View
2
Download
0
Category
Preview:
Citation preview
Are Institutions Informed About News?
Terrence Hendershott1 Dmitry Livdan1 Norman Schürhoff2
1University of California Berkeley2University of Lausanne, Swiss Finance Institute and CEPR
December 2014
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 1 / 31
Are institutions informed about news?
1 Stock prices aggregate dispersed informationfrom many different news sources: Firm; Media; Macro; Prop Researchthrough trading by many different participants: Institutions, insiders, retail(Grossman and Stiglitz, 1980; Kyle, 1985; Glosten and Milgrom, 1985)
2 Institutions play active governance role + institutional ownership >50%
0"
5"
10"
15"
0%"10%"20%"30%"40%"50%"60%"
1950"1960"1970"1980"1990"1995"2000"2005"2009"
Ins0tu0onal"equity"($"bn)" %"of"total"
This paper studies link between institutional trading, news & stock returns
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 2 / 31
Are institutions informed? Case of Martha Stewart
Martha Stewart was CEO and chairwoman of Martha Stewart LivingOmnimedia (MSO)She was indicted on charges of securities fraud, trial started on Jan 27,2004.
M. S
tew
art t
rial b
egin
s
Fra
ud c
harg
e di
smis
sed
Cou
rt v
erdi
ct
-30
-20
-10
010
Ret
urn
(%)
01feb2004 01mar2004 01apr2004Date
-20
-15
-10
-50
IOF
(%
)
01feb2004 01mar2004 01apr2004Date
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 3 / 31
Are institutions informed?
New York Times, “Can settlements actually level the playing field for investors?”, 23 Dec 2002
“For years, Wall Street’s dirty little secret was that its research was devisedexpressly for two key constituencies: its institutional investors and its corporateclients. If the individual investor wanted to join the party, well, caveat emptor.”
1 Better access to informationDirect line of communication with publicly traded firms & brokeragesEmploy buy-side analysts & enjoy relationships with sell-side analysts
2 Better information processingEconomies of scale allow monitoring many sources of informationEmploy professionals & technologies with superior data processing skills
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 4 / 31
No consensus on informed institutional tradingEmpirical evidence for YES:
1 High lead low institutional ownership stock returns - Badrinath, Kale, Noe (1995)2 Higher institutional holdings associated with more efficient pricing - Sias and
Starks (1997), Boehmer and Kelly (2009)3 Profitable institutional buying beginning 5 days prior to release of analysts’
initial reports with positive recommendations - Irvine, Lipson, Puckett (2007)4 Institutional trading predicts returns at firm, industry & market levels -
Boulatov, Hendershott, and Livdan (2011)
Empirical evidence for NO:1 Institutional not net buyers in target firms prior to acquisitions & trading not
related to future earnings announcement returns - Griffin, Shu, Topaloglu (2011)2 Institutions’ trading strategy around takeover announcements does not yield
significant abnormal returns - Jegadeesh and Tang (2010)3 Institutions are not able to differentiate between good and bad stock
recommendations - Busse, Green, Jegadeesh (2010)4 Short sellers do not anticipate news - Engelberg, Reed, Ringgelberg (2010)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 5 / 31
Our paper uses comprehensive news & trading data
Combine daily non-public data on buy & sell volume by all institutionswith Reuters news data feed & news sentiment for 2003–05
1 Comprehensive list of >1,600 NYSE-listed stocks
2 Complete trading records for all institutional investors (aggregate IVol & IOF)
3 All news announcements from Reuters news feed + textual sentimentanalysis
More complete picture and, more importantly, study joint behavior ofinstitutional trading, public news releases & stock returns
