Extending volatility models with market sentiment indicators
| Název práce v češtině: | Rozšíření modelů volatility pomocí ukazatelů tržního sentimentu |
|---|---|
| Název v anglickém jazyce: | Extending volatility models with market sentiment indicators |
| Klíčová slova: | volatility, Heterogeneous Auto-Regressive volatility model, market sentiment |
| Klíčová slova anglicky: | volatility, Heterogeneous Auto-Regressive volatility model, market sentiment |
| Akademický rok vypsání: | 2016/2017 |
| Typ práce: | diplomová práce |
| Jazyk práce: | angličtina |
| Ústav: | Institut ekonomických studií (23-IES) |
| Vedoucí / školitel: | prof. PhDr. Ladislav Krištoufek, Ph.D. |
| Řešitel: | skrytý - zadáno vedoucím/školitelem |
| Datum přihlášení: | 09.11.2016 |
| Datum zadání: | 09.11.2016 |
| Datum a čas obhajoby: | 31.01.2018 08:30 |
| Místo konání obhajoby: | Opletalova - Opletalova 26, O105, Opletalova - místn. č. 105 |
| Datum odevzdání elektronické podoby: | 05.01.2018 |
| Datum proběhlé obhajoby: | 31.01.2018 |
| Oponenti: | doc. PhDr. Ing. et Ing. Petr Jakubík, Ph.D., Ph.D. |
| Kontrola URKUND: | |
| Seznam odborné literatury |
| BAKER M, WURGLER J (2007) Investor Sentiment in the Stock Market. Journal of Economic Perspectives. Spring 2007. Volume 21. Number 2. Pages 129–151
CORSI, F. (2009) A simple approximate long memory model of realized volatility. Journal of Financial Econometrics 7. 174-196. RANCO G, ALEKSOVSKI D, CSLDARELLI G, GRČAR M, MOZETIČ I (2015) The Effects of Twitter Sentiment on Stock Price Returns. PLoS ONE 10(9): e0138441. doi:10.1371/journal.pone.0138441 UYGUR U, TAŞ O (2012) Modeling the effects of investor sentiment and conditional volatility in international stock markets. Journal of Applied Finance & Banking. vol.2, no.5. 239-260 ISSN: 1792-6580 (print version). 1792-6599 (online) Scienpress Ltd |
| Předběžná náplň práce |
| Understanding the volatility of stock returns is crucial for forecasting stock price movements. Although volatility is natural part of liquid market, excessive volatility leads to instability and potential crashes. Many economists have studied the nature and sources of volatility. Beside financial time series modelling, a behavioural approach to volatility has become popular during past decades. Scholars have been concerned with phenomena such as emotionality of investors, cognitive biases, or market sentiment.
There is no doubt that market sentiment affects stock price volatility. Baker and Wurgler (2007) outlined several approaches to behavioural finance and stock market. They conclude that investors' sentiment is an important factor of a stock market and showed that it can be measured. Economists have begun to incorporate market sentiment into volatility models and revealed novel findings. For instance, Uygur and Taş (2012) examined the effect of noise traders during high-sentiment and low-sentiment periods using EGARCH and TGARCH models. They provided an evidence that a mean-variance relationship is undermined during high-sentiment periods when noise traders are more active. Whereas Uygur and Tas used trading volume as a proxy for market sentiment, Ranco et al. (2015) took advantage of Twitter and used tweets about the companies as the proxy. They investigated 30 companies of Dow Jones index and found a significant dependence between the Twitter sentiment and market returns. This thesis aims to refine Heterogeneous Auto-Regressive (HAR) volatility model (Corsi, 2009) by market sentiment extension. I will use information about the set of stocks from Dow Jones Industrial Average (DJIA) index from the dataset, which use Ranco et al in their paper. For each of 30 companies, there are financial data containing open, high low and close price and the date and twitter data consisting of number of positive, neutral and negative tweets, number of total tweets and date. Furthermore, I will extract data of searching volume from Google Trends as another proxy of market sentiment. I will provide out-of-sample forecast performance comparison and compare performance of the model with other volatility models (e.g. GARCH). Methodology This thesis aims to refine the Heterogeneous Auto-Regressive (HAR) volatility model (Corsi, 2009) by market sentiment indicators. We are going to use the dataset compiled by Ranco et al. (2015). The panel captures DJIA companies for a period of 15 months. We have a financial data containing open, high low and close price, and a twitter data consisting of a number of positive, neutral and negative tweets, and a number of total tweets. Furthermore, we are going to extract searching volume from Google Trends as another proxy of market sentiment. We are going to derive an out-of-sample forecast and compare forecast accuracy of various specifications. Expected contribution This thesis intends to contribute to deeper understanding of the effect of market sentiment on volatility. Particularly, it strives to introduce an extension of the HAR model incorporating market sentiment indicators. Since this HAR specification has not been deeply studied, we hope that we could provide valuable novel insights. 