Brookings Papers on Economic Activity | Macro Paper Warehouse

Disentangling the Channels of the 2007-09 Recession

This 2012 Brookings Papers on Economic Activity paper by James Stock and Mark Watson asks whether the unusual severity of the 2007-09 recession and the weakness of the recovery that followed required some new economic mechanism – a “financial crisis factor” – or whether the episode can be understood as a larger, historically typical draw of familiar shocks propagating through the economy in its usual way. They address this with a high-dimensional dynamic factor model (DFM) fit to 200 quarterly U.S. macroeconomic series (1959Q1-2011Q2, 132 of them used to estimate the factors), extracting six common factors – a choice consistent with Bai-Ng (2002) information-criteria tests (which themselves gave mixed guidance, ranging from 3-4 to 12 factors depending on the criterion), visual inspection of a scree plot, and the number of distinct structural shocks examined – with factor loadings and dynamics estimated by principal components over the pre-crisis 1959Q1-2007Q3 subsample (the “old” factors) and then extended into the crisis period using a four-lag factor VAR fit over the full 1959Q1-2011Q2 sample. To identify structural shocks (oil, monetary policy, productivity, uncertainty, liquidity/financial risk, and fiscal), the paper develops what it treats as its most-cited contribution: a general “external-instruments” (proxy SVAR) identification strategy in which a shock is recovered as the population regression of an outside instrument – correlated with that shock and, under an exogeneity condition, uncorrelated with the other structural shocks – onto the reduced-form factor innovations, applied one instrument at a time across 18 candidate instruments spanning the six shock categories. The paper’s three headline findings are: (1) a DFM estimated only on pre-2007Q4 data, when fed the actual post-2007Q4 realizations of the “old” factors, tracks the 2007-09 downturn well, and formal tests find little evidence of a break in the factor loadings (rejected at the 5 percent level for only 15 percent of series against full-sample loadings, 12 percent against 1984Q1-2007Q3 loadings) or of a missing factor (a subsampling test gives p=0.59 at an 8-quarter post-2007Q3 horizon and p=0.90 at 15 quarters) – so the crisis reflects unusually large innovations to the same six factors that drove earlier postwar recessions, not a new factor or new dynamics; (2) those innovations were concentrated in financial and uncertainty-related series – the TED spread, VIX, and housing starts saw roughly 8-standard-deviation innovations in 2008Q4, while oil prices moved 1.7 standard deviations in 2007Q1 and 3.4 in 2008Q2 – and a composite uncertainty-liquidity shock (the first principal component of five estimated uncertainty and financial-risk shocks) is attributed roughly two-thirds of the 2007Q4-2009Q2 decline: 6.2 of a 9.2-percentage-point GDP shortfall and 4.5 of a 7.3-point employment shortfall, with oil and monetary-policy shocks contributing moderately and productivity and fiscal shocks contributing little; and (3) the slow recovery is mostly a story of secular trend decline rather than an unusually weak cyclical response – of the roughly 3-point shortfall in post-trough GDP growth relative to pre-1984 recoveries, about four-fifths (2.4 points) reflects slower trend growth rather than a weak cyclical rebound, and for employment the larger part of a roughly 6-point shortfall (3.3 points of trend versus 2.7 points cyclical) is trend-driven, traced to a decades-long decline in trend employment growth (trend GDP growth itself fell by roughly 1.2 percentage points from 1965 to 2005) attributed mainly to the plateauing of female labor-force participation and an aging-driven decline in male participation. Two scope conditions travel with these results and qualify how they should be used: the external-instrument shock estimates are frequently weakly identified (first-stage F-statistics below 5 in 10 of the 18 instrument cases, below 10 in all but 3) and are often substantially correlated with one another both within and across nominally distinct shock categories – most strikingly a -0.93 correlation between two ostensibly separate fiscal shocks whose underlying instruments correlate only -0.06 – which the authors say precludes treating the individual named shocks as a clean, mutually orthogonal decomposition; and because the DFM is linear and does not impose a zero lower bound, its monetary-policy shock estimates counterfactually permit negative interest rates and do not capture unconventional monetary policy, so the finding that monetary policy was “neutral or contractionary” during the crisis and recovery must be read subject to that caveat rather than as a claim about the effect of the actual, unconventional monetary-policy response.

