Great post Mark. You may also want to take a look at findings from Rick Johnston of Rice University in "Crowdsourcing Forecasts: Competition for Sell-Side Analysts" where he looks at the relative accuracy between the sell side and buy side (as represented by the Estimize data set).
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2333671
Also the paper from Vinesh Jha and myself titled "Generating Abnormal Returns Using Crowdsourced Earnings Forecasts from Estimize" where we look at a similar outcome of the buy side better representing the true expectation of the market vs the conflicted sell side.