Monetary-Policy-Shocks | Macro Paper Warehouse

Deciphering Federal Reserve Communication via Text Analysis of Alternative FOMC Statements

Thu, 01 Jan 2026 00:00:00 +0000

This paper proposes a text-based measure of monetary policy stance by modelling FOMC post-meeting statements as convex combinations of the staff-drafted dovish (“alternative A”) and hawkish (“alternative C/D”) versions that accompany each meeting, providing a transparent and adaptive reference spectrum. The authors fine-tune the Universal Sentence Encoder—a pre-trained language model—using synthetic examples that mirror numerical information in policy actions, enabling the model to capture both semantic tone and quantitative precision. Stance is defined as the product of tone (alignment with the dovish/hawkish alternatives) and novelty (semantic shift from the previous statement), and is decomposed into expected and surprise components using intraday financial data. Surprises arise from shifts in tone relative to market expectations or from statement novelty. The resulting surprise measure aligns closely (correlations of 70–80%) with established high-frequency measures (Swanson 2017, Nakamura-Steinsson 2018, Bauer-Swanson 2023), and the framework enables counterfactual analysis of how alternative communication could have moved markets.

[What Ends Recessions?]: Comment

Commenting on Christina and David Romer’s claim that active Federal Reserve policy ended postwar recessions, Cochrane argues that their historical narrative cannot distinguish a Fed that reacts to output from a Fed that causes it, that their regression-based policy multipliers are not separately identified from the Fed’s own feedback rule without an explicit, defended monetary theory, and that the size, permanence, and delay of their estimated multipliers do not match any known account of how money affects output. Cochrane places the Romers’ project squarely in the line of Friedman-Schwartz-style historical monetary analysis and revives the identification critiques Tobin and Kareken-Solow raised against that method thirty years earlier: a history of Fed officials perceiving recessions and lowering rates documents that the Fed reacted systematically to output and inflation, but says nothing about whether output reacted to the Fed, since predictable (“systematic”) policy actions are exactly the kind of correlation with dubious causal content that VAR methodology was built to sidestep. Working through the underlying regression algebra, Cochrane shows the Romers’ OLS and instrumental-variables “policy multiplier” estimates are observationally equivalent to the negative of the Fed’s own feedback rule unless one assumes, as they implicitly do, that anticipated and unanticipated monetary policy have identical real effects – an assumption he argues is hard to reconcile with the apparent neutrality of announced disinflations and the ends of hyperinflations. He further argues the estimated multipliers are implausibly large (requiring investment responses far outside what the investment literature supports), permanent, and delayed in a way no theory of monetary transmission predicts, and that a simple look at how little nondurables-and-services consumption moves across four recessions suggests households already expect recessions to be short-lived on their own, undercutting the premise that policy-induced “ending” is even needed. Cochrane closes by faulting the paper for engaging none of the intervening thirty years of VAR and monetary-theory literature that grew directly out of the very identification problems it re-encounters, while praising the Romers’ FOMC-minutes evidence as a potentially valuable input to resolving the Fed’s own reaction function.

