Content area
Full Text
Survey researchers face a fundamental cost-bias tradeoff in respondent recruitment. Fielding opinion polls on high-quality probability samples entail substantial costs that limit the scale and breadth of research activity. Recognizing the limits of using university-based convenience samples, many researchers have examined whether web-based crowd-sourcing tools like Amazon’s Mechanical Turk and Google Consumer Surveys can be cost-effective methods to survey diverse target populations (Berinsky, Huber and Lenz 2012; Huff and Tingley 2015; Santoso, Stein and Stevenson 2016). The empirical properties of these Internet-based samples remain a subject of active study.
In this article, we evaluate Facebook’s potential as a platform for survey recruitment. We propose a quota sampling method using Facebook advertisements to generate public opinion estimates that approximate national averages efficiently. Using a proof-of-concept study on US climate opinions, we demonstrate that researchers can cost-effectively recruit respondents through quota sampling using Facebook advertisements, at a fraction of the cost of hiring an online survey firm (we sampled over 2000 respondents at about $4 per response). Using Facebook for quota sampling also offers a comparative advantage because it gives researchers control over how they recruit subjects and allows researchers to target specific subpopulations.
Our method contributes to an emerging literature that uses Facebook to recruit respondents into research studies. For example, psychologists enrolled over 4 million Facebook users as subjects by creating a Facebook application that allowed users to take psychometric tests (Kosinski et al. 2015). The researchers subsequently linked these tests to users’ Facebook profiles to predict personal attributes and personality traits from users’ social media behavior (Kosinski, Stillwell and Graepel 2013; Youyou, Kosinski and Stillwell 2015). Medical researchers have similarly used Facebook to recruit subjects from specific subpopulations, such as young adults who smoke cigarettes (Ramo and Prochaska 2012; Ramo et al. 2014) or middle-aged American women (Kapp, Peters and Oliver 2013).
Political science researchers have used Facebook to recruit respondents in Brazil for survey experiments (Samuels and Zucco 2013, 2014), survey likely voters in state primaries (Hirano et al. 2015), deliver political advertisement treatments to diverse respondent pools (Broockman and Green 2014; Ryan 2012), and survey political activists in Germany and Thailand (Jäger 2017). In Samuels and Zucco (2013, 2014), researchers raffled off an iPad to one randomly selected survey-taker. For a cost...