Essays in Econometric Forecasting with Applications to Macroeconomics and Finance

This thesis comprises three chapters on econometric forecasting, with applications to macroeconomics and finance. The first chapter introduces a testing procedure for evaluating superior predictive ability (SPA) in unstable environments. Therefore, it allows one to detect time-varying superiority. A...

Descripción completa

Detalles Bibliográficos
Autor: Crespo Rey, Ignacio
Tipo de recurso: tesis doctoral
Estado:Versión publicada
Fecha de publicación:2024
País:España
Institución:CBUC, CESCA
Repositorio:TDR. Tesis Doctorales en Red
OAI Identifier:oai:www.tdx.cat:10803/691982
Acceso en línea:http://hdl.handle.net/10803/691982
Access Level:acceso embargado
Palabra clave:Econometric forecasting
Pronóstico econométrico
Macroeconomics
Macroeconomía
Finance
Finanzas
33
Descripción
Sumario:This thesis comprises three chapters on econometric forecasting, with applications to macroeconomics and finance. The first chapter introduces a testing procedure for evaluating superior predictive ability (SPA) in unstable environments. Therefore, it allows one to detect time-varying superiority. Applied to downside risk forecasts for the U.S. economy, it reveals substantial time-varying heterogeneity in the forecasting performance of the commonly used method (a quantile regression equipped with financial conditions). In the second chapter, we present a variant of the standard SPA test that assesses the predictive ability of one particular method against many alternatives when forecasting a panel of time series. Our empirical analysis assesses the forecasting performance of the factor model against different machine learning techniques when predicting macroeconomic variables in the U.S. (FRED-MD). We find that while the factor model dominates at short horizons, yet it is surpassed by simple methods at longer horizons, with results varying across different macroeconomic categories. Finally, the third chapter proposes a hybrid model combining factor models and largedimensional regularized regressions for intra-daily volume prediction in large panels of stocks. Applied to the STOXX 600 Index, our results demonstrate its superior performance over traditional univariate models.