1 Occurrence of news announcements
2 Direction & sentiment of news content
3 Stock returns on announcement days
4 What types of news are institutions more/less informed about?
5 How do institutions trade on news?
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 6 / 31
Our findings
1 Institutional trading predicts:1 Occurrence of news announcements (IVol rises before news)
2 Direction & sentiment of news (Net buying/selling before pos./neg. news)
3 Stock returns on announcement days, but also on non-news days
4 Real performance (IBES earnings announcement surprises)
2 True for specific news types, but less statistically significant
3 Institutions predict company crisis (BKRT, MNGISS, JOB, CRIM, JUDIC)
4 Press releases & sentiment reversals: Do not trade on company hype
5 Trade on firm-specific news more than on industry, market, macro news
6 Significant fraction of price discovery occurs before public news release
→ Consistent with model of strategic trading on long-lived asymmetricinformation with correlated public news releases (a la Kyle)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 7 / 31
Data description
1 Daily comprehensive institutional trading data (NYSE CAUD)Institutional volume in each NYSE stock
Institutional order flow imbalance in each NYSE stock
2 Take-by-take news releases on the Reuters Data Feed (RDF)News announcements (un-)scheduled & news generating process
News sentiment from Reuters NewsScope Sentiment Engine (RNSE)
3 CRSP stock returns + IBES accounting dataEarnings actual/surprises (SUE score)
Profitability & performance
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 8 / 31
Institutional trading data
NYSE Consolidated Equity Audit Trail (CAUD) for 2003-05Comprehensive daily trading data: 756 trading days in 1,667 NYSE stocksNon-public data; part of TORQ by J. Hasbrouck (144 stocks, Nov90–Jan91)Similar to TAQ, but 2 extra files: CD (audit trail) + SOD (system order data)Started after 87 crash; Used in other papers, e.g., Kaniel, Saar, Titman(2008, 2011) for “individuals”; Boehmer & Kelly (2009) and Boulatov,Hendershott, Livdan (2011) for institutions
Each transaction has code that identifies type of buyer & seller(proprietary/agency, program trade, index arbitrage, individual investor)
Institutional order flow imbalance is IOFi,t = IBuysi,t - ISalesi,t with
IBuysi,t =
# of Buysi,t∑n=1
$Buysni,t
MCi,t−250, ISalesi,t =
# of Salesi,t∑n=1
$Salesni,t
MCi,t−250.
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 9 / 31
News + sentiment data
Reuters Data Feed (RDF) – used by HFT & algo trading programs537K unique stories with 1-45 takes for 1-38 firms (126K story firm-days)Takes: alert, article, append, overwriteData: Unique story/take identifier, item type, time stamp, source, story text
Reuters NewsScope Sentiment Engine (RNSE)Stories classified into >70 topic codes (RES, MRG, HOT, MNGISS, CRIM)Most common topics: RES (Corporate Results, 46K), RESF (Results Forecasts,40K), STX (Stock Markets, 37K), MRG (Mergers & Acquisitions, 36K), DBT (DebtMarkets, 33K), NEWS (Major Breaking News, 30K), BACT (Business Activites, 28K)
RNSE uses 3-stage natural language processing (lexical analysis,linguistic parsing, machine learning) to measure news tone & relevance
