Title: A structurally motivated stochastic volatility model for equity returns
Authors: Alexander Back - Hanken School of Economics (Finland) [presenting]
Abstract: The conditional volatilities of equity returns often exhibit a long-memory property. In standard volatility models, this may lead to parameter estimates that suggest a highly persistent process, often bordering on non-stationarity. To deal with this, several researchers have proposed multiplicatively decomposing volatility into slow-moving and transient components. A recent vein in the literature has used a structural approach to identify the slow-moving component as potentially stemming from the amount of financial leverage that the firm under consideration has taken on. Consequently, the transient component is interpreted as a volatility model in assets rather than equity returns. This captures the well-known notion that leverage makes equity riskier. We use this backdrop to propose an extension of a seminal model in this literature. We argue that an autoregressive stochastic volatility model may be a better model for asset returns than its GARCH counterpart, and that the choice between the two may have non-trivial consequences in applications. We suggest a new parameterization and propose an estimation procedure based on indirect inference. A drawback of these models is that they are computationally expensive to estimate. We propose a machine learning approach that can drastically speed up estimation.