Slashdot Mirror
Will Cellular Phones Skew Survey Results?
FriedEgg writes "Recently, many people have started to forgo traditional landline telephones in favor of cellular phones only. This presents a problem for telephone pollsters who are prohibited by the FCC from calling cell phones with automated equipment, and from calling people for whom receiving the call costs money. While they recognize the exclusion of cellular only users can skew their results, they're not yet sure how much... because they're unable to survey cellular only users to find out their demographic information. Some evidence does indicate the frequency of cellular-only is highest among 18-24 year olds, traditionally the hardest to survey anyway. If the problem grows worse, it's possible we could end up with a "Dewey Defeats Truman"-like situation where the telephone poll results were skewed because Truman supporters were less likely to own a telephone."
...Of you who slept through History class, 'Dewey Defeats Truman' refers to the 1948 presidential election in which it was projected a man named Dewey would defeat encumbant president Truman, due to skewed mail-in poll results.
Link is to Historybuff for more info. Dewey Defeats Truman
Edward@Tomato - /home/Edward/ man woman
man: no entry for woman in the manual.
"Qua!?"
Haven't the Americans just got a 50 million list of folks who don't want to be called.
Do the folks who fit those two categories have a common demographic?
Sigs. We don't need no steenking sigs.
I would have to say that there is already an effect from the number of people with Caller ID or who screen their calls with answering machines. I know I don't bother picking up if there's no caller info or it says "GALLUP POLL".
The question for the market research is:
Does the universe of people who have cell phones correlate to the universe of respondents that I'm trying to reach or does it introduce bias? (I know just enough to get myself in trouble here...)
e.g. if 55% of your population is A, and 45% is B, and the cellphone population is also 55% A / 45% B, then it will have no impact on the results. OTOH, if the cellphone population is 75% A / 25% B, then there is going to be bias that will have to be corrected for in order to extrapolate back to the main population. (Guaranteed there is some wrong terminology there...)
Wolde you bothe eate your cake, and have your cake?
The 18-24 demographic is the one least likely to vote anyway
The problem is that many of the people in this business aren't mathematician and if they are then they are at best only second rates ones.
This poblem doesn't appear only in public survey - in fact most "statistics" in natural and political sciences are usually fucked. Most mathematicians don't care - we know that significance tests don't really give any viable results etc. So if people aren't interested in maths or aren't willing to invest time to understand it, then let them believe their rubbish.
Nevertheless all this whining about "skewed" surveys is meaningless, it's like complaing that the moon is made from bad tasting cheese.
Owner of a Mensa membership card.
The actual Dewey Defeats Truman headline was based more on early election returns than skewed polling. The famous error made concerning polling in the 1948 election was that Gallup simply stopped doing polls two weeks before the election and proclaimed that Dewey would win, not taking into account the massive sea change in public opinion that can occur over a two week span.
The famous example of a poll gone bad because of telephone ownership statistics was the poll used by the magazine The Literary Digest to predict the 1936 presidential election between FDR and Al Landon. The magazine mailed ballots to 10 million individuals with listed phone numbers and/or car registrations, and tallied up roughly 2 million returned ballots. Based on the results, the magazine predicted a Landon victory, while in fact the result was a 46-state landslide for FDR (and remember, there were only 48 states in 1936)!
The sample error in this survey was that telephone ownership and car ownership did not correlate to likeliness to vote in 1936. In 1936, at the height of the Great Depression, telephones and cars were luxury items that few people besides the wealthiest Americans could afford. The poor and lower class who were more likely to vote for FDR were not even sent ballots, so there was no way for their voices to be heard in this survey.
The impact of this flawed survey was such that within a short time after the publication of the survey, The Literary Digest would go out of business.
Blogging Weight Loss, Distance Education, and more at verlin.com