Johnson, Kenneth C, "Election Certification by Statistical Audit of Voter-Verified Paper Ballots" (October 31, 2004). http://ssrn.com/abstract=640943
One of the things I found was that the analysis of a conventional statistical recount is quite complicated - I only considered the case of a two-candidate race and wouldn't want to attempt more complicated cases. However, the alternative method that I advocate ("statistical error count") is much simpler and could be easily generalized to work with multiple candidates and alternative voting systems.
Ken Johnson
---- Election-methods mailing list - see http://electorama.com/em for list info