This is not to say that good, reproducible, research does get done--it does. Look at the research on IQ. If reproducible results is a sign of methodological rigor and valid findings.....IQ wins the day. It is to say, however, that a lot of social science research is influenced by sets of biases. Sometimes these biases are easy to identify and sometimes they are not.
Nonetheless, here are a couple of studies I found on certain biasing factors as well as a link to story on how business journals are now trying to manage these biases.
Confirmation Bias In Social Work Journals: http://sth.sagepub.com/content/15/1/9.short
Positive Outcome Bias in Peer Review
Business Journals Respond