Since arbitrage-free is a desirable theoretical feature in a healthy financial market, many efforts have been made to construct arbitrage-free models for yield curves. However, little attention is paid to review if such restriction will improve yield forecast. We evaluate the importance of arbitrage-free restriction on dynamic Nelson-Siegel term structure when forecasting yield curves. We find that it doesn’t help. We also compare these two Nelson-Siegel dynamic models with a benchmark dynamic model and show that Nelson-Siegel structure improve forecasts for long-maturity yields. / text

This thesis investigates the use of realized volatility features from high frequency data in com- bination with neural networks to improve forecasts of the yield curve of government bonds. I use high frequency data on futures of four U.S. Treasury securities to estimate the Nelson-Siegel yield curve and realized variance of its parameters over the period of 25 years. The estimated parameters are used in prediction of the level, slope and curvature of the yield curve using an LSTM neural network and compared to the Dynamic Nelson-Siegel model. Results show that the use of realized variance and neural network outperforms autoregressive methods in prediction of the level and curvature in daily and monthly forecasts. The yield curve of government bonds itself has a predictive power on multiple macroeconomic variables, therefore improvements in its forecastability may have broader implications on forecasting the overall state of the economy.

Neste trabalho, tínhamos por objetivo propor um modelo dinâmico de estrutura a termo de taxas de juros com variáveis macroeconômicas baseado na formulações de Diebold e Li (2006) e Nelson e Siegel (1987) (DNS). A estrutura de estimação proposta permite utilizar dados de frequências distintas, combinando observações diárias de curvas de juros e mensais de variáveis macroeconômicas de interesse através de uma estrutura MIDAS - Mixed Data Sampling. Também utilizamos uma estrutura de volatilidade estocástica multivariada para os fatores latentes e variáveis macroeconômicas e também permitimos que o parâmetro de decaimento do modelo DNS varie no tempo, permitindo capturar mudanças na estrutura de volatilidade condicional e no formato das curvas em períodos longos. O procedimento de estimação é baseado em métodos Bayesianos usando Markov Chain Monte Carlo. Aplicamos este modelos para a curva de juros de títulos do Tesouro Americano entre 1997 e 2011. Os resultados indicam que incorporação de informações diárias e mensais em um mesmo modelo permite ganhos significantes de ajuste, superando as estimativas usuais baseadas em modelos sem informações macroeconômicas e nos métodos usuais de estimação do modelo de Diebold e Li (2006) / In this present work, we propose a dynamic model for the term structure of interest rates with macroeconomic variables based on Diebold e Li (2006)\'s and Nelson e Siegel (1987)\'s researches. The estimation procedure we intend to build allows time series data sampled at different frequencies, mixing daily observations of yield curves and monthly observations of macroeconomic variable through a Mixed Data Sampling (MIDAS) regression. We also make use of a multivariate stochastic volatility structure for the latent factors and allow the parameter that governs the exponential decay rate to vary trough time, which enables us to capture changes both in the conditional volatility structure and in the curve\'s shapes during long periods. The estimation procedure is based on Baeysian inference trough the usage of of Markov Chain Monte Carlo (MCMC) method. We applied these models to the U.S. Treasure bonds\' yield curve from 1997 to 2011. The results denote that joining daily and monthly information into the same model allows significant gains on fitting these models to the term structure, overcoming the usual estimates based on models without macroeconomics information and on regular estimation methods of Diebold e Li (2006)\'s model.

This thesis uses Spline-based model, which was developed by McCulloch, and parsimonious model, which was developed by Nelson-Siegel, to estimate the yield curves of zero-coupon bonds in Turkey. In this thesis, we construct the data by using Turkish secondary government zero-coupon bond data, which contain the data from January 2005 to June 2005. After that, relative performances of models are compared using in-sample goodness of fit. As a result, we see that performance of McCulloch model in fitting yield is better than that of Nelson-Siegel model.

Neste trabalho, tínhamos por objetivo propor um modelo dinâmico de estrutura a termo de taxas de juros com variáveis macroeconômicas baseado na formulações de Diebold e Li (2006) e Nelson e Siegel (1987) (DNS). A estrutura de estimação proposta permite utilizar dados de frequências distintas, combinando observações diárias de curvas de juros e mensais de variáveis macroeconômicas de interesse através de uma estrutura MIDAS - Mixed Data Sampling. Também utilizamos uma estrutura de volatilidade estocástica multivariada para os fatores latentes e variáveis macroeconômicas e também permitimos que o parâmetro de decaimento do modelo DNS varie no tempo, permitindo capturar mudanças na estrutura de volatilidade condicional e no formato das curvas em períodos longos. O procedimento de estimação é baseado em métodos Bayesianos usando Markov Chain Monte Carlo. Aplicamos este modelos para a curva de juros de títulos do Tesouro Americano entre 1997 e 2011. Os resultados indicam que incorporação de informações diárias e mensais em um mesmo modelo permite ganhos significantes de ajuste, superando as estimativas usuais baseadas em modelos sem informações macroeconômicas e nos métodos usuais de estimação do modelo de Diebold e Li (2006) / In this present work, we propose a dynamic model for the term structure of interest rates with macroeconomic variables based on Diebold e Li (2006)\'s and Nelson e Siegel (1987)\'s researches. The estimation procedure we intend to build allows time series data sampled at different frequencies, mixing daily observations of yield curves and monthly observations of macroeconomic variable through a Mixed Data Sampling (MIDAS) regression. We also make use of a multivariate stochastic volatility structure for the latent factors and allow the parameter that governs the exponential decay rate to vary trough time, which enables us to capture changes both in the conditional volatility structure and in the curve\'s shapes during long periods. The estimation procedure is based on Baeysian inference trough the usage of of Markov Chain Monte Carlo (MCMC) method. We applied these models to the U.S. Treasure bonds\' yield curve from 1997 to 2011. The results denote that joining daily and monthly information into the same model allows significant gains on fitting these models to the term structure, overcoming the usual estimates based on models without macroeconomics information and on regular estimation methods of Diebold e Li (2006)\'s model.

This thesis enhances rare literature focusing on modeling and forecasting of term structure of crude oil markets. Using dynamic Nelson-Siegel model, crude oil term structure is decomposed to three latent factors, which are further forecasted using both parametric and dynamic neural network approaches. In-sample fit using Nelson-Siegel model brings encouraging results and proves its applicability on crude oil futures prices. Forecasts obtained by focused time-delay neural network are in general more accurate than other benchmark models. Moreover, forecast error is decreasing with increasing time to maturity.

This thesis is composed of three chapters which propose some novel approaches to model and forecast financial time series. The first chapter focuses on high frequency financial returns and proposes a quantile regression approach to model their intraday seasonality and dynamics. The second chapter deals with the problem of forecasting the yield curve including large datasets of macroeconomics information. While the last chapter addresses the issue of modelling the term structure of interest rates. The first chapter investigates the distribution of high frequency financial returns, with special emphasis on the intraday seasonality. Using quantile regression, I show the expansions and shrinks of the probability law through the day for three years of 15 minutes sampled stock returns. Returns are more dispersed and less concentrated around the median at the hours near the opening and closing. I provide intraday value at risk assessments and I show how it adapts to changes of dispersion over the day. The tests performed on the out-of-sample forecasts of the value at risk show that the model is able to provide good risk assessments and to outperform standard Gaussian and Student’s t GARCH models. The second chapter shows that macroeconomic indicators are helpful in forecasting the yield curve. I incorporate a large number of macroeconomic predictors within the Nelson and Siegel (1987) model for the yield curve, which can be cast in a common factor model representation. Rather than including macroeconomic variables as additional factors, I use them to extract the Nelson and Siegel factors. Estimation is performed by EM algorithm and Kalman filter using a data set composed by 17 yields and 118 macro variables. Results show that incorporating large macroeconomic information improves the accuracy of out-of-sample yield forecasts at medium and long horizons. The third chapter statistically tests whether the Nelson and Siegel (1987) yield curve model is arbitrage-free. Theoretically, the Nelson-Siegel model does not ensure the absence of arbitrage opportunities. Still, central banks and public wealth managers rely heavily on it. Using a non-parametric resampling technique and zero-coupon yield curve data from the US market, I find that the no-arbitrage parameters are not statistically different from those obtained from the Nelson and Siegel model, at a 95 percent confidence level. I therefore conclude that the Nelson and Siegel yield curve model is compatible with arbitrage-freeness.