Sentiment score ∈ [0, 1]: Pr(positive), Pr(neutral), Pr(negative)Relevance score ∈ [0, 1]: Calculated from firm mentions in headline & textRelevance-weighted news sentiment for each take, story & daily average
Net sentiment = 1No. Takes
∑Takes (Pr(positive) - Pr(negative))*Relevance
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 10 / 31
RNSE news announcements:
ticker
story_id
seqnofirm
0count
date
time
relevance
sent_pos
sent_neut
sent_neg
sent
item_type
attribtn
size
bcast_text
SLE 1 1 1 25Jan05 15:04:41 1 0.272 0.136 0.593 I1 ARTICLE RTRS 3197 UPDATE03ISara0Lee0cuts0fiscalIyear0forecast;0shares0dropSLE 2 1 1 25Jan05 15:10:56 1 0.218 0.611 0.171 0 ALERT RTRS 0 SARA0LEE0<SLE.N>0CEO0SAYS0CONSIDERING0FURTHER0"STRATEGIC0PORTFOLIO0INITATIVES"SLE 2 2 1 25Jan05 15:12:09 1 0.216 0.611 0.173 0 ALERT RTRS 0 SARA0LEE0CEO0SAYS0COMMODITY0COSTS0NOW0SEEN0$3500MILION0HIGHER0IN0FY05SLE 2 3 1 25Jan05 15:12:53 1 0.112 0.235 0.653 I1 ALERT RTRS 0 SARA0LEE0CEO0SAYS0PRICING0OPTIONS0"VERY0LIMITED"SLE 2 4 1 25Jan05 15:13:55 1 0.852 0.118 0.030 1 ALERT RTRS 0 SARA0LEE0CEO0SAYS0TO0DETAIL0PORTFOLIO0INITIATIVES0BY0FEBRUARY0ANALYSTS0MEETINGSLE 2 5 1 25Jan05 15:25:03 0.707 0.842 0.123 0.035 1 ARTICLE RTRS 866SLE 3 1 1 25Jan05 16:25:39 1 0.078 0.155 0.767 I1 ARTICLE RTRS 4034 UPDATE04ISara0Lee0cuts0fiscalIyear0forecast;0shares0dropSLE 4 1 1 26Jan05 15:23:32 1 0.056 0.125 0.819 I1 ARTICLE RTRS 495 RESEARCH0ALERTIJP0Morgan0lowers0Sara0Lee0Corp.0to0"neutral"SLE 5 1 1 27Jan05 17:30:09 1 0.826 0.130 0.044 1 ARTICLE BSW 1548 Sara0Lee0Corporation0Declares0236th0Consecutive0Quarterly0Dividend0<SLE.N>SLE 6 1 1 27Jan05 17:31:05 1 0.833 0.127 0.040 1 ARTICLE BSW 1511 REGISara0Lee0Corporation<SLEq.L><SLE.N>0Dividend0DeclarationSLE 7 1 1 27Jan05 17:31:51 1 0.596 0.397 0.007 1 ALERT RTRS 0 SARA0LEE0CORP0<SLE.N>0DECLARES0DIVIDENDSLE 7 2 1 27Jan05 17:32:07 1 0.853 0.117 0.030 1 ALERT RTRS 0 SARA0LEE0CORP0<SLE.N>0STOCK0PROVIDES0ANNUALIZED0DIVIDEND0YIELD0OF0MORE0THAN03.5%SLE 7 3 1 27Jan05 17:33:39 1 0.555 0.438 0.007 1 ALERT RTRS 0 SARA0LEE0CORP0<SLE.N>0DECLARES0DIVIDEND0$0.1975/SHR,0UNCHANGEDSLE 7 4 1 27Jan05 17:34:41 1 0.227 0.615 0.158 0 ARTICLE RTRS 183SLE 8 1 1 31Jan05 12:49:04 1 0.056 0.125 0.819 I1 ALERT RTRS 0 WACHOVIA0CUTS0SARA0LEE0<SLE.N>0TO0"UNDERPERFORM"0FROM0"MARKET0PERFORM"SLE 8 2 1 31Jan05 12:55:16 1 0.056 0.125 0.819 I1 ARTICLE RTRS 387 RESEARCH0ALERTIWachovia0cuts0Sara0Lee0to0"underperform"SLE 9 1 1 31Jan05 17:00:00 1 0.056 0.125 0.819 I1 ARTICLE RTRS 387 RESEARCH0ALERTIWachovia0cuts0Sara0Lee0to0"underperform"SLE 10 1 2 1Feb05 16:06:04 0.522 0.289 0.369 0.342 0 ARTICLE RTRS 4307 UPDATE04IAvon0quarterly0profit0climbs,0dividend0raisedSLE 11 1 2 1Feb05 21:41:00 0.500 0.267 0.359 0.374 I1 ARTICLE RTRS 4516 UPDATE05IAvon0quarterly0profit0climbs,0dividend0raisedSLE 12 1 5 3Feb05 20:23:07 0.132 0.057 0.127 0.816 I1 ARTICLE RTRS 4568 Weak0margin0growth0starts0to0worry0Wall0StreetSLE 13 1 6 11Feb05 7:12:17 0.158 0.021 0.197 0.782 I1 ARTICLE RTRS 3047 PRESS0DIGEST0I0Wall0Street0Journal0I0Feb011SLE 14 1 1 16Feb05 20:00:24 1 0.198 0.579 0.223 0 ARTICLE BSW 1493 Sara0Lee0Corporation0to0Webcast0Presentation0at0CAGNY;0Executives0to0<SLE.N>SLE 15 1 1 16Feb05 21:59:19 1 0.248 0.129 0.623 I1 ARTICLE