In his book *What Americans Really Believe*, Rodney Stark also made a critical, crucial mistake in his reading of the data on church attendance and membership. Or, just as important, the Baylor study itself made the mistake, and he followed it to prove his point.
The extremely critical mistake? They asked the respondents about it.
Why is that a mistake? Because even people who don't go to church want to be seen as if they're part of a church. Or, they want to see themselves as someone who goes. On image questions of this kind, there's just too much lying. And they'll do it even when the information is to be kept confidential; many people don't trust that, either. So it is not something that is accurately studied by way of self-reporting.
The truth? According to those who look at actual counts, actual church attendance and membership are down overall. Attendance has been dropping noticeably for the last 10 years or so, with a brief interruption for 9/11. (Before that there was about two decades of 'shuffling', largely of folks leaving mainline churches for Pentecostal ones.) Many people, even those with strong Christian beliefs, don't bother with church at all anymore. Being an actual member of an actual congregation or church body isn't as important to their identity as in past generations. The drop isn't as large or as deep as the doomsday folks are saying, but it is real, and the churches must take action about it now. They must give a clear answer to the question, "what good does it do for Christians to gather together?"
The best way to measure such subjects by surveys? Indirectly. By creating a scale of indirect questions and doing a cluster analysis of them in relation to the other questions on the survey. By running a cross-check on it by looking at an area's actual attendance counts. By having enough respondents on these specific questions that you can have a high enough number of respondents (Ns) to be in effect a statistically-significant survey for each issue breakdown.
Stark et al.'s sunny analysis about current church attendance has no ground in fact except for the distorted 'self-reported' attendance and membership. This is one of the two instances in the book where the Baylor data was, from the start, not what it seemed to be.