In his book (http://www.amazon.com/Statistical-Models-Causal-Inference-Di...) Professor Freedman says one of the reasons the social sciences have had virtually no beneficial impact on society (as opposed to the incredible progress in the hard sciences) is that their "studies" have unsupported mathematics. They make too many assumptions, and their methodologies don't support their conclusions. For example, you can do a study on the top reasons teens use drugs, and in a decade those reasons have completely changed because of the shift in culture.

The social sciences don't work well with the scientific method. You can't randomly select people, and you can't get them to make long-term changes to their lifestyle. Without experimental evidence, you are left with observational data, which is a nightmare to find causation instead of just correlation.