If you love data, take a peek at DataFerrett, a data analysis and extraction tool for US census data. On this site, you’ll be able to calculate for yourself that the US population is 51.1% female, 16.8% aged 12 or under, 12.2% single person households, and 82.7% non-Hispanic.
Obviously, the best marketing research panels should look like that too. Or should they…
Let’s first think about the kinds of people that brand managers and marketing researchers are interested in listening to. Brand managers want to listen to people who use their products, as well as people who ought to use their products but don’t. Brand managers want to understand why women use a certain brand of lipstick or why men prefer a certain type of razor. Or, they want to understand why younger people prefer certain video games or what older people like about certain brands of vitamins. Given that many products are designed for specific demographic and psychographic groups, brand managers are rarely interested in hearing from a census representative sample.
Perhaps political polling would be better suited towards a census rep sample? Perhaps yes, but still no. Pollsters only want to hear opinions from people who are old enough to legally vote and who have the citizenship rights to do so. That excludes children, teenagers, non-citizen immigrants, and a few other groups of people who are valuable, contributing members of our society.
If a research panel was built to census representative specifications, we might find that single person households rarely get invited to answer surveys whereas people in large households with significant purchasing power get more invitations than they could ever complete. Or, that older people rarely get invited whereas younger people, who are just starting to learn which brands they will use for the rest of their lives, receive more invitations than they can handle. Why would we aim to build panels comprised of many people who are rarely invited to complete surveys and few people who are massively over-invited to surveys?
So what does the ideal panel look like, if not census representative? It’s quite simple. The ideal panel is one where, regardless of their demographic characteristics, every panelist receives the same number of survey invitations. As the panel provider learns more about its clients’ research needs, they should grow the panel in the demographic areas that are in highest demand and slow the growth in the areas of lowest demand.
When most panels can and do easily create census representative samples via project specific sampling and targeting, it simply makes no sense to force it on the entire panel.