Home > Plasmid: Science and bio-tech research blog > Archives > 2006 > January > 02
Monday, January 2, 2006
Tall glass or Tumbler? Take our poll…
In our never-ending quest to make you think, Plasmid wants you to pour yourself a nice drink and ponder the art of the public opinion poll. Go ahead, grab a glass and pour yourself about an ounce.
Did you pick a short, wide tumbler, or a narrow champagne flute? Does it really matter? Indeed, it does, but more on that in a moment.
Now that you’re relaxing, let’s talk about informal polls, analysis and scientific polls — otherwise known as lies, damn lies and statistics.
If you clicked on Plasmid from The Palm Beach Post’s “Scripps” page, you may have noticed our informal public opinion poll at the bottom of the page. It asks: “Where is the best place to build The Scripps Research Institute?”
Many people have asked why we haven’t printed the results of this poll in the newspaper. Still have that drink? Take a sip. Here comes the statistics part.
In case you don’t live in South Florida, this question has become important in the wake of a court ruling that makes it hard to continue construction of the publicly financed institute on a former citrus grove. Environmental groups are battling the development because it pushes intense real estate development next to a preserve.
When I last checked, according to our internet poll, 26 percent of respondents preferred the orange grove remain Scripps’ home — it won 553 votes. Meanwhile, 28.67 percent preferred the similarly located — but cheaper — Florida Research Park (598 votes), and 28.81 percent preferred the more urban Abacoa development in the Town of Jupiter. (601 votes)
So why haven’t we published the results?
Polls tend to fall into two categories: scientific ones and informal, unscientific ones. Scientific polls can be tremendously useful. They’re good at showing trends and patterns. Unfortunately, they’re also easily misused and misinterpreted. If we want to know how the public feels about the Scripps site debate, will our informal web poll give us useful information? The answer is most definitely not.
At this point, those of you with the champagne flutes probably need to refill your glass. More on that later.
To be scientific, a poll must take a representative sample, and that can be difficult in the best of cases. In the case of the Post’s Scripps page, the people viewing the page have already selected themselves into a very special sub-set of the general population: Those who are terribly interested in Scripps. Furthermore, people can vote in the web poll multiple times. So now what do we have? A poll that shows conclusively that some people are interested enough in the Scripps question to participate in an on-line poll. And that’s about it.
So the next time you see a web-based poll, or one of those television call-in polls, ask yourself whether the information they purport to offer means anything at all. In nearly every case, the answer is no.
A scientific poll will draw from a representative sample — that usually means large numbers. It will tell you the margin of error. And finally, it will allow you to see which questions were asked, and how they were posed.
The way a question is posed can have an enormous impact on the results. Too often, poll questions are cast a specific way to generate a desired answer. In polling, sometimes perception is everything.
Which brings us back to your glass. Anything left? If you’re drinking from a wide, shallow tumbler rather than the flute, there may well be. In the current issue of the British Medical Journal, Cornell University Professor Brian Wansink shows that people asked to pour the same amount of a drink into different shaped glasses are apt to overpour into a wide, shallow glass to the tune of 20 to 30 percent. “People who pour into short, wide glasses consistently believe that they pour less than those who pour into tall, narrow glasses,” said Wansink, an economics and marketing professor. Wansink says the the classic vertical-horizontal optical illusion is likely the cause: People perceive equally sized vertical lines as longer than horizontal ones.
So if you’re looking for good information from a poll, better look for a scientific one. Pay attention to the margin of error, the sample size and the questions posed.
And if you’re looking for a drink with exactly one ounce of liquor, better stick with a jigger.
So what do you think? Are informal polls useful? Or are they misleading? Do you pay attention to them?
Have a happy New Year.
