Submitted by Rafael Alberto Valpaços de Lemos Leal (ravllmolon@gmail.com) on 2013-06-27T01:02:48Z No. of bitstreams: 1 Dissertação MFEE - Rafael Leal (enviado Brochura).pdf: 644715 bytes, checksum: 1db8bca53f2499cce20eeab2d72a3423 (MD5) / Approved for entry into archive by Vitor Souza (vitor.souza@fgv.br) on 2013-06-28T19:27:12Z (GMT) No. of bitstreams: 1 Dissertação MFEE - Rafael Leal (enviado Brochura).pdf: 644715 bytes, checksum: 1db8bca53f2499cce20eeab2d72a3423 (MD5) / Made available in DSpace on 2013-07-19T19:18:22Z (GMT). No. of bitstreams: 1 Dissertação MFEE - Rafael Leal (enviado Brochura).pdf: 644715 bytes, checksum: 1db8bca53f2499cce20eeab2d72a3423 (MD5) Previous issue date: 2013-05-13 / Esta dissertação investigou, no mercado brasileiro, se o atraso no ajuste de preços das ações de baixa liquidez gera previsibilidade do retorno dessas ações quando comparadas às mais líquidas. O interesse estava em confrontar os resultados com os existentes na literatura internacional que apresentavam esse efeito. Para tanto, utilizamos a metodologia proposta no artigo 'Trading volume and cross-autocorrelations in stock returns', de Chordia e Swaminathan (2000), onde foi analisada a Bolsa de Valores de Nova York (NYSE). Verificamos que, na Bolsa de Valores de São Paulo (BOVESPA), uma vez controlados pelo tamanho das empresas, os retornos, sejam diários ou semanais, de portfólios com maior liquidez antecipam os retornos dos portfólios de menor liquidez, mais explicitamente nos quartis com pequenas e médias empresas. Os efeitos de não sincronia nas negociações e as autocorrelações próprias não são suficientes para explicar os padrões de antecipação-defasagem observados nos retornos das ações, já que esses são mais significativamente influenciados pelo volume negociado. As diferenças na velocidade da incorporação de novas informações aos preços ocorrem porque as ações menos líquidas parecem responder mais lentamente a informações de mercado, pelo menos nos portfólios de empresas de menor tamanho. Dessa maneira, podemos afirmar que no Brasil, assim como nos Estados Unidos, a baixa liquidez induz um atraso no ajuste de preços das ações de pequenas e médias empresas capaz de gerar previsibilidade dos retornos dessas ações, sugerindo alguma ineficiência do mercado. Os resultados são interessantes, já que indicam que, tanto no mercado nacional quanto nos de países desenvolvidos, os volumes negociados possuem um papel significativo na velocidade em que os preços se ajustam, jogando uma luz sobre como eles podem se tornar mais eficientes. / This study investigates, in the Brazilian market, if the delay in the speed of price adjustment of low liquidity stocks generates predictability of its returns when compared to more liquid ones. The interest was to compare the results against those in international literature that showed this effect. We used the methodology proposed in the paper 'Trading volume and cross-autocorrelations in stock returns', by Chordia and Swaminathan (2000), which has analyzed the New York Stock Exchange (NYSE). We verified that, at São Paulo Stock Exchange (BOVESPA), once controlled for firm size, the returns, whether daily or weekly, of high volume portfolios lead returns of low volume portfolios, more explicitly in quartiles with small and medium companies. The effects of nonsynchronous trading and the own autocorrelations are not sufficient to explain the lead-lag patterns observed in stock returns, since these are more significantly influenced by trading volume. The differences in the speed of price adjustment to new information occur because the less liquid stocks seem to respond more slowly to market-wide information, at least in the smallest size firms portfolios. Thus, we can say that in Brazil, as well as in the United States, the low liquidity induces a delay in the adjustment of stock prices of small and medium firms capable of generating predictability in the returns of these stocks, suggesting some market inefficiency. The results are interesting, since they indicate that, in the Brazilian stock market as well as in the developed countries ones, trading volumes have a significant role in the speed at which prices adjust to new information, throwing a light on how they can become more efficient.

