Skip to content

Some Employers Ruin Surveys For the Rest of Us

2010 September 2

This recent Dilbert comic caught my eye:

One of the biggest problems in organizational survey research, especially about sensitive topics, is that many employees do not trust that the survey responses they give will be held confidentially.

It’s led to a great number of techniques for minimizing impressions that surveys have anything at all to do with the company.  For example, if you were conducting a research project on ABC Co. about theft, it would be important not to use ABC Co. in materials.  Initial e-mail invitations should come from an independent research team with e-mail addresses unrelated to, the web address should not contain, the first page of the survey should have a color scheme unlike that of or any familiar company colors, a single graphic representing the company should be on the first page only to establish that this is company-sponsored (only if it really is – if the research is from a truly independent source, DON’T DO THIS), clear explanations of the anonymity/confidentiality of survey respondents should be given, and no references to the company should be made after that point, if at all possible.

Even doing this, some respondents may feel their responses are being forwarded to management.  Some researchers try to address this by asking “Do you think management will be able to link your responses directly to you?” and eliminating such people from subsequent analyses.  But thus we hit the major problem – if you thought management would be reading your responses, you’d probably say “No” to that too!

So far, we have no better solutions to this problem for survey research.  In fact, as long as there’s a person’s mind creating survey responses, this will always be a problem – at best, we can discourage such deception.  This is what leads some researchers to seek more “objective” measures, like absenteeism, units produced, weekly sales, and so on.  But there only so many things that can measured “objectively,” and many of these aren’t really that objective in the first place.

Previous Post:
Next Post:
One Response leave one →
  1. September 3, 2010

    This issue is pretty common for survey research. More sensitive research topics required more clever techniques to confront the confidentially issue.

    In the research lab we use social desirability scales. While they didn’t eliminate respondent tendencies to self-sensor, they did allow us to estimate (and correct) how afraid people were to respond accurately. The idea here is that some people will self-sensor or self-aggrandize more than other people.

    This technique could be used for large-scale surveys within organizations, but I have found that corporate clients are not enthusiastic about the idea. In fact, clients seem to be very suspicious of statistical correction techniques of any kind (in my experience), so they have only one weapon left to make sure survey data is not influenced by distrust. They must be trustworthy.

    If employees trust their employers, survey data will be less distorted.

Leave a Reply

Note: You can use basic XHTML in your comments. Your email address will never be published.

Subscribe to this comment feed via RSS