RTRS 1344 Sara0Lee0barred0from0using0Jimmy0Dean0endorsementSLE 16 1 1 17Feb05 11:57:22 1 0.596 0.397 0.007 1 ALERT RTRS 0 PIPER0RAISES0SARA0LEE0<SLE.N>0TO0"OUTPERFORM"0FROM0"MARKET0PERFORM"SLE 16 2 1 17Feb05 12:05:10 1 0.556 0.437 0.007 1 ARTICLE RTRS 417 RESEARCH0ALERTIPiper0upgrades0Sara0Lee0to0"outperform"SLE 17 1 1 17Feb05 17:00:03 1 0.556 0.437 0.007 1 ARTICLE RTRS 417 RESEARCH0ALERTIPiper0upgrades0Sara0Lee0to0"outperform"SLE 18 1 1 17Feb05 20:21:11 0.913 0.095 0.211 0.694 I1 ARTICLE BSW 2206 Nutrition0Expert0to0Discuss0Heart0Health,0Importance0of0Whole0Grains0<SLE.N>SLE 19 1 3 23Feb05 2:45:59 0.667 0.207 0.117 0.676 I1 ARTICLE RTRS 2906 INTERVIEWIFlowers0Foods0sees0organics0in0futureSLE 20 1 1 24Feb05 0:30:02 1 0.513 0.476 0.011 1 ALERT RTRS 0 SARA0LEE0CORPORATION0REVIEWS0TRANSFORMATION0PLAN0AT0CAGNY'S0ANNUAL0CONFERENCESLE 20 2 1 24Feb05 0:31:34 1 0.214 0.610 0.176 0 ALERT RTRS 0 SARA0LEE0<SLE.N>0SAYS0TO0SELL0ITS0EUROPEAN0APPAREL,0EUROPEAN0PACKAGED0MEATS,0U.S.0RETAIL0COFFEE,0DIRECT0SELLING0BUSINESSESSLE 20 3 1 24Feb05 0:32:24 1 0.176 0.810 0.014 0 ALERT RTRS 0 SARA0LEE0<SLE.N>0SAYS0ALSO0PURSING0SPINIOFF0ITS0BRANDED0APPAREL0AMERICAS/ASIA0BUSINESSSLE 20 4 1 24Feb05 0:34:07 1 0.225 0.614 0.161 0 ALERT RTRS 0 SARA0LEE0<SLE.N>0SAYS0TO0SPEND0$2500MLN0A0YEAR0ON0ADVERTISING,0PROMOTION,0R&DSLE 20 5 1 24Feb05 0:35:08 1 0.171 0.811 0.018 0 ALERT RTRS 0 SARA0LEE0<SLE.N>0EXPECTS0CHANGES0TO0SAVE0$5750MLNI$8100MLN0ON0ANNUALIZED0BASISSLE 20 6 1 24Feb05 0:36:36 1 0.535 0.456 0.009 1 ALERT RTRS 0 SARA0LEE0<SLE.N>0SEES0DEBT0DOWN0TO0$1.50BLNI$20BLN0BY0FISCAL020100FROM0$4.80BLN0FISCAL02004SLE 20 7 1 24Feb05 0:44:00 1 0.700 0.279 0.021 1 ARTICLE RTRS 189 TEXTISara0Lee0outlines0transformation0plan
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 11 / 31
Descriptive statistics
Mean S.D. 5% 50% 95%
Panel A: News releases and sentiment (126,148 observations)News stocks per day (% of firms) 11.506 3.921 5.886 11.187 17.595News days per stock (% of days) 10.610 16.667 0.000 4.636 47.458Stocks per news release 7.197 9.729 1.000 4.000 24.000Sentiment per news release 0.007 0.421 -0.726 0.040 0.738
Panel B: Institutional trading (1,096,514 observations)IVol (% of market cap) 0.832 1.644 0.046 0.432 2.728IOF (% of market cap) 0.004 0.168 -0.176 0.002 0.189|IOF|/IVol 0.153 0.166 0.008 0.102 0.475IBuys (% of market cap) 0.418 0.829 0.020 0.216 1.375ISales (% of market cap) 0.414 0.824 0.019 0.213 1.367
Cross-sectional variation in news coverage (11% of firms) & intermittentnews (1 per 10 days) & news clustering (7 firms)News sentiment not obviously biased on averageInstitutional order flow varies substantially & 15% of institutional volume
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 12 / 31
Most common topic codesCode Topic Description News0days SentimentRES Corporate+Results All+corporate+financial+results;+accounts,+annual+reports 46,457+++++ 0.015+(0.412)RESF Corporate+Results+Forecasts All+forecasting+of+corporate+financial+results;+tabular+and+textual 40,216+++++ E0.030+(0.462)STX Stock+Markets All+news+about+equity+markets+operations,+regulations+&+structure 37,983+++++ 0.016+(0.376)MRG Mergers+and+Acquisitions All+corporate+stories+about+ownership;+stakes;+mergers;+acquisitions 