National Repository of Grey Literature 116 records found  1 - 10nextend  jump to record: Search took 0.01 seconds. 
Can Machines Explain Stock Returns?
Chalupová, Karolína ; Baruník, Jozef (advisor) ; Vácha, Lukáš (referee)
Can Machines Explain Stock Returns? Thesis Abstract Karolína Chalupová January 5, 2021 Recent research shows that neural networks predict stock returns better than any other model. The networks' mathematically complicated nature is both their advantage, enabling to uncover complex patterns, and their curse, making them less readily interpretable, which obscures their strengths and weaknesses and complicates their usage. This thesis is one of the first attempts at overcoming this curse in the domain of stock returns prediction. Using some of the recently developed machine learning interpretability methods, it explains the networks' superior return forecasts. This gives new answers to the long- standing question of which variables explain differences in stock returns and clarifies the unparalleled ability of networks to identify future winners and losers among the stocks in the market. Building on 50 years of asset pricing research, this thesis is likely the first to uncover whether neural networks support the economic mechanisms proposed by the literature. To a finance practitioner, the thesis offers the transparency of decomposing any prediction into its drivers, while maintaining a state-of-the-art profitability in terms of Sharpe ratio. Additionally, a novel metric is proposed that is particularly suited...
Power markets and the EU ETS: How volatility propagates across Central Europe?
Jurka, Vojtěch ; Baruník, Jozef (advisor) ; Čech, František (referee)
The thesis deals with connectedness in the uncertainty of the carbon and power markets in Central Europe. While the drivers of power price were extensively documented in the literature, we investigate how uncertainty propagates between the German power market and its production factors using a recently developed framework of connectedness measurement. The connections in uncertainty on markets are insightful for the decision of the agents that require a premium for undertaking risk. The empirical results suggest that connectedness in uncertainty significantly varies over the studied period. The interdependence of power with coal decreases while the spillovers between gas and power rise on importance reflecting the changes in generation mix of Germany. For most of the period, the volatility of carbon and power markets is highly correlated. However, the share of volatility transmission spikes several times during the period of 2016-2019. In reaction to the reform of the EU Emission Trading Scheme, the uncertainty about emission allowance prices propagates to the German power market, increasing the uncertainty about power prices on the long horizon.
Multi-horizon equity returns predictability via machine learning
Nechvátalová, Lenka ; Baruník, Jozef (advisor) ; Krištoufek, Ladislav (referee)
We examine the predictability of expected stock returns across horizons using machine learning. We use neural networks, and gradient boosted regression trees on the U.S. and international equity datasets. We find that predictabil- ity of returns using neural networks models decreases with longer forecasting horizon. We also document the profitability of long-short portfolios, which were created using predictions of cumulative returns at various horizons, be- fore and after accounting for transaction costs. There is a trade-off between higher transaction costs connected to frequent rebalancing and greater returns on shorter horizons. However, we show that increasing the forecasting hori- zon while matching the rebalancing period increases risk-adjusted returns after transaction cost for the U.S. We combine predictions of expected returns at multiple horizons using double-sorting and buy/hold spread, a turnover reduc- ing strategy. Using double sorts significantly increases profitability on the U.S. sample. Buy/hold spread portfolios have better risk-adjusted profitability in the U.S. JEL Classification G11, G12, G15, C55 Keywords Machine learning, asset pricing, horizon pre- dictability, anomalies Title Multi-horizon equity returns predictability via machine learning
Cyber risk modelling using copulas
Spišiak, Michal ; Teplý, Petr (advisor) ; Baruník, Jozef (referee)
Cyber risk or data breach risk can be estimated similarly as other types of operational risk. First we identify problems of cyber risk models in existing literature. A large dataset consisting of 5,713 loss events enables us to apply extreme value theory. We adopt goodness of fit tests adjusted for distribution functions with estimated parameters. These tests are often overlooked in the literature even though they are essential for correct results. We model aggregate losses in three different industries separately and then we combine them using a copula. A t-test reveals that potential one-year global losses due to data breach risk are larger than the GDP of the Czech Republic. Moreover, one-year global cyber risk measured with a 99% CVaR amounts to 2.5% of the global GDP. Unlike others we compare risk measures with other quantities which allows wider audience to understand the magnitude of the cyber risk. An estimate of global data breach risk is a useful indicator not only for insurers, but also for any organization processing sensitive data.
