Survey researchers are, of necessity, perennially concerned about the validity of survey questions. It is often difficult, if not impossible, to establish conclusively the validity of many attitudinal and behavioral questions, but sometimes invalidity is blatant. For example, in every National Election Study conducted by the University of Michigan Survey Research Center (SRC) there has been a consistent over-report of voting behavior. When the SRC's turnout rate is compared with the official election returns, the survey derived data continually exceeds the actual results, usually by more than 10 percent. In 1968 Aage Clausen said "the consistency, as well as the magnitude" of the discrpancy makes it difficult to explain away by sampling or resonse error."
"Where and Why are Misreporters Misreporting?,"
Sigma: Journal of Political and International Studies: Vol. 2
, Article 3.
Available at: http://scholarsarchive.byu.edu/sigma/vol2/iss1/3