Chapter 2 - Macroeconomic Scenario Analysis from a Bank Perspective

Abstract

Time series analysis is vital for stress testing and risk integration. Two main perspectives can be followed to forecast the evolution of an economy. On the one hand, a structural approach aims to apply the economic theory to real data. However, it may suffer from some statistical deficiencies. On the other hand, a pure statistical method solves the latter issue but may lack a genuine economic interpretation. The practical difficulties in dealing with a structural method suggest we focus on a more marked econometric mindset. As a first step, the study starts from the univariate autoregression and moving average processes. Moreover, the Box-Jenkins analysis provides an accurate and parsimonious framework for the investigation of real time series. A practical implementation in MATLAB and R helps readers familiarize themselves with these modeling techniques. Moving to the multivariate analysis, the traditional vector autoregression and vector error-correction models constitute the essential toolkit to deal with macroeconomic time series. Estimation, simulation, and forecast are components of a comprehensive process aimed at assessing the impact of external shocks on a bank’s business. When interdependencies increase, models capturing international linkages become crucial. In this regard, the global vector autoregression model facilitates a multicountry business stress test analysis as required in Bank Alpha’s example. Additionally, a bank usually requires variables other than those included in a regulatory scenario to run its internal stress testing models. Therefore an enrichment process concludes the chapter by expediting the scrutiny of a wider set of coherent macroeconomic forecasts.