Welcome to the Listening Post!

Age Question

Tweet

 

Yesterday's Pluggers cartoon made fun of the age-range demographic question (please go read it; I don't want to violate their copyright by embedding it here). 

You wouldn't think this was a good idea for a comic strip, but it actually made the Top 10 Plugger comic strips for 2008.  Clearly it hit a chord with many readers.  When counseling new survey authors, I always advise them to avoid showing any bias, but showing an age range of "56 and over" reveals an interesting lack of perspective, grouping 70 million U.S. adults (almost a third of all American adults) into one demographic.

As I've written before, in general, you should avoid asking a respondent to select their age range.

• Ranges can offend the respondent. A 56-year old would not think that he had much in common with a 90-year old, yet he is grouped into the same category in the survey referenced by Pluggers.
• Ranges are arbitrary.  I could easily subdivide the "56 and over" into four segments (56-61 years old, 62-67 years old, 68-75 years old, 76+ years old), each with different attitudes and approaches to work and retirement, but you could probably come up with your own reasons for dividing it differently. 
• Ranges limit your analysis.  You can't report the average age of respondents from such a question.  You can't group the respondents into generational cohorts using a standard list of age ranges; for instance, Baby Boomers are part of the 36-45 age group, all of the 46-55 age group, and only some of the 56+ age group.

So what should you do?  Check out "Asking the Age Question in Mail and Online Surveys" (PDF) by Benjamin Healey and Philip Gendall of Massey University, who tested three different methods of asking the age question: 

1. Asking for the respondent's date of birth
2. Asking for their current age in years
3. Asking for their age within a series of ranges. 

Here's their conclusion:

The best advice for survey research practitioners is to ask respondents when they were born; either their date of birth or the year in which they were born. This format appears to work well in all survey modes, is parsimonious of questionnaire space, is easy to administer, and generally produces low non-response and high accuracy. The other piece of advice specifically for online survey researchers is to avoid drop-down response menus.

I would take this advice one step further and only ask the year of birth:

In what year were you born?  ____

If this was a web survey, I'd use a text box rather than a drop-down list and validate that the answer was between 1895 and this year.  (The world's current oldest person, Tomoji Tanabe, was born in 1895.)

You lose a small degree of precision asking the year of birth over asking the actual birth date, but you remove a degree of risk about privacy.  Governor Palin's email account was hacked by someone who used her birth date and zip code to gain access.  Marketers can use birth date, zip code and gender to uniquely identify individuals in many zip codes.  As a result, I expect more and more respondents to opt out of providing full birthdates, no matter what their age.

Sometimes, as in this case, the simplest way to ask the question will give you the best results.  And it will keep your survey from being lampooned in a comic strip!