Great Moderation

Definition

The Great Moderation refers to the marked decline in macroeconomic volatility observed in the U.S. (and many other advanced economies) from approximately the mid-1980s through the mid-2000s. Real gross domestic product (GDP) growth, inflation, and unemployment all became substantially less variable after roughly 1984:Q1. McConnell and Quiros (1999) documented the structural decline in GDP volatility within a linear model. Kim and Nelson (1999b) use a Bayesian Markov-switching (MS) framework to show that stabilization has two distinct components — a decline in shock variance and a narrowing of the boom–recession growth rate gap — and that the latter is the dominant source.

Key Ideas

Break date: Posterior mode at 1984:Q1 in Kim-Nelson (1999b), consistent with McConnell-Quiros (1999). The date coincides with the end of the Volcker disinflation and the beginning of a sustained expansion.
Two sources of stabilization: (1) a decline in the innovation variance $\sigma^2$ ; (2) a narrowing of the regime gap $\mu^*_1 - \mu^*_0$ between boom and recession growth rates. Within a linear model these are observationally equivalent; the MS framework separates them.
Model evidence: Log marginal likelihoods — break in shift parameters: −247.01; variance-only: −253.70; both: −260.19; no break: −267.63. Model II (shift parameters only) wins under all three prior specifications.
Pre/post break estimates (Model II): Recession mean: $-1.025 \to -0.195$ ; boom mean: $0.486 \to 0.137$ . Recessions became shallower; booms became modestly less vigorous.
Absorbing-state changepoint: The break variable $D_t$ is modeled as a two-state Markov chain with $\Pr[D_t=1|D_{t-1}=1]=1$ (absorbing). This allows the unknown break date to integrate out naturally in the Bayesian framework; the posterior distribution of $\tau$ is a byproduct of the Gibbs sampler.

How It Works

Kim-Nelson (1999b) augment Hamilton's (1989) Markov-switching autoregressive (AR) model with a latent absorbing state $D_t$ : $\mu^*_{S_t,t} = \mu_{S_t} + \mu_{S_t,\text{post}}D_t, \qquad \sigma^2_t = (1-D_t)\sigma^2_0 + D_t\sigma^2_1$ $D_t$ transitions from 0 to 1 with probability $1-q_{00}$ and remains at 1 forever thereafter. Four nested models are estimated via seven-block Gibbs sampling extending Albert-Chib (1993); Bayes factors are computed via the Chib (1995, 1998) reduced-run marginal likelihood decomposition. The Hamilton benchmark is nested as the special case $\mu_{S_t,\text{post}}=0$ , $\sigma^2_0=\sigma^2_1$ .

Why It Matters

Monetary policy interpretation: The 1984:Q1 break is consistent with the improved policy credibility established under Volcker. If better policy shortens and shallows recessions (see Plucking Model), this shows up as a narrowing boom–recession gap rather than purely as smaller shocks.
Confounded sources in linear models: Variance tests that find a break in 1984 may be detecting the gap-narrowing rather than shock-size changes. The Markov-switching decomposition is necessary for correct inference.
Forecasting implications: Models estimated on post-1984 data implicitly condition on the moderation; out-of-sample performance during the 2008 crisis showed the moderation was not permanent.
Regime-dependent structural interpretation: The regime asymmetry result — recessions are shallower, not just rarer — complements the plucking interpretation in which monetary and demand-side models are most relevant during recessions.

Gordon (2005): Structural Sources of the Great Moderation

Gordon (2005) takes a structural decomposition approach rather than a statistical changepoint approach, using a four-equation macro model with explicit supply shock variables. Key findings:

Break date: Rolling 20-quarter SD of real GDP growth fell from 2.76% (1952–87Q4) to 1.25% (1988–2005Q1), a 55% drop. Inflation SD fell ~60%. Break coincides with end of 1981–82 recession.
Three-sector demand decomposition: Residential investment, inventory investment, and Federal government spending together account for ~50% of the GDP variance reduction, despite representing only 17%→13% of nominal GDP. Their stabilisation is attributed respectively to financial market deregulation, IT-enabled inventory management, and declining military share of GDP.
Share shifts vs. within-component change: ~80% of reduced output volatility reflects lower volatility within each GDP component; ~20% comes from compositional shifts toward stable consumer services.
Supply shocks dominate inflation: Five named supply shock variables (import price changes, food-energy effects, medical-care effects, Hodrick-Prescott (HP)-filtered productivity trend, Nixon controls) account for ~80% of pre-1984 U.S. inflation variance; Investment-Saving (IS) curve shifts via the output gap explain only ~20%.
IS shifts dominate output gap: The residual in the IS equation explains over two-thirds of output gap variance in both sub-periods. Supply shocks explain ~40% of pre-1984 output gap volatility, approximately zero post-1984 (beneficial and adverse shocks roughly cancel).
Greenspan ≈ Burns after serial-correlation correction: Without AR(1) correction, the Greenspan-era (1990–2004) Taylor Rule inflation coefficient = 1.43 (inflation-fighting). After Feasible Generalized Least Squares (FGLS) correction it drops to 0.57, identical to the Burns-era coefficient. Only Volcker (1979–90, ~1.46) differs meaningfully. Most of the apparent Greenspan inflation-fighting credentials are an artefact of serially correlated residuals.

Open Questions

Was the Great Moderation primarily due to good policy, good luck (smaller structural shocks), or structural change (inventory management, financial innovation)?
The 2008 financial crisis ended the moderation; did the structural break reverse, or was there a second changepoint?
Do other advanced economies exhibit the same 1984:Q1 break date, or country-specific dates reflecting different policy chronologies?
Does the narrowing-gap finding survive richer models with time-varying transition probabilities?
Gordon's supply shock variables (especially import prices) are treated as exogenous, but import prices partly reflect U.S. monetary policy through the exchange rate. Gordon attributes ~⅓ of the 1981–85 supply-shock reversal to Volcker-era tight money, partially re-crediting monetary policy for the disinflation.