Blanchard-Quah (1989) The Dynamic Effects of Aggregate Demand and Supply Disturbances

Summary

Blanchard and Quah interpret fluctuations in GNP and unemployment as generated by two orthogonal structural disturbances: one with a permanent effect on the level of output (labeled a supply disturbance) and one with only a transitory effect (labeled a demand disturbance). Working with a bivariate system in output growth and (detrended) unemployment, they show that this single long-run identifying restriction — demand shocks have zero cumulative long-run effect on output — is enough to recover the structural moving-average (impulse-response) representation from an estimated reduced-form VAR. The recovered demand disturbances produce hump-shaped, mirror-image responses of output and unemployment; supply disturbances move output to a permanent plateau after about five years.

Key Claims

• Two-shock permanent/transitory taxonomy. Output is treated as a unit-root process (following Nelson–Plosser 1982) driven by two shocks; the permanent-effect shock is interpreted as supply, the transitory-effect shock as demand. The interpretation is an economic labeling of a statistical decomposition, not a structural model of price-setting.
• Long-run identifying restriction. With structural shocks normalized to be mutually orthogonal with unit variance, the assumption that demand shocks have no long-run effect on the level of output makes the long-run (cumulative) response matrix lower-triangular. For the 2×2 system this exactly identifies the impact matrix $S$ (up to sign), recovering $S$ from the reduced-form Wold representation $x_t=C(L)u_t$ and the long-run matrix $C(1)$.
• No contemporaneous restrictions needed. Identification uses a restriction on long-run multipliers rather than on the contemporaneous timing of responses — an alternative to Sims/Cholesky recursive identification that leans on long-run neutrality rather than short-run exogeneity.
• Empirical dynamics. Demand disturbances generate a hump-shaped output response that returns to zero, with a mirror-image (opposite-signed) unemployment response; supply disturbances raise output steadily, peaking after about two years and plateauing after ~five years. Demand shocks account for most of output's short-horizon forecast-error variance; supply shocks dominate at long horizons.
• Sensitivity. Results depend on the maintained assumption of exactly two disturbances, on the treatment of the unemployment trend (they remove a slowly varying mean), and on output being difference-stationary; the "supply/demand" labels are only as good as those maintained hypotheses.

Concepts Introduced or Extended

• Structural Identification (Strategy 3 — long-run restrictions)
• Common Stochastic Trends
• Impulse Response Function
• Variance Decomposition

Entities Mentioned

• Olivier Blanchard
• Danny Quah

Quotes

"We interpret fluctuations in GNP and unemployment as due to two types of disturbances: disturbances that have a permanent effect on output and disturbances that do not."

"Demand disturbances have a hump-shaped mirror-image effect on output and unemployment. The effect of supply disturbances on output increases steadily over time, peaking after two years and reaching a plateau after five years."

My Take

This is the paper that made long-run restrictions a standard identification device in structural VARs. Its move — trade a contemporaneous exclusion restriction (which is hard to defend under rational expectations, since anything can react to news within a quarter) for a long-run neutrality restriction grounded in economic theory — is exactly the logic later formalized in structural-identification's Strategy 3 and reused by King–Plosser–Stock–Watson (1991), Galí (1999), and Shapiro–Watson (1988). The two well-known fragilities are also visible here: the identification is only as credible as the two-shocks-and-difference-stationary maintained hypothesis, and the long-run matrix $C(1)$ must be estimated from a truncated VAR, which is numerically delicate near unit roots and invalid under un-modeled cointegration. Still, the demand/supply decomposition it produces remains a canonical benchmark, and the bivariate output–unemployment application is the textbook illustration of the method.