We find no difference in the average accuracy of IVR and traditional human polls, but IVR polls conducted prior to human polls do significantly worse than traditional human polls even after controlling for characteristics of the states, polls, and electoral environment. These findings provide suggestive evidence that pollsters may take cues from one another given the stakes involved. If so, reported polls should not be assumed to be independent of one another and so-‐called “poll-‐of-‐polls” will therefore be misleadingly precise.
Robo-polls: Taking Cues from Traditional Sources?
Remember a certain GOS suing Research 2000 for adjusting the poll results to match "expectations"? And winning?
IVR means "Interactive Voice Recognition", a.k.a. "Computer calls not human calls". Based on this Vanderbilt University study, the whole IVR industry is doing something akin to that. More below the Orange Squiggle of Power.
The study looked at states where there were only IVR polls, and then at states where there were both IVR and human polls. In both types of states, poll results tended to converge on the actual results as the election date drew near. This is to be expected. But ...
In states with no human polls to use as benchmarks, the IVR polls did much worse on the average. In other words, IVR pollsters must be improving their results using the human poll results; they adjust what is reported to match the better polls.
For Nate Silver, this poses an interesting mathematical problem, not easily resolved unless you know exactly how much lying the various IVR firms are doing.
For the people who paid for these IVR polls, you may wish to contact Markos about the name of his lawyer.
IIRC, the justifiably infamous Rassmussen uses IVR techniques. And therefore the "Rasmussen is way out there and then starts reporting results similar to everyone else" effect may very well represent Rassmussen adjusting their adjustment to bring themselves inline with everyone else.
Looking at the TPM poll tracker for likely voters, using automatic and telephone polls gives Obama 48.7, Romney 44.3. Removing the automatic polls gives EXACTLY the same results.
Sometimes a difference which makes no difference suggests a difference.