What is a representative sample and why would I want one?
- If you were a psychologist investigating loneliness (or any other construct), you might use a representative sample in order to better estimate how loneliness affects the entire country.
- If you were a market researcher collecting feedback on product concepts, you might use a representative sample to ensure your survey reaches participants from a range of backgrounds, giving you a better idea of how the public as a whole might receive your ideas.
- If you were a political scientist interested in the outcome of an upcoming poll, you might use a representative sample to collect opinions from a group resembling the country’s demographic makeup.
Ultimately, drawing conclusions from an unrepresentative sample limits the extent to which you can say “my findings are likely to hold in populations other than the one I’ve tested”.
How does Prolific’s representative samples tool work?
We currently offer representative samples of two national populations: The United Kingdom and the United States.
When you use a representative sample, we take the intended sample size and stratify it across three demographics: Age, sex and ethnicity. We use census data from the or the to divide the sample into subgroups with the same proportions as the national population. This means, for example, that a representative sample contains the same proportion of 28-37 year old Asian Women as the national population (or at least, as close as we can deliver).
- Currently we estimate our maximum deliverable representative sample size to be around 1500 for the UK, and 1000 for the US.
- The minimum representative sample size you can request is 300.
- It is not currently possible to use any in combination with a representative sample.
If you want greater detail on precisely how we deliver representative samples, read our
What are the limitations of a representative sample?
- The size of a sample determines the precision with which we can match the national population. For example, if a subgroup is 0.2% of the national population, and the sample size is 300, then technically this subgroup should be represented by 0.6 of a participant. Our stratification algorithm ensures every subgroup is represented by at least one participant, but naturally this means that smaller representative sample sizes do not match the intended national population perfectly. Furthermore, consider that a smaller sample size may result in a single participant representing an entire subgroup…
- Though a representative sample matches the distribution of the national population on age, sex and ethnicity, it does not mean the sample is representative of the national population on all demographics. For example, the participants in the sample will all be members of Prolific, which is obviously not the case for every person in the United Kingdom or the United States… We hope to add further demographics to our representative samples feature in the future, such as ‘Socio-Economic Status’, or ‘Highest Level of Education’.