Polling for Dummies

Over the last couple months, there has been a great many poll numbers flying about, with this candidate up and that candidate down or vice versa.  Now, without getting into the actual mathematical and statistical details, these polls, contrary to popular belief, do not give you a precise answer.  They may not even give you an accurate answer.  To understand what they are really telling you, you  have to understand the lingo.

First of all, there is the part that everyone likes to tout (or not) and pretend they understand.  For example, today’s Rasmussen poll has Dick Sue-em all


Pardon… has Richard Blumenthal up on Linda McMahon, 53 to 46.  Now, what Rasmussen is *NOT* saying is that this is tomorrow’s outcome, guaranteed, don’t be afraid to wager the housepayment on this answer.  (Side note:  while American casinos won’t take these wagers, betting parlors in the UK will — for those interested in the line on this race (and others), click here).

That having been said, if you look at the bottom of the page, you will find the following text.

This statewide telephone survey of 750 Likely Voters in Connecticut was conducted on October 31, 2010 by Rasmussen Reports. The margin of sampling error is +/-4 percentage points with a 95% level of confidence. Field work for all Rasmussen Reports surveys is conducted by Pulse Opinion Research, LLC. See methodology.

Now, to put this into layman’s language, Rasmussen, throught their agent, asked 750 “likely voters” how they would vote.  Now, they determine the likelihood not by simply asking the individual “Do you plan to vote,” but based on their assumptions, their expertise and a series of screening questions.  From their “methodology” link above:

All Rasmussen Reports’ survey questions are digitally recorded and fed to a calling program that determines question order, branching options, and other factors. Calls are placed to randomly-selected phone numbers through a process that insures appropriate geographic representation. Typically, calls are placed from 5 pm to 9 pm local time during the week. Saturday calls are made from 11 am to 6 pm local time and Sunday calls from 1 pm to 9 pm local time.

After the calls are completed, the raw data is processed through a weighting program to insure that the sample reflects the overall population in terms of age, race, gender, political party, and other factors. The processing step is required because different segments of the population answer the phone in different ways. For example, women answer the phone more than men, older people are home more and answer more than younger people, and rural residents typically answer the phone more frequently than urban residents.

For surveys of all adults, the population targets are determined by census bureau data.

For political surveys, census bureau data provides a starting point and a series of screening questions are used to determine likely voters. The questions involve voting history, interest in the current campaign, and likely voting intentions.

Now, once they have their 750 “likely voters,” the poll their poll results are expressed with their caveats, including an error range, in this instance plus or minus 4, and a confidence level, in this case, 95%.

This means that there is a 95% chance that the answer lies somewhere between Blumenthal 57 (53+4) , McMahon 42 (46-4) and McMahon 50 (46+4), Blumenthal 49 (53-4) and a 5% chance that the answer is something else.

Ultimately, there is only one poll that matters, and that is the one that occurs today, where the only questions are in the polling “booth,” everyone who responds is a “likely voter” and, after the vote is tallied, we argue about which side hired the better lawyers.  Cynicism aside, go out and vote, people, if only to have a legitimate complaint about the outcome.  Those who don’t vote have no reason to complain and the joy of democratic elections is that the people always, but always get the results they deserve.

1 reply

Comments are closed.