Friday, August 7, 2009

Poll Methodology

Photo credit: Adam Tinworth

There's been an excellent discussion going on over at The Intersection about a piece of poll data that we posted last week regarding how many Americans could name a living scientist.

We released that single slide from the data set as a preview of more data that will be coming out from our June 2009 survey. When the rest of the data is officially released, we'll be able to discuss it in more detail.

In the meantime, however, we wanted to share a little about the methodology of the polling, which seemed to be the cause of some concern.

Research!America has been commissioning public opinion research for 17 years, including both national and state polls (in 45 states!). The polls have been conducted both by telephone and online with well-established firms such as Charlton Research and Harris Interactive.

Here's a look at the general methodology for this polling:
Telephone (random digit dialing) polls are conducted with a sample size of 800 to 1000 American adults for a sampling error of +/- 3.5%. Data are demographically representative of adult U.S. residents.

Online polls are conducted with a sample size of 1000 to 2000 from a randomly generated pool of American adults for a theoretical sampling error of +/- 3.1%. Data are demographically representative of adult U.S. residents.
We've asked "Can you name a living scientist?" and other similar questions throughout the years, and therefore have multiple "glimpses of public opinion in time" on this subject, so we feel confident that the 65%/35% split is a good representation.

For those who are concerned about the presentation of the data, the graphic indicated 278 total mentions and not the total number of respondents to the survey. Those 278 represent the total number of responses of the 35% of Americans who said they could name a living scientist.

The commenters brought up a number of other interesting points, and we are energized by this conversation. When we get the greenlight to share the rest of the polling data, we'll have more to say about how and why questions like this are asked in national polls, and some analysis of the data.


  1. You only listed a single n value in the graphic. The n value, which is what you said the 278 represents, typically includes all responses.

    Also, this post doesn't address a problem with Mooney's and a few other bloggers interpretation of the data. I'll repeat:

    All those who named deceased scientists should not be counted because their answers are outside the parameters of the experiment, thus their data cannot be interpreted. In other words, we cannot simply put them in the 82% category, quite simply because they may have misread the question.

    Who’s to say those who named Einstein don’t also know a living scientist?

  2. That is a valuable point. Polling does have limitations, one being that there is always the chance that respondents may not understand the question.

    This is why, although we editorialized that 80+% of Americans can't name a living scientist, the graphic clearly shows that only 65% of Americans responded that they could not name a living scientist.