Conditional quantile models for asset returns
Havel, Štěpán ; Baruník, Jozef (advisor) ; Fanta, Nicolas (referee)
The literature related to Value at Risk estimation is rich in general. However, majority of papers written on this subject concentrates on the unconditional non-parametric or parametric approach to VaR modelling. This thesis focuses on direct conditional VaR estimation using quantile regression. Thereby im- posing no restrictions on the return distribution. We use daily volatility mea- surements for individual stocks in S&P 500 index and quantile regress them on one-day ahead returns of the entire index. Depending on the quantile selected this estimation produces different confidence levels of Value at Risk. In order to minimize complexity of the final model, regularization methods are applied. To the author's knowledge this specific methodology has not yet been applied in any paper. The main objective is to investigate whether this approach is able to produce sound VaR estimates comparable with different methods usu- ally applied. Our result suggests that quantile regression extended with lasso regularization can be used to produce sound one-day-ahead Value at Risk es- timates. JEL Classification C22, C58, G15 Keywords volatility, quantile regression, VaR, GARCH Title Conditional quantile models for asset re- turns Author's e-mail Supervisor's e-mail
Asset Prices, Network Connectedness, and Risk Premium
Procházková, Vendula ; Baruník, Jozef (advisor) ; Kukačka, Jiří (referee)
This diploma thesis introduces the measures of network connectedness in the context of asset pricing. It proposes an asset pricing model in which the factor of connectedness is included as one of the risk factors together with the three Fama-French factors. The goal of the analysis is to examine whether the con- nectedness represents a signifcant risk factor that should be considered while determining the risk premium of the portfolio in diferent sectors in the market. Using the realized volatilities and returns of 496 assets of SP 500 index over the period 2005 - 2018, that are divided into 11 sectors, we frstly determine the linkages of connectedness between the assets in the same sector. Applying Fama-MacBeth two-step regression model, we explore the signifcance of the connectedness factor for the determination of the risk premium. We argue that the sector overall connectedness represents a signifcant risk in most of the sec- tors and should be therefore taken into account by the investors in all sectors. Moreover, the total directional connectedness that captures the spillover of shocks to one asset from the other assets in the sector, is a signifcant risk fac- tor that should increase the risk premium of the portfolio, especially in sectors such as the fnancial, health care, consumer...
Predicting purchasing intent on ecommerce websites
Vařeka, Marek ; Krištoufek, Ladislav (advisor) ; Baruník, Jozef (referee)
This thesis analyzes behavior of customers on an e-commerce website in order to predict whether the customer is willing to buy something or is just window shopping. In addition the secondary model predicts, if the customer is going to leave the e-commerce website in next few clicks. To answer this questions different frameworks are tested. The base model used is the Logit model. The base model is compared with more sophisticated methods in machine learning - with neural networks. The best results were yielded by Recurrent neural network - the Long Short-Term Memory (LSTM). The results of the analysis confirm importance of the click stream data and calculated features that track user behavior on the e-commerce website, type of the page (product, category, information), product variance and category variance. The thesis emphasizes practical implications of this models. Two possible practical implementations are presented. The models are tested in novel ways to see how would they perform if implemented on the real e-commerce website.
Does economic uncertainty spill across countries?
Skákala, Norbert ; Baruník, Jozef (advisor) ; Vácha, Lukáš (referee)
1 Abstract We study economic policy uncertainty spillovers on a panel of ten countries between April 1998 to September 2019. The analysis is performed on the Economic Policy Uncertainty indices data. To measure the spillovers, we utilize forecast error variance decompositions of VAR model. We found that approximately half of the forecast variance can be explained by spillovers shocks across countries. Further, we disentangle the spillover measure to short-, mid- and long-term cycles using frequency domain. Our results suggest that most of the spillovers are caused by shocks into low frequencies, hence with long persistence. Employing quantile regression on equation-by-equation basis to estimate the VAR model, we find that idiosyncratic uncertainty shocks do not propagate strongly at the median but that powerful spillovers occur in the right tail of distribution. Additionally, we perform rolling window estimates of the spillovers. The results indicate strong variation in time, especially during major geopolitical events, such as Iraq War (2003), Global Financial Crisis (2007-09), European debt crisis (2010-12) or Brexit (2016).
Volatility and Skewness Spillover Effects: Multiresolution Analysis
Frýd, Lukáš ; Vácha, Lukáš (advisor) ; Baruník, Jozef (referee)
The thesis investigates volatility and skewness spillover effects among seven world stock indices and WTI oil under the assumption of the presence of heterogeneous investors. The data sample covers the period from January 1990 to July 2016. The questions addressed in the thesis are twofold: firstly, the dependency of the spillover effect for both the moments-volatility and skewness-on different investments horizons is tested. Further, it is mea- sured whether the inclusion of skewness into has an impact on the volatility spillovers. The decomposition to the different investment horizons is per- formed by the wavelet transformation. Conditional volatility and skewness were estimated by GAS model, which is capable to dynamize static parame- ters from Skewed t distribution. Empirical results suggest significant spillover effects from both volatil- ity and skewness. Another important result is that skewness has a non- significant impact on the volatility spillover effects. Further, it has been found that spillover effects for both the moments are time-scale dependent: the higher investment horizons are associated with higher spillover effects. Additionally, our results support the evidence of the significant impact of the financial crisis in 2008 on the structure of markets. From 2008, there are stronger volatility...

National Repository of Grey Literature : 116 records found   1 - 10nextend  jump to record:
See also: similar author names
2 Baruník, Jozef,
Interested in being notified about new results for this query?
Subscribe to the RSS feed.