SignalTrace

Macroeconomic Regimes and Regime Shifts

John Montgomery4 min read

Macroeconomic Regimes and Regime Shifts

A survey chapter by James D. Hamilton, circulated as NBER Working Paper No. 21863 (January 2016) and published in the Handbook of Macroeconomics Vol. 2 (Elsevier, ch. 0, pp. 163-201). It is the authoritative modern statement of the Markov Regime-Switching Model family. Its abstract frames the subject squarely as macroeconomics: “Many economic time series exhibit dramatic breaks associated with events such as economic recessions, financial panics, and currency crises … This paper surveys the literature for studying regime changes and summarizes available methods.” The survey lays out the model as a time-series process that is linear conditional on a latent first-order Markov state but nonlinear in observables, and develops the filter, the EM algorithm, the Bayesian Gibbs sampler, and the structural-VAR and smooth-transition extensions.

The paper is the vault’s source for two methodological distinctions that matter for any trading use of the model. First, the difference between filtered (real-time) inference — Prob(s_t = i | data through t) — and smoothed (ex-post) inference, Prob(s_t = i | data through t+k), which uses later data to revise the estimate of an earlier regime. Hamilton shows in the recession illustration that even one-quarter-ahead smoothed probabilities track NBER dates well, but he is careful that smoothed estimates use information unavailable in real time. Second, the difficulty of testing the number of regimes: “the likelihood ratio does not have the usual asymptotic chi-squared distribution because under the null hypothesis, some of the parameters of the model become unidentified” — a nuisance-parameter problem — and popular information criteria such as Schwarz’s BIC “rely for their asymptotic justification on the same regularity conditions whose failure causes the likelihood ratio statistic to have a nonstandard distribution”.

For this vault’s overfitting concerns, the survey’s “Conclusions and recommendations for researchers” section is unusually direct. Hamilton warns that “a researcher might be tempted to use the most general specification possible, with all the parameters changing across a large number of regimes … in practice this is usually asking more than the data can deliver”, that “building a richly parameterized description of the transition into and out of recession could easily result in an overfitted and misspecified model”, and that “by contrast, using a simple time-invariant Markov chain is likely to give a reasonable and robust approximation”. He also flags an estimation hazard relevant to any backtest pipeline: “there can be multiple local maxima to the likelihood function … it is therefore good practice to begin the EM iterations from a large number of different starting points”. On forecasting, Hamilton repeatedly notes for richer specifications that “it’s not clear how to form out-of-sample forecasts” — confirming that the model’s forecasting use is not straightforward even before transaction costs enter the picture.

Like Hamilton 1989 and Hamilton’s earlier Palgrave survey, this paper contributes the model and its estimation theory but no profitability evidence: it tests no trading strategy, reports no return, Sharpe ratio, drawdown, or after-cost result, and its illustrative empirical exercise is recession dating, not asset trading. Its value to the vault is precisely as the source of the cautions — parsimony, the non-standard regime-count test, multiple local maxima, and the difficulty of out-of-sample forecasting — that explain why the regime-switching literature is best read as descriptive/explanatory econometrics. It therefore carries a profitability_evidence_grade of inconclusive: the survey is essential for understanding Regime Classification and the Markov Regime-Switching Model, but it neither demonstrates nor refutes a tradeable edge.

Macroeconomic Regimes and Regime Shifts [defines] Markov Regime-Switching Model James D. Hamilton [defines] Macroeconomic Regimes and Regime Shifts Macroeconomic Regimes and Regime Shifts [supports] State-Count Selection Macroeconomic Regimes and Regime Shifts [relates] Hamilton 1989

Connections

Sources

Graph View

Backlinks