No description available.

Dans cette thèse nous élargissons le champ d'application des modèles ARMA (AutoRegressive Moving-Average) vectoriels en considérant des termes d'erreur non corrélés mais qui peuvent contenir des dépendances non linéaires. Ces modèles sont appelés des ARMA faibles vectoriels et permettent de traiter des processus qui peuvent avoir des dynamiques non linéaires très générales. Par opposition, nous appelons ARMA forts les modèles utilisés habituellement dans la littérature dans lesquels le terme d'erreur est supposé être un bruit iid. Les modèles ARMA faibles étant en particulier denses dans l'ensemble des processus stationnaires réguliers, ils sont bien plus généraux que les modèles ARMA forts. Le problème qui nous préoccupera sera l'analyse statistique des modèles ARMA faibles vectoriels. Plus précisément, nous étudions les problèmes d'estimation et de validation. Dans un premier temps, nous étudions les propriétés asymptotiques de l'estimateur du quasi-maximum de vraisemblance et de l'estimateur des moindres carrés. La matrice de variance asymptotique de ces estimateurs est d'une forme "sandwich", et peut être très différente de la variance asymptotique obtenue dans le cas fort. Ensuite, nous accordons une attention particulière aux problèmes de validation. Dans un premier temps, en proposant des versions modifiées des tests de Wald, du multiplicateur de Lagrange et du rapport de vraisemblance pour tester des restrictions linéaires sur les paramètres de modèles ARMA faibles vectoriels. En second, nous nous intéressons aux tests fondés sur les résidus, qui ont pour objet de vérifier que les résidus des modèles estimés sont bien des estimations de bruits blancs. Plus particulièrement, nous nous intéressons aux tests portmanteau, aussi appelés tests d'autocorrélation. Nous montrons que la distribution asymptotique des autocorrelations résiduelles est normalement distribuée avec une matrice de covariance différente du cas fort (c'est-à-dire sous les hypothèses iid sur le bruit). Nous en déduisons le comportement asymptotique des statistiques portmanteau. Dans le cadre standard d'un ARMA fort, il est connu que la distribution asymptotique des tests portmanteau est correctement approximée par un chi-deux. Dans le cas général, nous montrons que cette distribution asymptotique est celle d'une somme pondérée de chi-deux. Cette distribution peut être très différente de l'approximation chi-deux usuelle du cas fort. Nous proposons donc des tests portmanteau modifiés pour tester l'adéquation de modèles ARMA faibles vectoriels. Enfin, nous nous sommes intéressés aux choix des modèles ARMA faibles vectoriels fondé sur la minimisation d'un critère d'information, notamment celui introduit par Akaike (AIC). Avec ce critère, on tente de donner une approximation de la distance (souvent appelée information de Kullback-Leibler) entre la vraie loi des observations (inconnue) et la loi du modèle estimé. Nous verrons que le critère corrigé (AICc) dans le cadre des modèles ARMA faibles vectoriels peut, là aussi, être très différent du cas fort.

