Confidence sets for asset correlations in portfolio credit risk

Asset correlations are of critical importance in quantifying portfolio credit risk and economic capitalin financial institutions. Estimation of asset correlation with rating transition data has focusedon the point estimation of the correlation without giving any consideration to the uncertaintyaround these point estimates. In this article we use Bayesian methods to estimate a dynamicfactor model for default risk using rating data (McNeil et al., 2005; McNeil and Wendin, 2007).Bayesian methods allow us to formally incorporate human judgement in the estimation of assetcorrelation, through the prior distribution and fully characterize a confidence set for the correlations.Results indicate: i) a two factor model rather than the one factor model, as proposed bythe Basel II framework, better represents the historical default data. ii) importance of unobservedfactors in this type of models is reinforced and point out that the levels of the implied asset correlationscritically depend on the latent state variable used to capture the dynamics of default,as well as other assumptions on the statistical model. iii) the posterior distributions of the assetcorrelations show that the Basel recommended bounds, for this parameter, undermine the levelof systemic risk.

Evaluating multivariate volatility forecasts : how effective are statistical and economic loss functions?

Multivariate volatility forecasts are an important input in many financial applications, in particular portfolio optimisation problems. Given the number of models available and the range of loss functions to discriminate between them, it is obvious that selecting the optimal forecasting model is challenging. The aim of this thesis is to thoroughly investigate how effective many commonly used statistical (MSE and QLIKE) and economic (portfolio variance and portfolio utility) loss functions are at discriminating between competing multivariate volatility forecasts. An analytical investigation of the loss functions is performed to determine whether they identify the correct forecast as the best forecast. This is followed by an extensive simulation study examines the ability of the loss functions to consistently rank forecasts, and their statistical power within tests of predictive ability. For the tests of predictive ability, the model confidence set (MCS) approach of Hansen, Lunde and Nason (2003, 2011) is employed. As well, an empirical study investigates whether simulation findings hold in a realistic setting. In light of these earlier studies, a major empirical study seeks to identify the set of superior multivariate volatility forecasting models from 43 models that use either daily squared returns or realised volatility to generate forecasts. This study also assesses how the choice of volatility proxy affects the ability of the statistical loss functions to discriminate between forecasts. Analysis of the loss functions shows that QLIKE, MSE and portfolio variance can discriminate between multivariate volatility forecasts, while portfolio utility cannot. An examination of the effective loss functions shows that they all can identify the correct forecast at a point in time, however, their ability to discriminate between competing forecasts does vary. That is, QLIKE is identified as the most effective loss function, followed by portfolio variance which is then followed by MSE. The major empirical analysis reports that the optimal set of multivariate volatility forecasting models includes forecasts generated from daily squared returns and realised volatility. Furthermore, it finds that the volatility proxy affects the statistical loss functions’ ability to discriminate between forecasts in tests of predictive ability. These findings deepen our understanding of how to choose between competing multivariate volatility forecasts.

Forecasting volatility and correlation : the role of option implied measures

Forecasts of volatility and correlation are important inputs into many practical financial problems. Broadly speaking, there are two ways of generating forecasts of these variables. Firstly, time-series models apply a statistical weighting scheme to historical measurements of the variable of interest. The alternative methodology extracts forecasts from the market traded value of option contracts. An efficient options market should be able to produce superior forecasts as it utilises a larger information set of not only historical information but also the market equilibrium expectation of options market participants. While much research has been conducted into the relative merits of these approaches, this thesis extends the literature along several lines through three empirical studies. Firstly, it is demonstrated that there exist statistically significant benefits to taking the volatility risk premium into account for the implied volatility for the purposes of univariate volatility forecasting. Secondly, high-frequency option implied measures are shown to lead to superior forecasts of the intraday stochastic component of intraday volatility and that these then lead on to superior forecasts of intraday total volatility. Finally, the use of realised and option implied measures of equicorrelation are shown to dominate measures based on daily returns.

On the efficacy of techniques for evaluating multivariate volatility forecasts

The performance of techniques for evaluating multivariate volatility forecasts are not yet as well understood as their univariate counterparts. This paper aims to evaluate the efficacy of a range of traditional statistical-based methods for multivariate forecast evaluation together with methods based on underlying considerations of economic theory. It is found that a statistical-based method based on likelihood theory and an economic loss function based on portfolio variance are the most effective means of identifying optimal forecasts of conditional covariance matrices.

