Saturday, December 12, 2015


So the big news out from Iowa comes from an unscientific survey conducted by Selzer & Co on behalf of The Des Moines Register and Bloomberg Media. The pollster claims Ted Cruz is in the lead for the Iowa Republican Caucus pushing past Donald Trump.

Named The Iowa Poll, the claimed results get based on on telephone interviews with 400 registered Iowa voters who say they definitely or probably will attend the 2016 Republican caucuses. However, the poll results are invalid and the method used, unscientific. Here is why.

The pollsters only completed 400 surveys at random a paltry too small sample size of 2,635. However, there are 1,682 precincts electing delegates to 99 conventions, one for each county in Iowa. It is at these conventions that delegates are bound to vote for candidates in proportion to the votes cast for each candidate at the caucus sites.

To conduct a scientific survey with results that can be projected to the universe of delegates in Iowa who must vote for the candidates running for the GOP presidential nomination, 99 surveys must be completed and not one. The execs at The Des Moines Register and Bloomberg have elected to use the wrong method likely because they do not know anything about public opinion polling using the scientific method.

For any survey of all Iowa adults, to get a margin of error ±4% at the 95% confidence level requires the completion of 784 surveys taken at random from a suitable sample. However, 99 surveys need to be done rather than one survey. Because each county in Iowa has a known adult population, is it is easy to adjust the necessary completed surveys (for example, see FluidSurveys for the math). For most counties, that number falls between 500 and 600.

The jokers at The Des Moines Register and Bloomberg claim their survey has a margin of error ±4% at the 95% confidence. Yet, they only completed 400 surveys. Even for a statewide survey, they would need to complete 784 since the adult population of Iowa is 2,380,059.

So for a scientifically conducted poll with results that can get projected to the universe of voters, a survey firm would need to complete a total of 56,322 surveys and do so all within the same time frame before the next news cycle to get results at the 95% confidence level with a margin of error ±4%.

So for example, if some of the survey respondents completed a questionnaire before a candidate said something controversial, say freeze all immigration from countries with muslims, while other survey respondents completed the same questionnaire afterward, the results could not be used since the opinions are collected on different facts. Time-sensitive surveys like voting surveys must be done on the same day for all respondents to be accurate.

Since few firms, if any have such resources, all should reject the thoroughly unscientific fraud being perpetrated by The Des Moines Register and Bloomberg.

Here is how the survey sizes break down for each county convention.

Disclosure: I used to work as the director of public opinion polling and marketing research for a Dow Jones owned newspaper. My grad work at S.I. Newhouse School of Public Communications included public opinion polling.

No comments:

Post a Comment