By now, many of you have no doubt seen more presidential polling results than you ever wanted to see! But too many people equate internet polls such as the one featured above (from cheezburger.com) with scientific polls conducted by major news organizations. How can you tell which of them are valid measures of public opinion, and which are not?
Start with Project: Conducting a Poll in WorldView Software’s U.S. Government. Try it out, and then critically evaluate the results. Here are some guiding points for questions students should be asking about their data:
- What was the difference between the total population of people whose opinion they wanted to know and the number of people who actually answered the poll?
- Did people have a choice to take the poll or not?
- Could people respond to the poll more than once?
- Was one group of people more represented in the poll than another? Why or why not?
- How were the questions asked — did they notice they got a better response rate from one method over another, such as internet over in-person?
- How were the questions asked — did the questions use loaded language that hinted at the “right” answer?
- How were the questions asked — were they yes-or-no, multiple choice, etc.?
- and so on…
Then apply this methodology to the polls out there. Count the differences between this poll from Breitbart (warning: overlay popup ad):
and these polls, aggregated at RealClear Politics:
For a deeper dive into how polling works, visit the Pew Research Center’s Methods page. There, you’ll find all sorts of information, such as how to figure out a poll’s margin of error. Especially recommended reading: the overview “Flashpoints in Polling” by Claudia Deane, Courtney Kennedy, Scott Keeter and Kyley McGeeney.
UPDATE 12/06/2016: Very nice video from Scientific American explaining the math behind polling.
Preview WorldView Software’s programs for free at http://www.worldviewsoftware.com/preview/