The systematic component of monetary policy in SVARs: An agnostic identification procedure

This 2019 Journal of Monetary Economics paper by Jonas Arias, Dario Caldara, and Juan Rubio-Ramírez proposes identifying monetary policy shocks in a structural VAR by placing sign and zero restrictions directly on the coefficients of the monetary policy reaction function itself — the “systematic component” of policy — rather than on the impulse responses the shock is supposed to produce. The approach is partial and set identified: only the monetary policy shock is pinned down, out of a six-variable monthly VAR (real GDP, GDP deflator, a commodity price index, total reserves, nonborrowed reserves, and the federal funds rate) estimated with 12 lags over January 1965-June 2007 using a Bayesian uniform-normal-inverse-Wishart prior. Two restrictions on the contemporaneous federal funds rate equation carry the identification: the funds rate reacts to output and prices but not contemporaneously to total or nonborrowed reserves (ruling reserves out of the systematic rule), and its reactions to output and prices are both restricted to be positive, consistent with Taylor-type rules; critically, neither restriction touches the contemporaneous response of output to the shock, the assumption Uhlig (2005) and Ramey (2016) identify as the reason most VAR evidence finds monetary policy expansionary. Under these restrictions the posterior median response to a contractionary shock is an immediate output decline that remains significant for about 18 months, a protracted fall in the price level, and an on-impact funds-rate increase of roughly 20 basis points, while commodity prices and reserves show little systematic movement; the posterior median contemporaneous coefficients imply the funds rate reacts nearly one-for-one to output (0.84) and more than one-for-one to prices (2.73), though the posterior intervals are wide (95% interval for the output coefficient: 0.04 to 5.25), reflecting the “double-edged sword” of set identification. The results are qualitatively robust to restricting the sample to the 1983-2007 Great Moderation period, where the estimated standard deviation of the policy shock falls from about 0.9 to about 0.3 and the output elasticity to the funds rate rises in magnitude to roughly -2, in line with Gertler and Karadi (2015). Applying the same systematic-component restrictions to evaluate Uhlig’s (2005) admissible set of structural parameters, the paper finds that Uhlig’s IRF-based sign restrictions alone imply, with posterior probability 1.00, a systematic reaction of the funds rate to reserves and, with probability 0.63, a negative reaction to output — both violations of the paper’s restrictions — so that combining Uhlig’s restrictions with the systematic-component restrictions collapses the admissible set toward the paper’s own contractionary and Taylor-rule-consistent conclusions. The authors caveat that the identified set remains wide because only one shock is identified out of many admissible structural VARs, that the pre-2007 sample excludes the zero-lower-bound and unconventional-policy period, and that a robustness exercise bounding the output and price coefficients to (0,4) is used to guard against implausible admissible models in the spirit of the critique in Kilian and Murphy (2012).

Dario Caldara | Macro Paper Warehouse

The systematic component of monetary policy in SVARs: An agnostic identification procedure