For many years, the gold standard of data collection in survey based research has been face to face interviews with a random sample of the target population (Yeager et al. 2009). A large dataset of such randomly selected respondents is still considered the most trusted source of information on a target population. This dominance of face to face research using probability samples has been steadily decreasing due largely to cost, practicality and speed of delivery. Increasingly there is pressure to move surveys to an online collection methodology. They are seen as more practical on these parameters. Some online surveys have been done with probability samples of the population of interest (e.g., Moskalenko and McCauley 2009; Skitka and Bauman 2008). Here evidence suggests that data collection from probability samples via the Internet can yield results that are equally or more accurate than RDD telephone interviews or face-to-face interviewing with area probability samples (e.g., Chang and Krosnick in press; Smith 2003) possibly as a result of more candid responses in the absence of the interviewer (Dayan et al 2007).

However, the majority of online surveys are carried out with respondents collected through non random probability methods. Specifically, most commercial online surveys are run through Online Access Panels. There has been much research published indicating problems encountered when trying to compare information collected via Access Panels to the general population. A large part of these problems are associated with the samples taking part in the surveys and can be seen by errors or biases in results.