本篇研究專注於首次公開發行公司上市櫃初始交易日之異常報酬與新聞情緒兩 者間之關係。本研究建立情緒字典以判別新聞之正負情緒,並過濾出與首次公開發 行有關之新聞,利用本研究建立之情緒字典以過濾出正負情緒之詞組。利用正負情 緒詞組數量計算出三種新聞情緒變數,並採實證研究方法檢測三種新聞情緒變數與 首次公開發行公司之初始交易日之異常報酬兩者間之關係。根據本研究之實證結果, 發現初始交易日之前的新聞能影響首次公開發行之異常報酬,而相關新聞之情緒語 調亦和異常報酬有關。此外,本研究亦檢測三種情緒變數和三種傳統變數之交乘項 對異常報酬之影響,發現公司規模大小與首日交易量與情緒變數之交乘項會對初始 交易日之異常報酬有影響。總言論之,本研究對新聞會影響首次公開發行初始交易 日之異常報酬提供了實證證據。 / This study focuses on the relation between IPOs’ abnormal returns on initial trading days and news sentiment. To identify the tone of news, sentiment dictionary was established for this study, and news regarding IPO firms was picked out to count positive and negative words and phrases based on the sentiment dictionary. Using quantities of positive and negative words and phrases, three news variables were adopted and calculated. And linear regression was utilized to investigate the relation between IPOs’ abnormal returns on initial trading days and news sentiment. According to empirical results, I find that news prior to the IPO’s initial trading day can affect IPOs’ abnormal returns. The number of negative words and phrases is negatively related to the abnormal returns; the tone of news is positively related to the abnormal returns. Furthermore, I also investigated whether interaction terms of news variables and three control variables are related to abnormal returns on IPOs’ initial trading days. I find that interaction terms of the natural logarithm of firm size and two news variables and interaction terms of the natural logarithm of first-day trading volume and two news variables are related to abnormal returns. Overall, there is evidence that news can influence IPOs’ abnormal returns on initial trading days.

Cette thèse comporte quatre essais dont les deux premiers, s’appuyant sur les fondements théoriques de la finance comportementale de l’entreprise, montrent dans quelle mesure les sociétés françaises tirent avantage des erreurs de jugement systématiques des investisseurs. Nous étudions dans le premier essai les conséquences du biais de familiarité sur la structure du capital. En particulier, nous montrons que les entreprises associées à un sentiment de familiarité élevé, notamment parmi les petites capitalisations, ont un actionnariat individuel plus large. Ce résultat souligne l’intérêt économique pour l’entreprise d’augmenter sa visibilité et de consolider sa notoriété dans le temps. Le deuxième essai traite de la politique de dividende sous l’hypothèse des « catering incentives ». Il s’agit de tester si les entreprises sont davantage incitées à distribuer du dividende lorsque les titres payeurs se négocient avec une prime par rapport aux non-payeurs. Nous validons cette hypothèse et mettons en évidence que les dirigeants français font preuve d’un opportunisme court-termiste accru en cas de faible contrôle familial ou de forte participation institutionnelle dans le capital. Les deux derniers essais s’intéressent aux comportements décisionnels des entreprises face aux enjeux du changement climatique. Ils se situent dans un contexte international. Le troisième essai, notamment, teste la pertinence des stratégies d’éco-efficience, via la réduction des émissions de carbone, sur la performance financière à l’occasion d’opérations de fusions-acquisitions. Les résultats tranchent en faveur de la rationalité économique de ces stratégies et affirment la possibilité pour la firme d’envisager une relation gagnant-gagnant avec son environnement. Le quatrième essai, consacré à l’étude de la communication environnementale volontaire, démontre l’intérêt croissant des parties prenantes pour ce type d’information : les entreprises les plus exposées médiatiquement, les plus endettées et celles qui entrent dans le cadre des nouvelles réglementations environnementales sont les plus transparentes en matière de reporting environnemental. De plus, les entreprises les moins éco-efficientes tendent à communiquer davantage sur leur empreinte écologique, traduisant une recherche de légitimité auprès des parties prenantes. / This thesis is composed of four essays. The first two essays draw on behavioral corporate finance and show to what extent French firms can take advantage of investors’ erroneous judgment. We first study the impact of the familiarity bias on ownership. We find that firms with higher notoriety level, mostly small-cap ones, have higher individual ownership. Second, we test the catering hypothesis in dividend policy, in that whether firms are more prone to pay dividend when payers trade at a premium relative to non-payers. The results validate this hypothesis and support short-term opportunistic behavior by French firms, especially when family control is low or institutional ownership is high. The last two essays examine the impact of climate change issues on firms’ decisions, in an international setting. Especially, the third essay demonstrates that eco-efficiency-based strategies significantly matters to the financial outcomes of mergers and acquisitions, which supports the economic rationality underlying carbon reduction investments, and claims for a win-win relationship between corporations and their environment. The fourth essay that deals with the environmental voluntary disclosure, emphasizes the increasing interest of stakeholders toward this kind of information: firms with higher exposure, higher leverage and those that are in the scope of regulators tend to be more transparent in terms of carbon reporting. Moreover, firms that are less eco-efficient show higher probability to report on their environmental performance, suggesting the search for legitimacy.

本研究嘗試建置一個包含兩個主要功能的系統，分別是網路評價搜尋以及情感分類。在網路評價搜尋的部份，我們使用Google搜尋並蒐集一攜帶型智慧裝置（智慧型手機、平板電腦與筆記型電腦）的網路評價搜尋結果；情感分類的部分則是將搜尋結果依照對該產品的意見分類為，共有正面／負面／中立、正面／負面、正面／非正面，以及負面／非負面等四種分類方式。為了建置此系統，我們首先從知名的網路論壇Mobile01和批踢踢蒐集和攜帶型智慧裝置有關的網路文章以及產品名稱，接著以人工的方式標記每篇文章，以及部分文章中的句子的情感。本研究設計了兩個層次的情感分類實驗，我們首先從語句層次出發，以監督式機器學習法訓練將句子分為正面／負面／中立等三個類別的分類模型後，再進入文章層次，將句子的意見彙整，並同樣以監督式機器學習法訓練四種不同文章層次的分類模型：正面／負面／中立、正面／負面、正面／非正面，以及負面／非負面。我們分別選出四種分類實驗中表現最佳的模型，並用於系統建置，其中表現最佳的是分類為正面／負面的分類模型，平均的F-measure為0.87；其次是分類為負面／非負面的模型，對負面類別的F-measure為0.83；接著是分類為正面／非正面的模型，對正面類別的F-measure為0.81；表現最差的是正面／負面／中立的分類，平均的F-measure為0.77。在正面／負面分類的準確率上，本研究的表現並不壞於過去以英文為主要語言的相關研究。最後，我們也以過去不經過語句層次的分類方法進行實驗並比較，其結果發現經過語句層次的情感分類比不經過語句層次的情感分類較佳。 / In this research, we implemented a system that retrieves the search results of mobile phones, tablets, and notebooks from Google, and then classifies them as: (1) positive, negative, or neutral, (2) positive or negative, (3) positive or non-positive, (4) negative or non-negative. To build this system, first we collected some documents about mobile phones, tablets, and notebooks on two popular web forums: mobile01.com and ptt.cc. Next, a sentiment label (positive, negative, or neutral) is attached to each document and each sentence of these documents. We designed a two-level supervised sentiment classification experiment. At sentence level, we trained classifiers that classify sentences as positive, negative, or neutral. The best sentence classifier was then used at document level. At document level, the sentiment labels of the sentences in documents are used. We trained classifiers in four different classification problems: (1) positive, negative, or neutral, (2) positive vs. negative, (3) positive vs. non-positive, (4) negative vs. non-negative. The best is the second classifier with an average F-measure of 0.87. The next is the fourth classifier with an F-measure of 0.83 on negative class, and then comes with the third classifier with an F-measure of 0.81 on positive class. The last is the first classifier with an average F-measure of 0.77. Our accuracy is not worse than the past English study on the classification of positive vs. negative. Finally, we conducted another classification experiment using document-level-only classification method, and the results showed that our two-level sentiment classification (first sentence level, then document level) outperforms document-level-only sentiment classification.

