Still Jumping to Conclusions

January 30, 2009

For years, too many of our public officials developed a bad habit of jumping to conclusions on insufficient facts.  Even after a return to private practice, it seems, that habit is hard to break.

Three weeks ago, Hans von Spakovsky wrote an op-ed about a South Carolina state house race ostensibly marred by fraud.  He claimed that the losing candidate, Wallace Scarborough, found more than 300 "illegal votes," more than half again the margin of victory.  And he accused the State Election Commission of violating state law, "fail[ing] in its sole responsibility: to protect the security and integrity of the democratic election process." 

These are strong words, and Mr. von Spakovsky is certainly entitled to his opinion.  But those opinions would be more persuasive if they were more soundly supported.

The allegedly "illegal votes" were apparently compiled by attempting to match the addresses on voter registration records to addresses on postal records, motor vehicle records, and car tax records: those that don't match are branded illegal.  We've seen this flawed technique before—for example, in Montana, where a reservist's registration was challenged because he wanted to send his mail to his mom while he was deployed overseas. 

The attempt to find fraud through list-matching was similarly flawed in South Carolina.  There may have been a handful of voters who had moved within the county, and voted legally under the NVRA and South Carolina state law, but were given the wrong ballot.  But there were also many "illegal voters" whose votes were perfectly proper, like those with office addresses on their driver's licenses, or those who forwarded their mail elsewhere.  Indeed, because Wallace Scarborough forwards his own mail to a P.O. box, he apparently made his own first list of frauds

The State Election Commission did its job, by reviewing the facts.  As it recognized, the evidence didn't support the conclusion that the election had been stolen.

Today, Mr. von Spakovsky is back in the news, with more conclusions that just don't follow.  This time, he's attempting to support one of his favorite causes, restrictive voter identification laws, by showing that they did not depress turnout in the 2008 election.  Even if that were the right measure of the laws' legal or policy merit—a disputed proposition—his evidence doesn't say what he says it says. 

Mr. von Spakovsky notes that voter turnout in both Indiana and Georgia—the two states with the strictest photo ID requirements—was much higher in 2008 than in 2004.  He compares these gains to the "neighboring" states of Illinois and Mississippi, which had more moderate identification laws but a smaller increase in turnout. 

These are facts: Indiana and Georgia had strict photo ID laws and high turnout compared to 2004, Illinois and Mississippi had moderate laws and relatively lower turnout.  But the most casual students of statistics can spot the flaw in Mr. von Spakovsky's conclusion.   

A common example makes the problem clear.  Where you find a lot of firemen on the street, you'll often find a lot of property damage.  That doesn't mean, of course, that firemen cause property damage.  Big fires do.  The strength of the fire is the "independent variable" that explains far more about the property damage than the number of firemen on the scene.

Mr. von Spakovsky is ignoring the fire.  The turnout figures he cites don't tell us much at all about whether ID laws had an effect on turnout, because there's a much bigger factor at work.  In the 2008 presidential race, massive resources poured into registering and mobilizing voters in Indiana and Georgia, states where far less effort was spent turning voters out in 2004.  Thus, turnout shoots up.  Far less attention was spent in 2008 on Illinois and Mississippi.  Thus, turnout gains are comparatively more modest.

Campaign mobilization—and lots of it—is an "independent variable" that explains far more about the change in turnout than Mr. von Spakovsky's ID laws.  Indeed, neither Georgia's nor Indiana's turnout increase was as large as the increase in other new battleground states, but it would be just as premature to say that the voter ID rules were the cause of the gap as it would to say that they had no effect at all.

We know that overly restrictive ID laws cause unnecessary problems for some eligible citizens.  It is difficult to gauge the precise magnitude of that impact: those who don't currently have the right ID are among the least visible populations.  Acknowledging that the problem is difficult may help us find a better solution.  Jumping to unwarranted conclusions does just the opposite.