Unless you’re trying to make a point about how “atheists are good people too” or you happen to despise the Catholic Church, it’s really a no-brainer: Go with God. Why is this so obvious? As the political scientist Dominic Johnson has argued, “If supernatural punishment is held as a belief, then this threat becomes a deterrent in reality, so the mechanism can work regardless of whether the threat is genuine or not.” In other words, from a psychological perspective, the ontological question of God’s actual existence is completely irrelevant; all that really matters in the above case is that the taxi driver is fully convinced that God doesn’t like it when he cheats his passengers.
Research from the social psychologist Ara Norenzayan, Bering says, supports the supposition. In one study, people were asked to play a word game, and then "donate" money to charity. The people exposed to God-related words donated more money, whether or not they were religious themselves.
Interestingly, in a separate study from Norenzayan and a colleague, people had divergent responses when asked to assess themselves on measures that wouldn't be immediately obvious to a third party (as opposed to an omniscient God). In that case, the religious participants were more likely to agree with statements like “No matter who I’m talking to, I’m always a good listener." Bering: "This means that while nonbelievers might feel “exposed” in the wake of receiving implicit God primes, just like believers, this feeling doesn’t influence how atheists attempt to portray themselves socially."
If I'm understanding this correctly, Bering is saying that both atheists and believers are more likely to show "pro-social" behavior if primed to do so (by the God-related rhetoric), but atheists are less likely to think about themselves that way, or to brag about their ethics to other people. Other studies, however, have found a "Sunday effect" among believers: if you make a pitch for a charity on a Sunday, for example, a person who's just been to church is much more likely to respond than an atheist, whereas if you make the pitch on any other day, believers and nonbelievers have comparative responses.
The fact that the Sunday effect abates by Monday makes the religious motivation seem fairly flimsy, but as a practical matter, Bering concludes that all other things being equal, you should hope to run into a religious person rather than an atheist if you need help. This is a version of John Rawls' 'veil of ignorance' heuristic, and I tend to agree (so, for example, although I share some of Andrew Sullivan's questions about Mormons, if I was, say, marooned on a west Texas highway in the middle of the night, I would probably rather encounter a couple of missionaries than a roving band of atheists). I'm also wondering if you could run these experiments in the opposite direction--that is, figure out if someone is religious, regardless of how they describe themselves, by seeing how they respond to these kinds of cues.