Statistical noise alert

9 October 2008, 1535 EDT


When a polling organization conducts a mere 300 interviews each day for a national tracking poll, this is what you get: the potential for one day of outlier polling to produce phony movement.

Or, to quote Brad DeLong (who we really should be linking to more):

…the Diageo/Hotline Tracking Poll [is] an undersampled daily poll designed to produce a whole bunch of spurious three-day climbs in one candidate’s relative vote share followed by a three-day decline so that reporters can trick readers into thinking that there are important pieces of news and trends in there.

Anyone want to bet on how long it takes for someone to declare, based on the Diego/Hotline poll, that the second debate swung support back behind Obama?