Dan Tokaji's Blog
Professor Dan Tokaji
Election reform, the Voting Rights Act, the Help America Vote Act, and related topics -- with special attention to the voting rights of people of color, non-English proficient citizens, and people with disabilities

Dan Tokaji's Blog Links Publications & Working Papers
Equal Vote
Monday, January 31
 
Mea Culpa Rejected
A couple weeks ago, the Edison/Mitovsky folks who brought you the now-famously erroneous election-day exit polls issued this report explaining how they went wrong. The exit polls overstated the Kerry vote significantly, nationally and in 26 states. Overall, the polls showed Kerry winning the popular vote by 3.0% when in fact Bush won by 2.5%. In a nutshell, the Edison/Mitovsky report attributes the error to Kerry voters participating in exit polls at higher rates than Bush voters. The report also concluded that "Exit polls do not support the allegations of fraud due to rigging of voting equipment," noting that the exit polls were about as far off in jurisdictions using paper-based optical scan equipment as they were in those using electronic voting machines.

Now, a group calling itself "US Counts Votes" is contesting Edison/Mitovsky's attempt to fall on their sword. USCV has issued this response suggesting that it wasn't the exit polls that were wrong in all those states but the vote totals -- and, more specifically, that a variety of "un-auditable" voting systems are to blame. The group, which describes itself as composed of "prominent statisticians," claims to have come up with statistical evidence that the exit poll reports were just the "tip of a national iceberg." USCV rejects the conclusion that Kerry voters participated in exit polls at a higher rate than Bush voters, noting that precincts with 80-100% Bush voters showed higher participation rates than those with 0-20% Bush voters. They point to the fact that precincts with hand-counted paper ballots showed no statistical discrepancy between exit polls and actual results, while those with other kinds of voting equipment -- including punch cards, optical scans, electronic machines, and lever machines -- all did. One of the statisticians responsible for the report describes this as a "coherent theory that must be explored."

My take: I'll mostly leave the explanation of the problems with the exit poll results to those more skilled in the art and science of polling, like Mark Blumenthal and his Mystery Pollster site. What I do want to respond to here is the suggestion that the statisticians who make up this group -- none of whom appear to have any particular knowledge of election administration -- have come up with any theory, coherent or otherwise, of how huge numbers of votes were shifted from Bush to Kerry in multiple states.

As an initial matter, the USCV group cites the fact that hand-counted paper ballot precincts showed no discrepancy between exit polls and actual results, while precincts using other technologies did. What they don't mention is the fact that hand-counted paper ballots were used by only about 0.6% of voters in 2004, mostly in smaller rural counties. Thus, the fact that the exit polls were close to accurate among the few voters who still use hand-counted paper ballots proves, well, pretty close to nothing.

More to the point, how in the world do they hypothesize that hundreds of thousands of votes were switched on election night, in the 13,000 or so different local jurisdictions that are responsible for administering elections, using different types of voting equipment manufactured by several different companies? Were all of these folks in on some grand conspiracy? There's no "coherent theory" offered here -- in fact, there's no theory offered at all. The decentralization of our election system has many disadvantages, but one of it's advantages is that it makes it practically impossible to pull off the sort of grand conspiracy that USCV hypothesizes. You may be able to bribe one election official or even a dozen. But how do you bribe thousands without being detected?

To be sure, some votes were undoubtedly lost due to the faulty equipment -- most notably punch cards -- still used by a significant percentage of voters in the United States. Others may have been discouraged from voting due to long lines. Still others were undoubtedly prevented from voting due to registration foul-ups and the failure properly to administer provisional voting. But none of these things even begin to explain the discrepancy between the exit polls and the actual results. Nor is there any plausible explanation offered as to how such a large number of votes could have been switched.

Put simply, the USCV folks may know statistics but they don't seem to know much about the administration of elections. Now I do know something about elections (or at least, in the eyes of my detractors, pretend to), but concede relatively little knowledge about statistics. Still, there's something else in the USCV report that seemed a little fishy even to my relatively untrained statistical eye. As noted above, they rely on a chart appearing at page 4 of their response showing that voters in 80-100% Bush precincts had slightly higher response rates than than those in 0-20% Bush precincts. But precincts at those two extremes are relatively rare. Thus, as Mystery Pollster Mark Blumenthal notes here, this doesn't really refute the argument that Kerry voters participated in exit polls at higher rates than Bush voters:
the two extreme precinct categories are by far the smallest (see the table at the bottom of p. 36): Only 40 precincts of 1250 (3%) were "High Rep" and only 90 were "High Dem" (7%). More than three quarters were in the "Even" (43%) or "Mod Rep" (33%) categories. Not that this explains the lack of a pattern - it just suggests that the extreme precincts may not be representative of most voters.
He also points out here that the exit polls were off by almost as much in 1992 and have consistently overestimated the vote for Democratic candidates. Why might that be? Blumenthal notes here that errors favoring Kerry tended to occur in places where the interviewer was under 35, inexperienced, and had a graduate degree -- and that "it is not hard to see the underlying attitudes and behaviors at work might create and exacerbate the within-precinct bias." As he notes:
Consider age, for example. What assumptions might a voter make about a college student approaching with a clipboard? Would it be crazy to assume that student was a Kerry supporter? If you were a Bush voter already suspicious of the media, might the appearance of such an interviewer make you just a bit more likely to say no, or to walk briskly in the other direction? Would it be easier to avoid that interviewer if they were standing farther away? What if the interviewer were forced to stand 100 feet away, among a group of electioneering Democrats - would the Bush voter be more likely to avoid the whole group?
Maybe this explains the higher response rate among Kerry voters, and maybe it doesn't. But at least it's a coherent and plausible explanation of how this might have happened. That's a whole lot more than the USCV folks offer

Powered by Blogger Site Meter


Moritz College of Law The Ohio State University