Summary
This is the paper that documented the Great Moderation. McConnell and Perez-Quiros establish a structural break in the volatility of U.S. real GDP growth in 1984Q1 — the variance of output fluctuations before 1984 is more than four times that after — and trace it to a specific source: a sharp reduction in the volatility of durable-goods production, with no comparable stabilization in nondurables, services, or structures, and no contemporaneous break in any other G7 country. They link the durables stabilization to a decline in the share of durable-goods output held as inventories, consistent with the spread of better inventory-management practices (e.g. just-in-time). (First circulated as FRBNY Staff Report No. 41, 1998/1999; published in the American Economic Review 90(5): 1464–1476, 2000.)
Key Claims
- A one-time volatility break in 1984Q1. The variance of U.S. GDP growth over 1953–1983 is over four times the variance since 1984 — a sudden, permanent decline, not a gradual trend.
- It is a variance break, not a mean break. A Hamilton-style regime-switching model that lets both the mean and the variance of output growth switch fails to recover a business-cycle (mean-switching) signal, because the dynamics are dominated by the one-time drop in the residual variance. Splitting the residual variance of a linear AR model at the date suggested by the smoothed probabilities, the switch is overwhelmingly a variance phenomenon.
- Endogenous break-date estimation. Because the break date is unknown, standard Wald/LR tests for equal variances across subsamples have nonstandard distributions (a nuisance parameter identified only under the alternative). The authors use the Andrews (1993) / Andrews–Ploberger (1994) sup-type tests to endogenously estimate and test a break in the residual variance of an AR specification for output growth, locating it at 1984Q1.
- The break is concentrated in durable goods. Decomposing output, the volatility reduction emanates from durable-goods production; nondurables, services, and structures show no increased stability. So the aggregate moderation is a sectoral story, not a uniform economy-wide calming.
- No other G7 country broke contemporaneously. The U.S. timing is not shared by the other G7 economies, arguing against a common global shock as the sole cause.
- Inventories as the mechanism. The break in durables volatility is roughly coincident with a break in the share of durable-goods output accounted for by inventories; the declining inventory share aligns with the adoption of inventory-management techniques (just-in-time) in U.S. manufacturing in the early-to-mid 1980s, itself a response to high interest-carrying costs and increased global trade.
- Methodological warning. A volatility break of this size affects any technique that assumes constant second moments — e.g. comparing calibrated-model moments to sample moments, or estimating models over samples that straddle 1984 — so it is not merely a descriptive curiosity.
Concepts Introduced or Extended
Entities Mentioned
Quotes
"We document a structural break in the volatility of U.S. GDP growth in the first quarter of 1984, and provide evidence that this break emanates from a reduction in the volatility of durable goods production."
"The reduction in durables volatility corresponds to a decline in the share of durable goods accounted for by inventories."
My Take
This is the empirical anchor of the entire Great Moderation literature: it fixed the date (1984Q1), the nature (a variance break, not a change in the business cycle's mean dynamics), and a concrete mechanism (durables/inventories) that later work — Kim–Nelson's (1999) Bayesian Markov-switching-variance dating, Stock–Watson's decompositions, the "good policy vs. good luck" debate (Boivin–Giannoni, Gordon) — all had to reckon with. Two methodological points make it durable: it shows that a naïvely specified regime-switching model can be hijacked by a variance break and mistake it for a mean signal, and it uses the Andrews–Ploberger unknown-break-date machinery rather than assuming the break location — the honest way to test for a break you eyeballed in a plot. The inventory story is suggestive rather than dispositive, and the "good luck vs. good policy" question it opened remained live for two decades (and was partly reopened by the 2008 crisis).