A ballot order victory?

There are, of course, all sorts of explanations for Hillary's dramatic NH victory on Tuesday (for a really nice analytical run-down, check out Mark Blumenthal at pollster.com), but I wanted to highlight one interesting factor that may be enough to account for Hillary's narrow victory margin– the order of the candidates on the ballot. 

Back when I was at Ohio State and had the distinct pleasure of knowing and even working with Political Scientist extraordinaire, Jon Krosnick, I remember him presenting some really interesting data on how the order of candidate on the ballot can affect results.  I seem to recall that he was an expert witness in an Ohio court case on the matter.  Anyway, Slate's Trailhead blog cites Krosnick's research in looking at how NH's ballot order advantaged Hillary Clinton:

Until this year, New Hampshire rotated the order of the candidates from precinct to precinct. An analysis of recent primary elections in New Hampshire by Stanford social psychologist Jon Krosnick,
an expert in polling methodology, found that candidates averaged 3
percent better than their overall performance if their names were
listed first among the leading candidates…

The new law
dictates that New Hampshire now set its ballot order by publicly
drawing a random letter of the alphabet to determine where the state
will begin listing names alphabetically. (For example, drawing an E
would have meant that, among major candidates, John Edwards? name would
be listed first, while Chris Dodd would be last.) This year Z was drawn,
effectively starting back at the beginning and listing Joseph Biden
first, even though he was no longer in the race. The system applies to
every ballot in the state uniformly. 

This method might appear to be equally unfair for
everyone, except that candidates? surnames are not equally distributed
throughout the alphabet. On the Democratic side, for example, the major
candidates where heavily skewed toward the front of the alphabet. The abecedarian
lineup of Biden, Clinton, Dodd, and Edwards meant that the latter three
had only a 1-in-26 chance of being first?that is, only if the letter of
their last name was drawn (since their last names were adjacent to each
other in the alphabet). With Biden out of the race, the advantage
effectively fell to Clinton. 

Clinton?s gain was Obama?s loss. As the last
candidate alphabetically (not counting the various fringe candidates
who were on the ballot), Obama faced a large probability of being last.
There are 11 letters after O in the alphabet, all of which would result
in Clinton being first among viable candidates still in the race once
the alphabet cycled back to the beginning. Add three for the odds of
drawing an A, B, or C, and Clinton had a 14-in-26 shot at being first.
(That?s 54 percent.)  

One might scoff at the idea that this really
matters, but Krosnick insists that the data is there to support it.
When I spoke with him last night at about 11 p.m. ET, he said that, had
the previous rotation of names been in effect, ?my guess is this race
would be too close to call.?

Even had Hillary lost by 1 point, instead of triumphed by 2, this would have been portrayed as a huge comeback, given how she far exceeded media expectations.  Nonetheless, it is much better to be the victor, and should she go on to win the nomination, wouldn't be interesting to wonder if NH's (stupid) ballot law played a role.