When undertaking research one usually aims to reach a particular sample size (n=#). This post will explain why, and a few key things to consider.
Note that this is only really relevant for quantitative research (e.g. surveys) where you aim to report the percentage of people who have an opinion or engage in a behaviour. This is not relevant for qualitative research, which focusses on stories (not percentages).
Why do we think about sample sizes?
The ‘sample size’ is the number of people who complete the survey / participate in the research. When doing a survey with the community you will rarely be able to speak to every single person (unless it is the Census). Because of this, there will naturally be opinions and behaviours missed in the results. The more people you speak to the more opinions and behaviours your findings will cover, resulting in increased accuracy.
The stats gurus worked out a formula to help identify how many people are needed in a sample size to enable you to be relatively confident that your findings cover most opinions and behaviours. This formula works by considering the population (total number of people who could complete the survey) and error margin/confidence interval (what sort of variance you are happy to have).
The best practice ‘error margin’ is +/-5% at 95% confidence. This means that if your survey shows 50% of people undertake a behaviour, you can be 95% confident that if you had surveyed everyone in your population the result would be between 45% (-5%) and 55% (+5%) showing that behaviour.
So what sample size should you aim for?
For any population over 200,000, you need a sample size of n=384 to have a +/-5% error margin.
You can figure out the number for a population by using the calculator on this site: https://www.surveysystem.com/sscalc.htm
Use the ‘Determine sample size’ box and put in the population of your area then set the confidence interval to 5 (this means the error margin is +/- 5% which is industry standard).
The number in itself is relatively meaningless if the method used to reach those 384 people isn’t representative. For instance, if your sample is made up of 350 females and 34 males, your %s are not going to be reflective of the whole population, even after considering error margins. One of the biggest errors Councils make is to just distribute their survey through comms channels (in which case you only get those engaged with Council / interested in the topic, who aren’t representative of the broader community).
At ASDF Research we have developed a really good method for getting proper representative samples for Councils. Get in touch to find out more!