Outlier?

Not content to limit their war to just the media, the McCain campaign has now taken to pushing back against polls as well.  In their refutation of the just released WaPo poll that put Obama up by nine points, Team McCain’s leading pollster dismissed the data point as merely an outlier.

But is it really an outlier?  That’s a trickier question than one may think, and one that can rarely be answered with 100% certainty in the present.  That’s to say that on occasion data that may appear to be an outlier because it doesn’t track well with known trends may not be an outlier but instead an early indicator of significant movement.

It is because every outlier carries with it the potential for real change in the mood of the electorate that we can’t definitively label every unkind data point as an outlier.  But we can ask a few questions to determine the likelihood that it is such a statistical animal.

The first question to ask is if it is in keeping with current trends.  I make a point of mentioning trends because statistical error gives one a wide berth or a narrow berth depending on your intent to compare the specific numbers.

So we have this WaPo poll that is suggested may be an outlier, but how does it stack up to the actual trends?  Heading on over to Pollster, what we see is that the two candidates are only split by about four points, but the current trends indicate that that gap is still widening.  Perhaps they aren’t with the WaPo poll, but they do show that the gap it shows is not unattainable in coming days.

Second, we want to see if there are any other recently released polls that might back up the data.  This is pretty simple; if you have one data point that seems off of the norm, but there are other data points that confirm, it doesn’t mean you have an outlier, but instead you have data that is coming in faster than the current trends.

In this regard it feels curiously unlike a coincidence that FOX News of all places just dropped a poll that has Obama up by six points.  Not quite as dramatic of a data point as the WaPo poll, but dramatic nonetheless.

At this point, you may be wondering why I’m not including the daily trackers in my analysis.  After all, Gallup daily only has Obama up by three points.  Daily trackers tend to be more useful for trend analysis, but because of the three day rolling average (that is, every data point is a compilation of three different interviews with the oldest being dropped off on a daily basis) are less useful in catching an early indication of a radical shift in public opinion–the old tends to mute new data, meaning that you’ll have to wait a few days for the daily tracker to catch up to the individual polls especially when the trend starts moving fast.

On another note, the daily trackers also are going to be difficult to include in this analysis because just as we’re about to get useful (for this discussion anyway) data we’re going to have another major election event in the debates.  Thus, in guaging whether this is an outlier or a real trend shift, there will be a rather small window during which the daily tracker data will be useful without including post debate influences.

The final question I would have to ask to determine if a poll is an outlier or if it is indicative of a real shift is if there has been any major revelations in the campaign, or major political event that has taken place that might result in significant shifts in the public’s mood.  If nothing significant happens, normal campaign events and efforts could slowly shift opinion one way or another, but the daily slog of campaigning should not result in radical shifts of public opinion.

On the other hand, if something huge were to happen, something that really grabbed the public’s attention, that could feasibly alter the state of a race.  Now, has there been anything at all over the past few weeks that might qualify as such an occurrence?

Think, Kyle, think!  Anything?  I don’t know, I’m coming up blank…

OH!  HOW ABOUT THE WALL STREET MELTDOWN?!?!

I don’t know about you, but that would seem to me to be a major event that could influence the state of the presidential election.  Indeed, after making the claim that the WaPo poll is an outlier, Chris Cillizza notes that Bill McInturff, the McCain pollster, actually starts walking back his justification a little:

Despite that difference of opinion, McInturff did acknowledge that in the last week he had seen a double-digit increase in the number of voters citing the economy as the most pressing issue facing the country — the result of the ongoing financial crisis on Wall Street.

Do I think that Obama really has a nine point lead over McCain?  Probably not, but neither do I think the WaPo poll is necessarily an outlier either.  What I think is happening is that the election is in a serious state of flux, which means that trends are shifting, and they are doing so in such a way that is difficult to measure accurately.  It’s easier to measure a trend at rest than a trend in motion, and the past week and a half has created cause for a great amount of motion.

What I think we can see in the near term is Obama continuing to climb, but then come Friday evening, all bets are off as the state of the race will hinge upon both the debate performances, and the after debate spin.

Leave a Reply

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

Connect with Facebook