Essays on volatility and liquidity in financial markets

The price formation of financial assets is a complex process. It extends beyond the standard economic paradigm of supply and demand to the understanding of the dynamic behavior of price variability, the price impact of information, and the implications of trading behavior of market participants on prices. In this thesis, I study aggregate market and individual assets volatility, liquidity dimensions, and causes of mispricing for US equities over a recent sample period. How volatility forecasts are modeled, what determines intradaily jumps and causes changes in intradaily volatility and what drives the premium of traded equity indexes? Are they induced, for example, by the information content of lagged volatility and return parameters or by macroeconomic news, changes in liquidity and volatility? Besides satisfying our intellectual curiosity, answers to these questions are of direct importance to investors developing trading strategies, policy makers evaluating macroeconomic policies and to arbitrageurs exploiting mispricing in exchange-traded funds. Results show that the leverage effect and lagged absolute returns improve forecasts of continuous components of daily realized volatility as well as jumps. Implied volatility does not subsume the information content of lagged returns in forecasting realized volatility and its components. The reported results are linked to the heterogeneous market hypothesis and demonstrate the validity of extending the hypothesis to returns. Depth shocks, signed order flow, the number of trades, and resiliency are the most important determinants of intradaily volatility. In contrast, spread shock and resiliency are predictive of signed intradaily jumps. There are fewer macroeconomic news announcement surprises that cause extreme price movements or jumps than those that elevate intradaily volatility. Finally, the premium of exchange-traded funds is significantly associated with momentum in net asset value and a number of liquidity parameters including the spread, traded volume, and illiquidity. The mispricing of industry exchange traded funds suggest that limits to arbitrage are driven by potential illiquidity.

On the benefits of equicorrelation for portfolio allocation

The importance of modelling correlation has long been recognised in the field of portfolio management, with largedimensional multivariate problems increasingly becoming the focus of research. This paper provides a straightforward and commonsense approach toward investigating a number of models used to generate forecasts of the correlation matrix for large-dimensional problems.We find evidence in favour of assuming equicorrelation across various portfolio sizes, particularly during times of crisis. During periods of market calm, however, the suitability of the constant conditional correlation model cannot be discounted, especially for large portfolios. A portfolio allocation problem is used to compare forecasting methods. The global minimum variance portfolio and Model Confidence Set are used to compare methods, while portfolio weight stability and relative economic value are also considered.

Statistical Inference for Computable General Equilibrium Models with Application to a Model of the Moroccan Economy

We study the problem of measuring the uncertainty of CGE (or RBC)-type model simulations associated with parameter uncertainty. We describe two approaches for building confidence sets on model endogenous variables. The first one uses a standard Wald-type statistic. The second approach assumes that a confidence set (sampling or Bayesian) is available for the free parameters, from which confidence sets are derived by a projection technique. The latter has two advantages: first, confidence set validity is not affected by model nonlinearities; second, we can easily build simultaneous confidence intervals for an unlimited number of variables. We study conditions under which these confidence sets take the form of intervals and show they can be implemented using standard methods for solving CGE models. We present an application to a CGE model of the Moroccan economy to study the effects of policy-induced increases of transfers from Moroccan expatriates.

Méthodes d’inférence exactes pour un modèle de régression avec erreurs AR(2) gaussiennes

Ce texte propose des méthodes d’inférence exactes (tests et régions de confiance) sur des modèles de régression linéaires avec erreurs autocorrélées suivant un processus autorégressif d’ordre deux [AR(2)], qui peut être non stationnaire. L’approche proposée est une généralisation de celle décrite dans Dufour (1990) pour un modèle de régression avec erreurs AR(1) et comporte trois étapes. Premièrement, on construit une région de confiance exacte pour le vecteur des coefficients du processus autorégressif (φ). Cette région est obtenue par inversion de tests d’indépendance des erreurs sur une forme transformée du modèle contre des alternatives de dépendance aux délais un et deux. Deuxièmement, en exploitant la dualité entre tests et régions de confiance (inversion de tests), on détermine une région de confiance conjointe pour le vecteur φ et un vecteur d’intérêt M de combinaisons linéaires des coefficients de régression du modèle. Troisièmement, par une méthode de projection, on obtient des intervalles de confiance «marginaux» ainsi que des tests à bornes exacts pour les composantes de M. Ces méthodes sont appliquées à des modèles du stock de monnaie (M2) et du niveau des prix (indice implicite du PNB) américains

Identification, Weak Instruments and Statistical Inference in Econometrics

We discuss statistical inference problems associated with identification and testability in econometrics, and we emphasize the common nature of the two issues. After reviewing the relevant statistical notions, we consider in turn inference in nonparametric models and recent developments on weakly identified models (or weak instruments). We point out that many hypotheses, for which test procedures are commonly proposed, are not testable at all, while some frequently used econometric methods are fundamentally inappropriate for the models considered. Such situations lead to ill-defined statistical problems and are often associated with a misguided use of asymptotic distributional results. Concerning nonparametric hypotheses, we discuss three basic problems for which such difficulties occur: (1) testing a mean (or a moment) under (too) weak distributional assumptions; (2) inference under heteroskedasticity of unknown form; (3) inference in dynamic models with an unlimited number of parameters. Concerning weakly identified models, we stress that valid inference should be based on proper pivotal functions —a condition not satisfied by standard Wald-type methods based on standard errors — and we discuss recent developments in this field, mainly from the viewpoint of building valid tests and confidence sets. The techniques discussed include alternative proposed statistics, bounds, projection, split-sampling, conditioning, Monte Carlo tests. The possibility of deriving a finite-sample distributional theory, robustness to the presence of weak instruments, and robustness to the specification of a model for endogenous explanatory variables are stressed as important criteria assessing alternative procedures.

Projeção de inflação no Brasil utilizando dados agregados e desagregados : um teste de poder preditivo por horizonte de tempo

O trabalho tem como objetivo comparar a eficácia das diferentes metodologias de projeção de inflação aplicadas ao Brasil. Serão comparados modelos de projeção que utilizam os dados agregados e desagregados do IPCA em um horizonte de até doze meses à frente. Foi utilizado o IPCA na base mensal, com início em janeiro de 1996 e fim em março de 2012. A análise fora da amostra foi feita para o período entre janeiro de 2008 e março de 2012. Os modelos desagregados serão estimados por SARIMA, pelo software X-12 ARIMA disponibilizado pelo US Census Bureau, e terão as aberturas do IPCA de grupos (9) e itens (52), assim como aberturas com sentido mais econômico utilizadas pelo Banco Central do Brasil como: serviços, administrados, alimentos e industrializados; duráveis, não duráveis, semiduráveis, serviços e administrados. Os modelos agregados serão estimados por técnicas como SARIMA, modelos estruturais em espaço-estado (Filtro de Kalman) e Markov-switching. Os modelos serão comparados pela técnica de seleção de modelo Model Confidence Set, introduzida por Hansen, Lunde e Nason (2010), e Dielbod e Mariano (1995), no qual encontramos evidências de ganhos de desempenho nas projeções dos modelos mais desagregados em relação aos modelos agregados.

Previsão da inadimplência bancária no Brasil através dos métodos FAVAR e FAVECM

O objetivo do presente trabalho é utilizar modelos econométricos de séries de tempo para previsão do comportamento da inadimplência agregada utilizando um conjunto amplo de informação, através dos métodos FAVAR (Factor-Augmented Vector Autoregressive) de Bernanke, Boivin e Eliasz (2005) e FAVECM (Factor-augmented Error Correction Models) de Baneerjee e Marcellino (2008). A partir disso, foram construídas previsões fora da amostra de modo a comparar a eficácia de projeção dos modelos contra modelos univariados mais simples - ARIMA - modelo auto-regressivo integrado de média móvel e SARIMA - modelo sazonal auto-regressivo integrado de média móvel. Para avaliação da eficácia preditiva foi utilizada a metodologia MCS (Model Confidence Set) de Hansen, Lunde e James (2011) Essa metodologia permite comparar a superioridade de modelos temporais vis-à-vis a outros modelos.

Forecasting Brazilian inflation by its aggregate and disaggregated data: a test of predictive power by forecast horizon

This work aims to compare the forecast efficiency of different types of methodologies applied to Brazilian Consumer inflation (IPCA). We will compare forecasting models using disaggregated and aggregated data over twelve months ahead. The disaggregated models were estimated by SARIMA and will have different levels of disaggregation. Aggregated models will be estimated by time series techniques such as SARIMA, state-space structural models and Markov-switching. The forecasting accuracy comparison will be made by the selection model procedure known as Model Confidence Set and by Diebold-Mariano procedure. We were able to find evidence of forecast accuracy gains in models using more disaggregated data

Desagregação e pesos estocásticos em projeções de agregados econômicos : uma análise para o PIB brasileiro

O presente estudo tem como objetivo comparar e combinar diferentes técnicas de projeção para o PIB trimestral brasileiro de 1991 ao segundo trimestre de 2014, utilizando dados agregados, e dados desagregados com pesos fixos e estocásticos. Os modelos desagregados univariados e multivariados, assim como os pesos estocásticos, foram estimados pelo algoritmo Autometrics criado por Doornik (2009), através dos níveis de desagregação disponibilizados pelo IBGE no Sistema de Contas Nacionais. Os modelos agregados foram estimados pelo Autometrics, por Markov-Switching e por modelos estruturais de espaço-estado. A metodologia de comparação de projeções utilizada foi o Model Confidence Set, desenvolvida por Hanse, Lunde e Nason (2011). Foram realizadas duas simulações, sendo a primeira com a análise fora da amostra a partir de 2008, e a segunda a partir de 2000, com horizonte de projeção de até 6 passos à frente. Os resultados sugerem que os modelos desagregados com pesos fixos desempenham melhor nos dois primeiros passos, enquanto nos períodos restantes os modelos da série agregada geram melhores previsões.

Descobrindo modelos de previsão para a inflação brasileira : uma análise a partir do algoritmo autometrics

O presente trabalho tem como objetivo avaliar a capacidade preditiva de modelos econométricos de séries de tempo baseados em indicadores macroeconômicos na previsão da inflação brasileira (IPCA). Os modelos serão ajustados utilizando dados dentro da amostra e suas projeções ex-post serão acumuladas de um a doze meses à frente. As previsões serão comparadas a de modelos univariados como autoregressivo de primeira ordem - AR(1) - que nesse estudo será o benchmark escolhido. O período da amostra vai de janeiro de 2000 até agosto de 2015 para ajuste dos modelos e posterior avaliação. Ao todo foram avaliadas 1170 diferentes variáveis econômicas a cada período a ser projetado, procurando o melhor conjunto preditores para cada ponto no tempo. Utilizou-se o algoritmo Autometrics para a seleção de modelos. A comparação dos modelos foi feita através do Model Confidence Set desenvolvido por Hansen, Lunde e Nason (2010). Os resultados obtidos nesse ensaio apontam evidências de ganhos de desempenho dos modelos multivariados para períodos posteriores a 1 passo à frente.

Comparação de previsões para a produção industrial brasileira considerando efeitos calendário e modelos agregados e desagregados

O trabalho tem como objetivo verificar a existência e a relevância dos Efeitos Calendário em indicadores industriais. São explorados modelos univariados lineares para o indicador mensal da produção industrial brasileira e alguns de seus componentes. Inicialmente é realizada uma análise dentro da amostra valendo-se de modelos estruturais de espaço-estado e do algoritmo de seleção Autometrics, a qual aponta efeito significante da maioria das variáveis relacionadas ao calendário. Em seguida, através do procedimento de Diebold-Mariano (1995) e do Model Confidence Set, proposto por Hansen, Lunde e Nason (2011), são realizadas comparações de previsões de modelos derivados do Autometrics com um dispositivo simples de Dupla Diferença para um horizonte de até 24 meses à frente. Em geral, os modelos Autometrics que consideram as variáveis de calendário se mostram superiores nas projeções de 1 a 2 meses adiante e superam o modelo simples em todos os horizontes. Quando se agrega os componentes de categoria de uso para formar o índice industrial total, há evidências de ganhos nas projeções de prazo mais curto.