Henry Farrell passes along the welcome news that James Vreeland has a blog, which is now in my daily rotation. The first substantive post discusses Rosendorff and Hollyer's "bad ass" theory of why dictators sign on to human rights agreements that they know they will violate:
So, if a dictator enters into the CAT, he better be darn sure that he will never fall from power. And this is precisely the point. He is sending a signal to his domestic audience that he intends on doing whatever is necessary to hold on to power, and he's quite confident in his ability to survive in office. Only the strongest resolve dictators will take such a step. Soft dictators can't rist faking it - if they sign the CAT and lose their grip on power, they're libel to end up in the slammer. So the signal the tough dictators send is unambiguous, and might even serve to quite opposition, who realize that resistance is futile.
A non-gated pdf of the article he's discussing is here, and Eric Voeten discussed it here. The upshot is that signing on to international agreements like the Convention Against Torture (CAT) incurs costs for leaders, which sends a credible signal to domestic audiences that they aren't going anywhere.
I agree with everyone that this is an interesting, creative theory. My first reaction to hearing about the theory was this: if the logic is right, then the domestic audience should recognize the signal as credible and thus we should see dissident movements, and thus the torture intended to combat them, dry up. In fact, the authors do find such an effect: torture continues in signing states, but seems to decrease after signing the CAT. They also find that dictators who sign the CAT stay in power longer than those who don't.
I haven't had time to fully digest the model, but I like what I've seen so far.