Sunday, August 23, 2009

Online Surveys: Are They Reliable?

There remains a great deal of debate in the public opinion research industry about the reliability of so-called online research. Online research is typically when a panel is recruited by a company or group of companies to regularly participate in opinion or market research. Each panelist volunteers to participate in research studies conducted by the company or companies. Then, a sampling of the people who have opted-in to the online research panel is selected to participate in any one of these surveys. I don't believe the industry should express paralyzing fear over this kind of methodology and at the same time I join many others who have significant concern about the methodology and how it is promoted.

There are two main reasons one should be skeptical about online panels. The first is when the researcher promotes the results as a representative sample of the broader audience and includes a margin of error. The American Association of Public Opinion Research has some very interesting thoughts on this subject. In fact, the first sentence of the organization's narrative on online surveys reads "The reporting of a margin of sampling error associated with an opt-in or self-identified sample (that is, in a survey or poll where respondents are self-selecting) is misleading" (http://www.aapor.org). Like AAPOR, I believe reporting the results of an online survey as if they are somehow reliant on a normal random sample of a (target) population is misleading. These online surveys are sampling a database of people who have volunteered to participate in these kinds of surveys and not a random sampling of the population whose opinions you seek (i.e. voters). If the survey is not a random sample survey then one can not, and should not apply random probability and calculate a margin of error. Most of these online panel surveys are not based on a scientific random sampling but they are promoted as if they are, which is troubling. At best such surveys will suffer from unknown bias and could possibly skew towards what some have said is a younger, more politically active audience unrepresentative of the broader population. Particularly as a stand-alone methodology, I tend to agree with noted pollster and regular contributor to www.pollster.com , Mark Blumenthal who said "I'd just be, I'd be more cautious about online surveys." Secondly, online surveys are notorious for producing wild results inconsistent with traditional survey methodologies. Again, I will quote AAPOR's website to conclude my concern with these kinds of surveys.

"Surveys based on self-selected volunteers do not have that sort of known relationship to the target population and are subject to unknown, non-measurable biases. Even if opt-in surveys are based on probability samples drawn from very large pools of volunteers, their results still suffer from unknown biases stemming from the fact that the pool has no knowable relationships with the full target population.
"AAPOR considers it harmful to include statements about the theoretical calculation of sampling error in descriptions of such studies, especially when those statements mislead the reader into thinking that the survey is based on a probability sample of the full target population. The harm comes from the inferences that the margin of sampling error estimates can be interpreted like those of probability sample surveys."

Now let me explain why I don't believe the industry should be incredibly fearful of this type of methodology, generally speaking. Technology is creating incredible accessibility to vast numbers of people which can only improve the way opinion and issues research is conducted. In other words, the industry is evolving and the eventual outcomes should be embraced after the reliability and efficacy of such advancements is determined. There will be a time when Blackberries, iPhones and other personal communications devices will have features and applications that allow for research of the user's opinions. We need to maximize the best way to use these technologies to serve our clients and the public while not sacrificing reliability or misleading the end consumers (clients, news media, general public). At MBE we are working on some exciting technologies that can enhance how we conduct our research, and anyone in the industry worth their salt should be doing the same thing. Online surveys are not completely "bad." For instance, I believe as a census survey of a particular audience they can be fairly efficient and cost effective. For example, if a membership organization has a great many members, the organization has access to their email addresses and the ability to design a questionnaire and analyze the results, then an online survey can be appropriate. Especially when the results are not promoted as a random sampling of some broader audience complete with a margin of error. For example the membership organization could conclude that X% of its membership answered a certain question a certain way and that would be appropriate. Also, I believe statistical analysis always calls for a degree of practical thought. The more completed interviews the membership organization received the more reliable the aggregate answers would be (i.e. 400 respondents is more reliable than 40). Making strategic membership decisions based on those 400 respondents would be appropriate.

In the end, online surveys promoted as a scientific random sample surveys are probably not a good idea.