35,911+++++ 0.248+(0.461)DBT Debt+Markets All+debt+market+news;+including+primary+issuance+&+trading+of+debt 33,660+++++ E0.019+(0.463)NEWS Major+Breaking+News Top+stories+of+major+international+impact 30,550+++++ E0.131+(0.475)BACT Business+Activites News+relating+to+business+activities 28,724+++++ 0.138+(0.283)LAW Legislation Legislation+affecting+budgets,+securities+laws,+lawsuits,+court+rulings 27,633+++++ 0.043+(0.562)CORA Corporate+Analysis Analysis+about+a+company+or+group+of+companies 17,364+++++ 0.039+(0.065)HOT Hot+Stocks News+about+stocks+on+the+move 16,840+++++ E0.097+(0.529)REGS Regulatory+Issues News+about+regulation 16,234+++++ E0.259+(0.433)WASH Washington/US+Govt All+US+Federal+government+politics,+policies+and+economics 14,432+++++ E0.230+(0.432)DIV Dividends Dividend+forecasts,+declarations+and+payments 13,809+++++ 0.398+(0.381)ISU New+Issues All+new+corporate+issues+of+debt+and+corporate+issues+of+equity 13,523+++++ 0.149+(0.393)RCH Broker+Research+&+Recomm. All+news+about+broker+research+and+recommendations 13,014+++++ 0.013+(0.576)AAA Ratings All+news+about+credit+ratings 11,455+++++ 0.027+(0.512)JOB Labour;+(Un)employment All+news+on+labour+issues;+(un)employment;+labour+disputes;+strikes 10,768+++++ E0.131+(0.478)MNGISS Management+issues/policy Management+issues+including+executive+pay,+bonuses,+governance 9,302+++++++ E0.106+(0.466)IPO Initial+Public+Offerings First+public+listing+of+a+company's+stock 4,574+++++++ 0.162+(0.433)CRIM Crime,+Law+Enforcement Civil+and+criminal+law;+law;+crime;+corporate+crime;+fraud;+murder 4,573+++++++ E0.464+(0.339)JUDIC Judicial+processes/cases All+stories+about+judicial+processes/court+cases/court+decisions 3,923+++++++ E0.462+(0.333)ECI Economic+Indicators News,+forecasts+or+analyses+of+such+indicators 2,935+++++++ E0.029+(0.495)MCE MacroEEconomics All+news+and+analysis+on+macroEeconomics 2,474+++++++ E0.083+(0.507)BKRT Bankruptcies Corporate+insolvencies+and+bankruptcies.+ 2,276+++++++ E0.346+(0.419)VIO Civil+unrest Stories+about+riots,+demonstrations+and+other+internal+disturbances 1,285+++++++ E0.360+(0.404)FED Federal+Reserve+Board FED+activities+and+news 1,024+++++++ E0.200+(0.451)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 13 / 31
Institutional trading around news releasesEvent Study on Returns and IVol
Return volatility around news days: IVol around news days:
.009
6.0
098
.01
.010
2.0
104
.010
6|R
etur
n|
−10 −5 0 5 10Day
.42
.44
.46
.48
IVol
−10 −5 0 5 10Day
Return volatility & institutional volume rise around news releases
More uncertainty & potential for speculation before news
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 14 / 31
Institutional trading around news releasesLogit regression with firm fixed effects
(A) (B) (C) (D) (E)
IVolt−1 1.025*** 1.017***|Returnt−1| 1.101*** 1.015***Newsdayt−1 2.474*** 6.261***|Sentimentt−1| 2.697*** 0.471***
Likelihood -2.405 -2.403 -2.343 -2.388 -2.378
Institutional volume predicts news releases, beyond the rise in volatility &clustering in news days
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 15 / 31
Institutional trading around news daysEvent Study on IOF and Returns
Abnormal returns around news days: Abnormal IOF around news days:
-1-.
