Sector-level Attention And Stock Returns

Traditional asset pricing models typically assume that the market is able to fully incorporate all new information into asset prices instantaneously.The underlying premise,however,is that investors are capable to pay close attention to all information and process it in a timely and effective manner.Recently,numerous studies have shown that investors have limited attention and information processing capacity(Hirshleifer and Teoh,2003;Feng,2017).With the development of the Internet technology,we now live in an era of information explosion.Therefore,how investors allocate their limited attention has become an important research topic.Rooted in studies on behavioral finance,scholars have proposed a variety of attention proxies to study the impacts of attention on investor behavior and asset prices,providing a theoretical and empirical framework for the study of the relationship between attention and asset pricing(e.g.,Barber and Odean,2008;Da et al.,2011;Yu and Zhang,2012;Yang et al.,2016;Yang and Guo,2019).However,most of these studies focus on individual stock-specific attention.Although it is well-known that in the reality investors with attention constraints,who may totally ignore stock-level information in extreme cases,are inclined to pay more attention to sector-wide information such as news about regions and industries(Peng and Xiong,2006),no scholars have explored the economic implications of attention to locations and industries.This paper seeks to fill this gap in literature.“Sector” in capital markets refers to a group of stocks that share similar characteristics.Classifying stocks into sectors then allocating capital at the level of sectors is an important way for attention-constrained investors to simplify investment decisions.This categorical thinking leads to “Sector Phenomenon”,that is,the returns of stocks that belong to the same sector will comove(Barberis and Shleifer,2003;Peng and Xiong,2006).The China’s stock market which is dominated by retail investors is characterized by weaker information environment(Jones et al.,2020).Retail investors usually have more limited attention thus are more likely to show categorical thinking and sector herding behavior(Jame and Tong,2014),intensifying the sector phenomenon in China’s market(He,2001;Hu et al.,2015).Region and industry are the two most intuitive sectors.Numerous studies have also shown that the China’s stock market is distinguished by significant industrial rotation effect and regional rotation effect(Peng and Zhang,2003;Li et al.,2009).Therefore,exploring the regional and industrial sector phenomenon from the perspective of attention is helpful for the understanding of the logic and pricing mechanism of China’s stock market.Referring to several theoretical hypotheses on investor attention,this paper extends previous studies that focus on individual stock-level attention to the analysis of sector-level attention,and empirically explores how sector-level attention affects the cross-section of stocks returns from the perspective of attention to locations and attention to industries.Additionally,this paper also compares whether it is stock-level attention or sector-level attention that is more important in determining asset prices.For attention to locations,this paper explores its implications from the perspectives of both media attention to a firm’s headquarter cities and media attention to a firm’s operationally relevant cities,because headquarter is the place where most of firm’s business happens,on the other hand,however,the economic interests of firms are reasonably spanned across several regions.For attention to industries,this paper investigates the influence of investor attention to industries using search volume index as a proxy for attention.Specifically,the research contains three parts.First of all,focusing on the asset-pricing implications of attention to locations,the first research topic of this paper is “Attention to locations and the cross-section of stock returns:From the perspective of attention to headquarters”.How media attention to individual stocks affects stock prices has been explored in several studies.According to Merton’s(1987)investor recognition hypothesis,media coverage may increase the exposure of a stock and thus helps expanding its investor base.Therefore,investors will demand higher returns for stocks that receive less media coverage to compensate for incomplete diversification of risk,leading to“no-media premium”.On the other hand,according to Barber and Odean’s(2008)attentiondriven buying pressure hypothesis,media coverage attracts investors to buy stocks,resulting in positive buying pressure for stocks covered by media.Combined with short-selling constraints,this implies stocks that are covered heavily by media shall have higher expected returns.Using news extracted from China News Service during the period between January 2008 and December 2017,this paper finds that stocks in cities with higher abnormal attention outperform stocks in cities with lower abnormal attention by 7.92% per year,and the positive relation between attention and stock returns holds even after controlling for firm characteristics,firmlevel attention proxies and local economic conditions.Moreover,the initial premium derived from attention to headquarters is completely reversed within three months,indicating that the price increase is a short-term phenomenon,which is consistent with the attention-driven buying pressure hypothesis.Further robustness checks rule out endogeneity problem,concerns for extreme market conditions and city size,problems associated with firm-level regressions,doubts about the measure of attention and the confounding influences of stock-related news.Secondly,the focus is still on the asset-pricing implications of attention to locations,combined with the citation fractions of cities in firms’ annual reports,the second part of this paper is “Attention to locations and the cross-section of stock returns: From the perspective of attention to firms’ operationally relevant cities”.The geographical distribution of listed companies forms an economic network connecting different regions.Scholars typically identify the economic linkage between firms and regions by conducting textual analysis of the annual reports of listed firms.The more a city is mentioned in the annual report,the stronger the linkage between the city and