How Hidden Markov Models Improve Volatility Regime Classification
Summary
The document compares a two-state hidden Markov model applied to currency-return volatility with a simple threshold that labels observations as high or low volatility. A threshold can produce a similar-looking state sequence, but it does not account for the persistence or evolving behavior of regimes. A Markov model estimates the path of hidden states jointly, which can distinguish brief fluctuations from actual regime changes and yield a transition matrix useful for forecasting or simulation.
One cited paper reports fewer regime classification errors for a calibrated switching model than for a highest-quartile threshold in its example. That result is specific to the paper’s setup and should not be read as a general performance guarantee. The discussion offers conceptual advantages rather than a direct comparative test on the questioner’s currency series; model calibration and suitability still matter.
Key ideas
- A threshold rule classifies volatility using a fixed cutoff, while an HMM infers hidden regimes from the sequence.
- Markov regime models can represent persistence and changing volatility.
- The inferred state path can help distinguish temporary fluctuations from regime transitions.
- An HMM’s transition matrix provides information useful for regime forecasting and simulation.
- The cited classification advantage is an example, not a universal guarantee.
Tags
Full text
# Is HMM of Volatility any different from a simple filter? # Is HMM of Volatility any different from a simple filter? I have constructed a simple HMM (Hidden Markov Model) with 2 states on the Vol (stdev) of a time series of currency returns. The state vector I produce looks reasonable, in the sense that it appears to have identified periods of high or low vol. However, I appear to be able to generate a very similar state vector just by applying a simple rule like: ``` if (Vol) > x then 1 else 0. ``` Is there any advantage/difference to using a HMM? ## Answer by vonjd (score 10) https://quant.stackexchange.com/a/8728 Have a look at the following paper: Regime Shifts: Implications for Dynamic Strategies by Kritzman, Page and Turkington From the paper (p. 25): > But why go through all the trouble? When dealing with regime shifts, we expect Markov-switching models to perform better than simple data partitions based on thresholds. For example, in Figure 1, if we had simply classified the observations that were in the highest quartile as being associated with Regime 2 (the high-mean regime), we would have misidentified the actual regime 40 times out of 200 observations. In contrast, a well calibrated Markov-switching model would have misidentified the actual regime only three times. Arbitrary thresholds give false signals because they fail to capture the persistence in regimes as well as changing volatilities EDIT There doesn't seem to be a free version available any more. If you find one, please let me know: I will then update the link. ## Answer by Ilya (score 1) https://quant.stackexchange.com/a/8751 HMM allows to get transition matrix that provides additional information itself about probabilities of switching. As HMM looks on complete state path it allows to identify, for example, short periods of low volatility in high volatility regime that were not a result of regime switching. If we apply some simple rule we have a larger number of switches and thus have biased transition matrix. If the results of this analysis are used in some forecasting or simulations, error in transition matrix would be crucial.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.