Election Musings: the 2000 U.S. Presidentials

Well, I've enjoyed this window into the U.S. political system. It has brought good discussion among parts of my email circle, from whom no doubt all my good ideas stem. So here are my musings, after some back and forth.

It seems pretty clear that the election is down to a coin toss. The margin of victory was lower than the margin of error, meaning the final result is essentially a random event. All elections have this sort of noise in them. What makes the 2000 US Presidential election stand out is that the margin was close enough that we get to see just how much noise there is.

We discover, for instance, that in Palm Beach County and Duval County, there were about 20,000 double-punched ballots each (and Duval county has only about 2/3 the population of Palm Beach County, is strongly Republican, and seems to have used a traditional ballot design). So while the candidates fight over 300 (now 900) votes, there are 40,000 people in two counties alone who took the time to vote, apparently meant for their vote to be counted, but made mistakes.

Keep that in mind: the margin of victory is possibly an order of magnitude or more less than the noise. That means that there is no "real result" of the election, all rhetoric aside. However, elections are contests, not just estimates of votes. They are games with rules and, usually, only one winner. Normally the victory is pretty clear. This year, however, the margin is narrow, and so the rules-lawyers step in for the endgame. It happens in sailboat races and court cases, so why not elections? In the end we have to choose a winner. The vote count did not make the choice clear, and the apparent loser was not gracious enough to concede. (Well, at first he was, but that's another story.) So we decide by technicalities. It is ugly and unsatisfying, but appropriate. That is what technicalities are for. It seems to me that the system is functioning as it should, even if the candidates are not.

As long as the eventual victor does not claim a moral victory or a mandate, and accepts the fact that his win is as valid and binding as a coin toss, we should be alright. Laura thinks it would be appropriate to form a coalition government. If I thought the people involved could work together in a civil manner, perhaps I would agree.

Differences of Opinion

I talked with some of my family over the past week. That has been an interesting experience. It had not occurred to me that some ordinary people could see Bush's "don't count" tactics as legitimate, or his reasons as genuine. Although it seemed to me that he was the most likely winner, I could see no justification for obstructing recounts, nor any basis for his rejection of hand recounts, given the known anomalies (and especially now that it seems he supported hand recounts in Texas). And of course the claims that the country would dissolve into chaos if the election weren't resolved this week were completely bogus: after all, the electoral college doesn't vote for weeks, and the votes are not tallied until January 3rd.

And yet my family thinks that it is clear that the Gore supporters have been altering ballots, and that they want manual recounts just so they can do so. They also seem to see Bush as the innocent and passive recipient of abusive court challenges.

Weird. Either we are hearing very different stories, or (one or both of) our interpretations of events are way too subjective for any kind of analysis of the actions of either side.

The Duval Data

So while the election battle goes to the courts, the results give us a good chance to look at the noise that is in the ballots. A website with election data was brought to my attention because of a mildly alarmist analysis available at: http://www.netrinsics.com/Duval/Duval.html

The "bottom line" of that analysis appears to be that there may have been selective vote counting in Duval County.

These charts clearly illustrate a strong correlation between Democratic precincts and missing votes for a presidential ticket. Almost all precincts with more than 50% Al Gore votes have more than 15% missing presidential votes. Almost all precincts with more than 50% George Bush votes have less than 15% missing presidential votes.

Interestingly, there were more valid votes for Senator than for President in all precincts, implying voter confusion across the board, but again, the discrepancy was far higher in predominantly Democratic precincts.

Now that would be disturbing. Election bungling is expected, although not quite on the scale it was observed. At any rate, it is forgivable and not culpable. If the ballots in Palm Beach County were misdesigned but not out of malice, that is unfortunate, but we can live with it. (It would of course be fairest to revote -- not just one county either -- but that seems to be impractical.) I'm sure it happens every year; it just happens to matter this time.

But if there were selective vote counting, that would be cause for concern. So what is the data which leads people to conjecture such things? You can see it on the Netrinsics site, above, but I found it a bit hard to interpret their graphs. I have re-plotted their data in a way I hope will be a little easier and make the effect of interest a little clearer. (Click on the chart to download the Excel file with the election data and several graphs, including this one.)

The data points are the individual precincts in Duval county. I calculated the number of missing votes by subtracting the (total number of Presidential votes) from the (total number of ballots cast) for each precinct. I don't know if that is how they calculated the numbers on the Netrinsics site. I suspect so: there is no "missing votes" column in the data. The formula is sensible, but to be clear: it says nothing about missing or invalid ballots, only about valid ballots with missing votes.

The effect is that there is a correlation between the percent of voters who voted for Gore, and the percent of ballots with no Presidential vote. (The correlation runs the opposite way for Bush and indeed for all the other candidates: the normal pattern is that the percentage of ballots missing a presidential vote either remains constant or declines.)

The sinister interpretation is that in predominantly Democratic precincts in Duval county, Presidential votes are being ignored. Well, that is one possible interpretation. It cannot be justified from the data alone, although the data may prompt a re-examination of the actual ballots. (Yes, that would amount to a manual recount. Obviously it has to be done by a bipartisan or neutral party. Maybe we could send it to Serbia.) There are other possible interpretations.

I sent this site to several friends, who responded with rather cogent reactions, I thought. I calculated the graph above after reading Robert and Gregg's comments, and give my reactions in the third column. I have paraphrased Robert and Gregg. Click on the headers for their full messages.