Investor Sentiment, Anomalies, and Macroeconomic Conditions

• Date: 01 December 2018
• Author: Dongcheol Kim,Haejung Na
• Published date: 01 December 2018
• DOI: http://doi.org/10.1111/ajfs.12237

Investor Sentiment, Anomalies, and

Macroeconomic Conditions*

Dongcheol Kim

Business School, Korea University, Republic of Korea

Haejung Na**

College of Business and Economics, California State University, Los Angeles, United States

Received 19 March 2018; Accepted 19 August 2018

Abstract

We examine whether the results supporting the sentiment-related overpricing story by Stam-

baugh, Yu, and Yuan (J. Financial Economics, v. 104, p. 288–302) are still valid after control-

ling for macroeconomic conditions. We no longer find the results consistent with the

sentiment-related overpricing story after adjusting for the effect of macroeconomic condi-

tions. The risk factors associated with macroeconomic conditions are mostly priced, and the

average return spread in the anomalies is largely accounted for by the expected return spread

implied by the risk factors. Their results might be a consequence of the use of an inade-

quately constructed sentiment index. It is premature to argue that the returns in the anoma-

lies are driven by investor sentiment.

Keywords Investor sentiment; Anomalies; Risk factors; Macroeconomic variables; Sentiment-

related overpricing

JEL Classification: G12, G14

1. Introduction

Investor sentiment is a psychological belief about future cash flows with excessive

optimism (high sentiment) or pessimism (low sentiment), and therefore may lead

to erroneous assessment of the value of assets, resulting in stock prices to deviate

from their fundamental values. According to Daniel et al. (1998), high (low)

*Kim was supported by the National Research Foundation of Korea Grant funded by the

Korean Government (NRF-2015-S1A5A2A01009547). This paper was a winner of the Best

Paper Awards at the 29th Asian Finance Association Annual Conference, July 2017. The

authors are grateful for the comments of the anonymous referee, and especially the Editor,

Kwangwoo Park.

**Corresponding author: California State University, Los Angeles, College of Business and

Economics, 5151 State University Drive, Los Angeles, CA 90032, USA. Tel: +1-323-343-2871,

email: hna5@calstatela.edu.

Asia-Pacific Journal of Financial Studies (2018) 47, 751–804 doi:10.1111/ajfs.12237

©2018 Korean Securities Association 751

investor sentiment is formed with good (bad) news related to the economy or

firms, which leads to overpricing (underpricing). In an efficient market, the mis-

pricing should be eventually corrected by arbitrageurs. Owing to short sales impedi-

ments, however, mispricing can persist. In particular, sentiment-driven overpricing

is more prevalent than underpricing. This could result in anomalies.

By combining market-wide investor sentiment with Miller’s (1977) argument

about the effect of short-sale impediments, Stambaugh et al. (2012) (hereafter,

SYY) suggest three hypotheses on the role of investor sentiment in affecting the

degree of mispricing in 11 anomalies widely known in the literature.

1

The first

hypothesis is that the anomalies should be stronger following high sentiment than

following low sentiment; the second hypothesis is that the returns on the short-leg

portfolio of each anomaly should be lower following high sentiment than following

low sentiment; and the third hypothesis is that investor sentiment should not

greatly affect the returns on the long-leg portfolio of each anomaly. In short, the

pattern in the returns of the anomalies is attributed to sentiment-related overpric-

ing, in particular for the stocks in the short-leg portfolio following high sentiment.

When sorting the returns in the anomalies according to the level (high or low) of

investor sentiment, SYY find empirical results consistent with their hypotheses and

conclude that sentiment-related overpricing is the main explanation for the 11 asset

pricing anomalies.

SYY use the investor sentiment index constructed by Baker and Wurgler (2006)

(hereafter, BW) as a measure of market-wide investor sentiment. The BW sentiment

index is constructed by estimating the first principal component of the six proxies

for investor sentiment orthogonalized by a set of macroeconomic variables to

remove business cycle variations from each of the proxies prior to principal compo-

nents analysis. Although BW attempt to remove business cycle variations from their

sentiment index, a substantial portion of the effects of macroeconomic conditions

could remain in the sentiment index. If this were the case, the sentiment index

would be correlated with the premium on the risk factor(s) associated with the

macroeconomic variables that are not included in the set of variables used by BW.

