Ambiguity and risk factors in bank stocks

• Published date: 01 December 2023
• Author: Luis García‐Feijóo,Ariel M. Viale
• Date: 01 December 2023
• DOI: http://doi.org/10.1111/jfir.12346

Received: 6 July 2021

|

Accepted: 22 June 2023

DOI: 10.1111/jfir.12346

ORIGINAL ARTICLE

Ambiguity and risk factors in bank stocks

Luis García‐Feijóo

1

|Ariel M. Viale

2

1

Department of Finance, Florida Atlantic

University, Boca Raton, Florida, USA

2

Marshall E. Rinker School of Business, Palm

Beach Atlantic University, West Palm Beach,

Florida, USA

Correspondence

Ariel M. Viale, Marshall E. Rinker School of

Business, Palm Beach Atlantic University,

901 South Flagler Dr., West Palm Beach,

FL 33416‐4708, USA.

Email: vialeam@pba.edu

Abstract

The determinants of banks' cost of equity are not well

understood. We depart from prior work assuming

rational expectations and instead explore the impact of

Knightian uncertainty or ambiguity on bank stocks. We

test a large set of asset pricing models and find that

investors' lack of confidence in both the drift and

correlation structure driving bank stock returns affects

banks' cost of capital. We also investigate the economic

relation among ambiguity, market liquidity, and banks'

capital shortfall, which reveals the transmission channels

through which ambiguity may increase the probability of

a systemic crisis. Our findings have implications for

macroprudential policy.

JEL CLASSIFICATION

C58, G01, G12, G21

1|INTRODUCTION

Identifying common asset pricing factors in the cross section of bank stock returns is important for both asset

pricing and macroprudential policy reasons. Banks make up a substantial fraction of the equity market, and

their stability is essential for the well‐being of the real economy. Achieving financial stability through an

effective macroprudential policy, however, requires understanding the factors driving banks' cost of capital.

Furthermore, recent influential literature argues that banks are the de facto marginal investor of asset pricing

theory (see, e.g., Adrian et al., 2014;He&Krishnamurthy,2012,2013;Heetal.,2017). Yet, empirical asset

pricing models for nonfinancial stocks perform poorly when tested against the “holdout”sample of bank

stocks.

Most of the few asset pricing studies focusing on banks stocks assume rational expectations (see, e.g., Adrian

et al., 2015; Barber & Lyon, 1997; Chira et al., 2013; Gandhi & Lustig, 2015; Schuermann & Stiroh, 2006; Viale &

J Financ Res. 2023;46:993–1019. wileyonlinelibrary.com/journal/JFIR

|

993

© 2023 The Southern Finance Association and the Southwestern Finance Association.

Madura, 2014; Viale et al., 2009).

1

Although the rational expectations hypothesis assumes that investors know the

probability law governing asset returns, authors such as Keynes (1921), Knight (1921), Shackle (1949), and Roy

(1952) have emphasized that decision makers form expectations based on vague information that cannot be

quantified precisely. Related evidence from experimental studies (Ellsberg, 1961) has confirmed that individuals are

averse not only to uncertainty regarding the outcome of events with known probabilities (risk) but also uncertainty

regarding the outcome of events with unknown probabilities (ambiguity). Consequently, there is an increasingly

influential body of literature that examines the implications of ambiguity for portfolio selection and asset pricing.

2

Moreover, as Hansen (2014) and Hansen and Marinacci (2016) point out, economic modeling needs to account

for three kinds of uncertainty faced by the decision maker: uncertainty within the model (risk), uncertainty across

models (ambiguity), and uncertainty about the model (model misspecification). Asset pricing models based on

rational expectations have been shown to perform poorly empirically when results are adjusted for model

misspecification (Harvey et al., 2016; Kan et al., 2013; Lewellen & Nagel, 2006).

We conjecture that ambiguity is priced in the cross section of bank stocks and hence it is relevant for

macroprudential policy. Work on ambiguity and bank stocks is scarce, and the few studies addressing it are based

on small samples. Boyarchenko (2012) tracks the dynamics of credit spreads for nine financial institutions around

the 2007–2008 crisis; she finds results consistent with the notion that spreads increased because of ambiguity.

Driouchi et al. (2020) use option data on 12 major US banks between 2003 and 2010; they find that ambiguity is

associated with banks' higher exposure to market and credit risks. We expand the ambiguity literature to the entire

cross section of bank stocks. Furthermore, Breuer and Csiszár (2013) argue for a systematic approach to searching

for the worst case scenario in bank stress tests. Their approach based on multiple plausible scenarios that are

described by an entropy ball is similar to the multiple‐priors approach in the ambiguity literature.

We start by deriving an asset pricing equation under ambiguity that will frame our investigation theoretically.

We adopt the multiple‐priors approach of Epstein and Wang (1994) and Epstein and Schneider (2010), in which

investors doubt the data‐generating process driving the drift and correlation structure of bank stock returns. The

impact of correlation ambiguity is unexplored in the empirical asset pricing literature; however, recent theoretical

work shows that both ambiguity about expected returns and correlation can affect asset prices (Epstein &

Halevy, 2019; Huang et al., 2017; Liu & Zeng, 2017).

In the empirical implementation, we adapt the learningunder ambiguity model of Viale et al. (2014)andusea

multifactor model that includes the market return and innovations in the level and slope of the term structure of

interest rates, which we augment with a joint dynamic latent asset pricing factor for investors' lack of confidence in

the drift and correlation structure of bank stock returns. We refer to this model as the robust ICAPM (R‐ICAPM).

Using the standardtwo‐pass methodology of Fama–MacBeth(1973), we find the model can explainthe cross section

of bank stock returns even after we consider model misspecification following the approach in Kan et al. (2013).

The literature shows that standard risk‐based asset pricing models, such as the Fama–French (1993)three‐factor

model, perform poorly on the cross section of bank stock returns (Viale et al., 2009).

3

Recently, He et al. (2017)findthata

two‐factor asset pricing model that includes the stock market return andan intermediary capital risk factor can explain the

cross section of many asset classes. We also find that a two‐factor intermediary‐based asset pricing model that includes

the market return and a proxy for banks' state‐dependent leverage can explain the cross section of bank stocks. However,

when we compare the relative performance of this model against the R‐ICAPM using the pairwise cross‐sectional R

2

comparative test developed by Kan et al. (2013)fornon‐nested models, we find that the R‐ICAPM is superior to the risk‐

only two‐factor intermediary‐based asset pricing model. Overall, the empirical results support our conjecture that both

dimensions of ambiguity (on the drift and correlation of stock returns) are important determinants of banks' cost of equity.

1

One notable exception is Baker and Wurgler (2015), who adopt a behavioral approach.

2

For excellent surveys, see Epstein and Schneider (2010) and Guidolin and Rinaldi (2013).

3

In unreported results, we confirm the underperformance of the three‐factor model, five‐factor model (Fama & French, 2015) plus momentum, and the

q‐model of Hou et al. (2015).

994

|

JOURNAL OF FINANCIAL RESEARCH