A defense of the five-point Likert scale

by Tom Ehrbar

When we are not looking for a simple yes or no answer in our surveys, we often use scaled-response questions (also known as Likert scales). These measure attitudes or opinions on an ordinal scale, and are very well suited to measuring degrees of satisfaction, likelihood, frequency, priority, agreement, etc.

It is very important in formulating such questions, however, to avoid scale-point proliferation, that is, presenting the respondent with too many answers across a linear scale. The following examples demonstrate the problems with this approach.

How would you assess the current degree of “bitcoin readiness” within your industry overall?
Please answer on a scale of 0–10 where = Poorly prepared and 10 = Fully prepared.

image

or

How often do you use bitcoin to pay your vendors?

Please answer on a scale of 1–8, where 1 = Never and 8 = Always.

(1)    Never

(2)    Rarely

(3)    Occasionally

(4)    Fairly often

(5)    Often

(6)    Very often

(7)    Almost always

(8)    Always

In both cases, these series of slightly different variations risks confusing (or annoying) respondents with hair-splitting differences. As they’re reading or listening to the questions, most respondents will not be able to distinguish more than six or seven response levels. For questions of this sort, we generally prefer to offer five scale points (and always starting at 1), with a “neutral” midpoint, splitting two “negative” and two “positive” responses. In the case of the second question above, we would likely simplify as follows:

How often do you use bitcoin to pay your vendors?

Please answer on a scale of 1–5, where 1 = Never and 5 = Always.

(1)    Never

(2)    Almost never

(3)    Occasionally/sometimes

(4)    Almost always

(5)    Always

Many types of questions can be formulated in this manner by asking respondents their level of agreement with a given statement, such as

Please indicate your level of agreement with the following statement:

Our company's workforce currently has adequate digital skills to cope with the demands of an interconnected customer base.

(1)    Strongly disagree

(2)    Disagree

(3)    Neither agree nor disagree

(4)    Agree

(5)    Strongly agree


. . . first in a series examining our survey methodology

Tom Ehrbar is a Senior Editor at Oxford Economics, where he oversees the editorial and production flow for all Thought Leadership studies, managing the department’s global editorial network as well as its survey teams.