Cellular Surprise for Political Polling?
Even as we enjoy new capabilities for gauging political opinion around the world, a nagging technology development haunts opinion polling here in the USA. The uncertainty suggests room for "surprises" come November.
Americans are increasingly ditching conventional telephones in favor of "cellular" and "internet phone" options. ("skype," vonage, etc.) 15% or more of the American voting population now uses cellular phones only. The trend away from land lines may be accelerating, now that major telecommunications carriers (like Verizon) permit customers to sign up for DSL (or FiOS) without having a conventional phone line.
Pollster are aware of the potential problems, but most polling has shied away from sampling cellular-only citizens. Why? Practically, it's considerably more expensive. First, there's the difficulty of accessing cell phone listings. Second, regulations forbid automated calling of cell numbers. Then too, what does a statistician do with those who keep their old cell phone numbers when they move to new locations? And most disconcerting, how do you convince someone to stay on the line to answer a survey, when that person may be paying 25-50 cents per minute for the call? Answer: -- you have to compensate them. (assuming subjects don't hang-up first!)
How do pollsters rationalize excluding 15% of the population? I gather that received wisdom deems the sampling problem to be theoretical -- that is, of no consequence, a "wash."
Yet we do have new data suggesting otherwise. Consider a recent Pew Research Center for the People & the Press analysis of three Pew presidential surveys that included cellular sampling:
In each of the surveys, there were only small, and not statistically significant, differences between presidential horserace estimates based on the combined interviews and estimates based on the landline surveys only. Yet a virtually identical pattern is seen across all three surveys: In each case, including cell phone interviews resulted in slightly more support for Obama and slightly less for McCain, a consistent difference of two-to-three points in the margin.
I'm not quite squaring the phrase "not statistically significant" with "virtually identical pattern." Furthermore, the study observes:
in each of the three polls, the cell-only respondents were significantly more supportive of Obama (by 10-to-15 percentage points) than respondents in the landline sample. For example, in the September survey Obama led McCain by a 55%-to-36% margin among cell only voters, but the candidates were tied at 45% in the landline sample.
Pew isolates "age" as the explanation for this considerable difference: "In large part, this reflects the fact that a substantial minority of the cell-only sample is younger than 30 - a demographic group that has consistently backed Obama this year."
If I were a pollster, I'd be starting to sweat. Some firms apparently are debating "adjustments" to account for the youth/cell factor - the one that supposedly is not yet "statistically significant." (If any jwn readers can explain that solution in "plain english," please chime in.)
By way of disclosure, I'm about to cancel my own land-line. I'm the last hold-out in my family. While I too am tired of the daily push-poll calls from this or that candidate (another subject!), my motivation is personal - I'm getting even with Sprint/Embarq for never delivering dsl. If the pollsters want to find me, they'll have to call my unlisted cell.
I got rid of my landline a long time ago, and have not missed it a bit. No one I wanted to talk to every called me on it anyway, so I never answered it.
What means statistically not significant ?
You have two polls here : one with cellular and one without. The results were different, but were they different because of the slight differences usually incuring during polls (aka sampling error in an assumed normal distribution) ? or do they correspond to a real difference between the two populations ? There are no definitive answers (the only definitive answer would be found in an exhaustive census or in the votes). So you have to base your decision on a statistical probability. Usually the results are considered to be really different (significant) if there are less than 5% chances that the differences are just obtained by case, that they result from the sampling error to be expected).
A difference is not significant if there are more than a 5% probability that the difference is due to sampling errors.
When the population is very large, the sampling error can be inferred from a simple formula (because the reference population is nearing infinite which allows simplifications).
In usual political polls, when nothing is precised, the confidence tests use the 5% limits. In other cases, where precision is much more critical, one could wish a greater precision of 1%.
To see the formula, look here :
http://en.wikipedia.org/wiki/Margin_of_error
You still believing in polls Scott?
So let read this and tell us if there is really financial crisis in US:
Buoyed by incentives that can be as high as $40,000 each, and coupled with the promise of thousands more for education or house down payments, the annual cost grew by 25 percent over last year's totals for the two services, The Associated Press has learned.
The money — particularly in these shaky economic times — has proven to be a strong recruiting tool, even as the U.S. death toll in Iraq surpasses 4,100 and violence in Afghanistan escalates.
As a result, the Army, Marines, Navy and Air Force all met their recruiting goals for the fiscal year that ended Tuesday. The Air Force did not provide bonus data.
Military recruiting bonuses grow by 25 percent
May be these polls also somehow had some money bonuses who knows this US, isn’t same nation believing in liars that they elected them twice, but they saying we oppose the war in Iraq?