50
.51
Ret
urn
(%)
-10 -5 0 5 10Day
-2-1
01
2IO
F
-10 -5 0 5 10Day
Good (Bad) news days = Top (Bottom) quintile of relevance-wt sentimentMean abnormal buy-and-hold return (BHAR) for i ∈ (Good,Bad):
BHAR(t0, t1) =∑
i
wi
t1∏t0
Ri,t −∑
Control
wControl
t1∏t0
Rcontrol,t
Controls = All stocks on news days; Abnormal IOF uses same methodHendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 16 / 31
Institutional trading predicts news day returnReturns on news day regressed on lagged IOF + controls (firm fixed effects & double-clustered SE)
NewsdayReturnt (A) (B) (C) (D) (E) (F)
IOFt−1 0.652*** 0.693*** 0.541***IBuyst−1 0.632*** 0.668*** 0.518***ISalest−1 -0.693*** -0.731*** -0.575***Returnt−1 -0.015 0.002 -0.015 0.002Sentimentt−1 0.034 0.037* 0.034 0.037*ln(Volume)t−1 -0.024 -0.004 0.001 0.018Industry returnt 0.715*** 0.715***Market returnt 0.266*** 0.266***N 126,148 126,148 126,148 126,148 126,148 126,148
Institutional trading predicts announcement day return
Both institutional buying & selling predict returns (ISales more so)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 17 / 31
Institutional trading predicts news sentimentSentiment on news day regressed on lagged IOF + controls (firm fixed effects & double-clust. SE)
Sentimentt (A) (B) (C) (D) (E) (F)
IOFt−1 0.079*** 0.052** 0.052**IBuyst−1 0.071*** 0.046** 0.046**ISalest−1 -0.095*** -0.061*** -0.061***Returnt−1 0.009*** 0.009*** 0.009*** 0.008***Sentimentt−1 0.167*** 0.167*** 0.167*** 0.167***ln(Volume)t−1 -0.013* -0.012* -0.007 -0.007Industry returnt 0.016*** 0.016***Market returnt -0.011*** -0.011***N 126,148 126,148 126,148 126,148 126,148 126,148
Institutional trading predicts news sentiment, controlling forpre-announcement runup & sentiment persistence
ISales more informative than IBuys (upward trend in institutional holdings)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 18 / 31
Institutional trading before & on news releasesIOF on news day regressed on lagged IOF + controls (firm fixed effects & double-clustered SE)
IOFt (A) (B) (C) (D) (E) (F)
IOFt−1 0.257*** 0.247*** 0.249***IBuyst−1 0.259*** 0.250*** 0.252***ISalest−1 -0.254*** -0.244*** -0.246***Returnt−1 0.004*** 0.004*** 0.004*** 0.004***Sentimentt−1 -0.001** -0.001** -0.001** -0.001**ln(Volume)t−1 0.002** 0.002** -0.001 -0.001Industry returnt 0.001*** 0.001***Market returnt -0.011*** -0.011***N 126,148 126,148 126,148 126,148 126,148 126,148
Institutional trading persistent (ρ ≈ 25%)
Institutions tend to trade in same direction on news days as before
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 19 / 31
Returns on news vs. non-news daysReturn on all days regressed on lagged IOF + controls (firm fixed effects & double-clustered SE)
Returnt (A) (B) (C) (D) (E) (F)
IOFt−1 0.393*** 0.285*** 0.288***IOFt−1 ∗ Newsdayt 0.408*** 0.398***IBuyst−1 0.385*** 0.280*** 0.288***ISalest−1 -0.406*** -0.294*** -0.288***IBuyst−1 ∗ Newsdayt 0.401*** 0.375**ISalest−1 ∗ Newsdayt -0.420*** -0.431***Returnt−1 -0.017 -0.017 -0.019 -0.017 -0.017 -0.019Sentimentt−1 0.048*** 0.048*** 0.079*** 0.049*** 0.048*** 0.080***ln(Volume)t−1 -0.007 -0.008 -0.012 0.002 0.001 -0.012Newsdayt 0.015 0.015 0.020 0.035Returnt−1 ∗ Newsdayt 0.005 0.005Sentimentt−1 ∗ Newsdayt -0.046 -0.046ln(Volume)t−1 ∗ Newsdayt 0.008 0.032
Institutional trading predicts returns on news & on non-news days
Institutional trading predicts returns more on news than non-news days
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 20 / 31
Joint behavior of IOF, return & sentimentPanel VAR & impulse responses
yit = (IOFit, Returnit, Sentimentit)′ jointly & endogenously generated
Panel VAR with unobserved heterogeneity (Holtz-Eakin et al., 1988):
