The other groups of respondents were asked these questions in an indirect fashion. In particular, we used an approach called a “list experiment”, which has been used in political polling for years. The approach asks people to indicate HOW MANY of a list of items relate to the respondent (not which one). This question is asked to a control group, and then a treatment group receives the exact same list plus one additional sensitive (or controversial) issue. By comparing how many items the respondent indicates in the treatment relative to the control, one can back out the aggregate percent of respondents to whom the additional issue applies. The essence of the approach is that people don’t have to actually tell you whether each issue corresponds to them, and thus it removes the potential for social desirability influencing respondents.
In the control group, respondents were asked, “Below are three activities; How many of the following things did you do last week? Went to a movie, Ate spaghetti, Bought toothpaste.” The treatment groups were the same except we added an additional fourth item, either “bought organic food”, “used a SNAP EBT card” or “Receive food assistance from a charitable organization like a food pantry, free community meal, or some other free grocery program?”
As the figure shows, the degree of affirmation inferred from the indirect question was lower for all three issues, particularly for organic food. The result suggests that respondents might face social pressure to indicate more support for organic food than they actually have, as the percent who said they purchased organic fell from 43% to just 11%. However, the results related to SNAP and charitable assistance are opposite of what was expected in that one might expect respondents to under-report these activities when directly asked if in fact respondents were remiss to reveal such information.