Lying to Gallup

Print More

There’s nothing new in Gallup’s new survey of church attendance by state, showing weekly attenders to range from 63 percent in Mississippi to 23 percent in Vermont. This is the classic Bible Belt-cum-Bible Suspender, stretching across the South and up the Plains to the Canadian border at the high end and with the Northeast and the West (always excepting Utah) at the low end. The national average of 41.6 percent weekly attenders, again, hews closely to the 40 percent mark that has been Gallup’s constant finding (give or take a couple of points) ever since it began asking the question in the 1940s.

The problem is, as Gallup now from time to time acknowledges, the number is not an accurate indicator of reality. Close studies of actual (as opposed to self-reported) attendance put the current number of weekly attenders in the mid-twenty-percent range. It’s a typical case of people “over-reporting” their good behavior, seen in self-reported voting behavior as well as church-going. The evidence is that Americans over-report now more than they did half a century ago.


In its current write-up, Gallup points to various factors to explain the regional variations. What would be interesting to know is if Americans in one part of the country are more likely to fib about going to church than those in another. My guess is that the more church-going is considered a norm of good behavior, the more fibbing there will be–and the less so, the less. In other words, guiltily non-attending Mississippians are more likely to tell Gallup they go to church than blithely non-attending Vermonters. If I’m right, then there’s not as much actual regional variation as Gallup shows.
  • Yes! You got it in the end! And, I think your final point is the correct interpretation of temporal variation in overreporting. We should actually expect LESS overreporting now than in the past. That is a commonly mistaken interpretation taken from the old Hadaway et al studies (and one that they tried to push). In any case, we don’t have any evidence on the estimates for historical overreporting.

  • Mark Silk

    I take your point, D. But I thought there were time-use data from the 1960s that suggested less over-reporting of church attendance. Not so?

  • No, not so, in my opinion. The time use stuff is a totally different measure, and whatever bias they might have measured is not comparable to what was done in the Hadaway et al study(s). There is nothing close to the Hadaway et al study for detail in measuring bias. They nailed it. Other studies suggesting bias have minimal basis for assessing the true population total and comparing it to sample estimates. And, whatever the case, time use tallies are WAY different from survey responses to 8 point ordinal scales.