Ziel der vorliegenden Arbeit ist es, das Preisach-Modell bezüglich stochastischer äußerer Felder und temperaturbezogener Aspekte zu untersuchen. Das phänomenologische Preisach-Modell wird oft erfolgreich angewendet, um Systeme mit Hysterese zu beschreiben. Im ersten Teil der Arbeit wird die Antwort des Preisach-Modells auf stochastische äußere Felder untersucht. Hier liegt das Augenmerk hauptsächlich auf der Autokorrelation; sie dient dazu den Einfluss des hysteretischen Gedächtnisses zu quantifizieren. Mit analytischen Methoden wird gezeigt, dass sich ein Langzeitgedächtnis, sichtbar in der Autokorrelation der Systemantwort, entwickeln kann, selbst wenn das treibende Feld unkorreliert ist. Im Anschluss werden diese Resultate, m.H. von Simulationen, auf äußere Felder ausgeweitet, die selbst Korrelationen aufweisen können. Der zweite Teil der Arbeit befasst sich mit dem Einfluss einer endlichen Temperatur auf das Preisach-Modell. Es werden unterschiedliche Methoden besprochen, wie das Nichtgleichgewichtsmodell in seiner mikromagnetischen Interpretation mit Temperatur als Gleichgewichtseigenschaft verknüpft werden kann. Eine Formulierung wird genutzt, um die Magnetisierung von Nickelnanopartikeln in einer Fullerenmatrix zu simulieren und mit Experimenten zu vergleichen. Des Weiteren wird die Relaxationsdynamik des Gedächtnisses des Preisach-Modells bei endlichen Temperaturen untersucht. / The aim of this thesis is to investigate the Preisach model in regard to stochastically driving and temperature-related aspects. The Preisach model is a phenomenological model for systems with hysteresis which is often successfully applied. Hysteresis is a widespread phenomenon which is observed in nature and the key feature of certain technological applications. Further, it contributes to phenomena of interest in social science and economics as well. Prominent examples are the magnetization of ferromagnetic materials in an external magnetic field or the adsorption-desorption hysteresis observed in porous media. Hysteresis involves the development of a hysteresis memory, and multistability in the interrelations between external driving fields and system response. In the first part, we mainly investigate the response of Preisach hysteresis models driven by stochastic input processes with regard to autocorrelation functions to quantify the influence of the system’s memory. Using rigorous methods, it is shown that the development of a hysteresis memory is reflected in the possibility of long-time tails in the autocorrelation functions, even for uncorrelated driving fields. In the case of uncorrelated driving, these long-time tails in the autocorrelations of the system’s response are determined only by the tails of the involved densities. They will be observed if there are broad Preisach densities assigning a high weight to elementary loops of large width and narrow input densities such that rare extreme events of the input time series contribute significantly to the output for a long period of time. Afterwards, these results are extended by simulations to driving fields which themselves show correlations. It is shown that the autocorrelation of the output does not decay faster than the autocorrelation of the input process. Further, there is a possibility that long-term memory in the hysteretic response is more pronounced in the case of uncorrelated driving than in the case of correlated driving. The behavior of the output probability distribution at the saturation values is quite universal. It is not affected by the presence of correlations and allows conclusions whether the input density is much more narrow than the Preisach density or not. Moreover, the existence of effective Preisach densities is shown which define equivalence classes of systems of input and Preisach densities which lead to realizations of the same output variable. The asymptotic behavior of an effective Preisach density determines the asymptotic correlation decay of the system’s response in the case of uncorrelated driving. In the second part, temperature-related effects are considered. It is reviewed how the non-equilibrium Preisach model in its micromagnetic picture can be related to temperature within the framework of extended irreversible thermodynamics. The irreversible response of a ferromagnetic material, namely, Nickel nanoparticles in a fullerene matrix, is simulated. The model includes superparamagnetism where ferromagnetism breaks down at temperatures lower than the Curie temperature and the results are compared to experimental data. Furthermore, we adapt known results for the thermal relaxation of the system’s memory in the form of a front propagation in the Preisach plane derived basically from solving a master equation and by the use of a contradictory assumption. A closer look is taken at short time scales which dissolves the contradiction and shows that the known results apply, taking into account the fact that the dividing line propagation starts with an additional delay time depending on the front coordinates in the Preisach plane. Additionally, it is outlined how thermal relaxation behavior in the Preisach model of hysteresis can be studied using a Fokker-Planck equation. The latter is solved analytically in the non-hysteretic limit using eigenfunction methods. The results indicate a change in the relaxation behavior, especially on short time scales.

Ziel der vorliegenden Arbeit ist es, das Preisach-Modell bezüglich stochastischer äußerer Felder und temperaturbezogener Aspekte zu untersuchen. Das phänomenologische Preisach-Modell wird oft erfolgreich angewendet, um Systeme mit Hysterese zu beschreiben. Im ersten Teil der Arbeit wird die Antwort des Preisach-Modells auf stochastische äußere Felder untersucht. Hier liegt das Augenmerk hauptsächlich auf der Autokorrelation; sie dient dazu den Einfluss des hysteretischen Gedächtnisses zu quantifizieren. Mit analytischen Methoden wird gezeigt, dass sich ein Langzeitgedächtnis, sichtbar in der Autokorrelation der Systemantwort, entwickeln kann, selbst wenn das treibende Feld unkorreliert ist. Im Anschluss werden diese Resultate, m.H. von Simulationen, auf äußere Felder ausgeweitet, die selbst Korrelationen aufweisen können. Der zweite Teil der Arbeit befasst sich mit dem Einfluss einer endlichen Temperatur auf das Preisach-Modell. Es werden unterschiedliche Methoden besprochen, wie das Nichtgleichgewichtsmodell in seiner mikromagnetischen Interpretation mit Temperatur als Gleichgewichtseigenschaft verknüpft werden kann. Eine Formulierung wird genutzt, um die Magnetisierung von Nickelnanopartikeln in einer Fullerenmatrix zu simulieren und mit Experimenten zu vergleichen. Des Weiteren wird die Relaxationsdynamik des Gedächtnisses des Preisach-Modells bei endlichen Temperaturen untersucht. / The aim of this thesis is to investigate the Preisach model in regard to stochastically driving and temperature-related aspects. The Preisach model is a phenomenological model for systems with hysteresis which is often successfully applied. Hysteresis is a widespread phenomenon which is observed in nature and the key feature of certain technological applications. Further, it contributes to phenomena of interest in social science and economics as well. Prominent examples are the magnetization of ferromagnetic materials in an external magnetic field or the adsorption-desorption hysteresis observed in porous media. Hysteresis involves the development of a hysteresis memory, and multistability in the interrelations between external driving fields and system response. In the first part, we mainly investigate the response of Preisach hysteresis models driven by stochastic input processes with regard to autocorrelation functions to quantify the influence of the system’s memory. Using rigorous methods, it is shown that the development of a hysteresis memory is reflected in the possibility of long-time tails in the autocorrelation functions, even for uncorrelated driving fields. In the case of uncorrelated driving, these long-time tails in the autocorrelations of the system’s response are determined only by the tails of the involved densities. They will be observed if there are broad Preisach densities assigning a high weight to elementary loops of large width and narrow input densities such that rare extreme events of the input time series contribute significantly to the output for a long period of time. Afterwards, these results are extended by simulations to driving fields which themselves show correlations. It is shown that the autocorrelation of the output does not decay faster than the autocorrelation of the input process. Further, there is a possibility that long-term memory in the hysteretic response is more pronounced in the case of uncorrelated driving than in the case of correlated driving. The behavior of the output probability distribution at the saturation values is quite universal. It is not affected by the presence of correlations and allows conclusions whether the input density is much more narrow than the Preisach density or not. Moreover, the existence of effective Preisach densities is shown which define equivalence classes of systems of input and Preisach densities which lead to realizations of the same output variable. The asymptotic behavior of an effective Preisach density determines the asymptotic correlation decay of the system’s response in the case of uncorrelated driving. In the second part, temperature-related effects are considered. It is reviewed how the non-equilibrium Preisach model in its micromagnetic picture can be related to temperature within the framework of extended irreversible thermodynamics. The irreversible response of a ferromagnetic material, namely, Nickel nanoparticles in a fullerene matrix, is simulated. The model includes superparamagnetism where ferromagnetism breaks down at temperatures lower than the Curie temperature and the results are compared to experimental data. Furthermore, we adapt known results for the thermal relaxation of the system’s memory in the form of a front propagation in the Preisach plane derived basically from solving a master equation and by the use of a contradictory assumption. A closer look is taken at short time scales which dissolves the contradiction and shows that the known results apply, taking into account the fact that the dividing line propagation starts with an additional delay time depending on the front coordinates in the Preisach plane. Additionally, it is outlined how thermal relaxation behavior in the Preisach model of hysteresis can be studied using a Fokker-Planck equation. The latter is solved analytically in the non-hysteretic limit using eigenfunction methods. The results indicate a change in the relaxation behavior, especially on short time scales.