A recent paper by Martinelli & Parker – “Deception and Misreporting in a Social Program” – shows just how much people lie in self-report questionnaires (using data from Mexico). The Freakonomics blog has a catchy write-up of it; we’d expect nothing less from those guys! What’s interesting (and problematic) is that people both underreport and overreport. (You expect people to underreport in a social program: if you look poorer, you might get more help.) This, from the Freakonomics summary:
Below is a list of household items that were overreported — i.e., which applicants said they had but in fact did not (again, followed by percentages):
Toilet (39.07 percent)
Tap water (31.76)
Gas stove (28.56)
Concrete floor (25.41)
So 4 out of 10 applicants without a toilet said they had one.
This reminds me of a favorite story. A good friend was with a survey enumerator who was gathering household data in rural Kenya. The enumerator asked the household head how many kettles they had. The head said, None! We don’t have even one kettle! And my friend asked, What about that kettle? [pointing to the one on the stove] Oh, that one! We borrowed it from the neighbor!
Nice. This illustrious tradition takes us back to Margaret Meade and her (self-admitted) lying informants in Samoa in the 1920s. Which makes me even more confident in most social science conclusions. (I say most, because my research is clearly different. Who would lie to me? I’m the Magic Man.)
Abstract of the academic paper below
We investigate empirically the extent of misreporting in a poverty-
alleviation program in which self-reported information, followed by a
household visit, is used to determine eligibility. In the model we pro-
pose and estimate, underreporting may be due to a deception motive,
and overreporting to an embarrassment motive. We nd that under-
reporting of goods and desirable home characteristics is widespread,
and that overreporting is common with respect to goods linked to
social status. Larger program bene ts encourage underreporting and
discourage overreporting. We also estimate the costs of lying and em-
barrassment for different goods, and show that the embarrassment
cost for lacking a good is proportional to the percentage of households
who own the good.