Any good student of Statistics 101 will tell you that correlation does not imply causation. Apparently, many voter ID supporters never got the memo.
Two and a half years ago, Justin Levitt wrote on this blog about how some proponents of voter ID requirements were asserting that stringent ID laws in Georgia and Indiana did not depress turnout in 2008. Those proponents thought they had found their magic bullet: turnout in Georgia and Indiana was higher in 2008 than in 2004, despite the implementation of strict ID laws in the interim.
Mr. Levitt gave them a simple statistics lesson. Even if turnout increases at the same time as the adoption of a new voter ID law, there may be something other than the voter ID law – Mr. Levitt identified campaign mobilization, in particular – that caused the turnout increase. In other words, correlation does not imply causation.
Bad statistical practices – like old habits – die hard. Supporters of voter ID requirements are at it again, this time misinterpreting a new set of election results in Georgia. In response to E.J. Dionne’s Washington Post column on vote suppression efforts across the United States,Georgia’s Secretary of State wrote to the Post’s editors about how an increase in black turnout between 2006 and 2010 showed that voter ID laws do not suppress turnout. Hans von Spakovsky repeated the assertion on NPR and in USA Today, and Ohio House Speaker William Batchelder picked up the same message in defending Ohio’s proposed voter ID requirement. Citing the Georgia statistics in a see-this-couldn’t-be-that-bad sort of way has become a central talking point among proponents of voter ID laws.
Once again, these proponents have mistaken simple correlation for causation. You don’t need to be a statistician to know that without controlling for other factors that might influence turnout, the assertion that Georgia’s voter ID requirement didn’t depress turnout is meaningless — at best unscientific, at worst just plain wrong.
By comparing Georgia’s turnout with turnout in other similar states that do not have voter ID requirements, it is possible to control for other factors that influence turnout. Of course, the comparison will never be perfect. But it can provide valuable context.