Survey Respondents Rarely Lie. They Skate By with “Satisficing.”
Satisfice is not a word you will find in most dictionaries, but it is a word you will hear often in the world of survey research. Satisficing is what survey respondents do when they “sort of” want to answer survey questions but not really. They are like lazy students taking a test. They dislike too much reading and thinking. “That’s good enough and close enough,” they say, so on to the next.
But giving the bare minimum is usually not good enough for those of us trying to measure opinions and behavior.
Examples of satisficing include:
(1) Picking the first “close enough” answer in a choice set, but not the optimal answer. Satisficing respondents don’t continue reading to see if there is an answer closer to what they really think or do.
(2) Agreeing or disagreeing to all questions presented, even though the questions are quite different and maybe even opposite. Typically, they’re not lying; they’re just too lazy or distracted to read the statements presented even though they want to take the survey.
(3) Consistently answering don’t know, neither, neutral, or the midpoint on answer scales even when they do, in fact, lean one way or the other. Instead of putting thought into one’s answer, saying “I don’t know” is the lazy way out.
There are two things you can do. First, write better surveys. Long, boring, repetitive, or complicated surveys encourage people to invest little effort in answering the questions. (And why should they? If the person who writes a survey can’t be bothered to write good questions, why should respondents be bothered to give good answers?)
Second, employ rigorous data cleaning protocols to flag satisficing behavior, potentially cutting some respondents from the data. Notice that I say potentially cutting. We strongly recommend against deleting satisficing respondents unless they demonstrate multiple instances or a consistent pattern of inattentive behavior. Why? Because satisficing is normal behavior coming form normal people, and you need normal people—not just superstars—reflected in your data.
—Joe Hopper, Ph.D.
