Spelling suggestions: "subject:"term structure"" "subject:"germ structure""
81 |
Previsão da estrutura a termo da taxa de juros brasileira usando redes neurais artificiaisArantes, Breno de Oliveira January 2013 (has links)
Avaliamos as previsões fora da amostra da curva de juros geradas por modelos de redes neurais artificiais e as comparamos com os modelos tradicionalmente usados para este fim. A curva de juros foi segmentada em três regiões distintas e para cada uma dessas regiões e horizontes de previsão, foram estimados duas classes de modelos de redes neurais (modelos dinâmicos e modelos estáticos), totalizando 198 modelos estimados. Afim de verificar a significância estatística das previsões dos modelos de RNA em relação aos outros modelos, foi realizado o teste Diebold-Mariano. Os resultados mostram que em média, os modelos estimados através de RNA conseguiram superar as previsões realizadas pelo random walk em todos os horizontes de previsão, sendo essas previsões em torno de 2% e 5% melhores para os horizontes de 1 dia e 1 semana a frente, e de 12% e 7% melhores para os horizontes mais longos, de 1 mês e 3 meses. Além disso, apresentou previsões em torno de 15% e 10% melhores que o modelo de NS para os horizontes de 1 mês e 3 meses a frente. Concluimos que os modelos de redes neurais são capazes de realizar previsões superiores para todos os horizontes testados, principalmente para região de curto prazo da curva, com destaque especial para as previsões com horizontes de 1 dia e 1 semana a frente. / We evaluate the out of sample forecasts of the yield curve generated by artificial neural network models and compare them with the models traditionally used for this purpose. The yield curve was segmented into three distinct regions and for each region and forecast horizons, we estimated two classes of neural network models (dynamic models and static models), totaling 198 models estimated. In order to check the statistical significance of the model predictions of RNA compared to other models, was performed the Diebold-Mariano’s test. The results show that on average, the models estimated using RNA overcame the predictions made by the random walk at all forecast horizons, and these forecasts around 2% and 5% better for horizons of 1 day and 1 week forward and 12% and 7% better for longer horizons, 1 month and 3 months. Moreover, forecasts showed around 15% and 10% better than the NS model for horizons of 1 month and 3 months ahead. We conclude that the neural network models are capable of superior forecasts for all horizons tested, especially for short-term region of the curve, with particular attention to the forecasts with horizons of 1 day and 1 week ahead.
|
82 |
Construção de um algoritmo para estimação da estrutura a termo da taxa de juros utilizando o método de taxas a termo constantes entre reuniões do Copom / Building an algorithm for implementing the term structure of interest rate adopting the flat forward rate between Copom meetingsFernando Odair Bristotti 04 April 2018 (has links)
Para que um operador de uma mesa proprietária de um banco consiga fornecer um preço competitivo e de forma a auferir lucro em uma operação é fundamental uma estimação adequada da estrutura a termo da taxa de juros. Afinal, cada uma dessas demandas e ofertas por liquidez exigem diferentes prazos e na grande maioria das vezes instrumentos utilizados para realizar a imunização de acordo com o prazo dessa operação não estão disponíveis para negociação no mercado financeiro. A construção de uma estrutura a termo de juros é uma forma de sintetizar em uma única curva toda a informação disponível de contratos negociáveis no mercado financeiro e que reproduzam o preço mais justo para a taxa de juros de um determinado prazo. O objetivo do presente trabalho é implementar a estimação da estrutura a termo da taxa de juros brasileira utilizando-se do método de taxas a termo constantes entre as reuniões do Comitê de Política Monetária (Copom). O algoritmo implementado deve ser capaz de resolver a estimação num tempo suficientemente rápido para que seja possível agregá-lo em um sistema de cotações de mercado em tempo real e fornecer aos operadores de mercado informações completas da curva de juros com as taxas zero cupom e as taxas a termo para cada prazo. Nesta dissertação serão apresentados detalhes da implementação do algoritmo e também do arcabouço teórico utilizado. Será apresentando também uma breve descrição da dinâmica do mercado de juros brasileiro e suas peculiaridades, além de apresentar alguns métodos de estimação da estrutura a termo comumente utilizados. / For an operator of a bank to be able to provide a competitive price and to make a profit in an operation, an adequate estimation of the term structure of the interest rate is essential. After all, each of these demands and offers for liquidity require different terms and in most cases the instruments used to carry out the immunization according to the term of this operation are not available for trading in the financial market. The construction of an interest rate term structure is a way of synthesizing in a single curve all the available information of contracts negotiable in the financial market and that best reproduces the fairer price for the interest rate of a certain term. The main purpose of this work is implement the estimation of the Brazilian term structure of interest rate using the flat forward rate method between Copom meetings. The implemented algorithm must be able to resolve the estimation in a sufficiently fast time so that it can be aggregated into a real-time market quotations system and provide to market operators information on the yield curve and forward rates. This dissertation will present the algorithm implementation in detais as well as the theoretical framework used. It will also present a brief description of the dynamics of the Brazilian interest market and its peculiarities, besides presenting some methods of estimation of the term structure commonly used.
