ABSTRACT
Using a very simple econometric framework, we identify two major changes in the dynamics of crude oil price volatility based on data from 1997 to 2017. More precisely, we model weekly West Texas Intermediate (WTI) crude oil price realized volatility in a two-regime setting, one where realized volatility evolves as a plain autoregressive (AR) process (static), and the other where the level, persistence and innovation volatility of the AR process are subject to changes (dynamic). We use a Markov chain to model the probability that the process is in the static regime. The post Great Recession period sees a longer duration of the dynamic regime as well as smaller changes in the level and conditional volatility of realized volatility when switching actually occurs. Crude oil volatility also responds more aggressively to changes in economic variables, such as the t-bill rate and equity market volatility in the dynamic regime.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 As stated above, we choose the sample, 1997–2017, to basically maintain the same sample size before and after the Great Recession. We also experiment with larger samples going back as far as 1986. However, we do not find any qualitative changes with regards to our overall conclusions in Section 3 worth mentioning. These additional estimation results are available upon request.
2 Daily crude oil price data, and weekly tb, vxo and twexm data are extracted from the Federal Reserve Economic Database (FRED): https://fred.stlouisfed.org/. With regards to the implied volatility series, we use the VXO index instead of VIX due to the change in the index calculation methodology in 2004. Weekly crude oil production, import and U.S. ending stocks of crude oil are extracted from the U.S. Energy Information Administration (EIA) website: https://www.eia.gov/.