The first questionnaire was published nearly 200 years ago, starting in England in 1838. Like everything else, surveys have changed quite a bit over the past 200 years. But how?

“The Era of Invention” 1930s-1960s

Using surveys and questionnaires for business and marketing purposes didn’t come around until the 1930s. This is when the idea of random samples to represent an entire population first came about. Jerzy Neyman, a Polish mathematician, first suggested in 1934 that sampling in this way could make accurate, unbiased predictions that were measurable. His article on the matter swept the statistics world by storm and led to the creation of the modern survey.

In those days, the samples were primarily separated by geography, such as rural versus urban, two different states, or different neighborhoods in the same city. Mailed questionnaires and face-to-face interviews were the most popular ways to administer surveys. Because taking surveys was so new, the response rate was through the roof. There was more of a concern about contact rate, or being able to reach enough people, than a rejection rate, or how often the surveys would be turned down. This was because response rates, or how many people answered a survey, were a whopping 90%. It was a golden age of sampling.

However, the science was still new and had its flaws. For face-to-face interviews, interviewers would sometimes classify respondents’ age, race, or economic status on sight without confirming with the respondents first, leading to biased data. They also found that interviewers tended to avoid houses and respondents who looked lower class, essentially rejecting their opinions. This bias led the Literary Digest in 1936 to mistakenly claim that Landon won the presidential election over Roosevelt because it had accidentally over-represented middle- and upper-class voters in its poll.

“The Era of Expansion” 1960s-1990s

The era of invention was ended by one critical invention: the telephone. As the telephone left upper-class households and spread across the country, it allowed surveys to spread further, wider, and easier than before. Samples could now be separated into lists of random phone numbers, and respondents could be reached with a single phone call. The invention of the computer also revolutionized how surveys could be conducted and their results compiled. By the late 1960s, many surveys were run by Computer Assisted Telephone Interviewing (CATI) programs, which led to a boom of small surveying companies and academic centers.

With this growth, though, came growing pains. Response rates plummeted, as respondents could simply hang up the phone if they didn’t want to participate in a survey. Costs began to skyrocket as government, academic, and private surveys all began to contend with these dropping rates and encourage respondents to be a part of their sample. At the same time, the culture of surveying shifted. Founders of private sector companies moved away from focusing on new methods of surveying and onto ways of making the company grow. Academic and government surveys saw an interest in analysis over sampling rise. In all sectors, it became more about the results than the process.

“The Era of Designed Data” 1990-present

In the 1990s and early 2000s, sampling became increasingly harder. Face-to-face interviews were nearly a thing of the past, and caller ID led response rates to drop even further. Cell phones made the problem worse, as portable phones meant the caller’s geographic location could be completely wrong. Mail-in surveys became the safest sampling method, but it was also met with low response rates.

The rise of the internet brought sampling back. At first, the new technology led to some confusing choices, such as mailing out URLs on letters to take specific surveys. Soon, though, the volunteer Internet panels, where groups of people agree to take surveys in exchange for compensation. Costs dropped, and turnaround was quick. However, with the internet’s faster, cheaper sampling comes more risk.

Fraud is at an all-time high, and the fear of hackers stealing personal information looms over everyone, not just survey companies. With advancements in AI technology being made every day, it’s getting harder to verify who and what online is real. Additionally, for the first time, sampling has to deal with separating useful information from basic data. With so many different ways to track things online, there’s never been more bulk data in history. Separating the useful data from the rest, though, can be a challenge, as more data is constantly pouring in.

What will the next era of sampling look like? With how much data is being compiled from the internet every second, it is possible that the concept of survey panels, like those first face-to-face interviews, becomes a thing of the past. Alternatively, we could see a crackdown on all that information being taken for free, and survey panels could boom. There could also be a rise of AI-generated respondents, taking humans out of surveying entirely. Who knows? Maybe one day, we’ll see a return to those first mail-in surveys.

The information in this blog was complied from "A Companion to Survey Research" by Michael Ornstein and "Three Eras of Survey Research" by Robert Groves.