Heterogeneous Intermediaries and Asset Prices

Abstract

Canonical intermediary asset pricing models assume a representative intermediary with an SDF linear in its leverage. Yet the leverage of broker-dealers is procyclical whereas the leverage of banking holding companies is countercyclical. I propose an empirically testable heterogeneous intermediary stochastic discount factor (HI-SDF) that depends non-linearly on aggregate leverage as well as net worth shares of individual intermediaries. I show that the HI-SDF emerges from general equilibrium models with heterogeneous intermediaries whereas its specific functional form hinges on different models' specifications of financing constraints. To address this issue, the HI-SDF is estimated semiparametrically. I provide empirical evidence that the wealth distribution among intermediaries is an important source of risk for pricing risky securities. The estimated HI-SDF is found to exhibit substantial explanatory power for cross-sectional variation in expected returns across a wide range of test assets. In contrast to representative intermediary empirical models, the HI-SDF exhibits lower pricing errors and explains larger fractions of the cross-section of expected returns.