Submitted:
17 November 2026
Posted:
18 November 2026
You are already at the latest version
Abstract
We decompose the volatility of a three-variable system for Korea (output growth, inflation and the Bank of Korea discount rate) over 232 quarters of data, 1966 Q1 to 2023 Q4. A time-varying parameter VAR(2) is estimated by classical maximum likelihood in two steps, with GARCH-DCC volatility and recursive identification. The likelihood prefers coefficient drift, but the maximum-likelihood drift failed our checks, so we calibrate it at 0.0001. The coefficient paths are then nearly fixed by construction, and the split between drift and volatility follows from this design rather than from estimation. In this model the per-point effects of discount-rate innovations barely change, while output volatility falls from 3.697 to 1.942 percentage points. After 1998 the discount rate was not the policy instrument, the identified shock produces a price puzzle throughout, and other orderings or faster drift reverse the response comparisons. A likelihood-ratio test rejects stability at 1998 Q1 (469.514).
Keywords:
time-varying parameter VAR
; monetary policy shocks
; shock volatility
; Asian Financial Crisis
; Korea
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.