We relate market stress to asset pricing by analyzing a large and systematic discrepancy among off-the-run Treasury securities: bond prices traded as much as five percent below otherwise identical notes, orders of magnitude more than we find concurrent special repo rates to explain. The relatively low lending revenue from holding the note begs the question of why its current holders would not trade it for cheaper yet identical cash flows. The pricing discrepancy persisted for months. We find that liquidity characteristics of Treasury securities explain a large share of the Treasury pricing anomaly. We relate insurers’ transactions in Treasuries to their characteristics. We find that the most highly levered insurers and those that transact most frequently tended to demand notes over bonds during the crisis, contributing to the anomaly.
Widening interest rate spreads in the recent financial crisis could represent deteriorating asset liquidity or concerns over issuer solvency. I construct new measures of market liquidity and credit risk to decompose these effects in spreads, and estimate the effect of liquidity risk; the possibility that asset liquidity could deteriorate precisely when an investor’s marginal utility is at its highest. My results show that market liquidity explains more than two-thirds of the widening of one- and three-month euro LIBOR-OIS and sovereign debt spreads over the first half of the financial crisis. The large role for market liquidity is partly due to the pricing of liquidity risk.
We examine the efficiency of using individual stocks or portfolios as base assets to test asset pricing models using cross-sectional data. The literature has argued that creating portfolios reduces idiosyncratic volatility and allows factor loadings, and consequently risk premia, to be estimated more precisely. We show analytically and demonstrate empirically that the smaller standard errors of beta estimates from creating portfolios do not lead to smaller standard errors of cross-sectional coefficient estimates. The standard errors of factor risk premia estimates are determined by the cross-sectional distributions of factor loadings and residual risk. Creating portfolios destroys information by shrinking the dispersion of betas and leads to larger standard errors.