Here's a study I found some weeks ago at Jeff Sackmann's blog, Heavy Topspin. He used data from the first 3 Slams of 2011. He found that while serves, obviously, confer an advantage upon the server, the advantage dissipates in long rallies. The exception he found was on grass. At Wimbledon, the server of a rally is still more likely to win the point than the receiver even when up to 14 shots have been made in the rally.
This is the study: http://heavytopspin.com/2011/08/17/h...dvantage-last/
And the follow-up, dividing the data into 1st and 2nd serve: http://heavytopspin.com/2011/10/14/s...second-serves/