How many statisticians does it take to change a light bulb? One, plus or minus three. And how many survey methodologists does it take to determine a person’s age? Well, no punch line here, but it takes many more than you might think. Recently, one member of the AAPOR-net discussion group, an online discussion forum hosted by the American Association for Public Opinion Research, posted a query about the best way to ask for a person’s age on a survey. The question generated more discussion than any other topic over the last four years—far more than complex and controversial topics like probability sampling, online surveys, new technologies, advanced statistical methods, and so on.

Along with gender, age is probably the most frequently asked survey question for all types of marketing and social research and across all industries. It seems that by now we ought to know how to ask about it, right? But there is a surprising degree of variation and disagreement among research professionals, and for good reasons.

Four Sources of Error

Our best practice at Versta Research has always been to ask for year born. But the extensive online discussion reminded us that any approach should depend, in part, upon which of four important sources of error need to be minimized for the study being designed. These four sources of error are: age heaping, non-response, erroneous data, and imprecise data.

1. Age heaping results when people give an approximate age rather than a specific age. Because so many of us round to magic numbers like “0” and “5” data will often show more people who are 40, 45, 50, and 55, compared to people who are 41, 47, or 52.

Why would anyone give an approximate age rather than a specific age? Lots of reasons. Some people, especially older people, do not know their age without doing a mental calculation. Some respondents prefer not to provide an exact age, as it could more easily compromise privacy. And even if you ask about year of birth rather than age, heaping is still a problem in cultures where people can only approximate the year in which they were born, or where, as one AAPOR colleague noted, “certain calendar years are auspicious or ominous.”

2. Age non-response occurs when people refuse to answer a specific question about age even though they are willing to answer most other survey questions.

If mental calculation is required, respondents may feel the burden is too great. Privacy and survey anonymity are important concerns as well, especially if a study is focused on a small population. Why? Because specific ages, if combined with other data such as gender, marital status, occupation, etc. can make it easy to identify individuals even if no personal identifiers are asked for.

3. Erroneous age data can stem from several sources, including both respondent error and survey design ambiguities.

For online surveys, drop-down menus are especially prone to error, and the errors have serious implications when respondents are offered numeric ranges rather than specific numbers. In addition, some respondents make calculation errors, and if a person has privacy concerns with no option to decline or skip the question, he or she may enter false data to get past it.

Respondents may also misinterpret a question about age. When asking year born, for example, some will interpret this as a request for age, and answer “47.” Or maybe they did not misinterpret, but instead left off the first two digits of 1947. It is impossible to know without following up or, if the survey is online, without constraining the possible answers within specific numeric ranges (easy to do, but rarely done!). Oddly enough, some respondents will even misread “In what year were you born?” as a request to provide the city, state, or country in which they were born.

4. Imprecise age data is the rule rather than the exception in nearly all surveys even if we manage to avoid other sources of error. The reason? Few surveys ask for an exact date of birth.

But not knowing an exact date means that some data will be one year off when classifying respondents into age groups or birth cohorts. For example, if we ask for year of birth and then calculate age, the calculation will be one year off for respondents whose birthday comes later in the year. Likewise if we ask for age and then derive year born, there are two possible years in which to classify each respondent because whether they were born earlier or later in the calendar year is unknown.

Options for Asking Age

Alas, there is no way to minimize all four sources of error at once. It is essential, therefore, to know during the design phase how the data are going to be used and analyzed (see Versta Research’s white paper, The Art of Asking Questions). A tracking study, for example, may require nothing but an approximate age grouping, whereas a cohort or generational analyses will likely require an exact year born. In short, each of these common ways of asking age will involve trade-offs:

“How old are you?” or “What is your age?”

Strengths

• Provides appropriate detail for most studies
• Aligns with our usual interest in numeric age (vs. generation or birth cohort)

Weaknesses

• A direct question may offend some
• Specific ages may raise privacy concerns, especially with small studies
• Higher item non-response than other ways of asking
• Subject to respondent error and age heaping

“In what year were you born?”

Strengths

• Provides appropriate detail for most studies
• Easier to answer and more accurate than providing age
• Feels less intrusive to respondents who are sensitive about age
• Directly identifies birth cohort

Weaknesses

• Age must be calculated, which will be imprecise for some
• Higher item non-response than asking for a range of ages
• Subject to age heaping in some cultures
• Requires careful instructions and programming to avoid data entry errors

“What is your birth date?”

Strengths

• Provides detailed data for all types of analysis
• Provides exact numbers to accurately derive age and birth cohort

Weaknesses

• High burden on respondents
• Significant concerns about privacy and data protection
• Highest item non-response
• Requires careful instructions and programming to avoid data entry errors

“What age range are you in?”

Strengths

• Respondents can answer quickly and accurately
• Least intrusive of all age questions
• Highest item response rates

Weaknesses

• Less detailed information limits data analysis
• Assumes knowledge beforehand of optimal age categories
• Inaccurate response will have bigger impact on data analysis
• Drop down menus will cause errors

Knowing Why Age Matters

Which age question should you choose? It all depends on details such as: (a) why the research is being done (b) how and why age data will be important (c) how age data will be used, analyzed, and reported (d) who will be responding to the survey, (e) whether the survey is being fielded online, by phone, in person, or on paper . . . the list goes on. The wisdom in the recent debate among survey methodologists is that how one asks depends on myriad aspects of the research that too many overlook.

Few questions seem simpler than asking “How old are you?” But anticipating the right approach and then untangling the errors, ambiguities, and missing data in the responses that come back can be amazingly complex. A wise researcher is one who poses hard questions about the research itself before feeling satisfied with even the simplest survey question.