Essays on high dimensional financial econometrics

• Autor(a) principal: Cunha, Ronan
• Data de Publicação: 2019
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: http://hdl.handle.net/10438/27781
• topic: Forecasting; Automatic model selection; Credit volume; Cross-sectional asset pricing; Covariance matrix; Autometrics; LASSO; Vector autoregression; Portfolio allocation; Previsão; Seleção automática de modelos; Volume de crédito; Precificação de ativos no cross-section; Matriz de covariância; Vetor autorregressivo; Alocação de portfólio; Economia; Econometria; Previsão econômica; Modelo de precificação de ativos; Algoritmos; Modelos econométricos

High dimensional models have gains relatively importance in several areas of economics due to advances in technology that has increased the set of information. They are useful to find and understand the relationship between agents, variables, to forecast and to generate inputs to economic decisions and policy making. The increase in dimensions, however, brings new challenges to overcome. Strategies to deal with highly parametrized models have been developed. This thesis aims to study three subjects in this literature. The first one is to evaluate the performance of macroeconomic indicators in forecasting out-of-sample behaviour of the aggregated and disaggregated at the State level credit volume series from May 2011 till April 2016. A variety of time series techniques are used to model level of credit such as Vector autoregressive models, Global Vector Autoregressive models and ARIMA. This study also uses the model selection algorithm called Autometrics to select parsimonious models in the forecast exercise. This work uses volume of credit from January 2004 to April 2011 as the first estimation growing sample. As macroeconomic indicator, we use the default indicator, industrial production, short-term interest rate (Selic) and inflation (IPCA). At State level, we collect data from employment and default rates. The best results at disaggregated level is obtained for Global VAR model. At aggregate level, VAR model with macroeconomic indicators has the best performance particularly at higher horizons. There is also some evidence that forecast combinations techniques help to improve prediction performance by reducing the Mean Absolute and Squared Forecast Errors. The second study revisits the empirical cross-sectional asset pricing literature analysing and comparing the explanatory power of 65 risk factors for Brazilian stock market from 2000 to 2017 using automatic model selection techniques. We apply two standard methodologies in the literature, time series and cross sectional approaches, for 234 portfolios. The results show that, for time series approach, excess market return and small minus big (SMB) are the most selected factors. Specifically, the former is selected to all tested portfolios. For cross-sectional approach, the average selection rates are very similar among factors, but some macroeconomic variables stand out. The third essay aims to compare the performance of three automatic model selection algorithms, Autometrics (Hendry and Krolzig, 1999; Doornik, 2009), LASSOVAR and adaptive LASSOVAR (Tibshirani, 1996; Zhao and Yu, 2006; Callot et al., 2017) for modelling and forecasting monthly covariance matrices. To do so, we compose a database with daily information for 30 Brazilian stocks (B3), which yields 465 unique entries in the matrix from July/2009 to December/2017. We apply three forecasting error measures, the model confidence set (Hansen et al., 2011) and Giacomini and White (2006) conditional test in the comparison. We also calculate the economic value for each of the forecasting strategy through a portfolio selection exercise. The results show that the individual models are not able to beat the benchmark, the random walk, but a weighted combination of them is able to increase precision up to 13%. The portfolio selection exercises find that there are economic gains for using automatic model selection techniques to model and forecast the covariance matrices. Specifically, under short-selling constraint, Autometrics VAR(1) with dummy saturation delivers the highest Sharpe-ratio and economic value. When the investor is able to short-sell, either Autometrics VAR(1) with dummy saturation or adaptive LASSOVAR(1) is preferable. This final choice depend on the risk aversion of the investor. If he is less risk-averse, he prefers the former, while the latter becomes his choice if his risk-aversion sensitivity increases.