Here is an excerpt from his book "Irreligion: A mathematician explains why the arguments for God just don't add up." (Undeline emphasis is mine.)
[A] problem associated with assigning ][ disproportionate payoffs to God's existence and the eternal happiness to be derived from obeying Him is that this assignment itself can serve to rationalize the most hateful of actions. Contrary to Dostoyevsky's warning that "if God doesn't exist, everything is allowed," we have the fanatical believer'S threat that "if God does exist, everything is allowed." Killing thousands or even millions of people might be justified in some devout believers' eyes if in doing so they violate only mundane human laws and incur only mundane human penalties while upholding higher divine laws and earning higher divine approbation.Page 139-141:
I would like to counter ][ the claim regularly made by religious people that atheists and agnostics are somehow less moral or law-abiding than they [are]. There is absolutely no evidence for this, and I suspect whatever average difference there is along the nebulous dimension of morality has the opposite algebraic sign.
[S]tudies on crime rates (and other measures of social dysfunction) showing that nonbelievers in the United States are extremely underrepresented in prison[,] suggest as much. So does Japan, one of the world's least crime-ridden countries, only a minority of whose citizens reportedly believe in God. And so, too, do those ][ monomaniacal true believers whose smiling surety often harbors a toxic intolerance. (Recall the physicist Steven Weinberg's happy quip "With or without religion, good people will do good, and evil people will do evil, but for good people to do evil, that takes religion.") Also worthy of mention are the garden-variety religious scoundrels, hypocrites, and charlatans in public life. Not quite evil, but also far from admirable, is the social opportunism that no doubt is the reason for many expressions of religious humbug. Like feigning an interest in golf to get ahead in business, mouthing the right pieties can often improve one's prospects in politics.
An atheist or agnostic who acts morally simply because it is the right thing to do is, in a sense, more moral than someone who is trying to avoid everlasting torment or, as is the case with martyrs, to achieve eternal bliss. He or she is making the moral choice without benefit of Pascal's divine bribe. This choice is all the more impressive when an atheist or agnostic sacrifices his or her life, for example, to rescue a drowning child, aware that there'll be no heavenly reward for this lifesaving valor. The contrast with acts motivated by calculated expected value or uncalculated unexpected fear (or, worse, fearlessness) is stark.
Still, people do often vigorously insist that religious beliefs are necessary to ensure moral behavior. Though the claim is quite clearly false of people in general, there is a sense in which it might be true if one has been brought up in a very religious environment. A classic experiment on the so-called overjustification effect by the psychologists David Greene, Betty Sternberg, and Mark Lepper is relevant. They exposed fourth- and fifth-grade students to a variety of intriguing mathematical games and measured the time the children played them. They found that the children seemed to possess a good deal of intrinsic interest in the games. The games were fun. After a few days, however, the psychologists began to reward the children for playing; those playing them more had a better chance of winning the prizes offered. The prizes did increase the time the children played the games, but when the prizes were stopped, the children lost almost all interest in the games and rarely played them. The extrinsic rewards had undercut the children's intrinsic interest. Likewise, religious injunctions and rewards promised to children for being good might, if repudiated in later life, drastically reduce the time people spend playing the "being good" game. This is another reason not to base ethics on religious teachings.