Episode 01 - Presented by Dr. Tim Ginker

Dr. Tim Ginker presented his recent research on interpreting conditional forecasts in reduced-form VAR models and introduced the accompanying R package, cforecast. The tutorial covered how to generate and interpret scenario-based forecasts using reduced-form VARs, with a focus on understanding which assumptions drive forecast revisions. Participants learned how conditional response functions (CRFs) extend generalized impulse response analysis to multi-period scenario conditioning, how to quantify variable importance under alternative scenarios, and how to decompose forecast revisions into contributions from specific conditioning paths. The session also demonstrated practical implementation in R using the cforecast package for policy analysis, stress testing, and macroeconomic forecasting applications. 

The Empirical Application

The empirical illustration uses U.S. quarterly data from 1986Q2 to 2015Q4, with a five-variable reduced-form VAR(2) selected by BIC. The variables are real GDP growth, core PCE inflation, the federal funds rate, the Moody's Baa–10Y corporate credit spread, and the WTI crude oil price. The forecast exercise begins at 2016Q1, looking 20 quarters (five years) ahead.

The Scenario Forecasting

The scenario combines two inputs that are common in policy and stress-testing work: Financial tightening: the corporate credit spread widens by approximately 200 basis points relative to baseline, remains elevated for several quarters, and then gradually normalises. Energy price disruption: the WTI oil price rises sharply above baseline, peaking around $75 per barrel within the first year, and then mean-reverts.