[pt] Sentiment Analysis é um problema de categorização de texto no qual deseja-se identificar opiniões favoráveis e desfavoráveis com relação a um tópico. Um exemplo destes tópicos de interesse são organizações e seus produtos. Neste problema, documentos são classificados pelo sentimento, conotação, atitudes e opiniões ao invés de se restringir aos fatos descritos neste. O principal desafio em Sentiment Classification é identificar como sentimentos são expressados em textos e se tais sentimentos indicam uma opinião positiva (favorável) ou negativa (desfavorável) com relação a um tópico. Devido ao crescente volume de dados disponível na Web, onde todos tendem a ser geradores de conteúdo e expressarem opiniões sobre os mais variados assuntos, técnicas de Aprendizado de Máquina vem se tornando cada vez mais atraentes. Nesta dissertação investigamos métodos de Aprendizado de Máquina para Sentiment Analysis. Apresentamos alguns modelos de representação de documentos como saco de palavras e N-grama. Testamos os classificadores SVM (Máquina de Vetores Suporte) e Naive Bayes com diferentes modelos de representação textual e comparamos seus desempenhos. / [en] Sentiment Analysis is a text categorization problem in which we want to identify favorable and unfavorable opinions towards a given topic. Examples of such topics are organizations and its products. In this problem, docu- ments are classifed according to their sentiment, connotation, attitudes and opinions instead of being limited to the facts described in it. The main challenge in Sentiment Classification is identifying how sentiments are expressed in texts and whether they indicate a positive (favorable) or negative (unfavorable) opinion towards a topic. Due to the growing volume of information available online in an environment where we all tend to be content generators and express opinions on a variety of subjects, Machine Learning techniques have become more and more attractive. In this dissertation, we investigate Machine Learning methods applied to Sentiment Analysis. We present document representation models such as bag-of-words and N-grams.We compare the performance of the Naive Bayes and the Support Vector Machine classifiers for each proposed model

This thesis merges the fields of Heterogeneous Agent Models (HAMs) and Be- havioural Finance in order to bridge the main deficiencies of both approaches and to examine whether they can complement one another. Our approach suggests an alternative tool for examining HAM price dynamics and brings an original way of dealing with problematic empirical validation. First, we present the original model and discuss various extensions and attempts at empirical estimation. Next, we develop a unique benchmark dataset, covering five par- ticularly turbulent U.S. stock market periods, and reveal an interesting pattern in this data. The main body applies a numerical analysis of the HAM extended with the selected Behavioural Finance findings: herding, overconfidence, and market sentiment. Using Wolfram Mathematica we perform Monte Carlo sim- ulations of a developed algorithm. We show that the selected findings can be well modelled via the HAM and that they extend the original HAM consider- ably. Various HAM modifications lead to significantly different results and HAM is also able to partially replicate price behaviour during turbulent stock market periods. Bibliographic Record Kukačka, J. (2011): Behavioural Breaks in the Heterogeneous Agent Model. Master thesis, Charles University in Prague, Faculty of Social Sciences,...

