Reply #23: well, it's an empirical question

[OnTheOtherHand]

whether the top-of-the-ticket undervote in a presidential year equals the top-of-the-ticket undervote in a midterm year. Honestly, I haven't seen much data that investigates this issue either way. (Full disclosure: I'm pretty darn sure some folks have written undervote papers that I haven't read yet.) In a while I will see if I have enough Ohio data to give an estimate for 2002 gov versus 2004 pres.*

I think your intuition here is basically sound: top-of-ticket undervotes should be pretty low regardless of the cycle. But I wouldn't assume that they ought to be the same. Or that they shouldn't. It's an empirical question. I guess I would expect more variability in House top-of-ticket undervote rates than in presidential top-of-ticket undervote rates, because there is more chance that someone who turns out in a midterm election to vote on a bond issue doesn't know about the House candidates than that someone who turns out in a presidential election doesn't know about the presidential candidates. But that is, how does one say?, a SWAG.

What I'm saying is that I find the comparisons on that Earthlink website more convincing, because they are comparing House top-of-ticket to House top-of-ticket. That doesn't mean that we need to ignore other comparisons.

(To reiterate, since there is probably some confusion, somewhere, on this point: I think the House undervote rate in Meck was around 4%, and I think that rate was somewhat higher than it 'should' have been. Southern Dem disagrees about that, apparently. It's a hard question to bang the table about. I think you've suggested that an ordinary DRE top-of-ticket undervote rate is 3% -- I think it's lower, actually. If it were 3%, well, the difference between 3% and 4% is hard to distinguish amidst ordinary variability.)

* EDIT TO ADD: The median presidential undervote in Ohio 2004 (precinct level) was 1.39%. The median gubernatorial undervote in Ohio 2002, for the precincts for which I have data, was 3.45%. Restricting the presidential analysis to those same precincts, the presidential undervote median is 1.51%. So I provisionally conclude that in fact, the top-of-ticket race in Ohio in 2002 had a higher undervote rate than the presidential race in 2004. Of course, the Ohio governor's race was not at all competitive; I wouldn't try to generalize from this result. Also, I'm working fast, so numbers are subject to correction.