Three essays in empirical asset pricing

Abstract

Inventory investment is volatile and procyclical, the inventory-sales ratio is persistent and countercyclical (Bils & Kahn, 2000; Blinder, 1986; Christiano, 1988), this is one of the stylized facts documented in the macroeconomic inventory behavior studies. Blinder (1986) states: “business cycles, are, to a surprisingly large degree, inventory cycles”. This thesis comprises three independent essays analyzing the inventory-based return predictors, inventory investment and the inventory-sales ratio, and the robustness of twelve fundamental-based anomalies. The first essay “Inventory growth spread: risk or mispricing?” empirically investigates the asset pricing implication of inventory growth (IG). The analysis suggests an alternative explanation on the IG spread from a liquidity-risk perspective, where the excess return generated from the long-short portfolio on IG reflects a liquidity risk compensation. Though also loads positively on value and investment risk factors, the IG spread cannot be persistently subsumed by other risks-based asset pricing models. In particular, the Fama and French (1993, 2015) three-factor and five-factor model, and Hou, Xue and Zhang (2015) q-factor model. The out-sample reversed sign of the IG spread also casts doubts on previous studies’ theoretical implication on the asset pricing role of inventory investment. The second essay “Inventory-sales ratio and the cross-sectional stock returns” documents that the industry-adjusted inventory-to-sales (IS) ratio, a measure that captures firm-level inventory to fulfill future sales, significantly predicts the monthly U.S stocks returns. Trading on a zero-cost portfolio that long the lowest IS decile portfolio and short the highest IS decile portfolio generates an annual abnormal return of 5% - 7%, the excess return remains strong after adjusting existing asset pricing models and various robustness tests. Empirical evidence suggest that the return spread of IS might potentially be driven by the stock-out risk. The third essay “Are robust return predictors really robust?” examines the robustness of the 12 prominent univariate return predictors identified by Green, Hand, and Zhang (2017) out of 94 firm characteristics. By applying both the time series portfolio analysis and Fama and MacBeth (1973) cross-sectional regression on pairs of sub-samples analyses, the results show that most characteristics fail to predict stock return persistently. In addition, I find potential statistical biases exist in testing the significance of the return spreads. Under the assumption that the model-adjusted performance of the lowest and the highest decile portfolios independently exist with no correlations, the corresponding t-values calculated are on average smaller in magnitude than those obtained from the regression tests, especially under the hurdle rate of 3.0 (Harvey, Liu, & Zhu, 2016).

Date of Award	8 Sept 2021
Original language	English
Awarding Institution	Univerisity of Nottingham
Supervisor	Weimin Liu (Supervisor) & Cherry-Yi Zhang (Supervisor)