In the past few weeks, I’ve explored how social media can be used in research planning and asked if social media mining was “real” research. Today, I turn my attention to using social networks as sources of research sample.
Can we use social networking sites as a source of research sample? Of course we can.
Should we use social networking sites as a source of research sample? That’s a more complicated question.
In many respects, it’s hard to see how using a social network’s membership is much different from using any of the large online research panels. Both provide a place to access large groups of people online. Neither social networks nor online panels are very likely to produce samples that are representative of the general population and any given panel (or social network) is going to have its own idiosyncratic biases.
On the issues of representativeness and bias, the most apt description of social networks and online panels is that they’re equal, but different. For any specific set of sampling requirements, it’s likely that one will be more suitable than another, but I don’t think we can say that any one social network or existing panel is a universally better option.
There is one aspect where social networks have a clear advantage.
Since people join a social network to interact with other people – often people they know in “real life” – there is a strong incentive to be truthful about the information they present. This leads to greater transparency with respect to respondent identity, the lack of which has been one of the major complaints about online panels.
When LinkedIn launched its research sample business, they touted exactly this differentiator. Because of its professional focus, we can probably assume that the incentive to share accurate profile information is particularly strong on LinkedIn. I think that the more “social” social networks like Facebook also have a strong built-in deterrent when it comes to fabricating profile details.
Another potential advantage of social networks over online research panels is size.
We have all learned from the early days of online panels that it’s not the number of panelists that matters, it’s the number of panelists who actually respond. The same is true if it’s a social network based “panel.” The difference is that the major social networks have orders of magnitude more users than the biggest panels. Assuming the same proportion of people who actually participate, social networks can provide a much larger pool from which to draw respondents.
Once we decide that social networks aren’t generally any worse than online panels when it comes to bias and representativeness, and may be better in terms of transparency and size, we need to address the issue of how to use social networks.
Three ways to use social networks for sample
One major obstacle to reaching people on many social networking sites is that they are “closed” communities, meaning that communication among people not connected in the network is generally restricted in some way. Three ways to get around that issue to recruit survey participants from social networks come to mind:
- Be part of their networks
- Build an app
Recruiting respondents from our own networks provides the greatest degree of transparency with respect to respondent identity, but it’s not realistic to think a sufficiently large network could be built to be an effective sample source for ongoing commercial research. An exception might be a company who can attract large numbers (thousands, at a minimum) of “fans” who could be used in research studies.
Perhaps the best method is to build some sort of application that network users could install on their page. I believe (though am not positive) that Peanut Labs uses this approach to build its social media sample. Most APIs allow applications to access some amount of user account information, so we can maintain transparency of respondent identity with this approach. Of course, the challenge becomes building an application that will be appealing enough for people to install and use.
The largest social networking sites offer the ability to purchase targeted ads based on user profile. This is, by far, the fastest, easiest way to recruit survey respondents from social networks. To the degree that the qualification criteria are targetable, you’ll have some confidence that only the “right” people are given the opportunity to enter the survey. However, much of the transparency is lost since profile information won’t be carried through to the survey. This approach largely nullifies the advantages of social networks.
Of course, none of these approaches do anything to control potential non-response bias and any bias that may be introduced because of the specific sub-set of the population found on any given social network. These same issues will be present regardless of whether you use an online panel or a social network.
What about you? What’s your experience sampling from a social network?
Next week, I’ll talk about using social media in qualitative research. Don’t miss is, subscribe to the RSS feed now.