Here's an article I wrote in Governing Magazine -- this has been a focus of mine for a long time.
I looked at eight years worth of recall results to try and find out whether special elections presents an accurate picture of the electorate. I believe that recalls are really the only place to see a pattern, as they sometimes take place as a special election and sometimes take place on a general or primary election day.
My original idea was that officials were much more likely to be removed by special elections and therefore special elections were inaccurate. But after looking at 861 recalls over the last eight years my theory was shot down.
Sitting officials, it turns out, are not disproportionately hurt -- and sometimes may actually be helped -- by when a recall election is held. Moreover, the more-motivated, angrier voters do not seem to skew special elections in an obviously predictable way.