Tuesday, June 12, 2007
Reporting from the Mathematically Challenged
One piece of big news over the weekend was the release of the Tennessee Bureau of Investigations (TBI) report on Hate Crime 2006. The report in the Nashville Tennesseean was typical, citing the good news that overall crimes went down 16% but also noting sharp rises is some categories, especially those with a religious or sexual orientation basis. They have an extensive quote from Chris Sanders, president of the Tennessee Equality Project:
Many have looked at the 21 reported cases of sexual-orientation bias in 2005 and compared it against the 60 cases in 2006 and cried the sky is falling. However, if you look at individual categories (racial bias, ethnicity bias, religious bias) everything went up from last year. But how can that be if overall hate crimes are down 16%? Well, in 2005 there we 258 "non-specific" hate crimes, over 60% of the total. The reporters, on a whole, fail to either understand or mention that with this margin of uncertainty in the 2005 numbers it's statistically impossible to discern any trends or perform any conclusive analysis. Granted, the report linked above mentions the better quality reporting of category in an off-hand way (a comment that was completely absent from the print version that ran Sunday) but then they seem to ignore what it means and just march down their narrative heedless of the fact they obliquely brushed against. But all is not lost!
I did find a report by Jake Jost of WBIR in Knoxville, TN. Jake not only recognizes the significance of the poor categorization in 2005, he clearly explains it to his readers:
[H]e believed that hate crimes targeting homosexuals may have been more frequent last year in part because of a constitutional amendment on the November election ballot defining marriage as being between a man and a woman.All fine and dandy, except the actual statistics don't support this analysis.
Many Christian groups rallied around the amendment to ban same-sex marriages.
"I think that created more hate discourse," particularly in fundamentalist congregations and small towns, Sanders said.
Many have looked at the 21 reported cases of sexual-orientation bias in 2005 and compared it against the 60 cases in 2006 and cried the sky is falling. However, if you look at individual categories (racial bias, ethnicity bias, religious bias) everything went up from last year. But how can that be if overall hate crimes are down 16%? Well, in 2005 there we 258 "non-specific" hate crimes, over 60% of the total. The reporters, on a whole, fail to either understand or mention that with this margin of uncertainty in the 2005 numbers it's statistically impossible to discern any trends or perform any conclusive analysis. Granted, the report linked above mentions the better quality reporting of category in an off-hand way (a comment that was completely absent from the print version that ran Sunday) but then they seem to ignore what it means and just march down their narrative heedless of the fact they obliquely brushed against. But all is not lost!
I did find a report by Jake Jost of WBIR in Knoxville, TN. Jake not only recognizes the significance of the poor categorization in 2005, he clearly explains it to his readers:
Despite the drop in the total number of hate crimes reported, several categories appear to have increased in numbers. However, 2006 showed a precipitous drop in the number of crimes reported with unknown bias motivation. Essentially, more hate crimes are now being reported to specific categories.To top it off, Jake provides a link to the source report on the TBI's web page right beside his article. Now this is what news reporting should look like!
TBI reports the change is the result of better training for law enforcement agencies to report bias-motivated incidents. As a result of this year's improvement in reporting accuracy, it is difficult to assess whether individual categories have increased or decreased from previous years.
