This paper presents a theoretical framework in which either long-term or short-term labor contracts arise endogenously. The fundamental trade-off is between firm specific and general human capital. While firm-specific human capital is more productive than general human capital, it cannot be reallocated in response to firm specific shocks. Firm-specific human capital is thus more attractive in environments where firms face less uncertainly about the quality of their projects. When firm-specific human capital is optimal and workers do not have access to perfect financial markets, long-term labor contracts are necessary to decentralize the first-best allocation. Within this framework, we show that `volatility,'' which has no aggregate consequences in economies with general human capital, has large and persistent negative effects in economies with long-term contracts. Volatility is modelled by an aggregate state which determines the accuracy of firms' signals about the quality of their projects. This model therefore sheds light on the sources of the recent stagnation of the Japanese economy. It also provides a rationale for Japan having adopted long-term labor contracts in the 1950's: since imitation is less uncertain than innovation, economies catching up to the frontier are well served by implementing long-term labor contracts
