Ron occasionally writes about religion (vs. atheism) on this blog. He's had at least two excellent insights: First, that "deconverting" a religious believer has much more in common with drug rehab for an addict than it does with a rational, scientific debate. And second, promoting the idea that perhaps there is a
In a related vein, Robin Hanson at Overcoming Bias suggests that a good analogy for the possibility of a (widespread) transition from religion to atheism is the changing of any existing widespread industry standard. And, as numerous startups have learned to their dismay, there is enough inertia behind any widespread standard, that it isn't sufficient that your new idea is objectively better. It must be enough better (perhaps an order of magnitude) that the effort involved in the change has enough payoff to make it worthwhile for the customer.
At the moment, atheism is "right", but the benefits of being "right" on this subject (vs. just being an average, typical member of your society) are so minor, that the cost is rarely worth it. QWERTY is a poor keyboard layout, but hardly anyone uses a different one. Metric is a much better measurement system than the old English units, but even though most of the world changed, the USA didn't quite make it over the bar (aside from the military and medicine, where it could be mandated top-down). In much the same way, religion and society have co-adapted to work well enough together, that "atheism is better" may be true, but it isn't (yet) enough better.
[Updated 9/3/2011: Incorporated Ron's corrections of his Loki suggestion.]