The checklist for sample quality
There are three basic principles, which should be adhered to when working with interviews and samples.
Basic Recruitment Principle
1) Random selection
All participants must be selected at random from the total population of the relevant target group. This means that if you need to conduct a social-economic opinion poll with 1,000 interviews, the respondents must be picked completely at random among Vietnam’s approx. 60 million citizens over the age of 18.
2) High participation rate
For an opinion poll to be accurate, it is a prerequisite that all types of people are being heard. The trick is to persuade as large a number as possible of the chosen candidates to participate in the survey. The remaining candidates fall into the categories “could not be reached”, “did not have time” or “did not wish to participate”.
An important factor is that the better the interviewers are trained, the fewer people will refuse to participate in the survey. Another important factor is that if the interviewers are paid on a fixed hourly wage and not on a piece-rate or bonus rates, the number of refusals can be kept down.
3) Declaration of statistical uncertainty
When basic principles 1 and 2 have been adhered to, i.e. when the selection is completely random and the participation rate is given, it is possible to mathematically state the statistical uncertainty.
Warning lamp – take your precautions here
Since 2000, a number of research companies have appeared, built up around a business idea, whose concept is to sell “quick and cheap” surveys. The idea is that you avoid the salaries for the interviewers and instead interview people via the internet. The respondents are invited through ads promising remuneration and prizes. The problem with this method is that you do not adhere to any of the basic principles.
1) Omnibus: Telephone interviews give the most representative picture, as everyone has had the same opportunity to be heard. The contact leads have been randomly selected among the entire population.
2) Panel and web: Web interviews conducted via a panel, which has been recruited by telephone, provides a reasonably representative picture, as everyone has been randomly selected and invited via a nationally representative omnibus conducted by telephone.
3) Web panel Web interviews conducted via a panel, which has been recruited through ads, give a skewed picture, which one can attempt to correct by using data weighting and special extracts through quotas on for example gender, age, geography and education. The all-important problem, however, is that the three basic principles (read above) cannot be kept, and that the respondents, who have signed up are motivated by the chance of winning a prize. This attracts a group of people who differentiate themselves from the rest of the population on several parameters.
Another significant problem, which needs to be taken into account, when you want to ensure the validity of your survey, is the design and application of the survey. You get answers to the questions you ask. The question design and interpretation of the numbers must be done correctly. It is important to know: How are the questions phrased? Which conclusions have been deduced from the numbers? Who paid for the survey?