Separate forecasts for rates, equity returns, and inflation can miss delayed relationships. A vector autoregression uses the joint history of those series. This lesson explains estimation, lag selection, predictive-content tests, and fitted shock responses. Reduced-form predictability does not identify the effect of a policy intervention; structural shock interpretations require additional restrictions. For cointegrated levels, retain the equilibrium term through a VECM.
Learning objectives. By the end of this note you should be able to:
- write a reduced-form VAR() and explain equation-by-equation OLS;
- select a common lag order with multivariate information criteria;
- formulate Granger non-causality and state what it does not establish;
- interpret IRFs and FEVDs, including the role of innovation ordering;
- distinguish reduced-form VAR, SVAR, and VECM by the question each answers.
In this lesson
- Reduced-form VAR
- Granger causality
- Impulse responses and FEVD
- Structural VAR (statement level)
- VECM cross-link
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Econometrics and Financial Time Series syllabus.