Negativity Bias in Attention Allocation: Retail Investors’ Reaction to Stock Returns

• DOI: http://doi.org/10.1111/irfi.12180
• Author: Tomas Reyes
• Date: 01 March 2019
• Published date: 01 March 2019

Negativity Bias in Attention

Allocation: Retail Investors’

Reaction to Stock Returns*

TOMAS REYES

Department of Industrial and Systems Engineering, Pontificia Universidad Católica de

Chile, Santiago, Chile

ABSTRACT

We argue that negative stock market performance attracts more attention

from retail investors than comparable positive performance. Specifically, we

test and confirm the hypothesis that retail investors pay more attention to

negative extreme returns than positive ones. We present a measure of atten-

tion at the aggregate and company-specific levels using Google’s internet

search volume indexes. These measures correlate with, but are different from,

existing proxies of attention. Our empirical results strongly support the posi-

tion that investors display a negativity bias in attention allocation with

respect to extreme stock returns. Across all specifications, lagged negative

extreme returns are stronger predictors of high attention at the individual-

stock and stock market levels than positive ones.

JEL Codes: G02; G14; D19

Accepted: 18 December 2017

I. INTRODUCTION

Psychology research supports the notion that bad is stronger than good. Bau-

meister et al. (2001) argue that in most situations, negative events will produce

larger, more consistent or more intense consequences than positive events of

comparable magnitude. Anecdotally, human beings usually ask to hear the bad

news first, and bad news sells more newspapers. Testing whether attentional

resources are automatically directed away from the current task when inessen-

tial good or bad traits are present, Pratto and John (1991) find that a bad extra-

neous stimulus attracts more attention in an automatic and nonintentional

* The author is grateful to Stefano DellaVigna, Simon Gervais, Isaac Hacamo, Ulrike Malmendier,

Thomas Mertens, Atif Mian, Terrance Odean, Richard Stanton, Adam Szeidl, Paul Tetlock, Hal Var-

ian, and Wei Xiong for helpful comments. He also acknowledges financial support from Fondecyt

Iniciación (No. 11130647), Fondecyt Regular (Nos. 1160048 and 1171894), and Nucleo Milenio

Research Center for Entrepreneurial Strategy under Uncertainty (No. NS130028). Part of the work

was completed while the author was at UC Berkeley.

© 2018 International Review of Finance Ltd. 2018

International Review of Finance, 19:1, 2019: pp. 155–189

DOI: 10.1111/irfi.12180

fashion than a good stimulus. And in a study of how long positive or negative

everyday events continue to impact a person’s mood, Sheldon et al. (1996) con-

clude that negative information takes longer to process and contributes more to

the creation of impressions than positive information.

In this paper, we relate the negative–positive attention asymmetry found in

psychology to stock market behavior. We argue that negative stock market per-

formance draws more attention than comparable positive performance. Specifi-

cally, we measure performance using stock returns and test the hypothesis that

retail investors pay more attention to extreme negative returns than to extreme

positive ones.

There is an inherent challenge in directly measuring attention and its alloca-

tion across tasks. We measure attention in the stock market using Google Trends,

which provides a search volume tool that is a powerful proxy of attention for

two reasons. First, it is common for internet users to search for information using

Google, so the results of Google Trends are truly representative of their interest

in a topic. Additionally, since searching for a term on Google obviously requires

paying attention to it, search volume from Google Trends is a better proxy for

attention than alternative instruments used in the literature previously.

The results of Google’s search volume tool are expressed in terms of the

search volume index (SVI). The SVI for a search term is the percentage of

searches for that term throughout 1 week within a geographical region, scaled

by its time-series maximum. Data are available from January 2004 for most

common terms used in Google searches.

Multiple authors have used this data in different areas for modeling and pre-

diction. Ginsberg et al. (2008) employ Google’s indexes to predict flu outbreaks

more efficiently than the Centers for Disease Control and Prevention (CDC).