Monetary Policy Alternatives at the Zero Bound: An Empirical Assessment

This 2004 Brookings Papers on Economic Activity paper by Ben Bernanke, Vincent Reinhart, and Brian Sack asks whether “nonstandard” monetary policies can still work once the short-term policy rate is at or near the zero lower bound (ZLB), applying “the methods of modern empirical finance” to the recent US and Japanese experience rather than relying on theory alone. The authors group nonstandard policies into three classes — using communications to shape expected future policy, increasing the size of the central bank’s balance sheet (quantitative easing), and changing its composition (targeted asset purchases) — and use two complementary empirical strategies to assess each. For communications, a one-hour event study around 116 FOMC decisions (July 1991-2004) decomposes market reactions via a Cholesky factorization into a “current policy surprise” factor and a second “path” factor tied to the wording of the post-meeting statement; the path factor turns out to dominate, accounting for roughly four-fifths of the variance of the year-ahead Eurodollar futures rate (versus about one-fifth for the current-surprise factor) and 68 percent of the variance of the five-year Treasury yield, and the “considerable period” language introduced in August 2003 roughly tripled the ten-year yield’s sensitivity to payroll surprises (from about 4 to about 11 basis points per 100,000 jobs). For quantity/composition effects, a five-factor no-arbitrage affine term-structure VAR (estimated on macro and policy factors, monthly 1982-2004) generates a counterfactual yield-curve benchmark, against which the paper attributes the 1999-2002 Treasury debt buyback with pushing the twenty-year yield from about 20 basis points above to about 80 basis points below the model’s prediction (bootstrap probability under 10 percent), concentrated at the long end with only limited evidence of pass-through to swap spreads at short maturities. Extending the analysis to Japan (110 BOJ decisions since independence in April 1998, two-day windows given data limitations), the paper finds little evidence the BOJ used statements to move near-term policy expectations the way the Fed did, but some evidence that quantitative-easing and current-account-balance-target surprises moved the Nikkei (3-6 percent on several occasions) and that a four-factor term-structure model — adjusted by simulation to respect the ZLB — implies actual JGB yields ran persistently below model predictions after the 1999 zero-interest-rate policy and 2001 quantitative easing, with the gap widening over time. The authors describe their overall evidence as “to some degree indirect,” since the US policy rate never actually fell below roughly 100 basis points during the sample, and caution that the size and reliability of quantity/composition effects remain “quite uncertain” and subject to identification problems (for example, the 2003 “deflation scare” episode overlapping with anticipated Treasury purchases); their bottom-line policy recommendation is that central banks are best off avoiding the ZLB altogether by maintaining an inflation buffer and easing preemptively.

The Effects of Quantitative Easing on Interest Rates: Channels and Implications for Policy

This 2011 Brookings Papers on Economic Activity paper by Arvind Krishnamurthy and Annette Vissing-Jorgensen evaluates the effect of the Federal Reserve’s large-scale purchases of long-term Treasuries and other long-term bonds – QE1 in 2008-09 and QE2 in 2010-11 – on interest rates, and argues that quantitative easing works through several distinct channels that affect different assets differently, so it is inappropriate to focus only on Treasury rates as a policy target and effects depend critically on which assets are purchased. The authors decompose the yield on a long-term risky, illiquid asset into the expected safe short rate, expected inflation, and premia for duration risk, illiquidity, lack of safety, default risk, and prepayment risk, and identify seven candidate channels (signaling, duration risk, liquidity, safety, prepayment risk, default risk, and inflation), each tied to instruments that should load on it (federal funds futures for signaling, CDS for default risk, inflation swaps/TIPS/swaption volatility for inflation). Identification comes from an event study around the QE1 announcement dates (Nov 25, Dec 1, Dec 16, 2008; Jan 28, Mar 18, 2009) and the QE2 announcement dates (Aug 10, Sep 21, Nov 3, 2010), combined with difference-in-differences comparisons across pairs of assets that share most characteristics but differ in one dimension (e.g., agency versus Treasury bonds isolating liquidity), and a 2SLS regression of the Baa-Treasury and Baa-Aaa spreads on the log ratio of ten-year-equivalent long-term Treasury supply to GDP using 1949-2008 annual data. For QE1, summed two-day yield changes around the five event dates were sizable and pervasive: Treasury yields fell 73 bp (30-year), 107 bp (10-year), 74 bp (5-year); Fannie Mae agency yields fell 144-200 bp across maturities; agency MBS yields fell 88-107 bp; the signaling channel alone (a roughly 6.3-month delay in anticipated rate hikes, a 40 bp drop in 24-month fed funds futures) is estimated to account for about 20-40 bp of the declines out to ten years, while duration effects only partly explain the pattern (they cannot account for agency bonds falling the most or for the absence of a within-rating maturity effect in corporates), and safety, MBS prepayment, default, and inflation channels are all found to be operative, with MBS purchases described as crucial to the fall in MBS yields and corporate credit risk. For QE2, which purchased only Treasuries, the results look different: signaling lowered 5-year yields by 11-16 bp and 10-year yields by 7-11 bp, safety lowered 10-year low-default-risk yields by a further 6-11 bp, and inflation expectations rose (implied 10-year inflation expectations up 14-16 bp), but the authors find no evidence of duration, liquidity, prepayment, or default channels for QE2, and conclude that its effect on the mortgage and lower-grade corporate rates most relevant to households worked mainly through signaling and inflation rather than a portfolio-balance effect. The safety-channel regression estimates an elasticity of -0.83 (Baa-Treasury) and -0.32 (Baa-Aaa) on log relative Treasury supply, implying safety effects of roughly 4-11 bp for QE1 and 8-20 bp for QE2 under “normal” demand conditions – substantially smaller than the safety effect the event study attributes to QE1, which the authors suggest reflects unusually elevated crisis-period demand for safety rather than an average relationship. The paper’s central mechanism claim is that the portfolio-balance channel operating in both QE1 and QE2 works through a preferred-habitat safety channel – a clientele demand for long-term assets with little or no default risk – rather than through the duration-risk channel emphasized in some contemporaneous work, a conclusion the authors state directly: they “do not find support for the operation of the duration risk channel.”

