I remember my freshman college political science textbook saying it takes two things to make a good poll: a representative sample and the right questions. It's easier said than done. The sample has to be broad enough to be statistically significant, and a truly representative one is almost impossible to create. Phone surveys, for instance, favor people who are at home during the day, own landlines, and tend to answer the phone. Pollsters try to compensate for this problem by using formulas that clean up the data and extrapolate trends, just like college professors who bell curve. But it's no better than tilting at windmills. Polls are often wrong, sometimes embarrassingly so.
Pollsters treat polling more like a science, when it's really an art. The question is, why pretend it's something it isn't?
Urtak, a collaborative public opinion site launched last year, offers polling without presumption. There's no mathematical posturing. The polls created on the site are no better than the professional ones, but they're also no worse. The genius of Urtak (pronounced OOR-talk) is that instead of resisting the inherent flaws of polling, it embraces them. It's a more honest way to poll.
The principle behind Urtak is simple: anybody can write questions, and anybody can answer them. All questions are yes or no, or don't care (so unpopular questions can be removed), and they appear in a random order, so nobody can cherry-pick. Instead of using the paternalistic approach of "scientific" polling, Urtak is crowdsourced.
You can find out all sorts of things on Urtak: how many people skateboard (10 percent), how many wear glasses or contacts (58 percent), how many are male (30 percent), how many think Obama will be a two-term president (78 percent), how many think Chicken McNuggets are delicious (53 percent), how many have vomited at the sight of other people vomiting (25 percent), how many have petted a llama (50 percent), and how many may be lying (33 percent).