The aim of this thesis is to provide an empirical evidence of multifractality in financial time series and to discuss the relevance of this concept for the current financial theory. We have applied two methods, the Multifractal Detrended Fluctuation analysis and the Generalized Hurst exponent method, on components of the Dow Jones Industrial Average. We analyzed daily data of 30 companies traded on U.S. stock markets from 2002 to 2012. We present results supporting presence of multiscaling in open-close returns. Contrary to published literature, we were not able to find any significant multiscaling in volatility. Moreover based on our analysis, multiscaling is not present in standardized returns and as multifractality requires relatively complicated models, this is our most valuable result. 1

M.Comm. / The aim of this study is to discuss, analyse and forecast market volatility. Financial liberalisation and technological innovation have taken place during the past twenty-five years, producing a highly integrated and competitive world financial system in which trillions of dollars are traded every day (Murray, van Norden & Vigfusson, 1996:1). These developments have been positive, but there are concerns about the problems that such unregulated capital flows might pose for the efficient pricing of financial assets and the stability of domestic and international financial markets. Speculation has increased and greater competition, information technology and new securities lead to excessive price volatility. Stocks, bonds and foreign exchange are more sensitive to sudden shocks and trade at prices that appear inconsistent with market fundamentals. It is important to point out the causes of market volatility in order to determine if any precautions can be taken to prevent the enormous impact of market volatility on economic performance. The study could be useful for investors and dealers. It might enable them to forecast volatility and use it as a risk management instrument.

A 50% research report to be submitted in partial fulfilment for the degree of: MASTER OF COMMERCE (FINANCE) UNIVERSITY OF THE WITWATERSRAND / Several studies have assessed the forward-looking characteristic of share prices and confirmed their resultant capability as leading indicators of economic activity, especially in advanced economies. Contention however exists when evaluating the role of stock markets as leading indicators for less developed countries. This study examines the validity of the stock market as a leading indicator of economic activity in South Africa using quarterly time-series data for the period January 1992 to June 2014. Causality and cointegration between the JSE All Share Index against Real GDP and Real Industrial Production is evaluated by employing Granger-causality tests and the Johansen cointegration procedure. The empirical investigation indicates that unidirectional causality exists between the nominal and real stock indices and economic activity in South Africa, and confirms a long-run relationship between the JSE and GDP and Industrial Production. Therefore, similar to the study by Auret and Golding (2012), in a South African context, the stock market is in fact a leading indicator of economic activity. / MT2017

Uma previsão de receitas precisa é muito importante para o administrador público na elaboração do orçamento anual, e para isso há a necessidade de se encontrar um modelo, econométrico ou não, que possibilite essa previsão com qualidade. Este trabalho apresenta uma forma inovadora para realizar a combinação de modelos de previsão. Seu objetivo foi criar uma metodologia para a obtenção de pesos para a combinação de modelos baseada no método de otimização de uma carteira de investimentos proposto por Markowitz. Para o estudo, foram utilizadas as estimações de três a cinco previsões individuais de um a cinco passos à frente, com os modelos Box-Jenkins SARIMA (Autorregressivo Integrado de Médias Móveis Sazonal), PLSR (Regressão com Mínimos Quadrados Parciais) e o Método não econométrico de Indicadores, como é denominado internamente na Receita Federal. A utilização da fronteira eficiente de Markowitz, que apresenta os pontos de mínima variância para cada retorno, é semelhante à minimização da variância da combinação, proposta no artigo seminal de Bates e Granger. O risco (desvio padrão), na teoria de portfólio de Markowitz, pode ser definido como a dispersão dos resultados e pode ser decomposto em risco sistemático e risco não sistemático. À medida que a quantidade de pesos das previsões a combinar cresce, a parte não sistemática do risco tende a zero, ficando o risco total representado somente pela parte sistemática. Por outro lado, observou-se que a curva de erros correspondente à fronteira eficiente apresenta quebras estruturais à medida que a quantidade de pesos não-zero varia. Selecionando-se trechos em que a quantidade de pesos é maior, minimiza-se a parte não sistemática, minimizando o erro. Dentro desses trechos selecionados, buscaram-se os pontos de menor erro, sendo a combinação encontrada chamada de Mínimo Erro Prim. O Mínimo Erro Seg foi o resultado da combinação com o menor erro, incluindo-se os trechos com a segunda maior quantidade de componentes diferentes de zero na combinação. Embora, na média, os pontos de Mínimo Erro Seg apresentem menor valor de erro que o Mínimo Erro Prim, como o segundo apresenta menor desvio padrão médio, optou-se pelo Mínimo Erro Prim para o ponto escolhido como a proposta de combinação deste estudo. Esse ponto apresenta resultados sistematicamente melhores que o da simples média, utilizada geralmente como benchmark. / A precise revenue forecast is very important for public administrators to draft an annual report. That is why there is a need to find a model, whether econometric or not, that makes it possible to have a quality forecast. This study proposes an innovative approach to executing a combination of forecasting models. The goal was to create a methodology to obtain weights in order to combine models based on the investment portfolio optimization method proposed by Markowitz. The estimates of three to five individual forecasts from one to five steps ahead were used for the study, with the Box-Jenkins SARIMA (Seasonal Autoregressive Integrated Moving Average) model, the PLSR (Partial Least Squares Regression) model and the non-econometric Method of Indicators, as it is called internally at the Brazilian Federal Revenue Service. The use of Markowitz\'s efficient frontier, which shows the points of minimum variance for each return, is similar to the minimization of the combination variance proposed in the seminal paper by Bates and Granger. The risk (standard deviation) in the Markowitz portfolio theory could be defined as a dispersion of results and could be broken down into systemic risk and non-systemic risk. Insofar as the amount of weights for the forecasts to be combined grows, the non-systemic part of the risks tends to move towards zero, with total risk only being represented by the systemic part. On the other hand, the error curve was found to correspond to the efficient frontier, showing structural breaks insofar as the amount of non-zero weights varies. By selecting parts where there is a greater amount of weights, the non-systemic part is minimized, thus minimizing error. Within these selected parts, the points of least error were sought, with the combination found being called the Prim Minimum Error. The Sec Minimum Error was the result of the combination with the lowest error, including the parts with the second highest amount of components different from zero in the combination. Although on average the Sec Minimum Error points show a lower error value than the Prim Minimum Error, since the second shows a lower standard deviation, the Prim Minimum Error was chosen as the point selected as the combination proposal of this study. This point shows systematically better results than the simple average generally used as a benchmark.

Os três artigos que compõem esta Tese possuem em comum a utilização de técnicas macroeconométricas para a investigação de problemas específicos. Embora no campo da macroeconometria os modelos do tipo Vetores Auto Regressivos (VAR) costumam ser uma das principais ferramentas utilizadas, muitas vezes precisamos complementar esta abordagem por outras técnicas para extrairmos conclusões mais apropriadas. É justamente isto que acontece com cada um dos artigos desta Tese. No primeir o artigo, foram construídas estimativas para a série do PIB mensal para o Brasil mediante aplicação de modelos de desagregação de séries temporais. Embora o método de estimação destes modelos seja o de mínimos quadrados generalizados (GLS), utilizou-se um modelo VAR para apurar o erro médio de projeção da taxa de inflação associado a cada uma das séries do PIB mensal encontrada para cada método de desagregação temporal empregado. No segundo artigo, foi avaliado se o fenômeno conhecido como price puzzle pode ser explicado pela presença de um canal de custo da política monetária. A investigação econométrica empregada para se verificar a existência de tal canal foi através de estimações, pelo método dos momentos generalizados (GMM), da Curva de Phillips Novo-Keynesiana Híbrida (HNKPC) estendida pela presença do canal de custo. Adicionalmente, ampliou-se a especificação do modelo VAR inicialmente considerado a fim de verificar se, na eventualidade de rejeição da existência do canal de custo para a economia brasileira, problemas de má especificação poderiam ser responsáveis pelo surgimento do price puzzle. Por fim, no terceiro artigo foi investigado se existe, ou não, complementaridade entre os investimentos públicos e privados no Brasil. Partindo de um modelo VAR inicial e efetuandose os testes de cointegração, conclui-se que é possível identificar um processo de mecanismo de correção de erros (VECM) entre o investimento público e privado no país, muito embora existam evidências de que a relação de complementaridade possa ter sofrido enfraquecimento ao longo do tempo. Assim, para verificar tal suspeita, conduziu-se uma análise via modelos de regressão com parâmetros variáveis no tempo (TVP), estimados mediante a aplicação do filtro de Kalman. Em suma, os três artigos desta Tese fornecem uma razoável combinação das principais técnicas macroeconométricas quando aplicadas sobre problemas específicos. / The three articles of this thesis have in common the use of macroeconometric techniques which investigate specific problems. Although in the field of macroeconometrics the VAR models are one of the most popular tools, very often we need to add other techniques to this framework, in order to achieve more appropriate conclusions. That is exactly what is applied to every article of this thesis. In the first article we construct estimates for the monthly GDP of Brazil by applying temporal disaggregation models. Although the method used in these models are the Generalized Least Squares (GLS), we estimate a VAR to find the average forecast error of the inflation rate associated to each of the GDP series found for each method of time-series disaggregation. In the second article, we evaluate if the so-called phenomenon \"price puzzle\" can be explained by the presence of a cost channel of monetary policy. The econometric investigation able to verify the existence of such channel, through the Generalized Method of Moments (GMM), is the Hybrid New -Keynesian Phillips Curve (HNKPC), extended by the presence of the cost channel. In addition, we amplify the specification of the VAR model considered in the beginning, in order to verify whether problems of misspecification could be responsible for the arise of the \"price puzzle\", in the case of rejection of existence of cost channel for the Brazilian economy. Finally, the third article investigates whether private and public investments in Brazil are complements. From an initial VAR model, we apply cointegration tests which are able to reveal that it is possible to identify a process of error correction (VECM) between public and private investment in Brazil, although there is evidence that their relationship has been losing strength over time. Therefore, in order to investigate such suspicion, we conduct an analysis of Time Varying Parameters models (TVP), estimated by a Kalman f ilter. In conclusion, the three articles of this thesis offer a reasonable combination of the main macroeconometric techniques when applied to specific problems.

Neste trabalho foi explorada a aplicação de redes neurais recorrentes simples, também conhecidas como Redes de Elman, na previsão de três séries temporais mensais do varejo de bens e serviços no Brasil. As variáveis destas séries estão relacionadas com a demanda de produtos farmacêuticos, adubos, e tráfego aéreo. As previsões com Redes de Elman foram comparadas com as realizadas por modelos lineares sazonais obtidos através da metodologia de Box-Jenkins. / In this work we explored the application of simple recurrent neural networks, also known as Elman networks, in the prediction of three series of retail goods and services in Brazil. The series are formed by variables related to the monthly demand for pharmaceuticals, fertilizers and domestic air traffic. The forecast with Elman networks were compared with those performed by seasonal linear models obtained by Box-Jenkins methodology.

Neste texto são revisados métodos de reamostragem de séries temporais discretas baseados em wavelets, como alternativas as abordagens clássicas, feitas nos domínios do tempo e da frequência. Tais métodos, conhecidos na literatura como wavestrap e wavestrapping fazem uso, respectivamente, das transformadas wavelet discreta (DWT) e wavelet packet discreta (DWPT). Existem poucos resultados sobre a aplicação da DWPT, de forma que este texto pode ser considerado uma contribuição. Aqui mostra-se também, a superioridade do wavestrapping sobre o wavestrap quando aplicados na estimação da densidade espectral de potência de séries temporais sintéticas geradas a partir de modelos autoregressivos. Tais séries possuem uma particularidade interessante que são picos, geralmente acentuados, em sua reapresentação espectral, de tal forma que grande parte dos métodos clássicos de reamostragem apresentam resultados viesados quando aplicados a estes casos. / This paper reviews resampling methods based on wavelets as an alternative to the classic approaches which are, made in the time and frequency domains. These methods, known in the literature as wavestrap and wavestrapping, make use, respectively, of the discrete wavelet transform (DWT) and of the discrete wavelet packet transform (DWPT). Since only few results are avaliable when the DWPT is applied, this text can be considered a contribution to the subject. Here we, also show the superiority of wavestrapping over wavestrap when they are applied to the estimation of power spectral densities of the synthetic time series generated from autoregressive models. These series have an interesting feature that are sharp peaks in their spectral representation, so that most of the traditional methods of resampling lead to biased results.

by Li, Man-kong, Vincent, Yee, Yat-pui, Yvonne. / Thesis (M.B.A.)--Chinese University of Hong Kong, 1990. / Bibliography: leaves 92-94. / Library's copy:Disk for circulation (3.5 in.) / TABLE OF CONTENTS / ABSTRACT --- p.ii / TABLE OF CONTENTS --- p.iii / LIST OF TABLES --- p.v / ACKNOWLEDGEMENTS --- p.vi / CHAPTER / Chapter I. --- INTRODUCTION --- p.1 / Importance of Earnings Forecast --- p.1 / Previous Research Findings --- p.3 / Objective --- p.5 / Chapter II. --- AN OVERVIEW OF EARNING FORECASTING --- p.6 / Methods of Forecasting --- p.6 / Financial Analyst's Model --- p.7 / Time-series Model --- p.9 / Common Belief Regarding Forecast Superiority --- p.10 / Chapter III. --- RESEARCH DESIGN AND METHOD OF ANALYSIS --- p.12 / General Framework --- p.12 / Choice of Companies and Time Period --- p.13 / Data Preparation --- p.16 / Actual Earning Data --- p.16 / Forecast Data by Financial Analyst --- p.17 / Adjustment of Data --- p.18 / Time Series Model Data --- p.20 / Generation of Time-series Data --- p.20 / The Time-series Model --- p.20 / Data Analysis --- p.25 / Definition of Forecast Error --- p.25 / Comparison of Forecast Accuracy --- p.26 / Grouping of Companies --- p.26 / Method of Comparison --- p.27 / Chapter IV. --- RESULTS OF ANALYSIS AND IMPLICATIONS --- p.30 / General Data Manipulation --- p.30 / Data For Analysis --- p.30 / Input Data --- p.30 / Computed Forecast Errors --- p.31 / Comparison of Forecast Accuracy --- p.31 / First Time Friedman Test Results --- p.31 / Full Cross-Sectional Dependence --- p.32 / Sectorial Dependence --- p.34 / Cross-Sectional Independence --- p.37 / Wilcoxon Signed Rank Test and Second Time Friedman Test --- p.41 / Full Cross-Sectional Dependence. --- p.42 / Sectorial Dependence --- p.43 / Cross-Sectional Independence --- p.44 / Chapter V. --- CONCLUSION AND DISCUSSION --- p.45 / Findings From Results of Study --- p.45 / Limitation of Study --- p.48 / Cost of Forecast --- p.48 / Sample Size and Length of Study Period --- p.48 / Limited Past Earning Data --- p.49 / Accounting Income Manipulation --- p.50 / Suggested Further Study --- p.50 / APPENDIX A --- p.52 / APPENDIX B --- p.53 / APPENDIX C --- p.56 / BIBLIOGRAPHY --- p.92

by Wong Siu Fun. / Bibliography: leaves 55-59 / Thesis (M.Ph.)--Chinese University of Hong Kong, 1986

Zhang Haiyan. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2001. / Includes bibliographical references (leaves 48-50). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 2 --- Literature Review --- p.6 / Chapter 2.1 --- An Introduction to the Uncovered Interest Parity (UIP) and previous works on UIP --- p.6 / Chapter 2.2 --- Previous empirical works applying Band Spectrum Regression(BSR) --- p.15 / Chapter 3 --- Basic Band Spectral Regression (BSR) Techniques --- p.20 / Chapter 3.1 --- BSR Based on the complex Fourier transform --- p.20 / Chapter 3.2 --- BSR based on the real-valued Fourier transform --- p.24 / Chapter 3.3 --- Testing for parameter stability in the frequency domain --- p.26 / Chapter 4 --- Data and Standard Time Series Analysis in the Time Domain --- p.29 / Chapter 5 --- Analyze the UIP relation in the frequency domain --- p.33 / Chapter 5.1 --- An overview of the UIP relation across frequency --- p.33 / Chapter 5.2 --- Testing parameter stability across different time horizons --- p.37 / Chapter 6 --- Test of UIP with the forward premium --- p.42 / Chapter 7 --- Conclusion --- p.45