About the Respondent :: Demographics

About the Respondent :: Demographics
(Return to Composing Questions )

How to select the right demographics
1. The key issue is always the value of the demographics in decision making. People should be prepared to make decisions based on differences due to demographics, otherwise there's no need to ask.
  1. For example, it is typical to ask about gender. But is your company prepared to take some action on attitudes or opinons that are more pronounced in men than in women? Are you ready to offer a training for female employees only? Are you considering a policy that targets one gender or the other? If you're just curious about whether men and women see the world the same, let me save you the trouble: they don't! Don't ask unless you're ready to take an action based on gender, which is a potentially delicate strategy.
  2. The same issues are true for asking about ethnicity. What will you do if Blacks have a different set of opinions than Latinos or Caucasians? For example, Black respondents are commonly more critical of their managers than are Latinos. Even discussing the potential cultural reasons for such a difference is likely to be a touchy debate. And for what purpose? Would you really consider a training program for just one ethnicity? Would you train managers to treat Blacks differently than Latinos? If you've asked about the experience of discrimination, then you would have to ask about ethnicity! But in that case, the survey itself has already put discrimination on the table.
  3. Asking about department or job level provides a distinctly different decision situation. It would make sense to find out if issues are common cross all departments or unique to 1 or 2 departments. It is easy to imagine targeting change efforts to only that segment of the organization effected. Problems more acute for a single job group could easily translate into developmental programs, coaching, or skill building.
  4. Just so you don't think the question is easy, let me offer a countervailing example. One of my clients was a company in the entertainment business. There was a large and vocal contingent of gays, lesbians, bisexuals, and transsexuals among the employees. There was a frequent and sometimes contentious debate about the treatment of people based on sexual orientation. To provide some empirical base for the debate, we asked employees to indicate whether they were straight, gay, or bisexual. Even though the bisexual group was a small minority that eventually had to be lumped in with the gays in the data analysis, it would have been politically incorrect to leave off that distinction in the demographics of the survey.
2. Be prepared to push back against the common request for seeing "just my data". Department heads or division heads often want to plan on a cut of the data that lets them focus on "just their group". What seems like a simple request, however, is sometimes a serious issue that should be addressed early on.
  1. It is seldom obvious on what level a survey issue should be addressed. Are Divison heads the right decision-makers for the survey data? Do you want each department pursuing separate strategies to deal with an issue? The answer is an empirical one, determined by the distribution of scores.
  2. Suppose, for example, several questions about employee satisfaction or morale show wide variance by department. Then perhaps each department should investigate further within their group to better understand what changes are needed. But if the scores were uniformly low across different departments, it would be best to pursue a company-wide solution (since the causes are more likely company-wide) rather than have each department develop a different approach. In that case, looking at the data by department would be a waste of time, since it would all be about the same.
3. Remember that you never want to display breakouts for small groups. If you have less than 10 members in a group, it would be dangerous to show their scores separately; it would violate their expectation of confidentiality. This often demands that you lump different levels of management, such as CXO's + VP's + Directors into one response option. If anyone asks for a crosstabulation (such as Managers in the Western Region), you could suddenly have a cell with very few members.
Some typical demographics
1. Job level: Executive | Manager or Supervisor | Individual Contributor | IT staff | Support staff
  NOTE: Notice the strategy here of identifying clumps of jobs rather than a long list of individual job categories. The idea is to identify a cluster of jobs that face a similar job situation, similar stresses, similar opportunities.
2. Site or region: Be sure to check the population numbers. You may have some sites or regions that are less than 20; a modest response rate could easily give you less than 10 respondents, which means you couldn't report their data separately. You may need to group sites or regions into some larger clusterings.
3. Department or Division
  NOTE: In hospitals department is often not appropriate; hospitals are notorious for dozens of tiny departments, sometimes with only 3-8 employees each. Consider some other dimension that might be more useful, such as level of contact with patients.
4. Tenure: less than 1 year | 1 to 3 years | etc.
  NOTE: You'll have to find the right ranges for your company. You want to create roughly equal groups that span the range of tenure. If there was a pivotal event in the company's history, try to use that as a cut point. For example, if you had a new CEO come on board 5 years ago, then use "1 to 5 years" and "5 to 10 years" as ranges. That way, you could cut the data by "Before our current CEO" vs. "Under our current CEO". Always include "Less than a year"; new employees are routinely happier about everything. They have a job! Life seems good.
  NOTE: Mergers create complex options for questions on tenure. Should I count my time with the merged company? If you don't allow that option, it seems like a discriminatory dismissal of the prior experience of some employees. If you do, then you will have people in the same tenure cluster (before the merger) that come from very different organizational situations.
5. After a merger, you might ask whether the respondent worked in one of the original company groups or the other.
  NOTE: This is a touchy question if you're trying to blend the cultures together. If you ask the question, people will argue you are reinforcing the divide; if you don't ask the question, you have no way of knowing if the two groups are really coming to a common view. You may need a preamble to the question, such as "This question will help us know how far along we've come in creating the single company we want".
How to introduce demographic questions
1. Put demographics at the end of the survey, never at the front. At the end of the survey, the respondent knows what they've said and whether they're uncomfortable being even potentially identified uniquely.
2. Establish a clear priority in favor of confidentiality and comfort. Some statement such as the following would be a good introduction:
  These questions will help us better understand the information you have already shared. None of this data will ever be used to uniquely identify an individual. If you feel at all uncomfortable answering any of these questions, feel free to skip it.