Markov Switching Regression Model

Markov switching regression is a time-series model in which the regression parameters change between a small number of unobserved regimes whose transitions follow a Markov chain.

마코프 전환 회귀모형은 관측되지 않는 소수의 국면(regime) 사이를 마코프 연쇄에 따라 전환하면서 회귀모수가 달라지는 시계열 모형이다.

How do I model a time series that switches between different regimes, such as booms and recessions?

호황과 불황처럼 국면이 바뀌는 시계열은 어떻게 모형화하나요?

A Markov switching model lets the intercept, slopes, or variance take different values in each hidden regime and estimates the probability of being in each regime at each time point along with the transition probabilities. It fits economic or financial series that alternate between states such as expansion and recession or calm and turbulent markets.

When to use Markov Switching Regression Model

• Business cycle data alternating between expansion and recession
• Financial series with calm and turbulent periods
• Break dates are unknown and regimes may recur

When to use something else

• There is one known break date: use a Chow test or regression with a break dummy
• Regime shifts are driven by an observed threshold variable: use a threshold autoregressive model

Assumptions and requirements

• Number of regimes specified in advance
• Regime transitions follow a first-order Markov chain
• Each regime is visited often enough to estimate its parameters
• Multiple starting values to avoid local maxima

What the analysis reports

• Regime-specific coefficients and variances
• Transition probability matrix and expected regime durations
• Smoothed and filtered regime probabilities over time
• Log-likelihood, AIC, BIC

How to report it

A two-regime Markov switching model identified a high-volatility regime (σ = [value]) with an expected duration of [value] periods; the probability of remaining in the low-volatility regime was [value].

Related methods

• Threshold Autoregressive Model
• Structural Break Test (Chow Test)
• Hidden Markov Model

Run Markov Switching Regression Model on Quorant