Sentiment and its asymmetric effect on housing returns

• Author: Ramya Rajajagadeesan Aroul,Sanjiv Sabherwal,Sergiy Saydometov
• Published date: 01 October 2020
• DOI: http://doi.org/10.1002/rfe.1097
• Date: 01 October 2020

580

|

wileyonlinelibrary.com/journal/rfe Rev Financ Econ. 2020;38:580–600.

© 2020 University of New Orleans

1

|

INTRODUCTION

The study of cognitive psychology sheds light on how people reason, make choices, and allocate their attention, among

other things. The psychology literature argues that people's emotions affect information processing and decision-making.

In particular, negative mood states significantly affect people's decisions. Buying or selling a home is one of the most

important economic decisions people face in their lives. Financial losses in housing can cause large and irreparable conse-

quences to homeowners. With the high stakes associated with financial decisions in the real estate market, it offers a unique

laboratory to assess the potential impact of sentiment. In financial markets, market sentiment, broadly referred to as the

overall expectations and beliefs of market participants towards a particular security or market, is believed to be an important

determinant of asset prices.1

In this paper, we propose to construct a sentiment measure to examine the impact of sentiment

on real estate—one of the most distinctive and important asset classes that typically comprises a significant portion of one's

overall wealth.

Unlike the stock market, the housing market features a relatively higher percentage of individual investors, extremely seg-

mented and localized markets, higher levels of information asymmetry, and lack of short-sale mechanisms, all of which make it

highly likely to be affected by investor sentiment (Clayton, Ling, & Naranjo, 2009; Hui & Wang, 2014). There have been justi-

fiable discussions pertaining to the impact of sentiment on residential real estate pricing. However, published empirical studies

that we are aware of fall short of providing a comprehensive way to measure sentiment of market participants in the residential

real estate market by directly observing their behavior.

Received: 23 June 2019

|

Revised: 18 November 2019

|

Accepted: 22 December 2019

DOI: 10.1002/rfe.1097

ORIGINAL ARTICLE

Sentiment and its asymmetric effect on housing returns

SergiySaydometov1

|

SanjivSabherwal2

|

Ramya RajajagadeesanAroul3

1College of Business, Dallas Baptist

University, Dallas, TX, USA

2Department of Finance and Real Estate,

University of Texas at Arlington, Arlington,

TX, USA

3Department of Economics and Finance,

Texas A&M University Commerce,

Commerce, TX, USA

Correspondence

Sanjiv Sabherwal, Department of Finance

and Real Estate, University of Texas at

Arlington, 701 S. West Street, Suite 434,

Arlington, TX 76019, USA.

Email: sabherwal@uta.edu

Abstract

We use Google search frequency to construct sentiment indices (positive and nega-

tive) for the housing market. We find that future housing prices are negatively related

to our measure of negative sentiment but not significantly related to that of positive

sentiment. These relationships are robust to controls for macroeconomic variables,

stock market return, and Housing Market Index, a survey-based housing sentiment

index. Furthermore, we find that an increase in negative sentiment results in a sig-

nificant negative response in housing prices, while a decrease evokes little response.

Thus, the housing market exhibits asymmetric responses to negative and positive

sentiment and to increases versus decreases in negative sentiment. We attribute these

asymmetric responses to the “negativity effect.” Finally, we find that home prices are

more sensitive to sentiment during recessionary periods.

KEYWORDS

housing returns, individual behavior, negativity effect, sentiment

JEL CLASSIFICATION

G10; G14; R30

|

581

SAYDOMETOV ET Al.

Prior studies do show that behavioral factors play an important role in explaining changes in the housing markets (Diaz &

Hansz, 2007; Mayer & Sinai, 2009). Price changes in housing markets depend on the behavior and attitudes of millions of indi-

vidual homebuyers who are directed by their beliefs (Case, Shiller, & Thompson, 2012). Cen, Lu, and Yang (2013) argue that

investors subject to market sentiment hold a biased belief in the aggregate. However, finding an accurate empirical measure of

people's sentiment towards housing markets at the aggregate level is not an easy task. Homebuyers' beliefs are unobservable,

not easily quantifiable and are very difficult to obtain.

