Predictive regressions are subject to two small sample biases: the coefficient estimate is biased if the predictor is endogenous, and asymptotic standard errors in the case of overlapping periods are biased downward. Both biases work in the direction of making t-ratios too large so that standard inference may indicate predictability even if none is present. Using annual returns since 1872 and monthly returns since 1927 we estimate empirical distributions by randomizing residuals in the VAR representation of the variables. The estimated biases are large enough to affect inference in practice, and should be accounted for when studying predictability.

The authors examine the optimal design of a risk-adjusted deposit insurance scheme when the regulator has less information than the bank about the inherent risk of the bank's assets (adverse selection) and when the regulator is unable to monitor the extent to which bank resources are being directed away from normal operations toward activities that lower asset quality (moral hazard). Under a socially optimal insurance scheme: (1) asset quality is below the first-best level, (2) higher-quality banks have larger asset bases and face lower capital adequacy requirements than lower-quality banks, and (3) the probability of failure is equated across banks.

An apparent pricing anomaly exists in the market for U.S. Treasury strips: zero‐coupon strips created from principal payments typically trade at significantly higher prices than otherwise identical zero‐coupon strips created from coupon payments. In addition to documenting this phenomenon, this study demonstrates that differences in liquidity and differences in reconstitution characteristics explain much of this price variation.

Predictive regressions are subject to two small sample biases: the coefficient estimate is biased if the predictor is endogenous, and asymptotic standard errors in the case of overlapping periods are biased downward. Both biases work in the direction of making t ‐ratios too large so that standard inference may indicate predictability even if none is present. Using annual returns since 1872 and monthly returns since 1927 we estimate empirical distributions by randomizing residuals in the VAR representation of the variables. The estimated biases are large enough to affect inference in practice, and should be accounted for when studying predictability.

This paper formulates and estimates a finite-horizon, structural dynamic model of agricultural investment behavior that incorporates the major features of low-income agricultural environments: income uncertainty, constraints on borrowing and rental markets, and the use of investment assets to generate income and smooth consumption. The model is fit to longitudinal Indian household data on farm profits, bullock stocks, and pump sets. The estimated structural parameters are used to assess th e effects on the life-cycle accumulation of bullocks, agricultural profits, and welfare associated with complete markets and bullock liquidity and with second-best policies that provide assured sources of income to farmers and weather insurance.

We develop a simple model in which the presence of portfolio insurers in a market of risk‐averse traders leads to multiple equilibria for the pricing of financial assets and can cause an increase in volatility, including insurance‐induced price drops. We demonstrate, however, that centralized portfolio insurance firms may actually reduce, not increase, volatility, even if the existence of these firms increases the total amount of funds under insurance.

The methods of Gibbons and Ferson (1985) are extended, relaxing the assumption that expected returns are linear functions of predetermined instruments. A model of conditional mean-variance spanning generalizes Huberman and Kandel (1987). The empirical results indicate that more than a single risk premium is needed to model expected stock and bond returns, but the number of common factors in the expected returns is small. However, when size-based common stock portfolios proxy for the risk factors, we reject the hypothesis that four of them describe the conditional expected returns of the other assets.

The methods of Gibbons and Ferson (1985) are extended, relaxing the assumption that expected returns are linear functions of predetermined instruments. A model of conditional mean‐variance spanning generalizes Huberman and Kandel (1987). The empirical results indicate that more than a single risk premium is needed to model expected stock and bond returns, but the number of common factors in the expected returns is small. However, when size‐based common stock portfolios proxy for the risk factors, we reject the hypothesis that four of them describe the conditional expected returns of the other assets.