Further, if this risk factor is more sensitive to the stocks in each short-le g portfolio

but less to those in each long-leg portfolio, the results could be consistent with the

predictions of SYY’s hypotheses, whether or not investor sentiment plays a role in

explaining the returns of the anomalies. Thus, SYY’s results supporting their

1

The 11 anomalies examined in SYY are: (i) the failure probability of Campbell et al. (2008);

(ii) Ohlson’s (1980) O-score; (iii) the net stock issuance of Ritter (1991) and Loughran and

Ritter (1995); (iv) the composite equity issuance of Daniel and Titman (2006); (v) the total

accruals of Sloan (1996); (vi) the net operating assets of Hirshleifer et al. (2004); (vii) the

momentum of Jegadeesh and Titman (1993); (viii) the gross profitability premium of Novy-

Marx (2013); (ix) the asset growth of Cooper et al. (2008); (x) the return on assets of Fama

and French (2006) and Chen et al. (2011); and (xi) the investment to assets of Titman et al.

(2004) and Xing (2008).

752 ©2018 Korean Securities Association

D. Kim and H. Na

hypotheses might be mixed with the effect of such a risk factor. In that, SYY’s

results are obtained from using the BW sentiment index, it would be premature to

conclude that the return spreads between the long-leg and short-leg portfolios in

the anomalies are attributed to sentiment-related overpricing rather than to the

spread of risk.

The purpose of this paper was twofold. We first examine whether the results

supporting SYY’s hypotheses are maintained even after separating the effect of the

risk factor from SYY’s original results. We then examine to what extent the returns

of the anomalies can be accounted for by the expected returns implied by the risk

factor(s) associated with some macroeconomic variables. In the context of Merton’s

(1973) Intertemporal Capital Asset Pricing Model, we use a set of macroeconomic

variables as state variables that proxy for such a risk factor and that are possibly

linked to the investor sentiment index. As state variables, we select the following

five macroeconomic variables: the Lettau and Ludvigson (2001) consumption-

wealth ratio (cay), term spread, default spread, the 3-month Treasury-bill yield, and

inflation rate. These variables are not the ones BW used to remove business cycle

variations from the proxies for investor sentiment, although they are often used as

stock return predictors in the literature.

To examine the first purpose, we decompose the BW sentiment index into the

predicted sentiment index by the 1-month-lagged five macroeconomic variables and

the residual sentiment index comprising the BW sentiment index component

remaining after removing the component linked to the macroeconomic variables.

We find that when the predicted sentiment index is used, results consistent with

SYY’s three hypotheses are obtained. The BW sentiment index component predicted

by the five macroeconomic variables has as much predictive power for the returns

of the short-leg, long-leg, and long-short portfolios in the anomalies, as does the

BW sentiment index itself. However, results that are consistent with SYY’s hypothe-

ses can no longer be obtained once the residual sentiment index is used. Our results

are contradictory to those of SYY, since they report that even after controlling for

these five macroeconomic variables, the results are consistent with their hypotheses.

We find that the discrepancy in the results between the two studies is attributed to

the difference in the estimation window used to control the macroeconomic vari-

ables for the sentiment index. We use the 60-month rolling window due to nonsta-

tionarity of the parameters, while SYY use the whole-period window.

To examine whether components of macroeconomic conditions are successfully

removed in the BW sentiment index used in SYY, following BW, we construct a

new investor sentiment index by estimating the first principal component of BW’s

six sentiment proxies orthogonalized by the five macroeconomic variables used in

this study. We find that when this new sentiment index is used, the predictability

of the sentiment index for the returns in the anomalies disappears. Note that these

five variables are not used in BW to remove components of macroeconomic condi-

tions. These results indicate that the BW sentiment index still contains components

of macroeconomic conditions, and a substantial portion of the predictability of the

©2018 Korean Securities Association 753

Investor Sentiment, Anomalies & Macroeconomy