Friday, December 29, 2006

musings and AM radio

Not that you ought to care, but when growing up my family's first car was a used 1973 Ford Pinto. That may seem like a straight line for a joke, but it's true. The price of Pinto's dropped through the floor when the gas tanks started to blow up. But one thing I remember was that it had just an AM radio.. and an AM radio with rather limited reception at that. Thus one of the few stations you would be able to listen to was KQV. That memory came back today just because I was listening to KQV for the first time in a long time today. Their news preamble must be the same as it has been for decades. What really caught my attention was their phone survey which they seem to do every day.

The topic today was "What do you think of Governor Rendell's performance?". Fair enough question. The results when I heard this were something like 480 or so saying "poor", a handfull saying "fair" and something in the 40's saying "good". So worse than 10-1 against the Governor. Now I know that these types of call in surveys are not surveys in any real sense and I think they had an honest disclaimer to that effect. But no matter what you actually think of the governor's performance, he did win state-wide and locally by 20 percentage points give or take. Some self selection in who listens and calls in is understandable, but supposedly measuring a 10-1 sentiment against a politician that won by a large margin is pretty screwey. Compared to a random draw from the population that voted 60% for Rendell, to get 90% against him in a sample of 500 or so people would be a zero probability event.

Purely speculation, but I have a hypothesis that I will never be able to test. This call-in survey was measured by calling into one number to register your vote for "poor" and another number for "good". Your 'vote' gets tabulated by calling in automatically, you dont need to talk to anyone. What I wonder is whether the numbers that represent which side of a question are consistent. i.e. would the 'liberal' answer always have one number and the 'conservative' answer a different but consistent number. If so I would wonder if some people are just in a routine of calling one number or the other without actually taking account of the question.


Blogger Amos_thePokerCat said...

The bias in the "survey", such that it is a survey, is in allowing multiple calls/votes for the same originating number. Another is that the audience for news/talk radio is shewed to the right, and older. Finally, KQV's ratings for Summer '06 are small, very small (0.9).(You have to select "Pittsburgh" from the pull down.)

My theory is the outcome of the "survey" is probably the result of one or two older people at home, rather than on a cell phone, venting after an election that made them unhappy, hitting redial until they got the desired result. Those "voting" on the other side, had no motivation to stuff the fantasy votes, since they were not discontent with the outcome of the actual election.

Wednesday, January 03, 2007 4:27:00 PM  

Post a Comment

<< Home