Skip to main content

Stock Market Liquidity and Return Predictability: A Bayesian Nonparametric Approach

Kalli, Maria, Ellington, Michael (2019) Stock Market Liquidity and Return Predictability: A Bayesian Nonparametric Approach. Review of Financial Studies, . ISSN 0893-9454. (In press) (The full text of this publication is not currently available from this repository. You may be able to access a copy if URLs are provided)

The full text of this publication is not currently available from this repository. You may be able to access a copy if URLs are provided. (Contact us about this Publication)

Abstract

Predicting stock returns is a central issue in finance. Seminal work by, for example: Keim and Stambaugh (1986); Fama and French (1988); and Campbell and Shiller (1988), demonstrate predictive power of an array of financial variables for future returns. Insights by Ang and Bekaert (2007) challenge this view and stress that the debate is far from settled. Among many others, including Welch and Goyal (2008) are pessimistic regarding the predictive power, and robustness, of financial variables. In particular, Kostakis et al. (2014) conform to this view by using a novel estimation procedure and show that return predictability becomes weaker as the horizon increases.However, there is an emerging literature utilising (aggregate) measures of illiquidity in predictive regression analysis, and out-of-sample forecasting exercises. A liquid stock market permits the ease of trading assets which investors and market participants rely on to ensure fair pricing.The main contribution of our paper is to quantify the out-of-sample forecasting perfor- mance of aggregate stock market illiquidity for future stock returns using novel Bayesian nonparametric vector autoregressions (BayesNP-VAR) to US data from 1927 to 2018 (Kalli and Griffin, 2018). The benefits of using these methods are threefold. First, we make no assumption regarding the underlying distribution stock returns, or errors of the model. Existing studies examining return predictability fit linear models using OLS such as Næs et al. (2011). Second, we do not presume a ́priori that the data are homoskedastic. Typ- ically, papers studying return predictability presuppose this as a result of model choice (see e.g. Chen et al. (2018)). Third, our approach permits different aspects of nonlinear- ity such as time-varying relationships without specifying variable and model evolution.

Item Type: Article
Uncontrolled keywords: Stock Market Liquidity, Return Predictability, Bayesian Methods, Bayesian Non-parametrics, Vector Autoregression, Statistics
Subjects: H Social Sciences > HG Finance
Q Science > QA Mathematics (inc Computing science) > QA273 Probabilities
Divisions: Faculties > Sciences > School of Mathematics Statistics and Actuarial Science > Statistics
Depositing User: Maria Kalli
Date Deposited: 13 Nov 2019 13:04 UTC
Last Modified: 15 Nov 2019 14:31 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/78658 (The current URI for this page, for reference purposes)
  • Depositors only (login required):