Gallup, the Pew Research Center, and other US survey firms have historically employed all manner of strategies to gather vast amounts of opinion data. While mail-in and door-to-door polling were standard in the early- to mid-1900s, the 1980s saw an explosion in telephone surveys. According to Pew, telephone survey responses in the United States sat at a comfortable 36 percent in 1997

Over the past few decades, however, pollsters in this and other wealthy countries have been forced to confront a drastic decline in the public’s willingness to pick up the phone and answer survey questions. From 1997 to 2012, average response rates in the United States dropped by 27 percent and now stand shakily at a minuscule 9 percent. Add to this the rising cost of using telephones to conduct surveys, and it makes sense that most polls in the U.S. and other wealthy countries are now conducted online.

The advent of the Internet has dramatically shaped how pollsters do their work. People are still willing to answer questions online, particularly if they have already “opted in” to a pre-existing survey panel. Once substantial internet-based data is gathered, however, it becomes more difficult to compare with past data collected via telephone, as online and phone survey responses differ

One of the key issues is how to identify respondents and non-respondents. If you can figure out who is systematically more likely to answer an Internet survey, you can get a better sense of how representative your online survey really is. 

Response to internet questions is driven, first and foremost, by rates of web users worldwide. In the United States, for example, approximately 90 percent of adults use the Internet on at least an irregular basis. That’s almost 300 million users. Compare this to Uganda, whose 18.5 million Internet users comprise only 43 percent of the nation’s population. 

Using the same survey methods in the United States and Uganda will thus yield vastly different r results. If over half of a nation lacks reliable Internet access, web-based polls could significantly misrepresent the national population’s views. 

For this reason, door-to-door surveys and telephone polling are still used in countries with low online penetration. From 2011 to 2016, I commissioned a series of face-to-face surveys all across the world, asking survey companies to knock on doors and interview respondents in their doorways. We collected responses from 6100 people and published a book on the results.  

Internet surveys are still the wave of the future, however, and innovative companies such as Riwi Corp, a Toronto-based polling firm, are coming up with new ways to randomly sample online populations around the world. In my next post, I’ll talk about Riwi, which offers an exciting, proprietary method based on “random Internet interception.”