As discussed in the Daily Caller, I believe that pollsters are under-sampling Republican leaning constituencies. However, at the same time they are going to incredible lengths to ensure that no Democratic leaning groups are left out. A demonstration of this can be found in Mark Blumenthal’s complaint that Gallup has a pro-Romney house effect because they are excluding non-telephone households, which decreases the minority sample.  Blumenthal’s piece is instructive in the care that pollsters take to get an accurate sample. It is also instructive in the institutional bias created when pollsters pay special attention to  ensure that Democratic leaning groups are fairly represented.

For Blumenthal, losing a fraction of one percent from a minority sample was an example of a questionable poll. Cutting the the evangelical vote by 25% from 2008, however, does not appear to concern him.

Many pollsters also believe that a portion of their interviews must be conducted using cell phones to capture voters who do not use landlines. This makes sense. However, as “cell phone only” voters have a pronounced Democratic lean this consideration also helps fortify the number of Democrats in the sample.

While it is understandable that pollsters would use census data to obtain accurate samples, a problem occurs when they only fortify Democratic leaning demographics. Republicans can be under-sampled because pollsters refuse to build a sample basement for GOP constituencies. In fact, the buttressing of Democratic leaning groups likely ensures that Republican leaning groups are under-sampled. For example, Marist did not seem to notice or care that registered Republicans actually outnumber registered Democrats in Iowa.

When you add everything together you get the very real possibility that pollsters are talking to a false electorate.