RC-Macro Model
Global VAR forecasts and stress scenarios by region
The problem
A macroeconomic stress scenario has to be internally consistent: a shock in one economy should move the others, and the paths should follow from estimated relationships rather than be set by hand. RC-Macro Model estimates those relationships region by region, combines them into a global model, and lets a shock to one variable carry through to the rest via the estimated covariance and the model's dynamics. Analysts can still fix chosen paths with overrides and bounds.
What it does
RC-Macro Model is an implementation of the Global VAR (GVAR) model of Pesaran, Schuermann and Weiner (2004). It fits a model for each region, links them into one global model through foreign weights, and produces a base macroeconomic forecast and stress scenarios built from shocks and overrides. Published scenarios are used by the stress-testing application.
What’s inside
- Region-specific error-correction models estimated by reduced rank regression (Johansen 1988)
- Domestic, foreign and global variables per region, with foreign variables built from a foreign weights matrix
- Cointegration rank set by trace statistics against critical values supplied with the data
- Region models combined through link matrices into global intercept, trend and coefficients
- Residual covariance with an optional Ledoit-Wolf style adjustment (Bailey, Pesaran and Smith 2014)
- Annual, quarterly or monthly data; sample start and forecast horizon set as options
- GVAR base forecast, with optional base shocks, overrides and lower bounds
- A shock to one variable carried to the others through the estimated covariance and the model's dynamics
- Market variables derived from a forecast rate plus a risk premium
- FX variables derived from the difference between two rates plus a risk premium
- Stress scenarios defined in the input workbook or created later against a stored model
- Each scenario reported in levels and against the base forecast
- Input workbook validated before calculation, with messages naming the sheet and region/variable at fault
- Calculation status and errors kept with each model
- Models and scenarios published for use by other applications
- Scenarios locked against accidental edits; a scenario still in use by a calculation cannot be deleted
- Input data and results exported to Excel
What changes for your risk team
- Regional economies are modelled together, so a shock in one region reaches the others through the foreign weights
- Stress scenarios start from an estimated base forecast and are reported against it
- Each estimation step follows a published method, referenced to the paper's equations
- The input workbook is checked before anything is calculated
How it feeds the rest of the suite
- RC-Stress Testing System: consumes published scenarios and serves the model to users
See RC-Macro Model on your own data
A demo takes an hour. A pilot on your portfolio takes weeks, not quarters.