The Tightening

The conventional wisdom is that Presidential races (and perhaps most races at all levels) get closer in the final week of an election. With Barack Obama holding a rather significant lead coming into today, exactly one week before election day, are we beginning to see that tightening? Certainly, if you look at Pollster.com and turn the "sensitivity" up to high, you get this picture:





It does look like John McCain is closing in a bit, and Obama is falling off somewhat. That seems to suggest that there is, at the least, a bit of tightening happening at the national level (the state level is another matter). But I want to point out a few things about the national "race."

Before get there, though, let's look at more fun graphs (one of my loyal readers, in particular, has been very positive about these graphs I make). I took all the national polls from Pollster.com and plotted them here based on a five-day rolling simple average (just like my state graphs):


The five day averaging smooths out quite a bit of the variation, and leaves me with the unmistakable impression of relative stability. There may, indeed, be a bit of tightening today but, even so, the range for each candidate has been really narrow. Obama's low of 49% came 27 days out from the election, and his high of 51.6% was 9 days out. That's a range of less than 3 percentage points (and in fact every day except for his high point was within the range 49-51%). McCain's support has ranged from 41.7% to 43.5%, also remarkably stable.

Now, as I mentioned, part of that stability comes from the fact that I'm using a five day average. For state polls, that's useful because there are, in general, fewer polls. For the national numbers, it may be better to use a shorter time span, say 3 days. Here's that picture, with the 5 day lines in there just for comparison's sake:

Not a whole heck of a lot of difference, though the recent tightening is a bit more pronounced. My point in posting this second version, with a shorter rolling average time-span is to underline the basic stability of this race over the past 20 days or so.

I have one other point to make. Ignore Senator McCain's level of support for a moment and focus on Senator Obama's. Using either the 3 day or the 5 day average*, its clear that Obama's support hovers right around 50%. That is actually the key takeaway from all this, and it's a point that doesn't get mentioned enough in the media.

The media and the talking heads usually focus on the spread, the difference between Obama and McCain in any given poll, and that's ok. That margin is a good indicator, especially in the early days of a campaign. But as we get closert to election day, if one candidate is consistently at 50% or above, the margin no longer matters. Don't forget, the final margin is actually irrellevant. We operate in a winner-take-all system, so what matters is who gets to 50% (or more like 49.5% assuming 1% to minor candidates). The bottom line here is that if Obama has a solidified support of 50% to McCain's 44 or 45%, and there only remains about 4 or 5% of voters who have yet to decide, then there's not enough "undecideds" left for McCain to win over. He would essentially have to win all of the remaining undecideds to have shot at pulling even (I'll write about my feelings on how the undecideds will break down in a later post).

This principle, to look and see if one candidate has crossed the 50% threshold, both nationally and on a state-by-state basis, is how you should read the polls for the next seven days. I'll apply my own advice to current polling in battle ground states in my next post.


*Incidentally, in addition to the 3 and 5 day simple averages, I also produced a few graphs using 3 and 5 day weighted averages wherein polls are weighted according to sample size (bigger samples are more accurate). For those of you who are interested, here's that graph: