|
|
Entropy as a Measure of Predictability in Financial Time Series
Nahodil, Vladimír ; Krištoufek, Ladislav (advisor) ; Wang, Yao (referee)
This work studies stock markets efficiency and predictability using the information-theoretic concepts of approximate entropy (ApEn) and sample entropy (SampEn) and compares them to the estimates of the Hurst exponent. This is assessed together with the property of distinguishing between developing and developed markets. Moreover, an investment strategy based on the value of the sample entropy is tested. ApEn shows very weak relationship with other measures and performs poorly as a measure of efficiency. SampEn and the Hurst exponent clearly confirm lower overall efficiency of developing markets. The sample entropy also forms quite strong downward linear relationship with hit-rates of forecasting models. ARMA shows highest hit-rates in periods with SampEn values around 1.6 - 1.7. This could be considered as an investment strategy with lower risk; however, also as one with potentially lower accumulated returns due to smaller investing windows.
|
|
|
Event Study on Financial Announcements: New Evidence of Stock Sensitivity and Post-Earnings-Announcement Drift
Čonka, Matěj ; Krištoufek, Ladislav (advisor) ; Habiňák, Ladislav (referee)
This thesis investigates the presence of abnormal returns after the companies announce their earnings (earnings-price anomaly) on 23 companies listed on STOXX 50 Europe index. Weuse the event studies framework and we summarize main models for abnormal returns' estimation with closer look on the Market Model and CAPM. We do not find considerable value added when using more complex CAPM compared to the Market Model. The results show significant abnormal returns for good news and bad news earnings surprises with bigger market reaction on good news earnings surprises. The findings also provide the evidence of market inefficiency and the possibility of pre-announcement leakage of information. We find post-earnings-announcement drift for good news earnings surprisesandthepresenceofcontrarianreturns.
|
|
|
Strategies for Spread Trading using Futures Contracts
Gottlieb, Oskar ; Krištoufek, Ladislav (advisor) ; Čech, František (referee)
The focus of this thesis are futures spreads, more specifically trading strategies based on two approaches - cointegration tested on inter-commodity spreads and seasonality observed amongst calendar spreads. Commodity pairs which we identify to be cointegrated are tested for four mean reversion strategies, three of them being based on fair value approach, the fourth on the relative value approach. Similarly calendar spreads exhibiting seasonality are optimized for naive buy and hold trading strategies. Both approaches are tested on in-sample and out-of-sample data. Amongst seasonal strategies we have not found a pattern yielding sufficiently profitable signals in both in-sample and out-of-sample periods. Inter-commodity spreads on the other returned profitable strategies on cointegrated spreads which were also similar in physical nature. The exception to that rule were spreads known well in the industry, which failed to deliver positive results in the out-of-sample period.
|
|
|
Scale of Market Movements for US stock market
Kašpárek, Radim ; Krištoufek, Ladislav (advisor) ; Smutná, Šarlota (referee)
Currently, there is no singular, codified, and widely accepted approach to wards measuring the depth of financial crises. One of the approaches ap plied towards this problematic has been to build on the observed similarity between financial markets and dynamic systems in physics and to create analogous systems. The Scale of Market Shocks originally proposed for foreign exchange markets has been adapted for the US stock market in or der to provide US policy makers with a tool to assess the severity of such crises. Using methodology adapted from relevant research and literature we used volatilities calculated with different sampling resolution as the basis for our scale as we believe that these capture the behavior of different market agents. The resultant scale correctly identifies sharp movements and assign them a numerical value that denotes the importance of a crash. This scale is applicable for US policy makers to assess outcomes of proposed policies, however, the use of Principal Component Analysis to ease the computational complexity proved to not yield required results.
|
|
|
Using the log-periodic power-law model to detect bubbles in stock market
Kožuch, Samuel Maroš ; Krištoufek, Ladislav (advisor) ; Nevrla, Matěj (referee)
Stock market crashes were considered as an chaotic even for a long time. However, more than a decade ago a specific behavior was observed, which accompanied most of the crashes: an accelerating growth of price and log-periodic oscillations. The log-periodic power law was found to have an ability to capture the behavior prior to crash and even predict the most probable time of the crash. The log-periodic power law requires a complicated fitting method to find the estimated values of its seven parameters. In the thesis, an alternative simpler fitting method is proposed, which is equally likely to find the true estimates of parameters, thus generating an equally good fit of log-periodic power law. Furthermore, four stock indices are fitted to log-periodic power law and examined for possible log-periodic oscillations in different time periods, including a very recent period of 2017. In all of the analyzed indices, a log-periodic oscillations could be observed. One index, analyzed in past period, was fitted to log-periodic power law, which was able to capture the oscillations and predict the critical time of crash. In the rest of the selected stocks, which were analyzed in a recent period, the critical time was estimated with varying results.
|
|
|
Stock market prediction using Twitter
Hynek, Jan ; Krištoufek, Ladislav (advisor) ; Křehlík, Tomáš (referee)
In this work I examine the short-time relationship of Twitter on the markets. I had been downloading English tweets in the period between 9th March and 4th April and also tweets containing words and hashtags "apple", "microsoft", "boe- ing", "cocacola". Afterwards, I investigate the predictive power of frequency of individal words on the marke using multinomial and binomial penalised logistic regression. I conclude that this method cannot be used for prediction, but can provide interesting insight ex-post. 1
|
|
|
Do crypto-currencies form a new asset class?
