This is the kind of paper where approaching it with a kind of Bayesian orientation is really helpful. As in: what's your prior on movies cutting the divorce in half? It's almost certainly very small. So either this paper has to have amazingly compelling evidence, or something is wrong.