…here at The Moderate Voice, an interesting analysis of the “house effects” titled “Which way do polls lean?” is well worth reading.
A second, longer discussion can be found at this link, with the following listed as the conclusions:
Specific polls do appear to have consistent biases on specific questions.
The amount of lean can exceed the nominal MoE (margin of error), which only measures sampling error, not errors from systematic biases.
One should focus on overall patterns of movement in the polls rather than fixate on any single poll.
The last conclusion is sensible advice for anyone who wants to comment on polling results, but I suspect that the tendency to “see only the good news for my side” will continue to rule the commentary in weblogs.
—
Cross-posted to Random Fate.