Don’t freak out over the Gallup spike (with graphs)
In a previous post, I explained why you can’t compare the daily tracking releases from Gallup and Rasmussen. To recap: Gallup uses a 7 day sample while Rasmussen uses a 3 day sample, so the daily results are for two different time periods. As a result, Gallup lags behind Rasmussen by 2 days when reacting to events. However, comparing the two polls by their midpoints is more useful, and helps interpret spikes like the 6 point Gallup margin from the last 2 days. For example, the Gallup sample from 20-26 Sept should be compared with the Rasmussen results from 22-24 Sept. The results from the two polls track closely when compared that way, as shown in the graphs below:
Looking at the results from that perspective, a few things jump out. First, if you ignore the convention swings, both Obama and Romney have returned to the same level of support they had before the convention. Second, the Gallup spike from the last 2 days is a huge departure from previous Gallup trends. Third, the Rasmussen swing over the same period is much smaller, suggesting that the Gallup spike is a statistical outlier and not a true representation of the population. Fourth and most important, because Gallup lags Rasmussen by 2 days and Rasmussen has stayed flat, it is likely that Gallup will return to a much closer margin over the next few days. Statistical outliers normally regress to the mean over time.
However, because Gallup uses a 7 day sample, it will take at least 6 more days for the huge spike over the last few days to filter out. So until then, ignore the bump and remember it will likely all be moot after the debate next week.