Though there are multiple sentiment measures available that gauge general public's level of optimism or pessimism, most of

these indices rely heavily on conducting periodic surveys. Therefore, while some of these indices work well in some circum-

stances, they suffer from the usual limitations of survey-based indices.2

In addition, the few sentiment proxy measures available

for the stock market (see Baker & Wurgler, 2007 for a comprehensive list) are not suitable for housing markets, because of its

unique characteristics such as low liquidity, high transaction costs, and short selling constraints among others. In view of this

gap, one of the main goals of this study is to construct a new measure of housing market sentiment that captures general public's

expectations and beliefs, and then examine the relationship between this measure of sentiment and housing prices.

Specifically, we use internet search data (search terms related to real estate) on aggregate search frequencies obtained via

Google Trends to capture the behavior of market participants, and then construct a monthly national sentiment index based on

that data over our study period of January 2004 to December 2014. Then, we investigate the ability of this index to predict the

changes in the national home price index. Following prior studies such as Da, Engelberg, and Gao (2015) and Kogan, Levin,

Routledge, Sagi, and Smith (2009), we use a historical, regression-based approach to identify the most relevant search terms to

be included in the sentiment measure. This approach has several benefits, such as objectively letting the data speak for itself.

We restrict the search interest data to the United States because most of the variables of interest in this study are related to the

U.S. As a result, our measure of sentiment gauges the sentiment of American households towards the real estate market in the

U.S. We call this measure the Real Estate sentiment index, and in short, the Sentiment index.

In this paper, we objectively separate the most relevant real estate search terms into “negative” and “positive” based on their

relationship with the real estate market, and construct separate indices using terms with negative or positive sentiment. We also

report the main regression results for both negative and positive indices. However, we focus more on the negative index be-

cause we observe that negative terms have a much stronger economically significant relationship with the housing market than

positive terms. In fact, the sentiment index constructed using negative terms is significantly negatively related to future hous-

ing prices, whereas the sentiment index constructed using positive terms is not significantly related to future housing prices.

This finding is similar to Tetlock (2007) and Da et al. (2015). Tetlock constructs a media factor based on a popular Wall Street

Journal column to examine the interactions between the media and the stock market. He finds that negative or pessimistic terms

in the media have the most stock market predictability. Similarly, Da et al. (2015) find that negative terms are the most useful

for identifying sentiment and they use only the negative terms in constructing their stock market sentiment index.

Prior studies such as Edmans, Garcia, and Norli (2007) and Akhtar, Faff, Oliver, and Subrahmanyam (2011), Akhtar, Faff,

Oliver, and Subrahmanyam (2012) have shown that the stock market has a negative response to bad sentiment news but a neg-

ligible or weaker positive response to good sentiment news. Such an analysis has not been done for the real estate market. In

this study, as mentioned above, we find that an increase in negative sentiment is associated with a decrease in housing prices

but an increase in positive sentiment is not associated with an increase in housing prices. This asymmetric response of housing

prices to increases in negative versus positive sentiment suggests the presence of a negativity effect in the real estate market.

We further analyze this negativity effect in the housing market by focusing on the negative sentiment and examining whether

the magnitude of the impact of an increase in our negative sentiment index on housing prices is greater than the impact of a

decrease in the index. To the best of our knowledge, this is the first study to directly examine and provide evidence of the

negativity effect in the real estate market. We find that an increase in negative sentiment index results in a stronger response in

housing prices than a decrease in this index. When market participants become more pessimistic about the real estate market, as

reflected in an increase in the negative index, housing prices decline over the next month. In contrast, when there is a decrease

in the negative index, housing prices do not show a significant increase. This finding of a strong negative reaction to a deterio-

ration in negative sentiment but a negligible reaction to an improvement in negative sentiment provides further evidence of an

asymmetric response to negative versus positive changes in sentiment.

Our conclusion of a negativity effect in the real estate market, which is based on a comprehensive analysis of hundreds of

real estate terms, differs from the finding of Beracha and Wintoki (2013). Beracha and Wintoki look at two keywords “real

estate” and “rent.” They propose that “real estate” reflects positive sentiment whereas “rent” reflects negative sentiment. They

find that the search volume for “real estate” is predictive of future home prices but not for “rent.” They conclude that a change

in positive sentiment is associated with home prices but a change in negative sentiment is not. In contrast, we find evidence of

a negativity effect in the housing market, similar to prior evidence for the stock market.