Table 1.

Summary of Techniques to Address Nonresponse Bias

Technique	Description	Pitfalls to Avoid	Reference/Further Reading
Comparison of sample and population	Comparison of the sample to known characteristics of the population. Often only demographic characteristics will be compared	If demographics are not the core variables tested in the model, simply comparing demographics of the sample to the population may not provide a valuable evaluation of nonresponse bias	Armstrong and Overton (1977) (See also Groves 2006; Beebe et al. 2011)
Follow-up analysis	A sample of nonrespondents are surveyed on substantive variables in the study to determine if any meaningful differences exist between those who did complete the survey and those who did not	Nonresponse may also be high in the follow-up survey, making it difficult to assess nonresponse bias	Sosdian and Sharp (1980) (See also Groves 2006)
Wave analysis	Respondents who completed the survey prior to the deadline are compared with those who completed the survey after the deadline (or in response to a reminder)	This technique does not actually assess nonrespondents, as the second (or later) wave participants did indeed respond to the survey. Thus, although assumptions can be made regarding nonresponse bias, it cannot truly be assessed	Ellis, Endo, and Armer (1970) (See also Filion 1976; Fitzgerald and Fuller 1982; Lin and Schaeffer 1995; Mazor et al. 2002; Yessis and Rathert 2006)
Passive and active nonresponse analysis	Meaningful differences exist between those who actively choose not to participate in a survey and those who are passively nonrespondent. For active nonrespondents, researchers can examine a random sample of the population through focus groups, interviews, and a very brief survey to assess whether they intend to complete the survey, and if not, why not. For passive nonrespondents, simply resending the survey might address this issue. Researchers may also consider including items in the survey that might assess causes of passive nonresponse (e.g., workload)	Those who were opposed to the survey to start (active nonrespondents) may not be likely to participate in an interview or focus group. Repeated surveys may help with passive nonrespondents but will probably not lead to full participation	Rogelberg and Stanton (2007) (See also Beebe et al. 2008; Peiperl and Baruch 1997; Rogelberg et al. 2003; Roth 1994)
Interest-level analysis	Interest in the survey topic is associated with a greater likelihood of responding. Researchers may consider asking questions about the level of each participant's interest in the survey and, assuming an adequate level of variation in those responses, statistically control for interest level when conducting the analyses	This technique does not actually examine nonrespondents and thus may not directly address the issue of nonresponse bias	Rogelberg and Stanton (2007) (See also Groves, Singer, and Corning 2000; Groves and Peytcheva 2008; Rogelberg et al. 2000)
Benchmarking	Benchmarking findings against other published data, examining whether the descriptive statistics (e.g., means, standard deviation, etc.) for those measures are consistent with previously published studies	If one is looking at a specific group because he or she believes the group is unique and will differ from the population, then such a comparison becomes problematic	Rogelberg and Stanton (2007) (See also Asch, Jedrziewski, and Christakis 1997; Cummings, Savitz, and Konrad 2001; Sitzia and Wood 1998)
Replication	Researchers may attempt to replicate the findings. Similar findings in multiple samples would suggest that nonresponse bias is not a significant concern	Replication may be more costly than trying to raise the original response rate and it is difficult to determine if nonresponse bias is acting in the same way across multiple samples	Rogelberg and Stanton (2007)