Friday, September 07, 2007

Cell Telephones and Polls

The primary season is over and the general election campaigns began in earnest after Labor Day. Now we are going to be bombarded with polls touting how all the candidates are doing. Indeed, has put up the first general election poll I have seen on any general election race, the race between Jamie Franks and Dewey Bryant.

Back in the 1940s, 1950s and 1960s, pollsters actually used to go to the person's residence and interview him (or her). (Indeed, Gallup Poll interviewers had to lug with them briefcases and wear a big button on the left side of their shirt or dress saying "Gallup Poll."). Except for 1948, when pollsters stopped polling weeks before the Presidential election, the polls were very close to the actual result. In the 1970s and beyond, the pollsters stopped the personal interviews and started using telephone interviews. After all, almost everybody had a listed telephone number and mobile phones were rare. But it was until the end of the 1990s and the beginning of the 2000s that poll results were far different among different pollsters.

One reason is the rise of cell telephones. As you know, cell phones are NOT listed in any telephone directory (Well, at least not yet.). You don't get the telephone solicitors or any advertisements when you have a cell phone. I have a Cingular cell phone under the family plan and an Alltel cell phone under the zone plan. I have a land phone but I very rarely use it. I thought of getting rid of it. However, I live right near the airport. If there is bad weather and the cell phones don't work (Alltel doesn't work even in the best of circumstances. Cingular works very good.), I need the land phone.

So let's say you are a pollster and you want to contact me. You can't. The phone calls on my land line are all telemarketers. I have an answering machine that screens the calls. If it's somebody I know, I'll call him on my cell phone (I'm under the measured dialing plan under Ma Bell for my land phone. Why waste the minutes?). I give my clients and political friends my cell phone numbers. I receive no important calls on the land phone.

The bookkeeper/accountant in our firm has only a land phone. Ditto for her husband. Almost all of her friends have cell phones and no land phones. When I worked contract at a large CPA firm doing very complex personal tax returns, ALL the twenty somethings have cell phones and no land lines. How does a pollster contact them?

The demographics are very simple. The young have cell phones and don't have land lines. The thirty somethings have cell phones but may have land lines, but mainly cell phones. The forty somethings and fifty somethings have cell phones and most have land lines. The retired may have cell phones but most of them have land lines. Another problem is people may have land lines only for emergency purposes but use their cell phones as their main telephone. And what if you have an answering machine that screens the calls? They may not want to talk to a pollster. So any telephone poll is going to be skewed towards oversampling the elderly.

Oh, sure, I hear the pollsters adjust the samples because of the cell phone usage. But in the next ten years, estimates show five to fifteen million users will dump their land lines and use only cell phones. How do you adjust the polls for them? It won't be the young people in that batch, but the forty somethings on up that will be dumping their land lines. Just like the pay phone, the land phone might be a relic in the Smithsonian in about twenty years.

I read in about a pollster that called Dewey's victory over Charlie Ross by a two point variation. I wrote in my diary what I THOUGHT it would be and I was off by one point. That same pollster said he called it for Fordice's first run in 1991. There were a couple of polls that showed Fordice within the margin of error.

Review the national polls from 1996 on for Presidential, gubernatorial and Congressional races. This is about the time cell phones started hitting the market. You will see how radically divergent the polls are. I'm not going into the details because I don't have the time to research the polls. But I remember reading in many newspapers and political publications how wildly divergent the polls were.

I very strongly suggest you take polls with a grain of salt and put very little stock in them. For pollsters to be taken seriously in their data, two things will have to occur: All cell phone numbers will have to be listed in a telephone directory (That's not going to happen in the near future. One reason for having cell phones is keeping the phone number away from solicitors and "junk" calls. There'd be a lot of protest from cell phone users if their numbers were listed.) or the pollsters would have to revert to polling techniques they used in the 1950s--actually going to the person's home and interviewing him.

No comments: