Some Thoughts Before the Analysis

Long night (I entered my last post at 5 a.m.) – late start this morning. I’ll need a couple of days to fully digest the numbers coming out of last night’s events (and to catch up on sleep).  So for this morning let me just begin the post-election analysis with a few observations.

As most of you know, I take a bit of pride in not getting emotionally invested in any candidate, in order to analyze elections (and the presidency more generally) as objectively as possible. This sometimes drives my students (and readers of this blog) nutty, because they want me to post comments that reaffirm their gut feelings regarding which candidate is better, why their opponent is scum, etc.  By now, I hope you’ve developed a bit of a grudging respect for my perspective, even if you don’t agree with it.

However, this doesn’t mean I’m dispassionate about the presidency, or about elections.   It will be said over and over again in the coming days, and much more eloquently than I can say it. But I’ll say it anyway.

We, as Americans, should take pride in what happened last night.  First, we affirmed an ideal that for far too long in this country has been more often talked about than acted upon: that the only qualifications one needs under the Constitution to become president are the ones related to age, citizenship and residency (subject to term limits).  Last night we took a step closer to realizing that ideal. We aren’t all the way there, of course, but it was a huge symbolic victory.

Second, and more important I would suggest, is that we once again transferred power from one party to the other without threat of coup, bloodshed, intimidation, vote stealing (well, ok, my guess is that there was not a little vote stealing going on).  We take this for granted now.  But it’s not the norm across the world, and it has not even always been the norm in our country (although we have a pretty decent track record dating back to 1800.)

Democracy, American-style, is an on-going experiment.  We did pretty well last night.

Ok. Now on to the important things: how well did our forecast models work?  Pretty darn well. I’ll address that, and related election issues, in the next post.

PS. I know that some of you had trouble logging in to post on the live blog last night. None of the tech people were able to explain what happened. I’m very sorry you weren’t able to participate, but I do hope you were at least able to read the exchange. Lots of great comments – Andy outlasted me, and was still posting excellent updates at 4 in the morning – I finally had to tell him to go to bed at 5 am.  I’m sorry that I couldn’t provide the more detailed election night analysis of exit polls that you’ve become accustomed to, but my effort to do simultaneous election night commentary at the college and the online blogging  meant that I didn’t do either particularly well!  A lesson for next time…

And let me say special thanks once again for Professor Bert Johnson’s willingness to cohost the Election Night event – as always, his insights and comments made it a better night for everyone.


  1. The only thanks I need is the vindication that I’m getting from the national exit poll data, which are showing that late deciders split about evenly between McCain and Obama.

    Seriously, though, I had a fine time at the Grille last night. Thanks to the organizers, and also to all who showed up and contributed.

  2. Actually, as I’ll demonstrate in a more extended post, the exit polls show conclusively that the “late deciders” broke for McCain and in almost the exact proportion that I suggested in my college wide talk. But don’t take our word for it – go to the national exit polls and look at voters making up their mind in the last two weeks – the NES definition of late deciders – to see how they broke. Or look at those making up their minds in the last three days.

    I’ll give Bert the small number of people who made up their mind in the last day.

    But I claim vindication.

  3. Four percent of people made up their minds on the last day, and Obama won these folks by five percentage points (50-45). Three percent of people made up their minds in the last three days, and these went for McCain by only six percentage points (52-46). If you aggregate these groups, the breakdown among the latest deciders is almost exactly 50-50. Even throwing out the people who made up their minds on the last day (and on what principle would we do that?), a six-point margin is not a decisive shift to McCain. (I heard something about a 5 to 1 pro-McCain margin being a good benchmark for how late deciders might go.) Kerry did better than this among late deciders in ’04.

    If we broaden our definition of late deciders to the “last week,” McCain does a bit worse, winning this particular group (3% of the electorate) by only 2 points, 50 to 48 percent.

  4. I am shocked – shocked! – by Bert’s blatant attempt to manipulate the data to support his argument, while not granting me the equal privilege of doing so to support my argument. Rather than my transparent effort to simply present the exit poll data as it stands, with multiple categories of possible “late deciders” showing McCain winning at least 2 of 3 of those categories, Bert instead decides we should aggregate the categories into one version of “late decider”.

    But I will suggest that, in doing so, Bert has hoisted himself on his own petard. He suggests that we if aggregate exit polls categories to determine a single class of late deciders, it proves his assertion – following the Gopoian article – that late deciders make their choice randomly. If we use Bert’s definition, however, we find that late deciders have NOT broken randomly in recent elections (although they may have done so in 2008). Take 2004: late deciders, in the aggregate (those making up their mind in the last week – Bert’s definition)) – broke for Kerry 52-46% In fact, according to Brian Schaffer, (see late deciders, using Bert’s definition, have broken for the Democrat in every election since 1992! Hardly 50-50!

    But of course that is NOT the NES definition, nor the one I used in my previous posts. I (mistakenly it seems) followed the initial definition provided in the Gopoian article. To refresh the readers’ memory: “We define late deciders as those voters whose final candidate choice decision materialized during the last two weeks of the campaign.”

    So, we either follow the original definition, consistent with the NES data, or we allow Bert to arbitrarily define a new category that fits his argument. If the latter, I assert an equal right to do so.

    If we follow the NES definition, I stand by my claim, pending refutation by the NES data – McCain won the majority of voters who made up their mind in the last two weeks of this election.

    Bert- your turn!

  5. I’ll have to go back and read what I wrote, but I think I listed every category of “late decider” separately, in addition to providing a rough estimate of the aggregate. I call shenanigans on the “arbitrary aggregation” accusation.

    It’s true that late deciders in past elections have gone marginally for one candidate or the other. The questions I think we’re debating are 1) whether late deciders are systematically predictable, and 2) whether this predictability could lead us to the conclusion that they would overwhelmingly support McCain this year. As I understand Matt’s (pre-poll closing) position, it was that late deciders would go overwhelmingly for McCain. Exit polling shows that they haven’t – Obama won one category of late decider by five percent, and McCain won the other two categories by six percent and two percent. Hardly an overwhelming margin, and one that might be characterized as appearing random or unpredictable.

Leave a Reply

Your email address will not be published. Required fields are marked *