the firm.Following existing studies,this paper constructs an economic relevance weighted attention proxy based on firm-city citation fractions.The empirical results show that the long-short strategy based on this indicator generates an annualized return of5.72%.Comparing the relative return predictive power of this new attention index and the attention index of the headquarter cities,this paper finds that media attention to firms’ economically relevant cities shows stronger impacts on stock prices than attention to headquarter cities,which in turn has a stronger impact on stock prices than attention to individual stocks.Therefore,economic relevance weighted attention index seems contains more information than the unweighted one.Consistent with the attention hypothesis,of course,attention to firms’ economically relevant cities does not exhibit significant persistence.What’s more,the positive correlation between attention and stock returns will completely reverse within three months.A series of tests reveal that the main results are robust after removing extreme samples and interfering cities,as well as altering the way to derive the attention index.Finally,focusing on the asset-pricing implications of attention to industries,the third part of this paper is “Attention to industries: Based on Baidu search volume index”.Using aggregate search frequency from the Baidu Index based on the six-digit CITIC industry names as a proxy for industry-specific attention,this paper finds that an increase in the Baidu search volume index for an industry predicts higher stock returns the next day.A strategy that goes long on stocks in industries with higher abnormal search volume index values and simultaneously short on stocks in industries with lower abnormal search volume index values generates a valueweighted average return of 4.28% per year.Within the framework of Barber and Odean’s(2008)attention hypothesis,this paper compares the degree of price pressure for companies with different short-selling constraints,and finds that the price pressure is stronger among companies with more binding short-selling constraints,which is consistent with the attention hypothesis.Further results show that the initial price increase will reverse in the following few days,suggesting that the price increase is driven by attention rather than fundamental information.More importantly,this part compares the relative strength of industry-and stock-level attention proxies in predicting stock returns and finds that it is the industry-level attention rather than the stock-level attention that will dominate.Specifically,the Fama-Mac Beth regression results show that the positive relationship between Baidu search volume index for individual stocks diminishes once controlling for industry-level attention,but not vice versa.Further research indicates that there is a lead-lag relationship between industry-wide attention and stock-level attention,and it is the leading of industry-wide attention to stock-level attention that attributes to the relatively stronger stock return predictability power of industry attention.Finally,additional robustness tests show that the main conclusions are not affected by extreme market conditions and how the abnormal search volume indexes are constructed.This paper contributes to previous studies in several ways.First of all,this paper enriches research on investor attention.It is the first work that proposes and quantifies attention given to locations and industries,and also the first paper that directly compares the relative strength of more macro attention indexes such as location-and industry-wide attention and stock-level attention indexes in predicting stock returns.Secondly,this paper complements research on the relationship between regional factors and stock returns.Previous studies have documented various regional factors that affect stock returns,such as the degree of financial development,local macroeconomic conditions,political uncertainty,economic policy uncertainty,and so on.Compared with these studies,this paper proposes a new channel that impacts the price dynamics of stocks within an area—that is,attention-induced buying pressure.Thirdly,this paper also adds to the rapidly growing literature that employs disclosure-based measures to investigate the economic implications of the geographical segmentation of the financial markets.Previous studies explore the influence of the economic activity forecasts of relevant regions on stocks.This paper distinguishes from these studies linking geographical dispersion with media attention and offer the first evidence that media attention to a firm’s relevant regions,which is more important than information of the mere headquarter state,carry significant predictive power for average stock returns.Fourthly,this paper is related to studies that explore the assetpricing implications of industry properties.Despite the clear insights about the critical role of industries in asset allocations,the empirical evidence on the link between industry characteristics and stock returns is rather limited.This paper seeks to fill this gap in the literature by offering evidence of a previously uncovered dimension: the relationship between industrywide attention and the cross-section of stock returns.Additionally,the results of this paper carry significant practical lessons.The empirical results of this paper show that investors are able to obtain an excess annual return of 7.92%,5.72%,and 4.28%,respectively,by employing strategies based on media attention to a firm’s headquarter city,media attention to a firm’s operationally relevant cities,and investor attention to industries.Therefore,investors can keep track of these indicators so as to identify hot regions and industries to seek a reference for their investment decisions.For financial practitioners,the attention indexes proposed in this paper are helpful for them to analyze the market and understand stock valuations.For regulators,attention to locations and industries indicate market sentiment,therefore can be employed to predict capital flows so as to guide investors to invest in a rational manner,and in turn improve supervision quality.The Chinese stock market is dominated by retail investors.A profound understanding of the impacts of attention on investor behavior is of great significance for understanding the mechanism of China’s securities market and promoting its stability.