yit = α0 +
L∑l=1
λlyit−l +αi + εit,
where α0 is 3x1 vector of intercepts, λl are 3x3 coefficient matrices, αi isvector of unobserved individual (fixed) effects, and εit are innovations
Coefficient estimates obtained using GMM estimation
Impulse responses:Dependent variables ordered in sequence: sentiment, IOF, return
Correspond to 1 S.D. shock in model with L = 2 lags
Standard error bands at 5% level for impulse responses (dashed lines)generated using Monte-Carlo simulations with 1,000 draws
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 21 / 31
Panel VAR: Joint behavior of IOF, return & sentiment
IOFt Returnt Sentimentt
IOF t − 1 0.240*** 0.169*** 0.003***(0.008) (0.016) (0.001)
t − 2 0.077*** 0.032** -0.001(0.006) (0.016) (0.001)
Return t − 1 0.003*** -0.005*** 0.001***(0.000) (0.001) (0.000)
t − 2 0.001*** 0.000 0.000(0.000) (0.001) (0.000)
Sentiment t − 1 0.000 0.090*** 0.098***(0.001) (0.014) (0.002)
t − 2 -0.001 0.018 0.030***(0.001) (0.013) (0.002)
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 22 / 31
News types: Macro news
MACRO categories
MACRO ECI FED GVD MCE WASHN 17,586 1,812 990 3,682 1,918 13,075
Panel A: Returnt
IBuyst−1 0.756** 0.518 0.872 0.736 -0.028 1.069**ISalest−1 -0.810** -0.787 -0.883 -0.999 -0.406 -1.084**Returnt−1 0.018 0.017 -0.052 0.028 0.022 -0.005Sentimentt−1 0.056 0.048 0.221 0.273*** 0.154 -0.031
Panel B: Sentimentt
IBuyst−1 0.121 0.481** -0.224 0.066 -0.061 0.174ISalest−1 -0.122 -0.467** 0.221 -0.005 0.047 -0.192Returnt−1 0.011*** 0.022 0.040*** 0.023*** 0.033** -0.001Sentimentt−1 0.173*** 0.189*** 0.256*** 0.234*** 0.211*** 0.135***
Panel C: IOFt
IBuyst−1 0.210*** 0.140** 0.212*** 0.191*** 0.159*** 0.239***ISalest−1 -0.201*** -0.149*** -0.217*** -0.167*** -0.157*** -0.238***Returnt−1 0.004*** 0.003*** 0.004*** 0.003*** 0.003*** 0.004***Sentimentt−1 -0.002** -0.006 -0.004 -0.003 -0.001 -0.003**
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 23 / 31
News types: Crises
CRISIS categories
CRISIS BKRT MNGISS JOB CRIM JUDIC REGS CDV WEAN 31,633 2,069 7,338 8,766 3,925 3,407 14,270 2,466 2,234
Panel A: Returnt
IBuyst−1 0.738*** 1.350 0.455 0.788 1.278* 1.047 1.127** -0.162 -0.190ISalest−1 -0.872*** -1.652 -0.621 -0.978* -1.444** -1.303-1.287*** -0.265 0.226Returnt−1 -0.003 0.069 -0.007 -0.008 -0.029 -0.025 0.009 0.070***-0.095**Sentimentt−1 0.047 0.107 0.062 0.106 -0.027 0.036 0.013-0.478*** 0.086
Panel B: Sentimentt
IBuyst−1 0.156** 0.718** 0.224 0.118 0.038 0.236 0.094 0.058 0.382ISalest−1 -0.171*** -0.807** -0.226 -0.117 -0.019 -0.226 -0.107 -0.072 -0.383Returnt−1 0.003 -0.010 0.005 -0.001 -0.002 0.001 0.006 0.008 -0.006Sentimentt−1 0.151*** 0.095** 0.185*** 0.139*** 0.131*** 0.009 0.128*** 0.221***0.197***
Panel C: IOFt
IBuyst−1 0.241*** 0.192*** 0.256*** 0.206*** 0.210*** 0.248*** 0.243*** 0.191*** 0.206**ISalest−1 -0.237***-0.206*** -0.257***-0.207***-0.203***-0.248***-0.243***-0.179***-0.198**Returnt−1 0.004*** 0.004*** 0.004*** 0.004*** 0.004*** 0.004*** 0.004*** 0.003***0.005***Sentimentt−1 -0.001* 0.004* -0.002 -0.000 -0.001 -0.002-0.003*** -0.008* 0.002
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 24 / 31
News types: Earnings surprisesQuantifiable event related to fundamental real performanceStandardized unanticipated earnings score (SUE):
SUEi =ERi − E[ERi]
σ(ERi)
SUEt (A) (B) (C) (D)
IOFt−1 0.980** 0.838*IBuyst−1 1.058** 0.853*ISalest−1 -0.980** -0.837*Returnt−1 0.059** 0.058**Sentimentt−1 0.104 0.103ln(Volume)t−1 0.066 0.058
Contemp. correlation positive w/ returns & sentiment, negative w/ IOF:
Correlations IOFt Returnt Sentimentt
SUEt -0.038*** 0.338*** 0.211***
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 25 / 31
News or Hype?
