Stock market returns and the content of annual report narratives

Abstract

JEL classification: M41 G12 G14Keywords: Content analysis Annual reports Stock market returns a b s t r a c t This paper uses the tools of computational linguistics to analyze the qualitative part of annual reports of UK listed companies. More specifically, the frequency of words associated with different language indicators is used to forecast future stock returns. We find that two of these indicators, capturing 'activity' and 'realism', predict subsequent price increases, even after controlling for a wide range of factors. Elevated values of these two linguistic variables, however, are not symptomatic of exacerbated risk. Consequently, investors are advised to peruse annual report narratives, as they contain valuable information that may not yet have been discounted in the prices.

The Influence of CEO Letters’ Tone on Financial Indicators of the Company

“…Иначе говоря, в зависимости от ряда факторов письмо генерального директора будет по-разно-му влиять на формирование репутации. В [Wisniewski, Yekini, 2015] проведено исследование наличия взаимосвязи меж-ду содержанием письма генерального директора и отдельными финансовыми показателями компании. Авторы про-веряли гипотезу о наличии взаимосвязи между содержанием письма генерально-го директора и изменением курса акций компании, выделяя основные лингви-стические характеристики текста, спо-собствующие их росту.…”

“…[Mosteller, Turkey, 1968], а в нейросетевых исследованиях она приме-няется с конца прошлого века (см., напр. : [Wisniewski, Yekini, 2015]). …”

Combining bag-of-words and sentiment features of annual reports to predict abnormal stock returns

“…The experiments in this study were carried out in R 2.12.0, Statistica 12 and Weka 3.7.13 using the MS Windows 7 operation system. [34] earnings press releases Positive, negative and tone word categories from Diction [24] LR Net optimism generates a significant market response in a short window around the earnings announcement date [39] news stories [1], and Diction [24] PCA+LR The effect of textual features is greater when text is accompanied by a pro forma earnings figure, managerial earnings forecast, and more numerical data [33] earnings press releases Positive, negative and tone from Loughran and McDonald [1] LR Abnormal positive tone positively affects the immediate stock price reaction to earnings announcements [13] news stories BoW SVM Market feedback is used to select text features [17] analyst reports BoW NB+LR Investors react more strongly to negative than to positive text [43] news stories BoW SVM A simulation trading return was up to 166.11 % [41] news stories BoW SVM Firm-specific news articles can enrich the knowledge of investors and affect their trading activities [48] news stories BoW, positive and negative sentiment from Harvard General Inquirer and Loughran and McDonald [1] SVM Models with sentiment word categories outperform the BoW model [36] news stories BoW SVM Integrating market news and stock price information improves the prediction accuracy of stock returns [38] news stories BoW, sentiment score, business event categorization MLP, DT MLP performed best using market data, simple news item counts, categorization into business events and calibrated sentiment scores as predictors [49] 10-Ks Word categories from Diction [24] LR Activity and realism predict subsequent price increases [35] earnings press releases Tone…”

Investor and Firm Perspectives on Impression Management in Earnings Press Releases: Insights from TSX Venture Exchange Firms

“…More in-detail discussion is provided in Section 2.3 of this study. Wisniewski andYekini, 2015, Arslan-Ayaydin et al, 2016). Measuring investor reaction to apparent impression management by analyzing stock price assumes that, if there is impression management, then investors have paid attention to it.…”

“…For example, investors react to incremental information about firms' current and future performance contained in disclosure documents (Abrahamson and Amir, 1996;Baginski et al, 2000;Botosan and Plumlee, 2002;Matsumoto et al, 2006). Therefore, firm performance is controlled for in most studies that investigate investor reaction to apparent impression management (Henry et al, 2008;Huang et al, 2014;Feldman et al, 2010;Aerts and Cheng, 2011;Blankespoor et al, 2013;Prokofieva, 2015;Wisniewski and Yekini, 2015;Arslan-Ayaydin et al, 2016). These studies also consider benchmark stock returns, firm size, and industry as important variables that influence investor reaction by including them as control variables in their analyses.…”