I am a research analyst and recently conducted an employee satisfaction survey at the organization I work for. Ideally, I should have received completed questionnaires from all 380 employees (making it more like a census rather than a survey), but I got responses from 204 people, that is, just slightly above 50%. Am I right that I cannot treat the received responses as a "random sample" of all employees?

  • 1
    $\begingroup$ Welcome to CV. Survey nonresponse is a well-developed topic in sampling theory, particularly as it relates to corrections for this. The key question is the original sampling frame for your survey and how employees were selected for participation. If you know this in a probabilistic sense and if participants were randomly selected from that list, then the received set of responses remains a random sample subject to the correction(s) for nonresponse. In the absence of a probabilistic sampling frame, you have a convenience sample no matter how you look at it. $\endgroup$ – Mike Hunter Nov 17 '15 at 15:20
  • 1
    $\begingroup$ Yuri, do you have descriptive information about the employees and can you compare it to your sample? By making the comparison you can at least know if your sample is skewed in one way or another. Also, can you tell us more about how your survey was distributed? That would help address some of the questions posed by DJohnson. $\endgroup$ – Ben Nov 17 '15 at 15:37
  • $\begingroup$ DJohnson and Ben, thank you very much for your comments/clarifying questions. The sampling frame consisted of all employees of the organization (excluding the top manager and his 3 deputies), and the goal of the survey was to collect responses from (ideally) all employees - so no random selection was performed. The questionnaire was uploaded to the Google Forms and then employees were provided with a link to complete it online. Three reminders were sent via email throughout the data collection stage (resp. 1, 2 and 3 weeks after the survey had started). $\endgroup$ – Yuri Nov 18 '15 at 16:09
  • $\begingroup$ Ben: yes, some basic descriptive information on employees' is accessible - such as gender, job position (manager vs. rank-and-file), employee's job location (central office vs. regional units). $\endgroup$ – Yuri Nov 18 '15 at 16:12
  • $\begingroup$ A little unclear what you want as an answer here. The obvious one is "You can so treat them, but your conclusions will be liable to non-response bias" - but I suspect you already know that. As @DJohnson says, it's a well-developed topic, & you may find it useful to search the site to see what's been asked before, e.g. Addressing Non-response in a Convenience Sample, & then, if necessary, formulate a more specific question. $\endgroup$ – Scortchi - Reinstate Monica Nov 19 '15 at 13:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.