesting how these numbers don’t jive with what the police numbers are,” offered one news man. The other points out that the debate seems to be about looking at “months or years”, but doesn’t offer much to help figure out what anyone did to get their numbers.

I’d like to think that our public servants just forgot about the Mann-Kendall test for significant trends in time series data after consulting statisticians before making a presentation, and are not trying to sway public sentiment with highly immoral, cherry-picked data. A method of statistical inference, like the Mann-Kendall test, would definitely help settle this debate, because it doesn’t care about the length of your time-series as long as you have at least a few data points! Of course, if you look over smaller periods of time, you’ll need to have a stronger trend to be confident. Are you worried about the month vs. year debate that the newscaster pointed out? You can just model that with the seasonal version of the Mann-Kendall test! See no problem. Now let’s take a look at what The Honorable Jeff Sessions said in his comments.

But violent crime is rising. The murder rate, for example, has surged nearly 11 percent nationwide in just one year — the largest increase since 1968. Per capita homicide rates are up in 27 of our 35 largest cities.

That includes right here in a great city like Minneapolis. According to Minneapolis police, violent crime has increased six years in a row. Through the first half of this year, the preliminary data shows that violent crime is up 17 percent and homicides are up over 40 percent from this time last year.

These numbers are deeply troubling—and especially since they represent a sharp reversal of decades of progress. My best judgment is that this rise is not an aberration or a blip. We must take these developments seriously and consider carefully what can be done about them.

First, the percentage increases don’t tell us very much without knowing some accompanying statistical inference to know if those changes are worth thinking about. At the end of June 2016, there had been 12 homicides in Minneapolis that year. At the end of June 2017, there had been 17 homicides in Minneapolis. A whopping 5 more people (and yes, a 41.6666% increase). But in June 2015, there had been 22 homicides. It’s interesting that Sessions ignored this 45% decrease. Instead he focused on how violent crime had increased six years in a row.

Let’s look at that claim. First, I’ll ignore the poor reasoning that simply increasing six years in a row is at all meaningful. I’ll assume that what he means is that there is a significant trend of increased violent crime over the last 6 years. I conducted the Mann-Kendall test on the raw crime reports from the past 6 years. I couldn’t easily find the population to adjust for the total city population, which is typically done in this type of analysis, so take it all with a grain of salt.

For murder, no significant trend. For rape, no significant trend. For robbery, there is not a significant trend, but it’s close. For assault, there is indeed a significant positive trend. Since assault constitutes ~46% of all violent crimes, when I total violent crime, there is also a significant positive trend. The increase in total violent crime is 18% over 6 years. That means that if Minneapolis hasn’t grown much in that time, the violent crime rate is increasing. I couldn’t find the population growth of Minneapolis over that time, but I can tell you that it would take an equivalent and hefty 18% change in population in the last 6 years to make the trend non-significant, so Jeff Sessions is likely correct to be confident about that claim.

Overall, I felt like Jeff Session’s presentation of the data was purposely misleading to create more fear than the situation actually warrants in order to fulfill an agenda. That to me, is especially distasteful given that his assertion about violent crime increasing in Minneapolis appears to be true, and he could have fortified his claims with better statistics.

But perhaps I should share the attorney general’s worry. The last murder in Minneapolis saw increases in crime up one infinity percent from the previous second.