Exchange rate forecasting is hard issue for most of floating exchange rate economies. Studying exchange rate is very attractive matter since almost no model could beat random walk in short run yet. Relative yields and information in relative yield curves are contemporary topics in empirical literature and this study follows Chen and Tsang (2009) who model exchange rate changes with relative factors obtained from Nelson-Siegel (1987) yield curve model and find that relative factor model can forecast exchange rate change up to 2 years and perform better than random walk in short run. Analysis follows the methodology defined by Chen and Tsang (2009) and TL/USD, TL/EUR exchange rate changes are modeled by the relative factors namely relative level, relative slope and relative curvature. Basically, 162 weekly datasets from 09.01.2007 to 16.03.2010 are used and the relative factors for each week are estimated. Afterwards, regression analysis is made and results show that relative level and relative curvature factors are significant up to 4-6 weeks horizon but relative slope does not provide any valuable information for exchange rate prediction in Turkish financial market. Length of forecasting horizon of relative factor model is too short when compared to other exchange rate models. Since it is accepted that exchange rates follow random walk, we provided some tests to compare performance of the model. Similar to the literature, only short run performance of relative factor model is compared to random walk model and concluded that the relative factor model does not provide better forecasting performance in Turkish financial market

This thesis focus on a yield curve modelling. It estimates unobserved "global" yield curve factors which drives changes in individual real yield curves. Yield curves of USD, GBP, JPY and EUR are considered and global factors are able to explain substantial part of their variances. The method is built on the Nelson-Siegel model which is implemented in a state-space form to be able to extract the unobserved yield factors. The estimated global yield factors are further used for explaining the evolution of the Czech yield curve. Their impact to the Czech yield curve is estimated in a time-varying regression which results show that the impact of the global factors is stronger during the years of the interventions of the Czech National Bank and thus suggests that the interventions help to transmit the global low interest rates to the Czech economy.

This thesis investigates the modelling and forecasting of multivariate volatility and dependence in financial time series. The first paper proposes a new model for forecasting changes in the term structure (TS) of interest rates. Using the level, slope and curvature factors of the dynamic Nelson-Siegel model, we build a time-varying copula model for the factor dynamics allowing for departure from the normality assumption typically adopted in TS models. To induce relative immunity to structural breaks, we model and forecast the factor changes and not the factor levels. Using US Treasury yields for the period 1986:3-2010:12, our in-sample analysis indicates model stability and we show statistically significant gains due to allowing for a time-varying dependence structure which permits joint extreme factor movements. Our out-of-sample analysis indicates the model's superior ability to forecast the conditional mean in terms of root mean square error reductions and directional forecast accuracy. The forecast gains are stronger during the recent financial crisis. We also conduct out-of-sample model evaluation based on conditional density forecasts. The second paper introduces a new class of multivariate volatility models that utilizes high-frequency data. We discuss the models' dynamics and highlight their differences from multivariate GARCH models. We also discuss their covariance targeting specification and provide closed-form formulas for multi-step forecasts. Estimation and inference strategies are outlined. Empirical results suggest that the HEAVY model outperforms the multivariate GARCH model out-of-sample, with the gains being particularly significant at short forecast horizons. Forecast gains are obtained for both forecast variances and correlations. The third paper introduces a new class of multivariate volatility models which is easy to estimate using covariance targeting. The key idea is to rotate the returns and then fit them using a BEKK model for the conditional covariance with the identity matrix as the covariance target. The extension to DCC type models is given, enriching this class. We focus primarily on diagonal BEKK and DCC models, and a related parameterisation which imposes common persistence on all elements of the conditional covariance matrix. Inference for these models is computationally attractive, and the asymptotics is standard. The techniques are illustrated using recent data on the S&P 500 ETF and some DJIA stocks, including comparisons to the related orthogonal GARCH models.