How to Create Effective Scales - Keep it simple. Given a choice between a very short, concise scale and a more
elaborate, sophisticated one, the less complex scale should be used. Even after
identifying a scale to be used for an item, ask, “Is there an easier, simpler scale
or way of asking this question?”
- Respect the respondent. While respondents are ordinarily cooperative and
helpful, response is a favor. They have little involvement with the task. Select
scales that will make it as quick and easy as possible for them. That will reduce
nonresponse bias and improve accuracy.
- Dimension the response. In what dimensions do respondents and sponsors
think about the issue? They’ll not always be the same, so some commonality
must be discovered. The dimensions along which the respondents are to answer
must not be obscure or difficult, and they should parallel respondents’
thinking.
- Pick the denominations. Always use the denominations that are best for respondents.
The data can later be converted to the denominations sought by information
users. Feet and inches can be changed to metric or time converted to
a 24-hour clock during processing.
- Choose the range. Categories or scale increments should be about the same
breadth as those ordinarily used by respondents. Normally, respondents classify
things into a range from about 2 to about 7 or 8 categories, and seldom
more than 10. Respondents often can’t be as precise as researchers would like.
- Group only when required. Never put things into categories when they can
easily be expressed in numeric terms. People think in years, not decades or
centuries. Data can always be grouped during processing, but if obtained in broad categories,
it can’t be desegregated later, no matter how desirable that might be.
- Handle neutrality carefully. If respondents genuinely have no preference,
they’ll resent the forced choice inherent in a scale with an even number of alternatives.
If feelings aren’t especially strong, an odd number of scale points
may result in fence-riding or piling on the midpoint, even when some preference
exists.
- State instructions clearly. Even the least capable respondents must be able to
understand. Use language that’s typical of the respondents. Explain exactly
what the respondent should do and the task sequence they should follow. List
the criteria they should judge by and use an example or practice item if there’s
any doubt.
- Always be flexible. Scaling examples provided here are only that. They can
and should be modified to fit the task and the respondents. The instructions,
format, vocabulary, and number of scale points can all be changed to suit the
needs of the survey. Scales should fit the task, not conform to the original authors’
specifications.
- Pilot test the scales. When there’s any doubt about the ability of respondents to
use the scales, a brief, informal pilot test is quick, inexpensive insurance. Don’t
wait until the entire questionnaire is written. Individual parcels can be checked
with a few typical respondents.
|