Mayr, Samuel ; Krištoufek, Ladislav (advisor) ; Hanus, Luboš (referee)
This paper examines statistical properties of crypto-currencies' price variations in comparison with statistical properties of price variations in common financial markets. Price data of Bitcoin, ripple and Litecoin have been directly compared with price data of euro currency and stock index S&P500. Additionally, and compared with set of stylized facts of asset returns. The properties in scope of this work include an autocorrelation of day-to-day returns, a shape of return distributions, a volatility clustering, a leverage effect and a volume/volatility correlation. To answer the question of this thesis, we have tried to find unique differences in the way prices of crypto-currencies behave. After every point of the data analysis has been checked, we have concluded that the only major difference is in the shape and the significance of autocorrelation in day-to-day returns. While crypto-currencies seem to autocorrelate, there has been no such a cross-autocorrelation found in the benchmark values. Therefore, we argue that it is the most distinctive sign of crypto-currencies and the reason for crypto-currencies to be regarded as separate asset class. Powered by TCPDF (www.tcpdf.org)
|
|
|
Electricity market: Analysis and prediction of volatility
Kunc, Vladimír ; Krištoufek, Ladislav (advisor) ; Hájek, Jan (referee)
Electricity market: Analysis and prediction of volatility Abstract Vladimír Kunc July 30, 2015 The last two decades can be characterized by restructuring of energy industry and the creation of new, competitive energy markets, where accurate forecasts of elec- tricity prices and price volatility are valuable both to consumers and producers. The aim of this work is to analyse several models for prediction of the price volatility of electricity on the Czech Electricity Day-ahead market on price data provided by OTE, a.s. for years 2009-2014. This work compares 144 different models' configura- tions for three distinct classes of models - autoregressive models, GARCH models, and artificial neural network models. This work provides comparison based on five different criteria, each describing the model in different way. Keywords: price prediction, volatility prediction, GARCH, neural networks, LSTM 1
|
|
|
Forecasting Jump Occurrence in Czech Day-Ahead Power Market
Hortová, Jana ; Krištoufek, Ladislav (advisor) ; Kukačka, Jiří (referee)
The very specific features of the spot prices, especially occurrence of severe jumps, create a spot price risk for retailers who purchase electricity at unregulated highly volatile prices but resell it to consumers at fixed price. Therefore, it is of high im- portance to forecast whether jump is likely to occur during the next hour. However, to the best of our knowledge, such research has not been devoted to the Czech power market yet. Therefore, the aim of this thesis is to forecast the jump occurrence in the Czech day-ahead market. For this purpose we suggest four logit model spec- ifications, each containing various independent variables (for example, electricity demand, outside temperature, lagged price and various dummy variables) where the variable selection is supported by the previous literature and by the characteristic features of the spot prices. Within the in-sample period we compare the suggested models based on the values of pseudo-R squared and Bayesian information criterion. When evaluating the out-of sample performance of suggested models we apply jump prediction accuracy and confidence, but opposed to the previous literature we sug- gest a kind of sensitivity analysis which, to the best of our knowledge, has not be proposed by any other power research. JEL Classification C25, C32, C51,...
|
|
|
Practical usage of optimal portfolio diversification using maximum entropy principle
Chopyk, Ostap ; Krištoufek, Ladislav (advisor) ; Kraicová, Lucie (referee)
"Practical usage of optimal portfolio diversification using maximum entropy principle" by Ostap Chopyk Abstract This thesis enhances the investigation of the principle of maximum entropy, implied in the portfolio diversification problem, when portfolio consists of stocks. Entropy, as a measure of diversity, is used as the objective function in the optimization problem with given side constraints. The principle of maximum entropy, by the nature itself, suggests the solution for two problems; it reduces the estimation error of inputs, as it has a shrinkage interpretation and it leads to more diversified portfolio. Furthermore, improvement to the portfolio optimization is made by using design-free estimation of variance-covariance matrices of stock returns. Design-free estimation is proven to provide superior estimate of large variance-covariance matrices and for data with heavy-tailed densities. To asses and compare the performance of the portfolios, their out-of-sample Sharpe ratios are used. In nominal terms, the out-of- sample Sharpe ratios are almost always lower for the portfolios, created using maximum entropy principle, than for 'classical' Markowitz's efficient portfolio. However, this out-of-sample Sharpe ratios are not statistically different, as it was tested by constructing studentized time-series...
|
guest ::
RSS feed.