Hamilton 1989

James D. Hamilton’s paper “A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle” (Econometrica 57(2), 1989, 357-384) is the founding reference of the modern regime-switching literature. The paper “models occasional, discrete shifts in the growth rate of a nonstationary series”: the parameters of an autoregression — in particular the mean growth rate — are treated as “the outcome of a discrete-state Markov process”. Applied to quarterly US real GNP growth (1951-1984), the model lets the data alone identify a recurrent switch between a positive-growth expansion regime and a negative-growth contraction regime, which Hamilton argues “could be used as an objective criterion for defining and measuring economic recessions”. With more than 10,000 citations, it is the single most-cited entry in this vault’s literature.

The technical contribution Hamilton credits to himself in his later surveys is the inference machinery, not the idea of a Markov state per se. Markov-switching regressions “were introduced in econometrics by Goldfeld and Quandt (1973)” (Goldfeld and Quandt 1973), themselves building on Quandt’s switching-regression work (Quandt 1958 1972); the likelihood function was “first correctly calculated by Cosslett and Lee (1985)“. Hamilton’s 1989 advance was to handle dependent data — an autoregression whose conditional density depends on current and several past regimes — and to compute the optimal inference about the unobserved state, the parameter likelihood, and forecasts as the by-product of a single recursive filter “similar in spirit to a Kalman filter”. Kim, Piger and Startz (2008) summarise the lineage exactly: “In an influential article, Hamilton (1989) extended Markov-switching models to the case of dependent data, specifically an autoregression.” This filter-based formulation is what made the model practical and is the reason the whole family is now usually called the “Hamilton model”.

Crucially for this vault’s research goal, Hamilton 1989 is a paper of descriptive and explanatory econometrics, not of trading. It tests no strategy, holds out no out-of-sample period, and reports no return, Sharpe ratio, hit rate, or drawdown. Its “benchmark” is the NBER recession chronology, used only to show that the statistically estimated regime probabilities line up with independently dated business-cycle turning points. The model classifies the macroeconomic state; it does not forecast a tradeable asset price, and the paper makes no claim that knowing the regime generates an investment edge. For a vault asking whether Markov-based models are profitable, the founding paper supplies the model and the estimation method but contributes no profitability evidence either way — hence a profitability_evidence_grade of inconclusive.

That distinction is not merely a gap in scope; it foreshadows a documented weakness. The very inference problem Hamilton solved — recovering an unobserved state from noisy observations — means the real-time (filtered) regime estimate is uncertain, and later work showed this caps the model’s forecasting usefulness. Dacco and Satchell 1999 proved analytically that a true regime-switching model can have higher mean-squared forecast error than a random walk once realistic regime-misclassification is allowed. Hamilton’s own later survey Macroeconomic Regimes and Regime Shifts reinforces the caution from the modelling side, warning that richly parameterised switching specifications “could easily result in an overfitted and misspecified model” and that for some specifications “it’s not clear how to form out-of-sample forecasts”. The 1989 paper is therefore best read as the rigorous origin of the model family, with its tradeability left entirely to be established (or not) by the applied finance literature catalogued elsewhere in this vault.

Hamilton 1989 [defines] Markov Regime-Switching Model Goldfeld and Quandt 1973 [precedes] Hamilton 1989 James D. Hamilton [defines] Hamilton 1989 Dacco and Satchell 1999 [contradicts] Hamilton 1989

Connections

• Markov Regime-Switching Model — proposes_model, source: https://www.jstor.org/stable/1912559
• James D. Hamilton — proposes_model, source: https://www.jstor.org/stable/1912559
• Goldfeld and Quandt 1973 — replication_available, source: https://s3.amazonaws.com/real.stlouisfed.org/wp/2003/2003-015.pdf
• Quandt 1958 1972 — relates, source: https://econweb.ucsd.edu/~jhamilto/palgrav1.pdf
• Macroeconomic Regimes and Regime Shifts — relates, source: https://www.nber.org/system/files/working_papers/w21863/w21863.pdf
• Dacco and Satchell 1999 — lacks_live_evidence, source: https://www.nber.org/system/files/working_papers/w21863/w21863.pdf
• State-Count Selection — relates, source: https://www.nber.org/system/files/working_papers/w21863/w21863.pdf

Sources

• Hamilton, J.D. (1989), “A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle”, Econometrica 57(2), 357-384 — https://www.jstor.org/stable/1912559
• Hamilton, J.D. (2016), “Macroeconomic Regimes and Regime Shifts”, NBER WP 21863 — https://www.nber.org/system/files/working_papers/w21863/w21863.pdf
• Hamilton, J.D. (2005), “Regime-Switching Models”, Palgrave Dictionary of Economics — https://econweb.ucsd.edu/~jhamilto/palgrav1.pdf
• Kim, Piger & Startz (2003), “Estimation of Markov Regime-Switching Regression Models with Endogenous Switching”, FRB St. Louis WP 2003-015 — https://s3.amazonaws.com/real.stlouisfed.org/wp/2003/2003-015.pdf