Our Services | Flashpoll | Westpoll | Other Services

About Westpoll | Current Month’s Questions | History of Voting Intent

WestPoll

How Westpoll is produced | The Survey Theory | Question Reliability | Sample Bias | Sample Error

How Westpoll Is Produced

The WESTPOLL material is generated exclusively for the West Australian Newspaper. It is conducted as part of Patterson Market Research’s twice monthly telephone Omnibus survey - FLASHPOLL. That survey is conducted on the first and third Mondays and Tuesdays of each month throughout the year (24 surveys are conducted a year, 12 of which contain WESTPOLL questions). Each survey covers 300 metropolitan and 100 country interviews with people aged 18 yrs or over. Top

The Survey Theory

Probability surveys are just that. Estimates of the "true" world by means of a probability survey in which a sample of the population is asked for their opinions by means of the survey questionnaire.

The three broad issues that we have to consider when constructing our surveys are:

Question Reliability

The first and most difficult issue to quantify is the survey error caused by poor understanding of the question by respondents. That is, the survey does not measure what it purports to measure due to some miscommunication between interviewer and respondent. Patterson Market Research’s senior consultants have a combined 30 yrs experience in questionnaire design. During the Company’s last 14 years, our guiding philosophy has been the principle of keeping the questions simple, with minimal respondent load. We believe that our questioning technique clearly communicates with respondents, enabling us to accurately test public opinion on a wide range of subjects.Top

Sample Bias

The second major factor in survey design is the management of possible sample bias, considering that only a proportion of the population is surveyed. If there were a significant bias in the way that people were selected for the survey, then there would be some prospect of sample bias affecting the result.

The survey conducted for Westpoll has developed techniques over the last decade of on-going monthly opinion polling, to achieve what we believe to be the highest standard commercial polling format possible. The format is set out below:

Firstly, the sample is generated by a random selection from the white pages CD-ROM. Then numbers from that list are phoned in an attempt to make contact with the target household. Up to three calls are made to an unanswered number before it is substituted with another number.

The survey is generally conducted on the first Monday and Tuesday of each month. Monday and Tuesday evenings have the highest "at home" rates of the week. The survey is over a minimum of two nights to enable the "call backs" to unanswered numbers to be spread over at least two work sessions.

Where more than one eligible voter lives at a contacted number, we use a simple random selection technique to choose the target respondent from the household. If that person is not available when we call, up to two further calls are made to appointment, to contact that person when he/she will be at home.Top

Sample Error

The third component of survey design, which has the potential to influence survey accuracy, is in the area of sample size. Our sample uses 400 interviews each month, (300 metropolitan and 100 country) which produces results that are accurate to within about +/- 2% depending on the actual survey estimate.

The table below shows the high and low ranges for our survey estimate of 82% "yes" at the 70% 95% and 99% levels of confidence.

The first row of figures above means that we are 70% confident that a survey estimate of 82% is between 84% and 80% in the "real world".

The second row indicates that we can be 95% confident that the survey estimate of 82% is between 86% and 78% in the real world, and the third row indicates that we may be 99% confident that the "real world" figure is between 87% and 77%.Top