|
83 |
Estudo comparativo do poder preditivo do PIB brasileiro por variáveis macroeconômicas em relação ao conjunto de variáveis da estrutura de taxa de jurosSinzato, Arthur Yukio 16 August 2012 (has links)
Submitted by Arthur Yukio Sinzato (arthursinzato@gmail.com) on 2012-10-03T16:53:57Z
No. of bitstreams: 1
Dissertacao - Arthur Y Sinzato.pdf: 1540863 bytes, checksum: 58f233b710218ecced2ad0bea5cbe63a (MD5) / Approved for entry into archive by Suzinei Teles Garcia Garcia (suzinei.garcia@fgv.br) on 2012-10-03T17:09:30Z (GMT) No. of bitstreams: 1
Dissertacao - Arthur Y Sinzato.pdf: 1540863 bytes, checksum: 58f233b710218ecced2ad0bea5cbe63a (MD5) / Made available in DSpace on 2012-10-03T17:12:04Z (GMT). No. of bitstreams: 1
Dissertacao - Arthur Y Sinzato.pdf: 1540863 bytes, checksum: 58f233b710218ecced2ad0bea5cbe63a (MD5)
Previous issue date: 2012-08-16 / O trabalho busca comparar dois conjuntos de informações para a projeção das variações do PIB brasileiro: através de modelos econométricos aplicados sobre a série univariada do PIB, e a aplicação dos mesmos modelos, mas contemplando adicionalmente o conjunto de informação com dados da estrutura a termo de taxa de juros de swap PRÉ-DI. O objetivo é verificar, assim como descrito na literatura internacional, se informações de variáveis financeiras tem a capacidade de incrementar o poder preditivo de projeções de variáveis macroeconômicas, na medida em que esses dados também embutem as expectativas dos agentes em relação ao desenvolvimento do cenário econômico. Adicionalmente, o mesmo procedimento aplicado para os dados brasileiros é aplicado sobre as informações dos Estados Unidos, buscando poder fornecer ao estudo uma base de comparação sobre os dados, tamanho da amostra e estágio de maturidade das respectivas economias. Como conclusão do resultado do trabalho está o fato de que foi possível obter um modelo no qual a inclusão do componente de mercado apresenta menores erros de projeção do que as projeções apenas univariadas, no entanto, os ganhos de projeção não demonstram grande vantagem comparativa a ponto de poder capturar o efeito de antecipação do mercado em relação ao indicador econômico como em alguns casos norte-americanos. Adicionalmente o estudo demonstra que para este trabalho e amostra de dados, mesmo diante de diferentes modelos econométricos de previsão, as projeções univariadas apresentaram resultados similares. / This study intends to compare two groups of data in order to forecast the Brazilian GDP growth: through the use of econometric models applied on a univariate time series of the GDP, and the use of the same models but additionally incorporating the group of data containing the term structure of Brazilian interbank swap rates information. The objective is to test, as already described on abroad academic literature, the possibility of financial market data being capable of increasing predicting power of forecasting macroeconomic data, as these financial information possess agents expectations over the development of the economic scenario. Additionally, the same procedure applied over the Brazilian data is also applied over North-American information, in order to provide the study a relative basis comparison over the data, sample size, and development stage of each economy. As conclusion of the results of this work is the fact that it has been possible to obtain one model that inserts the market component and presents smaller forecast errors than the predicted by univariate models, however, the benefits of the projections do not present a great comparative advantage in order to capture the market anticipation effect when compared to the economic indicators as in some North American cases. Additionally the study shows that for this work and sample of data, even within different econometric predicting models, the univariate projections presented similar results.
|
84 |
O poder preditivo da estrutura a termo da taxa de juros: uma abordagem com indicadores não linearesCunha, Filippe Santa Fé 03 July 2013 (has links)
Submitted by Filippe Santa Fé (fsantafe@gmail.com) on 2013-07-31T22:49:28Z
No. of bitstreams: 1
Poder_preditivo_versao_final.pdf: 1657453 bytes, checksum: 6504132a61068c66a84f8b59fd5518ae (MD5) / Approved for entry into archive by Suzinei Teles Garcia Garcia (suzinei.garcia@fgv.br) on 2013-08-01T17:31:22Z (GMT) No. of bitstreams: 1
Poder_preditivo_versao_final.pdf: 1657453 bytes, checksum: 6504132a61068c66a84f8b59fd5518ae (MD5) / Made available in DSpace on 2013-08-01T17:32:34Z (GMT). No. of bitstreams: 1
Poder_preditivo_versao_final.pdf: 1657453 bytes, checksum: 6504132a61068c66a84f8b59fd5518ae (MD5)
Previous issue date: 2013-07-03 / Based on the methodology idealized by Frankel and Lown (1994), developed with the intent to perfect the one used by Mishkin (1990a, 1990b) by allowing, as opposed to that one, the long term real interest rate to vary, this paper intends to verify the existence, as far as brazilian data is concerned, of the empirical evidence already found in countries such as the United States of America and German regarding the predictive power of the slope of the yield curve on the future trajectory for inflation data. The marginal changes carried on by Frankel and Low are important because they allow the final results to be interepreted within a context that is more in line with reality, that is, where monetary policy decisions have the capability to influence the behavior of the long term interest rate for a given economy. This paper will assess, for the period between january 2003 and january 2013, the predictive power exhibited by a variety of indicators whose task is to measure the slope of the yield curve when faced with data for the brazilian consumer price índex (IPCA). In a nutshell, it is found that, in line with what is reported in the international literature, there is evidence that the several measures used to depict the slope cointain information about the future variation of inflation. Despite the fact that some measures perform this task better than others, this characteristic is common to all, regardless of the way inflation data is presented in, i.e. with or without seasonal adjustments, and also regardless of the way in which intereste rates variables are represented. / Com base em uma metodologia desenvolvida por Frankel e Lown (1994), que surge para aperfeiçoar o arcabouço utilizado por Mishkin (1990a,1990b) ao permitir, em contraposição a este, a variação ao longo do tempo da taxa de juros real de longo prazo de uma economia, o presente trabalho se propõe a verificar o existência, para dados brasileiros, de evidências empíricas que suportem a aceitação da hipótese de que a estrutura a termo da curva de juros possui informações que auxiliam na previsão da trajetória futura da inflação, característica já encontrada em dados de algumas economias desenvolvidas, como Estados Unidos e Alemanha. A modificação marginal implementada por Frankel e Lown é importante porque permite que os resultados finais sejam interpretados dentro de um contexto teórico mais próximo da realidade, isto é, onde as decisões de política monetária são capazes de influenciar a variável acima mencionada. Analisa-se, para o período de janeiro de 2003 a janeiro de 2013, o poder explicativo que diversas medidas de inclinação da curva de juros local têm sobre a variação futura do índice de preços ao consumidor ampliado (IPCA), o mais importante da economia brasileira. Em linhas gerais, encontramos que, em consonância com a experiência internacional, independente de tratarmos os dados na forma com ou sem ajustes sazonais, ou de tratar as variáveis relativas à taxa de juros como logaritmo de um fator mais suas taxas ou de apresentá-las em seu próprio nível, existem evidências empíricas que indicam a existência de poder preditivo desta variável explicativa sobre a referida variável dependente, ou ao menos tornam difícil rejeitar, em um primeiro momento, esta hipótese inicial.
|
85 |
Modeling of Foreign Exchange Swap Distributions : A statistical evaluation of two stochastic modelsEhrenpreis, Ludvig, Oscar, Eriksson January 2023 (has links)
The global foreign exchange (FX) market is one of the world's largest financial markets and a significant part of this market concerns the trading of FX swaps. For banks and other financial institutions, it is of great interest to model these swaps as accurately as possible, as this could improve their risk management. Numerous methods exist for modeling FX swaps, but it is not always clear if one model is superior to another. The purpose of this thesis is therefore to analyze, evaluate and compare different models that represent the stochastic processes in the FX swap market. To accomplish this, the thesis employs the reality model evaluation methodology developed by \citet{Blom_fx_pdf}. With this methodology, likelihood values for an out-of-sample period can be determined for a model, thereby enabling a statistical comparison to ascertain which model more accurately reflects the true distribution. This thesis will compare two models for FX swap prices: an interest rate model and a PIP-model. The PIP-model is constructed by determining a multivariate distribution based on in-sample observations of pips. The likelihood values for the out-of-sample observations can therefore be determined directly. The interest rate model, on the other hand, will be implemented using Blomvall's reality model evaluation in order to determine the likelihood values. It is constructed by evaluating risk factors of the FX swaps, rather than historical pips. The risk factors evaluated in this thesis are forward curves, the spot price and spikes in the supply and demand curve at certain dates. The results show that the interest rate model better represents the true distribution of FX swaps compared to the PIP-model. The statistical test of the out-of-sample likelihood values shows that the probability of the interest rate model outperforming the PIP-model is approximately 100 \%. Additionally, the result suggests that an implementation of the interest rate model using a Student's t-distribution is more advantageous than using a normal distribution, a conclusion also supported by a statistical test. Moreover, the effectiveness of Blomvall's reality model evaluation in determining likelihood values is confirmed, thus enabling the statistical comparison of different models.
|
86 |
[pt] ENSAIOS EM ECONOMETRIA / [en] ESSAYS IN ECONOMETRICSPRISCILLA BURITY 12 August 2020 (has links)
[pt] Este trabalho consiste de quatro artigos e esta dividida em duas partes distintas. A primeira esta relacionado a modelos de estrutura a termo estimados para países cujos títulos soberanos sao considerados sujeitos a default. A segunda parte é sobre crescimento e convergencia. No primeiro artigo da primeira parte, usamos um modelo de estrutura a termo afim sem arbitragem e escolhemos tres países da zona euro para esta analise: Espanha, Grécia e Itália. Notamos que a dívida do proprio país tem desempenhado um papel importante na ampliac¸ao recente dos spreads, sobretudo para a Grecia e Itália. Para a Espanha, o recente aumento dos spreads está sendo impulsionado principalmente pelas variaveis relacionadas com a Alemanha (entre as quais a dívida alemã e a mais importante entre eles), e estresse do mercado (representada por um índice alto rendimento). No segundo artigo da primeira parte, usamos um modelo de sem arbitragem para investigar os determinantes da estrutura a termo dos spreads soberanos brasileiros e as expectativas da taxa de cambio no Brasil. Nossos exercícios indicam que fatores ligados a economia brasileira tiveram uma contribuic¸ao positiva para os spreads 1999-2006 e, na maior parte deste período, os fatores ligados a economia dos EUA tiveram uma contribuição negativa. Desde o final de 2007, esse cenario inverteu. Além disso, apresentam-se evidências de que a melhoria na nossa indicador de solvência externa, juntamente com a melhoria no cenario de inflação, e o principal responsável pela redução de spreads a partir do meio de 2007. O primeiro artigo da segunda parte propoe uma abordagem semi-parametrica para controlar a heterogeneidade não observada em modelos de regressao linear. Utilizamos estimadores extremos de redes neuras e o modelo e ilustrado em uma aplicação para a convergência de renda per capita entre os municípios brasileiros. Finalmente, no segundo artigo da segunda parte, vamos explorar os desenvolvimentos recentes sobre a convergencia para a produtividade industrial. Encontramos que o crescimento da produtividade esta negativamente relacionado com a produtividade inicial, mas que países com instituições piores têm um nível de equilíbrio de produtividade mais baixo. A taxa de crescimento da produtividade tambem tem uma relação não monotonica com a abertura comercial e educação, sendo mais rápida nos extremos. / [en] This work consists of four chapters and is divided into two distinct parts. The first is related to models of the term structure estimated for countries whose sovereign bonds are considered defaultable. The second part is about growth and convergence. In the first article of the first part, we used a no- arbitrage term structure model and chose three eurozone countries for this analysis: Spain, Greece and Italy. We noted that the country s debt has played an important role in the recent expansion of spreads, especially for Greece and Italy. For Spain, the recent increase in spreads has been driven primarily by variables related to Germany (Germany s debt is the most important among them), and market stress (represented by a highyield index). In the second article of the first part, we use a no-arbitrage model to investigate the determinants of the term structure of sovereign spreads and the expectations of the exchange rate in Brazil. Our exercises suggest that factors related to the Brazilian economy had a positive contribution to the 1999-2006 spreads and, in most of the period, factors related to the U.S. economy had a negative contribution. Since late 2007, this scenario has inverted. Furthermore, we present evidence that the improvement in our external solvency indicator along with the improved inflation scenario are the main responsible for the reduction of spreads from the middle of 2007 on. The first article of the second part proposes a semi-parametric approach to control unobserved heterogeneity in linear regression models. We use extreme neurotic networks estimators and the model is illustrated in an application to the convergence of per capita income across Brazilian municipalities. Finally, in the second article of the second part, we explore recent developments on industrial productivity convergence. We find that productivity growth is negatively related to initial productivity, but that countries with worse institutions have lower level of
productivity in equilibrium. The rate of productivity growth also has a nonmonotonic relationship with trade openness and education, being faster at the extremes.
|
87 |
Essays on forecast evaluation and financial econometricsLund-Jensen, Kasper January 2013 (has links)
This thesis consists of three papers that makes independent contributions to the fields of forecast evaluation and financial econometrics. As such, the papers, chapter 1-3, can be read independently of each other. In Chapter 1, “Inferring an agent’s loss function based on a term structure of forecasts”, we provide conditions for identification, estimation and inference of an agent’s loss function based on an observed term structure of point forecasts. The loss function specification is flexible as we allow the preferences to be both asymmetric and to vary non-linearly across the forecast horizon. In addition, we introduce a novel forecast rationality test based on the estimated loss function. We employ the approach to analyse the U.S. Government’s preferences over budget surplus forecast errors. Interestingly, we find that it is relatively more costly for the government to underestimate the budget surplus and that this asymmetry is stronger at long forecast horizons. In Chapter 2, “Monitoring Systemic Risk”, we define systemic risk as the conditional probability of a systemic banking crisis. This conditional probability is modelled in a fixed effect binary response panel-model framework that allows for cross-sectional dependence (e.g. due to contagion effects). In the empirical application we identify several risk factors and it is shown that the level of systemic risk contains a predictable component which varies through time. Furthermore, we illustrate how the forecasts of systemic risk map into dynamic policy thresholds in this framework. Finally, by conducting a pseudo out-of-sample exercise we find that the systemic risk estimates provided reliable early-warning signals ahead of the recent financial crisis for several economies. Finally, in Chapter 3, “Equity Premium Predictability”, we reassess the evidence of out-of- sample equity premium predictability. The empirical finance literature has identified several financial variables that appear to predict the equity premium in-sample. However, Welch & Goyal (2008) find that none of these variables have any predictive power out-of-sample. We show that the equity premium is predictable out-of-sample once you impose certain shrinkage restrictions on the model parameters. The approach is motivated by the observation that many of the proposed financial variables can be characterised as ’weak predictors’ and this suggest that a James-Stein type estimator will provide a substantial risk reduction. The out-of-sample explanatory power is small, but we show that it is, in fact, economically meaningful to an investor with time-invariant risk aversion. Using a shrinkage decomposition we also show that standard combination forecast techniques tends to ’overshrink’ the model parameters leading to suboptimal model forecasts.
|
88 |
擔保債權憑證選擇權之評價與避險:跨期因子相關結構性模型之運用 / On the pricing and risk characteristics of options on CDO tranches陳文萱, Chen,Wen Hsuan Unknown Date (has links)
這篇論文主要利用信用價差的時間結構與信用投資組合的損失分配評價擔保債權憑證選擇權。利用跨期因子相關結構性模型找到信用價差的動態過程及損失分配跨期相關性。這篇論文也探討了擔保債權憑證選擇權的風險值。最後,我們發現遠期生效擔保債權憑證與其選擇權對跨期損失相關性有高度敏感性。 / This article tries to find the term-structure of credit spread and portfolio loss distribution to price an option on CDO tranche. Our solution is based on a multiple period of factor copula model proposed by Andersen to fit the dynamic credit spread process by considering inter-temporal loss correlations through time. We also extend the model of valuing European options on CDO tranches presented by Hull and White and discuss the Greeks of the option formula. We numerically test the dependence of forward-starting CDOs on the correlation of loss across time. With the results, we price the options on CDO tranches. Finally, we find forward-starting CDOs and options on CDO tranches can have strong sensitivity to inter-temporal loss correlation.
|
89 |
Une méthode d'inférence bayésienne pour les modèles espace-état affines faiblement identifiés appliquée à une stratégie d'arbitrage statistique de la dynamique de la structure à terme des taux d'intérêtBlais, Sébastien January 2009 (has links)
Thèse numérisée par la Division de la gestion de documents et des archives de l'Université de Montréal.
|
90 |
Essays in macro finance and monetary economicsSomé, Modeste Yirbèhogré 01 1900 (has links)
Les questions abordées dans les deux premiers articles de ma thèse cherchent à comprendre les facteurs économiques qui affectent la structure à terme des taux d'intérêt et la prime de risque. Je construis des modèles non linéaires d'équilibre général en y intégrant des obligations de différentes échéances. Spécifiquement, le premier article a pour objectif de comprendre la relation entre les facteurs macroéconomiques et le niveau de prime de risque dans un cadre Néo-keynésien d'équilibre général avec incertitude. L'incertitude dans le modèle provient de trois sources : les chocs de productivité, les chocs monétaires et les chocs de préférences. Le modèle comporte deux types de rigidités réelles à savoir la formation des habitudes dans les préférences et les coûts d'ajustement du stock de capital. Le modèle est résolu par la méthode des perturbations à l'ordre deux et calibré à l'économie américaine. Puisque la prime de risque est par nature une compensation pour le risque, l'approximation d'ordre deux implique que la prime de risque est une combinaison linéaire des volatilités des trois chocs. Les résultats montrent qu'avec les paramètres calibrés, les chocs réels (productivité et préférences) jouent un rôle plus important dans la détermination du niveau de la prime de risque relativement aux chocs monétaires. Je montre que contrairement aux travaux précédents (dans lesquels le capital de production est fixe), l'effet du paramètre de la formation des habitudes sur la prime de risque dépend du degré des coûts d'ajustement du capital. Lorsque les coûts d'ajustement du capital sont élevés au point que le stock de capital est fixe à l'équilibre, une augmentation du paramètre de formation des habitudes entraine une augmentation de la prime de risque. Par contre, lorsque les agents peuvent librement ajuster le stock de capital sans coûts, l'effet du paramètre de la formation des habitudes sur la prime de risque est négligeable. Ce résultat s'explique par le fait que lorsque le stock de capital peut être ajusté sans coûts, cela ouvre un canal additionnel de lissage de consommation pour les agents. Par conséquent, l'effet de la formation des habitudes sur la prime de risque est amoindri. En outre, les résultats montrent que la façon dont la banque centrale conduit sa politique monétaire a un effet sur la prime de risque. Plus la banque centrale est agressive vis-à-vis de l'inflation, plus la prime de risque diminue et vice versa. Cela est due au fait que lorsque la banque centrale combat l'inflation cela entraine une baisse de la variance de l'inflation. Par suite, la prime de risque due au risque d'inflation diminue.
Dans le deuxième article, je fais une extension du premier article en utilisant des préférences récursives de type Epstein -- Zin et en permettant aux volatilités conditionnelles des chocs de varier avec le temps. L'emploi de ce cadre est motivé par deux raisons. D'abord des études récentes (Doh, 2010, Rudebusch and Swanson, 2012) ont montré que ces préférences sont appropriées pour l'analyse du prix des actifs dans les modèles d'équilibre général. Ensuite, l'hétéroscedasticité est une caractéristique courante des données économiques et financières. Cela implique que contrairement au premier article, l'incertitude varie dans le temps. Le cadre dans cet article est donc plus général et plus réaliste que celui du premier article. L'objectif principal de cet article est d'examiner l'impact des chocs de volatilités conditionnelles sur le niveau et la dynamique des taux d'intérêt et de la prime de risque. Puisque la prime de risque est constante a l'approximation d'ordre deux, le modèle est résolu par la méthode des perturbations avec une approximation d'ordre trois. Ainsi on obtient une prime de risque qui varie dans le temps. L'avantage d'introduire des chocs de volatilités conditionnelles est que cela induit des variables d'état supplémentaires qui apportent une contribution additionnelle à la dynamique de la prime de risque. Je montre que l'approximation d'ordre trois implique que les primes de risque ont une représentation de type ARCH-M (Autoregressive Conditional Heteroscedasticty in Mean) comme celui introduit par Engle, Lilien et Robins (1987). La différence est que dans ce modèle les paramètres sont structurels et les volatilités sont des volatilités conditionnelles de chocs économiques et non celles des variables elles-mêmes. J'estime les paramètres du modèle par la méthode des moments simulés (SMM) en utilisant des données de l'économie américaine. Les résultats de l'estimation montrent qu'il y a une évidence de volatilité stochastique dans les trois chocs. De plus, la contribution des volatilités conditionnelles des chocs au niveau et à la dynamique de la prime de risque est significative. En particulier, les effets des volatilités conditionnelles des chocs de productivité et de préférences sont significatifs. La volatilité conditionnelle du choc de productivité contribue positivement aux moyennes et aux écart-types des primes de risque. Ces contributions varient avec la maturité des bonds. La volatilité conditionnelle du choc de préférences quant à elle contribue négativement aux moyennes et positivement aux variances des primes de risque. Quant au choc de volatilité de la politique monétaire, son impact sur les primes de risque est négligeable.
Le troisième article (coécrit avec Eric Schaling, Alain Kabundi, révisé et resoumis au journal of Economic Modelling) traite de l'hétérogénéité dans la formation des attentes d'inflation de divers groupes économiques et de leur impact sur la politique monétaire en Afrique du sud. La question principale est d'examiner si différents groupes d'agents économiques forment leurs attentes d'inflation de la même façon et s'ils perçoivent de la même façon la politique monétaire de la banque centrale (South African Reserve Bank). Ainsi on spécifie un modèle de prédiction d'inflation qui nous permet de tester l'arrimage des attentes d'inflation à la bande d'inflation cible (3% - 6%) de la banque centrale. Les données utilisées sont des données d'enquête réalisée par la banque centrale auprès de trois groupes d'agents : les analystes financiers, les firmes et les syndicats. On exploite donc la structure de panel des données pour tester l'hétérogénéité dans les attentes d'inflation et déduire leur perception de la politique monétaire. Les résultats montrent qu'il y a évidence d'hétérogénéité dans la manière dont les différents groupes forment leurs attentes. Les attentes des analystes financiers sont arrimées à la bande d'inflation cible alors que celles des firmes et des syndicats ne sont pas arrimées. En effet, les firmes et les syndicats accordent un poids significatif à l'inflation retardée d'une période et leurs prédictions varient avec l'inflation réalisée (retardée). Ce qui dénote un manque de crédibilité parfaite de la banque centrale au vu de ces agents. / This thesis consists of three essays in the areas of macro finance and monetary economics. The first two essays deal with the analysis of the term structure of interest rates in dynamic and stochastic general equilibrium (DSGE) models. The third essay explores inflation expectations formation across different economic groups in South Africa.
Interest rates are one channel through which monetary policy affects the real economy. Typically, central banks implement monetary policy by influencing short term interest rates. Theoretically, the interest rate on a long-term bond is the average of expected future short term interest rates over the maturity period, plus a risk premium demanded by the holder of the bond to compensate for the risk involved in holding a longer maturity bond. Therefore, any changes in the target rate of the central bank and the risk premium affect long -- term interest rates, such as mortgage rates and interest rates on certain durable goods. It is then important for the central bank to understand the economic factors that affect both components of long - term interest namely the market expectations about the short - term rates and the risk premium. For example, recently in the U.S. economy, between June 2004 and June 2006, the ineffectiveness of monetary policy to affect long - term interest rates has been attributed to a decline in risk premium over this period, which has offset the effect of the increase in the target rate of the Federal Reserve (Fed). In the implementation of its monetary policy, the central bank can more or less control agents' expectations through transparent communication. However, the risk premium is endogenous and unobservable and therefore can not be fully controlled by the central bank. On the other hand, achieving the goal of prices stability in an inflation targeting framework depends on the credibility of the central bank.
