Pollster’s trendlines are making a splash on twitter today, since they now show Romney and Obama moving into a true tie at 46.1 percent apiece. But there haven’t been very many recent polls, and several seem to show Obama doing pretty well. So what’s moving Pollster’s trendlines into a tie? Oddly enough, it’s the absence of recent polling that’s allowing Romney to surge. Without new surveys, the daily or weekly tracking polls—which as a matter of coincidence happen to show a tighter race than the pollsters releasing surveys every month—dominate the Pollster trendline, which is designed to respond to the most recent numbers.
Check out the last 11 polls entered into the Pollster dataset. Out of those eleven, nine are tracking polls and four are from Rasmussen. Although this isn’t exactly how it works, you can essentially think of the Pollster average as consisting of nine trackers and two more traditional pollsters. As I wrote this week, I don’t think this balance is appropriate. The four tracking polls are just four out of fifteen or twenty firms conducting national polls every month, and so they should only represent approximately one-quarter of a given polling average. Since there haven’t been very many polls recently, they’re essentially representing nine-elevenths of the pollster average. Just for good measure, Rasmussen is four of the eleven polls. They’re the most pro-Romney pollster, and they should probably represent only one-fifteenth or twentieth of a polling average, not one-forth.
|Most Recent Polls In Pollster Dataset|
The tracking polls haven’t been great for Obama recently. PPP and YouGov/Economist have shown a tighter race than they did a few months ago. But most other pollsters are showing Obama with a clearer lead. For instance, a tighter race in PPP and YouGov/Economist should be balanced out by Pew and Democracy Corps. Given the stability of the race, it doesn’t make sense to give Rasmussen or Gallup more weight, just because they poll more often.
Correction, 8/3: Pollster only represents a firm once a week, which is inconsistent with my characterization that Rasmussen represents 4/11ths of the Pollster average. While this measure does limit the influence of a single pollster to an extent, it is not enough to prevent what we're seeing today.