How to create a representative sample
Gaining a clear understanding of a particular audience’s views is important for many types of research, from psychology to marketing. But consulting an entire population is impractical.
The solution is to use a smaller sample. But how can you be sure this sample has the same demographic makeup? If the sample doesn't reflect the target population, the results won't be generalizable to the population at large.
That’s where representative samples come in. So, why are representative samples so important – and how can you gather them for your study?
This refers to a sample that matches the demographic distribution of the population being studied. So, for example, the sample has roughly the same percentage of women over 60 as the population you’re studying. This means the results are more likely to give an accurate snapshot of the population. In other words, using a representative sample makes your findings more generalizable.
Let’s say you’re conducting market research on a product. You want to know what the UK public might think about it. The problem is that a lot of people live in the UK. Creating a survey to reach every single person living in the UK would be an insurmountable task. Using a smaller sample seems like the best solution. However, this comes with risks...
The sample group might love the product, but when it comes out in stores, it barely sells. Why would this happen? Well, it’s likely that the sample didn’t reflect the target population. The sample was unrepresentative. Perhaps 80% of the sample group were men – which isn’t representative of the entire population of the UK.
A small, unrepresentative sample is likely to result in data that isn’t reflective of the population. Meanwhile, if your results are generalizable, you can begin to make inferences about the entire population.
Representative sampling can have real-world consequences. Take car safety, for example. Traffic researchers found that women are 73% more likely to be seriously injured in a vehicle crash than men. Why? Because the crash test dummies used in car crash testing were designed around the proportions of an average-sized man. Men and women have different physiques. But the test sample skewed male. So, the results don’t generalize towards the wider population – which leads to potentially dangerous outcomes.
AI research is another area where representative sampling is crucial. Using an unrepresentative sample to train AI could be disastrous. The impact of prejudice and bias on AI has already been widely documented. The more representative your sample is, the less likely that bias from sample participants will affect the data.
If you want to know how to create a representative sample, there are some key steps you should follow...
First, you need to understand your target population size. This will give you an idea of the kind of sample size you’ll need. For national populations, this is obviously a very large number. As you add more demographic factors to your study, you’ll need more groups of participants. Each of those participant groups needs to be big enough to represent the entire population. At Prolific, we offer representative samples of two national populations – the United States and the United Kingdom.
For our representative samples, we’ll take your chosen sample size and split it down into three key demographics. These are age, sex, and ethnicity. Then, using census data, we’ll divide the sample into a set of subgroups. We'll ensure these meet the same proportions as the national population.
Think carefully about the sample size you’ll need. The sample size will affect how accurately the sample reflects the population.
For example, in a sample size of 300, a subgroup that represents 0.2% of the population should technically be represented by 0.6 of a participant. In practice, Prolific uses an algorithm that assigns at least one participant to each subgroup. However, this means that smaller samples will not precisely match the national population.
Our minimum sample size is 300, while the maximum deliverable sample size we offer is around 2,000 for the UK and 1,500 for the US.
Once you have a clear understanding of the representative sample needed for your study, it’s time to launch it on Prolific. When you publish your study on the platform, eligible participants will be informed via email and the Prolific Studies page. Participants take part in a representative sample study on a first-come, first-served basis, if they fit the criteria. There are clear processes in place to ensure that studies are fairly distributed.
Representative samples can take longer to recruit for. This is because they specifically require participants from demographics with a smaller population. However, this type of study still only takes around 2-4 days to complete.
Another important consideration is reallocating. If a representative sample study is still awaiting submissions 48 hours after launch, there is the option for reallocation. This removes age bracket restrictions for any unfilled places within the sample. This means that candidates of any age can fill the role, if they meet the sex and ethnicity criteria. In some cases, reallocated places will be reassigned to participants of a different sex or ethnicity. This speeds up the data collection process but will result in a less representative sample overall.
Wondering how to calculate sample accuracy? You'll need to calculate the total number of subgroup requirements that participants met in the final sample. Next, divide this by the maximum number of requirements that could have been met. You can find more about how this works in our representative samples FAQ.
Representative samples mean that your studies can be generalized to the broader population. If you’re ready to get started, why not set up your study on Prolific today? With over 130,000+ participants on the platform, you’ll find trusted, professional candidates for your research. Plus, there are plenty of tools to make creating a representative sample easy.
Want to find out more about best research practices before you get started on your study? Check out our complete best practice guide to online research. This handy guide will help you get to grips with everything you need to know to create ethical, generalizable studies.
Fresh out of YC's Summer 2019 batch, we want to share some of our most interesting learnings. If you're a startup founder or enthusiast and want to learn about product-market fit, growth experimentation and culture setting, you're in the right place!
Today Prolific is turning 5 years old – Happy Birthday to us! 🥳 It's been a remarkable journey so far. 3000+ researchers from science and industry have used Prolific last year, we have 45,000 quarterly active participants, and we've seen 200% year-on-year growth. But we're only getting started. In this post, I'll tell you a little bit about our journey, give credit where it's due!, and tell you about our exciting plans for the future.