Michael Huemer, professor of philosophy at the University of Colorado:
The problem of political irrationality is the greatest social problem humanity faces. It is a greater problem than crime, drug addiction, or even world poverty, because it is a problem that prevents us from solving other problems. Before we can solve the problem of poverty, we must first have correct beliefs about poverty, about what causes it, what reduces it, and what the side effects of alternative policies are. If our beliefs about those things are guided by the social group we want to fit into, the self-image we want to maintain, the desire to avoid admitting to having been wrong in the past, and so on, then it would be pure accident if enough of us were to actually form correct beliefs to solve the problem.
Based on the level of disagreement, human beings are highly unreliable at identifying correct political claims. This is extremely unfortunate, since it means that we have little chance of solving social problems and a good chance of creating or exacerbating them. The best explanation lies in the theory of Rational Irrationality: individuals derive psychological rewards from holding certain political beliefs, and since each individual suffers almost none of the harm caused by his own false political beliefs, it often makes sense (it gives him what he wants) to adopt those beliefs regardless of whether they are true or well-supported.
The beliefs that people want to hold are often determined by their self-interest, the social group they want to fit into, the self-image they want to maintain, and the desire to remain coherent with their past beliefs. People can deploy various mechanisms to enable them to adopt and maintain their preferred beliefs, including giving a biased weighting of evidence; focusing their attention and energy on the arguments supporting their favored beliefs; collecting evidence only from sources they already agree with; and relying on subjective, speculative, and anecdotal claims as evidence for political theories.
Full paper in PDF here.