Le XVIIIe siècle n’est pas seulement le siècle de la raison, il est aussi celui où le sentiment s’impose dans l’esprit de certains philosophes pour rendre compte de la façon dont certains objets sont connus. Le moi ainsi que les valeurs morales et esthétiques sont, par excellence, des objets qui échappent à une analyse rationnelle ainsi qu’aux perceptions issues des sens externes. Ils se donnent dans cette expérience intérieure qu’est le sentiment. La particularité de celui-ci est que, alors qu’il est une impression d’ordre affectif, constituée de perceptions de plaisir et de douleur, il est amené à représenter autre chose que l’état purement subjectif de l’âme. Tout le problème est de déterminer à quel point le sentiment constitue un mode de connaissance irréductible : est-il un principe de connaissance à part entière, à côté de la sensation et de la réflexion, ou simplement la manière de connaître de celui qui, ayant développé des habitudes de penser et de sentir, a l’impression de juger de façon immédiate ? Reconnu comme fait mais n’ayant pas de fondement clairement assignable, il est sujet aux interprétations les plus contradictoires. Placé au croisement d’un courant métaphysique et d’un courant empiriste radical, il incarne une des notions qui manifestent le plus fortement la diversité des écoles qui perdurent au siècle des Lumières. / The 18th century is not only the age of reason, it is also the time when the sentiment becomes very important in the mind of some philosophers to explain how a certain kind of objects are known. The self as well as the moral and esthetic values are, par excellence, objects that escape both the rational analysis and the perceptions derived from external senses. They are given in an internal experience called sentiment, whom particularity is to represent something different from the pure subjective state of mind, although it is an affective impression, made of perceptions of delight and pain. The problem is to determine in what extent the sentiment represent an irreducible way of knowing: is it a source of knowledge of its own, next to sensation and reflection, or is it just an impression one’s get of judging immediately which occults a succession of unconscious judgments? Acknowledged as a fact, but lacking obvious foundation, it is likely to receive the most contradictory interpretations. At the intersection of a metaphysical current and an empiricist one, it embodies one of the notions that exhibit the diversity of schools which remains in the Enlightenment.

No description available.

Цель исследования – на основе теоретического описания и практической реализации в других исследованиях, провести сравнительную оценку методов машинного обучения для выявления оптимального решения при анализе тональности текстов. Объектом исследования выступают тексты, публикуемые в средствах массовой информации. Научная новизна исследования состоит в совершенствовании существующих методов для выявления наиболее универсального решения. Практическая значимость исследования заключается в том, что полученные результаты исследования могут быть полезными для других ученых, занимающихся анализом тональности текстов в средствах массовой информации. В результате сравнительного исследования был найден наиболее эффективный и точный метод для решения задачи. Полученные результаты и выводы исследования могут служить основой для последующих исследований в этой области и применяться в практических разработках и приложениях, требующих анализа тональности текстов. / The purpose of the study is, based on the theoretical description and practical implementation in other studies, to conduct a comparative assessment of machine learning methods to identify the optimal solution when analyzing the sentiment of texts. The object of the study is texts published in the media. The scientific novelty of the research lies in the improvement of existing methods to identify the most universal solution. The practical significance of the study lies in the fact that the results obtained may be useful for other scientists involved in the analysis of the sentiment of texts in the media. As a result of a comparative study, the most effective and accurate method for solving the problem was found. The obtained results and conclusions of the study can serve as the basis for subsequent research in this area and be used in practical developments and applications that require sentiment analysis of texts.