In the first two essays I explore the economic factors of the term structure of interest rates and risk premiums. I build a non-linear dynamic stochastic general equilibrium (DSGE) models whereby I incorporate a range of bonds with different maturities. Specifically, the goal of the first essay is to understand the relationship between macroeconomic factors and the level of risk premium in a New Keynesian general equilibrium framework. Uncertainty in the model comes from three sources: productivity, monetary policy and, preferences shocks. The model has two types of real rigidities namely habit formation in preferences and adjustment costs in capital stock. The model is solved by perturbation method up to second order and calibrated to the U.S. economy. Since the risk premium is by nature a compensation for risk, the second - order approximation implies that the risk premium is a linear combination of the volatility of the three shocks. Results show that at the calibrated parameters, real shocks (productivity and preferences) play a more important role in determining the level of the risk premium relative to monetary shocks. I show that, contrary to previous work (where production capital is fixed), the effect of habit formation on the risk premium depends on the degree of capital adjustment cost. When capital adjustment costs are so high that the capital stock is fixed in equilibrium, an increase in the parameter of habit formation leads to an increase in the risk premium. However, when agents can freely adjust the capital stock without cost, the effect of the habit formation parameter on the risk premium is negligible. This result is explained by the fact that when the capital stock can be adjusted without cost, it opens an additional channel to the agents for consumption smoothing. Therefore, the effect of habit formation on the risk premium is reduced. In addition, the results show that the way the central bank conducts its monetary policy has an effect on the risk premium. The more aggressive the central bank vis-à-vis inflation, the lower the risk premium and vice versa. This is due to the fact that when the central bank fights against inflation it leads to a decrease in the variance of inflation. As a result, the risk premium due to inflation risk decreases.
In the second essay, I extend the analysis of the first essay by using recursive preferences (as those proposed by Epstein - Zin) and by allowing the conditional volatility of the shocks to be time - varying. The use of this framework is motivated by two reasons. First, recent studies (Doh, 2010, Rudebusch and Swanson, 2012) showed that these preferences are appropriate for the analysis of asset prices in general equilibrium models. Second, heteroscedasticity is a prominent feature of economic and financial data. This implies that, contrary to the first essay, the uncertainty here is time - varying. Thus, the framework in this essay is more general and realistic than in the first essay. The main objective of this paper is to examine the impact of uncertainty due to conditional volatility of the shocks on the level and the dynamics of interest rates and risk premiums. Since the risk premium is constant at second order approximation, the model is solved by the perturbation method with an approximation of order three in order to get a time - varying risk premium. The advantage of introducing shocks conditional volatilities is that , it induces additional state variables that provide an additional contribution to the dynamics of the risk premium. I show that the risk premiums implied by the third -- order approximate solution have an ARCH-M (Autoregressive Conditional Heteroscedasticty in Mean) type representation as that introduced by Engle, Lilien and Robins (1987). The difference is that in this model the parameters are structural and the volatilities are conditional volatility of economic shocks and not those of the variables themselves. I estimate the model parameters by Simulated Method of Moments (SMM) using U.S. data. The estimation results show that there is evidence of stochastic volatility in the three shocks. Moreover, the contribution of conditional shocks volatility to the level and the dynamics of the risk premium is significant. In particular, the effects of the conditional volatility of productivity and preferences shocks are important. The conditional volatility of the productivity shock contributes positively to the means and standard deviations of risk premiums. These contributions vary with the maturity of the bonds. Conditional volatility of the preferences shock contributes negatively to the averages and positively to the variances of risk premiums. As for the impact of volatility of monetary policy shock, its impact on the risk premium is negligible.
The third article (coauthored with Eric Schaling and Alain Kabundi, revised and resubmitted to the journal of Economic Modelling) deals with heterogeneity in inflation expectations of different economic agents and its impact on monetary policy in South Africa. The main question is to examine whether different groups of economic agents form their inflation expectations in the same way and if they perceive the central bank (South African Reserve Bank) monetary policy in the say way. We specify an inflation expectation model that allows us to directly test whether inflation expectations are anchored or not to the inflation target band (3% - 6%). The data used are inflation expectations data from surveys conducted by the central bank. There are three groups of agents: financial analysts, businesses and trade unions. We therefore exploits the panel structure of the data to test the heterogeneity in inflation expectations and derive their perceived inflation targets. Results show that there is evidence of heterogeneity in the way the three groups form their expectations. The expectations of financial analysts are well anchored to the central bank target band while those of businesses and trade unions are not. In fact, businesses and trade unions put a higher weight on lagged realized inflation in their expectations. This Indicates a lack of full credibility of the central bank.
|
Page generated in 0.0794 seconds