Choi and Varian (2012) use it to predict sales and tourism. Da et al. (2011) pro-

vide support for the Barber and Odean (2008)‘s price pressure hypothesis using

search data on ticker symbols. Da et al. (2015) use search volume to measure

investor sentiment and show that decreases in search volume are correlated

with price increases, which then reverse in the short term. Campos et al. (2017)

use search volume to model and predict the oil’s VIX, finding that search data

significantly increases the returns of volatility-exposed portfolios.

In this paper, we use three aggregate measures of investor attention to the

stock market based on Google search volume data. The first measure is Atten-

tion to the Stock Market (StockMarket), defined as the sum of the SVI values for

a series of terms such as “stock market”or “best stocks”that investors typically

search when seeking general information about the whole stock market. The

second measure, attention from potential market entrants (OnlineTrading), is

the sum of the SVI values for a series of terms such as “online trading,”“online

brokerage account,”or “best brokerage account,”and captures the tendency of

potential investors to enter the market, that is, to search for information on

opening a brokerage account. Finally, attention from existing investors (Etrade)

is a measure of retail investors who already have a brokerage account and use

Google to access its webpage and log in. It is defined as the sum of the SVI

© 2018 International Review of Finance Ltd. 2018156

International Review of Finance

values for a series of terms such as “etrade”or “ameritrade.”In addition to the

previous three indicators, we implement a measure of attention at the stock

level. Following Da et al. (2011) we use the SVI for the ticker symbols of large

companies in the S&P 500 (e.g., “YHOO”for Yahoo or “WMT”for WalMart).

After constructing these measures, we proceed as follows. First, we study

what drives SVI and how it relates to other indirect proxies for attention. We

find that aggregate US-level SVI measures have positive contemporaneous and

lagged correlations with trading volume and volatility. When we explore the

relationship between lagged returns and attention, we find that the measures

StockMarket and OnlineTrading display the greatest amount of attention to

extreme positive and negative returns. More importantly, lagged negative

extreme returns are stronger predictors of attention in the stock market than

positive extreme returns.

Second, we get SVI data from Google for each US state and construct in-state

versions of our attention variables. We then sort the companies by state using

location codes from Compustat. For each state and week, we construct a portfo-

lio of in-state firms with a high-market capitalization. As with the US-level

results, we find that individual investors show a negativity bias and pay more

attention to negative than positive extreme returns.

Third, we focus on attention to specific stocks and its relationship to individ-

ual stock returns. We use SVI for ticker symbols and stock-level market data to

create a panel with the 100 largest companies in the S&P 500 for which we

have complete data. Here again we find patterns supporting a negativity bias,

even after controlling for other proxies of attention.

Overall, our empirical results strongly support the view that investors display

a negativity bias in attention allocation with respect to extreme stock returns.

Across all specifications, lagged negative extreme returns are stronger predictors

of retail investors’attention than positive extreme returns. To the best of our

knowledge, this is the first time such a pattern has been documented for US

investors.

We perform several robustness checks. First, we show that the negativity bias

is present in subsamples of companies with both low- and high-institutional

ownership; however, this bias is amplified among companies with a larger frac-

tion of individual investors. This is consistent with previous findings that sup-

port SVI as a proxy for attention from individual investors (Da et al. 2011).

Second, when splitting the sample between groups of stocks with low and high

(lagged) abnormal trading volume, we find that investors pay more attention to

extreme negative returns than positive ones in both subgroups; however, they

display a greater amount of sensitivity to extreme returns that are accompanied

by high volume. Third, we find no evidence to support the possibility that our

results are driven by a negative media bias (that is, more media attention to

negative events than positive ones). Finally, we rule out the possibility that neg-

ative returns are stronger predictors of attention simply because they are more

unusual or because negative and positive returns are not symmetrical events to

stockholders.

© 2018 International Review of Finance Ltd. 2018 157

Negativity Bias in Attention Allocation