Hype = Predictions in news articles that are ultimately wrong
Measuring hype:1 Press releases by companies from PR Newswire & Business Wire2 Sentiment reversals within x = 5, 10 days
Press releases:PR Newswire: 76,394 obs, +0.290 average sentimentBusiness Wire: 51,147 obs, +0.317 average sentimentContemporaneous correlation between returns & sentiment close to 0
Hypotheses:1 Naive: follow trend by buying during & selling after hype2 Short-run informed/creator: buy before & sell during hype3 Long-run strategic: anticipate reversal & abstain from trading
Institutions do not fall for hype; they do not react significantly
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 26 / 31
How do institutions respond to hype?PR Newswire (PRN) press releases
0.0
2.0
4.0
6.0
8F
ract
ion
-1 -.5 0 .5 1Sentiment
0.0
1.0
2.0
3.0
4S
entim
ent
-10 -5 0 5 10Day
-.25
-.2
-.15
-.1
-.05
0R
etur
n (%
)
-10 -5 0 5 10Day
-1-.
50
.51
IOF
(%
)
-10 -5 0 5 10Day
PR Newswire (PRN) => IOF does not follow hypeHendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 27 / 31
How do institutions respond to hype?Business Wire (BSW) press releases & regulatory filings
0.0
2.0
4.0
6.0
8F
ract
ion
-1 -.5 0 .5 1Sentiment
0.0
1.0
2.0
3.0
4.0
5S
entim
ent
-10 -5 0 5 10Day
-.15
-.1
-.05
0.0
5R
etur
n (%
)
-10 -5 0 5 10Day
-1-.
50
.5IO
F (
%)
-10 -5 0 5 10Day
Business Wire (BSW) => IOF does not follow hypeHendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 28 / 31
How do institutions respond to hype?Sentiment reversals between t and t + 5
Sentiment reversal = Top decile of |Sentimentt − Sentimentt+5|
IOF does not follow hype
Long-run strategic behavior
-.5
0.5
Ret
urn
(%)
-10 -5 0 5 10Day
-2-1
01
2IO
F (
%)
-10 -5 0 5 10Day
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 29 / 31
How do institutions respond to company hype?Sentiment reversals between t and t + 10
Sentiment reversal = Top decile of |Sentimentt − Sentimentt+10|
IOF does not follow hype
Long-run strategic behavior
-.5
0.5
1R
etur
n (%
)
-10 -5 0 5 10Day
-10
12
3IO
F (
%)
-10 -5 0 5 10Day
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 30 / 31
Conclusion
NYSE audit trail data on buy & sell volume by institutions (IOF)
Reuters NewsScope Sentiment Engine (News sentiment)
Institutional trading predicts:1 News announcements
2 Sentiment of the news
3 Return on announcement days
4 Earnings announcement surprises + crisis news
Institutions do not trade based on company hype
Institutional trading does not predict industry, market, macro news
→ Institutions are informed about various news sources & contributesignificantly to price discovery
Hendershott, Livdan, and Schürhoff (2014) Are Institutions Informed About News? December 2014 31 / 31
Recommended