A number of studies support this hypothesis, at least on the surface. But many of these studies are weak because they are based on self report (i.e. what individuals say about themselves in questionnaires). Self-report is notoriously unreliable as a measure of what a person actually thinks. Not only do people sometimes not tell the truth, but frequently they don't actually know what's going on inside their own head (in other words, their beliefs about how they behave can differ quite a lot from how they actually behave). In his 2003 book The Psychology of Religion, Professor Bernard Spilka (Purdue University) concluded (p422):
In the end, although more religious people apparently tend to say that they are more honest than less religious persons, such findings seem to be contradicted by other research showing no relationship, or even a positive relationship between lie scale scores and religiosity. More importantly, there is not much evidence from studies of actual behavior to support the supposition that religious people are somehow more honest, or less likely to lie or cheat, than are their less religious or nonreligious peers. In view of the clear teachings of most faiths on such issues, we are left to ponder why religion does not have a significant impact in reducing cheating behavior.There is, however, some more evidence that religious or superstitious beliefs can make people act more honestly, although by mechanisms that proponents of religion will find surprising (and probably won't like). For example, subconscious priming with religious messages can make you more honest, whether you believe in god or not (what's more, priming with non-religious but 'wholesome' messages has the same effect).
And there's some interesting evidence that a belief in the supernatural can make you more honest by convincing you that someone is watching you. Kevin Haley at UCLA has shown that, in an economic game (the anonymous dictator game), showing stylised eyespots on the computer monitor increases honesty (see refs for this and other papers below). In similar study but this time conducted in the 'real world', Melissa Bateson at the University of Newcastle Upon Tyne has shown the same thing. In their study, coffee-drinkers in a university common room were asked to make a voluntary, anonymous contribution to the cost of milk. They found that, on the weeks when they pasted a pair of eyes above the 'honesty box', people contributed more (see fig). The scary eyes at the end seem to be particularly effective!
Even more striking is the work of Jesse Bering at the University of Arkansas, who has shown that simply telling people that the lab is haunted will make test subjects more honest. The subjects were given a task to do, with the opportunity to cheat (they were told that the newly-developed computer program that administered the test sometimes malfunctioned and gave the answer ahead of time). Some were also informed that the study was dedicated to a recently dead grad student. And some of this group were further told by the experimenter (as a "casual but serious aside") that the ghost of this dead student had been seen around the lab. Sure enough, those who had been spooked were significantly less likely to cheat.
So it seems that there are good psychological reasons to expect that people exposed to religious concepts (exposure to ethical messaging, fear of being found out by the 'policeman in the sky') might be more honest. On the other hand, it's also true that people who are convinced that they are holders of moral truths are more likely to behave disreputably. Which may explain why empirical evidence that religious people really are more honest remains lacking.
A last thought. These studies are done in individuals within a society. But what happens if society as a whole becomes less religious? Does corruption and dishonesty increase, or decrease? This is an extremely important question for humanists, and one that is addressed in a follow up post.
Hood, R.W., Spilka, B., Hunsberger, B., & Gorsuch, R. L. (2003). The psychology of religion: An empirical approach. 2nd Edit., New York: Guilford.
Bateson, M., Nettle, D., Roberts, G. (2006). Cues of Being Watched Enhance Cooperation in a Real-World Setting. Biology Letters, 12, 412-414. PDF
Haley, K., & Fessler, D. (2005). Nobody’s Watching? Subtle Cues Affect Generosity in an Anonymous Economic Game. Evolution and Human Behavior, 26(3), 245-256.
Bering JM, et al. Reasoning about Dead Agents Reveals Possible Adaptive Trends. Human Nature, Winter 2005, Vol. 16, No. 4, pp. 360-381.