We propose a theory to explain the choice between nominal and indexed labor contracts. We find that contracts should be indexed if prices are difficult to forecast and nominal otherwise. Our analysis is based on a principal-agent model developed by Jovanovic and Ueda (1997) in which renegotiation can take place once the nominal value of the agent's output is observed. Their model assumes that agents use pure strategy, with the strong result that only nominal contracts can be written without being renegotiated. But, in reality, we do observe indexed contracts. We resolve this weakness of their model by allowing agents to choose mixed strategies, and find that the optimal contract is indeed nominal for certain parameters. For other parameters, however, we show that the optimal contract is indexed. Our findings are consistent with two empirical regularities: that prices are more volatile with higher inflation, and that countries with high inflation tend to have indexed contracts.