Survey USA -- cheap polls for TV

The SF CBS affiliate breathlessly reported last night that Bill Simon had pulled ahead of Davis in the CA governor's race. This new "lead" was about 1/2 a percentage point, well within the margin of error. But the real story is not the poll results, which run counter to every respectable non-partisan poll taken thus far, but the veracity of the polling company itself -- Survey USA.

The Toricelli campaign offered this devastating critique of the poll and its methodology (computerized response audience polling):

Social scientists have been sharply critical of CRAP polls for several justifiable reasons. One problem is the identity of the respondent. Your five year old can punch the keys on the telephone and voice her opinion, or your favorite uncle from Tulsa who happens to be around when the phone rings can have his say as well. Who really knows who’s on the other end of the phone? Real survey researchers call the practice the uncontrolled selection of respondents, and it can be a real source of error. It’s tantamount to standing on the street corner and haphazardly asking questions of anyone who happens by…

A quick Google search indicates that Survey USA targets its services mostly to low-budget local television news stations, and even promises same-day turnaround on polls (which should automatically ring alarm bells). The results bear witness -- poll predictions were 21 points off in the Texas senate primary, and 19 points off the California guvernatorial primary.

(I got the Toricelli link from MyDD's excellent critique of a Survey USA poll for the Texas governor's race.)

Posted July 31, 2002 08:42 AM | Comments (0)