This paper estimates the incumbency effects in elections to state legislatures in India. I use the Regression Discontinuity Design (RDD) technique to overcome the bias that arises in previous empirical methodologies. RDD exploits the comparability of winners and losers in closely fought elections to net out the effect due to incumbency. Using a large dataset on legislative elections of 25 Indian states between 1975 and 2003, I find that incumbent candidates are at a significant disadvantage as compared to non-incumbents. Variation in incumbency effects across Indian states can be explained in part due to differences in degree of competitiveness of elections and in levels of public good provision.