noise convert

Anyone who checks pollster, fivethirtyeight and realclearpolitics a minimum of 15 times a day will have heard a lot of talk about "sampling noise" over the last week or so. The concept is that, shockingly enough, random samples of several hundred of people or so may show slight differences that do not actually reflect any national trend. As this idea is primarily invoked to dispute the notion that Obama's lead is in the polls is shrinking, my immediate reaction to any references to such "noise" is something along the lines of "SHIT, the lead IS narrowing and the liberally bent websites like pollster and fivethirtyeight are trying to deny the reality of the imminent doom the impending election will bring when McCain some how manages not only to hold on to all the Republican states, but also miraculous to win Pennsylvania!" However, after yesterday's panic (spurred primarily by the Rasmussen poll shrinking to 3 points, the closest it's been in over a month), the polls seem to have returned more or less to normal. Obama's back to leading McCain 51-46 in Rasmussen, bringing it back to the range where it's been for the past 35 days or so. Zogby shows Obama ahead by 7 (50-43), up from their 5 point margin yesterday. his lead in the traditional voters model on Gallup is also back up to 5 points (50-45).

The fact that Rasmussen shows Obama's lead returning to a 5-6 point margin is particularly relieving. They are one of the more reputed pollsters, they use the largest sample size, have had the least fluctuations and Nate Silver of fivethirtyeight considers them to be one of the best (certainly the most reliable daily tracker), which surely has some effect on my opinions. Moreover, Rasmussen remains the only site I have visited recently (perhaps ever) that regularly displays McCain/Palin advertisements, giving me a strange sense of confidence in the accuracy of reports I read of Obama gaining or holding steady in the polls. So, in conclusion, I am beginning to have some degree of confidence in this whole noise theory and will try, in future posts, not to react in sheer panic to slight day-to-day changes which occur only in some portion of the polls.