The Zero Bound on Interest Rates and Optimal Monetary Policy

This 2003 Brookings Papers on Economic Activity article by Gauti Eggertsson and Michael Woodford builds a fully dynamic New Keynesian general-equilibrium model — Calvo (1983) staggered pricing, money in the utility function with a satiation level so the zero lower bound (ZLB) can actually bind, complete financial markets, and a central bank balance sheet that can hold any of several assets with arbitrary state-contingent returns — to ask two questions raised by Japan’s near-zero call rate and the US funds rate’s approach to 1%: does expanding the monetary base (quantitative easing) give a central bank an additional policy instrument once the short rate is stuck at zero, and how should optimal monetary policy be redesigned when the ZLB can bind? It is framed as a dynamic extension of Krugman’s (1998) one-period flexible-price treatment of the same problem. The paper is a theory paper; all quantitative results below are model propositions or numerical illustrations from a calibrated log-linearized model (quarterly, relative-risk-aversion-type intertemporal elasticity sigma=0.5, Phillips-curve slope kappa=0.02, discount factor beta=0.99, long-run real rate 4% a year), not empirical estimates, and depend on that calibration. First, the authors prove an irrelevance proposition: with complete markets, a representative household, and no change in expectations about future monetary or fiscal policy, the equilibrium paths of prices, output, the interest rate, and total government liabilities are independent of the central bank’s base-supply rule, its portfolio-composition rule, or the debt-composition rule — so open-market purchases of long-term bonds or other assets, on their own, have no effect (a result in the spirit of Wallace 1981), and any real-world effect of QE must run through the way such operations change expectations about future policy rather than through mechanical portfolio-balance channels. Second, they show the ZLB is a genuine binding constraint: under a strict zero-inflation target, when the natural rate of interest falls to -2% a year and is expected to stay negative for about ten quarters, the calibrated model produces a 14% output gap and 10% annual deflation, and even a positive constant inflation target only partially mitigates this (a 1% target still leaves roughly a 7% output gap and 4% annual deflation when the trap binds). Third, and centrally, they show that optimal policy — minimizing a quadratic loss in inflation and the output gap subject to the New Keynesian IS and Phillips-curve relations and the ZLB — is history-dependent: it commits the central bank to engineer a future output boom and above-target inflation once the natural rate turns positive again, and to hold the nominal rate at zero for longer than a purely forward-looking (including strict inflation-targeting) policy would — five additional quarters beyond the point the natural rate itself turns positive, in their illustrative 15-quarter trap. Fourth, they show this optimal commitment is implementable as a history-dependent price-level targeting rule expressed in a gap-adjusted price index, requires no estimate of the natural rate, and dramatically dominates any strict inflation target in a calibrated welfare comparison (expected discounted loss relative to a strict zero-inflation target normalized to 100: strict 1% target 24.1, strict 2% target 32, a simple constant gap-adjusted price-level target 0.0725, the fully optimal history-dependent rule 0.036); a simpler constant price-level target captures most of this gain because it automatically commits to undoing deflation with later inflation, while a rule written in inflation terms performs worse than even strict zero-inflation targeting because it mandates deflation during the recovery. Finally, the paper shows optimal policy responds to anticipated future ZLB episodes (driving the nominal rate to zero even before the natural rate turns negative once a future shock is foreseen) but not to a mere increase in the assessed probability of a future binding ZLB, and discusses how a self-fulfilling permanent deflationary trap — which the model does not otherwise rule out — must be excluded by pairing the price-level commitment with a fiscal or base-supply commitment that prevents the nominal value of government liabilities from contracting without bound.