Companies play an important role in the process of sustainable development, and thus investors have increased their focus on companies' sustainability-related activities. These activities are often measured through ESG scores, which mostly are based on biased documents reported by the companies themselves. A company can be considered ESG-compliant when looking at the ESG scores, but its underlying sustainability profile is not entirely investigated. Thus, there is a lack of transparency in ESG scores as well as in the process of evaluating companies' sustainability performance. Therefore, this thesis aims to explore the possibilities of incorporating automatic text analysis, specifically sentiment analysis, to analyze news articles. In that way, a broader part of a company's sustainability profile is covered, and potential controversies or other involvements could be detected. To investigate whether sentiment analysis would be useful to increase the transparency an explorative approach was used. Specifically, companies' ESG scores and sentiment scores from news articles were analyzed and compared. A lower sentiment score would reasonably indicate a lower ESG score, and thus indicate transparency in the evaluation method. The study finds a mixed result of positive and negative sentiment scores for each company, within each industry. A lower sentiment score does not necessarily indicate a lower ESG score, and no clear correlation between the scores was found. Interestingly, the study also identifies previous studies which indicate a correlation between the sentiment scores from biased company documents and the ESG scores.The findings strengthen the problem of lack of transparency in ESG scores, and further conclude that sentiment analysis would be useful in the context of identifying negative and positive articles and thus increase the transparency. However, it is also concluded that sentiment analysis cannot ensure that the calculated sentiment score is of relevance to a specific company and its' sustainability-related activities. Therefore, it can only be used as a complementing tool in the evaluation of companies' sustainability performance. / Företag har en viktig roll i processen av hållbar utveckling, och därför har investerare riktat ett större fokus på företags hållbarhetsrelaterade aktiviteter. Dessa aktiviteter mäts ofta genom ESG-poäng, vilka för det mesta baseras på partiska dokument som rapporteras av företagen själva. Ett företag kan anses vara ESG-kompatibel när man tittar på deras ESG-poäng, men deras underliggande hållbarhetsprofil undersöks inte helt. Således finns det en brist på transparens i ESG-poäng och även i bedömningsprocessen av ett företags hållbarhetsarbete. Därför syftar den här forskningsuppsatsen till att undersöka möjligheterna med att använda automatisk textanalys, specifikt sentimentanalys, för att analysera nyhetsartiklar. På så sätt kan en större del av ett företags hållbarhetsprofil undersökas, och potentiella kontroverser eller annan inblandning kan upptäckas. För att undersöka om sentimentanalys är lämpligt för att öka transparensen användes en utforskande metod. Specifikt, så analyserades och jämfördes företags ESG-poäng och sentimentpoäng från nyhetsartiklar. Ett lägre sentimentpoäng borde rimligtvis indikera ett lägre ESG-poäng, och därigenom indikera på en transparens i bedömningsprocessen. Studien hittar ett blandat resultat med både positiva och negativa artiklar för varje företag inom varje industri. Ett lägre sentimentpoäng indikerar nödvändigtvis inte ett lägre ESG-poäng, och ingen korrelation mellan poängen hittades. Intressant nog, identifierar studien tidigare studier som har hittat en korrelation mellan sentimentpoäng från partiska dokument och ESG-poäng. Resultaten förstärker problemet med bristen på transparens i ESG-poäng, och kan vidare dra slutsatsen om att sentimentanalys är användbart i kontexten att identifiera positiva and negativa artiklar, och således öka transparensen. Dock dras också slutsatsen att sentimentanalys inte kan säkerställa att det beräknade sentimentpoänget är relaterat till det specifika företaget och dess hållbarhetsrelaterade aktiviteter. Därför kan det bara användas som ett kompletterande verktyg i bedömningsprocessen av företags hållbarhetsarbete.

We present results of a sentiment annotation study in the context of historical German plays. Our annotation corpus consists of 200 representative speeches from the German playwright Gotthold Ephraim Lessing. Six annotators, five non-experts and one expert in the domain, annotated the speeches according to different sentiment annotation schemes. They had to annotate the differentiated polarity (very negative, negative, neutral, mixed, positive, very positive), the binary polarity (positive/negative) and the occurrence of eight basic emotions. After the annotation, the participants completed a questionnaire about their experience of the annotation process; additional feedback was gathered in a closing interview. Analysis of the annotations shows that the agreement among annotators ranges from low to mediocre. The non-expert annotators perceive the task as very challenging and report different problems in understanding the language and the context. Although fewer problems occur for the expert annotator, we cannot find any differences in the agreement levels among non-experts and between the expert and the non-experts. At the end of the paper, we discuss the implications of this study and future research plans for this area