Essays on macroeconomic forecasting

• Autores: Alex Tagliabracci
• Directores de la Tesis: Luca Gambetti (dir. tes.)
• Lectura: En la Universitat Autònoma de Barcelona ( España ) en 2018
• Idioma: español
• Tribunal Calificador de la Tesis: Gabriel Pérez-Quirós (presid.), Stefano Neri (secret.), Alberto Russo (voc.)
• Programa de doctorado: Programa Oficial de Doctorado en Análisis Económico
• Materias:
  • Ciencias económicas
    • Econometría
      • Proyección económica
• Enlaces
  • Tesis en acceso abierto en:  TESEO  TDX 
• Resumen

  • This thesis is a collection of three empirical essays with a focus on forecasting. The first chapter focuses on an important policy task as forecasting inflation. The work aims to investigate how the dynamics of the business cycle may impact the distribution of inflation forecasts. The second chapter considers two econometric models used in the nowcasting literature and propose a comparison with an application to the Italian GDP. The last chapter is centered around forecasting the effects of macroeconomic data releases on the exchange rates.

    The first chapter studies how the business cycle affects the conditional distribution of euro area inflation forecasts. Using a quantile regression approach, I estimate the conditional distribution of inflation to show its evolution over time allowing for asymmetries across quantiles. I document the evidence of downside risks to inflation which vary in relation to developments of the state of the econ- omy while the upside risk remains relatively stable over time. I also find that this evidence partially characterizes the corresponding distribution derived from ECB Survey of Professional Forecasters.

    The second chapter proposes two multivarite econometric models that consider two important characteristics in the nowcasting literature, as timely and high frequency data, to predict Italian GDP, namely a dynamic factor model and a mixed-frequency Bayesian VAR. A pseudo out-of-sample exercise shows three main results: (i) both models considerably outperform a standard univariate benchmark; (ii) the dynamic factor model turns out to be more reliable at the end of the forecasting period while the mixed-frequency BVAR appears superior with an incomplete information set; (iii) the overall forecasting superiority of the dynamic factor model is mainly driven by its ability in capturing the severity of recession episodes.

    Finally, the third chapter, jointly written with Luca Brugnolini and Antonello D’Agostino, inves- tigates the possible predictability of macroeconomic surprises and their effects on the exchange rates. In particular, we analyze two of the most important data releases that impact the US financial mar- ket, namely the change in the level of non-farm payroll employment (NFP) and the manufacturing index published by the Institute for Supply Management (ISM). We examine the unexpected compo- nent of these two, as measured by the deviation of the actual release from the Bloomberg Consensus. We label it as the market surprise, and we investigate whether its structure is partially predictable and in which cases. Secondly, we use high-frequency data on the eurodollar as a laboratory to study the effect of these surprises. We show in a regression framework that although the in-sample fit is sufficiently good, the performance deteriorates in an out-of-sample setting because a naive model can hardly be beaten in a sixty-minute window after the release. Finally, we demonstrate that under certain circumstances there is some structure that can be exploited and we provide a framework to take advantages of it.