Mathias Dewatripont

Credit and Efficiency in Centralized and Decentralized Economies

We study a credit model where, because of adverse selection, unprofitable projects may nevertheless be financed. Indeed they may continue to be financed even when shown to be low-quality if sunk costs have already been incurred. We show that credit decentralization offers a way for creditors to commit not to refinance such projects, thereby discouraging entrepreneurs from undertaking them initially. Thus, decentralization provides financial discipline. Nevertheless, we argue that it puts too high a premium on short-term returns. The model seems pertinent to two issues: Soft budget constraint problems in centralized economies, and differences between Anglo-Saxon and German-Japanese” financing.

Contract renegotiation in models of asymmetric information

In practice, parties to a contract can usually rewrite the contract’s terms if they so choose; courts do not typically interfere with a mutual agreement to rescind an old contract and replace it with a new one. A recent body of theoretical work has demonstrated, however, that the ability to renegotiate can considerably constrain parties’ ex ante welfare. In this short paper, we review this and other implications of renegotiation when contracting parties have asymmetric information.’ In a world of complete contracts (subject to informational constraints) the possibility of future renegotiation can only hurt parties. It can never help, since if it did, the renegotiated outcome could simply be written into the original contract. Moreover, it may strictly reduce welfare because a contract that is ex ante optimal may no longer remain optimal at a later date (i.e., it may fail to be sequentially optimal). And if so, the contract will presumably be renegotiated, thereby losing its initial optimality. Sequential optimality tends to fail because parties’ objectives change over time. There are two reasons why they do. First, with time, parties may acquire information. Suppose that, as is common in models with asymmetric information, parties initially try to strike a contractual balance between risksharing and allocative efficiency. If at a later date they learn the realization of the random variable giving rise to the risk, they will then no longer be