The Akron Beacon Journal reports that a trial on whether punch card ballots were constitutional has been delayed. Why? Read on:
ACLU lawyers complained Wednesday that the state's last-minute filing hadn't given them enough time to study the evidence---a report comparing punch cards with other voting methods. ...
The ACLU had wanted the judge to declare punch cards unconstitutional.It argues that the ballots---the bane of the 2000 presidential election---aren't uniform and don't alert voters to a mistake, as electronic machines do. The ACLU also contends that the state's "unequal" balloting system violates the rights of blacks, who live in predominantly punch-card counties.
And who wrote this last minute report?
The report in question Wednesday generally supports the use of punch cards. It was written by John Lott Jr., a resident scholar with the American Enterprise Institute, a conservative think tank.
Lott revised the report earlier this year and submitted a new version to the state weeks ago. However, the ACLU didn't get the new version until last week.
A spokeswoman for the Ohio Attorney General's Office couldn't say why the revised report was filed only Friday.
But she said that's irrelevant.
"We believe that the conclusions in both reports are identical," Kim Norris said for her office, which represents the Ohio Secretary of State's Office in court.
ACLU attorney Dan Tokaji said Lott's revised report has 11 additional pages. The original version, he said, contained "serious flaws."
"The report contradicts what the secretary of state has been saying for months, that punch cards need to be replaced," he said. "... The state is talking out of both sides of its mouth."
Most experts seem to think that punch cards are a lousy system, but they obviously don't have Lott's way with statistics. Lott's report is here. Even Lott is forced to concede that punch cards have a higher undervote rate in the presidential race. However, Lott argues that this is cancelled out by a lower undervote in the state races. I can't say that I'm impressed with this argument. The state races aren't as important, so undervotes there aren't equivalent to undervotes in the presidential election. Also, the undervote rate is much higher in the state races, so that most of the undervotes must be intentional---Lott is comparing intentional undervotes with unintentional ones.
Lott also includes 36 demographic variables for age, sex and race in his analysis. He included these in his "More Guns, Less Crime" analysis as well. As Ayres and Donohue have pointed out, these variables are strongly correlated with each other. (This is called multicollinearity.) This can cause spurious results. Ayres and Donohue found that Lott's MGLC results,
are incredibly sensitive to the inclusion of various seemingly unimportant demographic controls.
and
Apparently, then, Lott and Mustard's thirty-six demographic variables mimic time trends in crime that we can control for directly with our controls for state trends.
Lott's results on punch card ballots must similarly be regarded as suspect. Lott, of course, has read Ayres and Donohue and is well aware of the multicollinearity problem, but chose to repeat his previous mistake.
Incidently, Dan Tokaji has a blog where he writes about voting technology.