What Does Monetary Policy Do?

This 1996 Brookings Papers on Economic Activity paper by Eric Leeper, Christopher Sims, and Tao Zha asks what monetary policy actually does, and argues that the answer the identified-VAR literature had been giving was fragile because different papers used different samples, variable sets, and identification schemes; the authors instead estimate an escalating sequence of Bayesian structural VARs – first small 4- and 5-6-variable systems that reinterpret existing recursive and non-recursive identifications (Strongin’s NBR/TR system, Christiano-Eichenbaum-Evans’s NBR system), then integrated 13- and 18-variable systems – all on one common monthly U.S. data set (January 1960-March 1996, six lags, quarterly series interpolated via Chow-Lin) so that conflicting conclusions can be checked for robustness on equal footing. Identification proceeds through a mix of exact zero restrictions, probabilistic (“soft-zero”) priors, and informal plausibility screening on impulse responses, organized around a sectoral block structure that splits variables into slow-moving private (“P”), fast-moving auction-market information (“I”), Federal Reserve policy (“F”), and banking (“B”) blocks, with the Fed’s reaction function restricted to respond within the month only to fast-moving financial information, not to CPI or GDP (which are measured with a lag); estimation uses a Bayesian “reference prior” (Sims-Zha) that downweights explosive, poorly-identified dynamics, and published error bands are 68% (about one standard error) probability bands from posterior simulation, not classical confidence intervals. Across the model sequence, a policy-tightening shock in the integrated 13- and 18-variable systems produces a broadly coherent contractionary picture – short and long rates rise, reserves and M1 fall smoothly, output and its components fall, unemployment rises, commodity prices fall, and the dollar appreciates – but the CPI response is very small (briefly slightly positive, only weakly negative after several years), and policy shocks account for only a modest share of output and (in the 13-variable model) a “negligible” share of CPI variance, while a separate private-sector shock – one the model allocates to the private sector with nothing in its estimated form suggesting misallocation, and which the authors judge it “unlikely” is “mistakenly incorporat[ing] much of an expansionary monetary policy disturbance” – is nonetheless the dominant source of M1 and reserves variation, showing that most of the observed variation in these aggregates is genuinely nonpolicy in origin and so is unsatisfactory as a one-dimensional policy indicator. Specifications that imply large real effects of policy (notably Strongin’s NBR/TR system, where policy shocks explain about half of output variance at three-plus-year horizons) tend to come bundled with an implausible sustained price-level rise after a contraction (a severe price puzzle), which the authors read as a sign of policy misspecification rather than a genuine large policy effect; the CEE NBR system, by contrast, produces believable but small real effects. The authors’ robust cross-specification conclusions are that only a modest portion (sometimes essentially none) of U.S. output and price variance since 1960 traces to monetary policy shocks, that most of the observed variation in policy instruments and monetary aggregates is the systematic, endogenous response of policy to the state of the economy rather than exogenous disturbance, that treating reserves or a monetary aggregate as moving mainly in response to policy is therefore unreliable, and that specifications implying large real effects tend to be exactly the ones with implausible price responses. The authors caution that their identified policy shocks may still be absorbing some adverse-supply-shock variation (which would understate price effects and overstate output effects), that the published hard-zero restrictions are a computationally tractable substitute for soft-zero priors the authors initially found more plausible, and that with small changes in specification a shock formally labeled “monetary policy” can rotate into an information-sector shock and vice versa – a fragility warning central to the paper’s message.