1) Introduction: We will introduce volatility, market sentiment and our motivation. 2) Literature review: We will summarize existing literature on market sentiment, measuring volatility, the HAR model and its extensions. 3) Data description: We will describe data sources and transformations. 4) Methodology: We will outline used methodology and HAR specifications. 5) Empirical findings: We will present outcome of the models and compare their performance using out-of-sample forecasts and benchmarking. 6) Conclusion: We will summarize the most remarkable findings, if any, and suggest further research opportunities. |
| Předběžná náplň práce v anglickém jazyce |
| Understanding the volatility of stock returns is crucial for forecasting stock price movements. Although volatility is natural part of liquid market, excessive volatility leads to instability and potential crashes. Many economists have studied the nature and sources of volatility. Beside financial time series modelling, a behavioural approach to volatility has become popular during past decades. Scholars have been concerned with phenomena such as emotionality of investors, cognitive biases, or market sentiment.
There is no doubt that market sentiment affects stock price volatility. Baker and Wurgler (2007) outlined several approaches to behavioural finance and stock market. They conclude that investors' sentiment is an important factor of a stock market and showed that it can be measured. Economists have begun to incorporate market sentiment into volatility models and revealed novel findings. For instance, Uygur and Taş (2012) examined the effect of noise traders during high-sentiment and low-sentiment periods using EGARCH and TGARCH models. They provided an evidence that a mean-variance relationship is undermined during high-sentiment periods when noise traders are more active. Whereas Uygur and Tas used trading volume as a proxy for market sentiment, Ranco et al. (2015) took advantage of Twitter and used tweets about the companies as the proxy. They investigated 30 companies of Dow Jones index and found a significant dependence between the Twitter sentiment and market returns. This thesis aims to refine Heterogeneous Auto-Regressive (HAR) volatility model (Corsi, 2009) by market sentiment extension. I will use information about the set of stocks from Dow Jones Industrial Average (DJIA) index from the dataset, which use Ranco et al in their paper. For each of 30 companies, there are financial data containing open, high low and close price and the date and twitter data consisting of number of positive, neutral and negative tweets, number of total tweets and date. Furthermore, I will extract data of searching volume from Google Trends as another proxy of market sentiment. I will provide out-of-sample forecast performance comparison and compare performance of the model with other volatility models (e.g. GARCH). Methodology This thesis aims to refine the Heterogeneous Auto-Regressive (HAR) volatility model (Corsi, 2009) by market sentiment indicators. We are going to use the dataset compiled by Ranco et al. (2015). The panel captures DJIA companies for a period of 15 months. We have a financial data containing open, high low and close price, and a twitter data consisting of a number of positive, neutral and negative tweets, and a number of total tweets. Furthermore, we are going to extract searching volume from Google Trends as another proxy of market sentiment. We are going to derive an out-of-sample forecast and compare forecast accuracy of various specifications. Expected contribution This thesis intends to contribute to deeper understanding of the effect of market sentiment on volatility. Particularly, it strives to introduce an extension of the HAR model incorporating market sentiment indicators. Since this HAR specification has not been deeply studied, we hope that we could provide valuable novel insights. 1) Introduction: We will introduce volatility, market sentiment and our motivation. 2) Literature review: We will summarize existing literature on market sentiment, measuring volatility, the HAR model and its extensions. 3) Data description: We will describe data sources and transformations. 4) Methodology: We will outline used methodology and HAR specifications. 5) Empirical findings: We will present outcome of the models and compare their performance using out-of-sample forecasts and benchmarking. 6) Conclusion: We will summarize the most remarkable findings, if any, and suggest further research opportunities. |