It turns out that when people are asked by phone about the regularity of their attendance at religious services, they report higher figures than when they are asked on the internet. It is speculated that the internet with its anonymity creates a greater willingness to tell the truth. According to the survey conducted by the Public Religion Research Institute:
The research shows that every subgroup of Americans inflates their levels of religious participation, with young adults, Catholics and white mainline Protestants particularly likely to inflate the frequency of their attendance at religious services.
In their report titled I Know What You Did Last Sunday: Measuring Social Desirability Bias in Self‐Reported Religious Behavior, Belief, and Identity, the authors suggest reasons for this discrepancy, such as ‘social desirability bias’. In the US religion is associated with morality and pro-social behavior, and respondents may interpret the question as being a proxy for “Am I a good Christian” and answer aspirationally rather than realistically. They think that eliminating the presence of the interviewer, as online surveys do, takes away that pressure.
What I did find particularly interesting is Table 2 that looks at just the religiously unaffiliated. There have been studies that suggest that many in this group hold somewhat traditional religious beliefs and some have referred to them as ‘unchurched believers’ rather than nonbelievers. This study says that this too might be an artifact of the way those interviews were done because compared to the telephone interviews, there was a large increase in the online numbers who said that ‘Religion is not important’ (73% vs. 49%) and those who never attend religious services (68% vs. 36%).