A Model of Monetary Policy Shocks for Financial Crises and Normal Conditions

This 2019 Journal of Money, Credit and Banking paper by John Keating, Logan Kelly, A. Lee Smith, and Victor Valcarcel asks whether replacing the federal funds rate with Divisia M4 (DM4) – a user-cost-weighted monetary aggregate – as the policy indicator in a Christiano-Eichenbaum-Evans (1999)-style block-recursive structural VAR can identify monetary policy shocks that behave sensibly in both normal times and financial-crisis/near-zero-lower-bound conditions, when standard funds-rate-based shocks are prone to implausible (“puzzle”) responses. The empirical model partitions macro variables into three ordered blocks – a slow-moving block (real GDP, the GDP deflator, a commodity price index), the policy indicator (DM4), and a fast-responding money-market block (the monetary base and DM4’s user cost) – estimated by Bayesian Gibbs sampling (following Sims and Zha 1999) with a noninformative prior and 90% probability intervals, at 5 lags quarterly (13 lags in a monthly robustness check), across three overlapping samples: 1967:Q1-1995:Q2 (replicating the original CEE funds-rate sample), 1967:Q1-2007:Q4 (precrisis), and 1967:Q1-2015:Q4 (full sample including the 2008-09 crisis and the ZLB period). A complementary small-scale New Keynesian model with money shows that when the Fed’s estimated interest-rate rule displays high inertia (rho approximately 0.94, estimated by nonlinear GMM on 1967-2007 quarterly data), its policy dynamics can be closely replicated by a non-inertial money-growth rule reacting to output and inflation – providing theoretical grounding for using DM4 in a Cholesky-ordered VAR even though the Fed never formally targeted it – whereas a low-inertia Taylor (1993) rule is not well replicated by a money-growth rule. Empirically, the DM4-based VAR produces no output, price, or liquidity puzzles in any of the three samples: GDP falls in a significant, hump-shaped (U-shaped) pattern for three years after a contractionary shock, the price level declines with a lag (becoming significantly negative after about two years), DM4 itself falls on impact with a significant liquidity effect, and its user cost rises on impact – a pattern the authors describe as “remarkably stable” across samples and broadly similar to the CEE funds-rate benchmark in the precrisis sample. By contrast, VARs built on three shadow short rates (Lombardi-Zhu, Krippner, Wu-Xia) all generate persistent, statistically significant price puzzles, with two of the three also generating liquidity puzzles. The cumulative identified DM4 shocks crest around QE1-QE3 and register the 2013:Q2 taper tantrum as a large negative shock, and a historical decomposition finds monetary policy shocks after 2008 were mildly net expansionary (real GDP 0.48% and the GDP deflator 0.50% higher than they otherwise would have been), with monetary policy shocks accounting for 8.44% (90% interval 1.73%-18.78%) of real GDP forecast-error variance at 8 quarters and 6.47% (0.66%-17.04%) of GDP-deflator variance at 20 quarters. A Great Depression counterfactual – applying the 1929:Q3-1937:Q3 pace of M2 collapse to DM4 starting in 2007:Q4 – implies a hypothetical 12.65% peak-to-trough GDP decline and 4.1% deflation, versus the actual -4.35% GDP decline and +0.26% inflation, which the authors interpret as evidence that post-2008 Fed actions, while not large departures from its historical policy rule, prevented a much deeper contraction. Long-run policy-rule coefficients on inflation are negative and significant in all three samples (-1.83, -2.26, and -0.88 respectively across the three samples) while the output-growth coefficient stays small and insignificant throughout, which the authors read as the Fed behaving “as if” it stabilized inflation via broad money growth. Results are robust to a monthly-frequency VAR (only 1 of 102 responses falls outside the quarterly model’s 90% bands, though the monthly liquidity effect is not statistically significant) and largely, though not entirely, to substituting DM2 for DM4 (DM2 shows a small, statistically insignificant price puzzle and a rising monetary base in the full crisis-inclusive sample). The authors attribute DM4’s puzzle-free performance in the crisis period to two features: its Divisia expenditure-weighting, which (per a DSGE decomposition) avoids the aggregation bias that afflicts simple-sum aggregates like M1 and M2 when asset substitutability is time-varying, and its breadth, since DM4 includes institutional money-market funds, repos, commercial paper, and Treasury bills – assets specifically targeted by the Fed’s 2008-09 liquidity facilities (PDCF, TSLF, CPFF, ABCPMMMFLF).

A New Measure of Monetary Shocks: Derivation and Implications

This 2004 American Economic Review paper by Christina D. Romer and David H. Romer constructs a new narrative-based measure of monetary policy shocks designed to be free of both the endogenous response of policy to current economic conditions and the Federal Reserve’s anticipatory response to expected future developments – two contaminations the authors argue bias the federal funds rate as a policy indicator, pushing estimated output effects toward zero and generating a spurious “price puzzle.” Construction proceeds in two steps: first, reading the FOMC’s Record of Policy Actions and internal expected-rate memos for each scheduled meeting from March 1969 through December 1996 to identify the “intended” federal funds rate change at that meeting, excluding intermeeting moves and moves based on incorrect preliminary data; second, regressing these intended changes on the pre-meeting funds rate level and the Fed’s own Greenbook forecasts of output growth, inflation, and unemployment (levels and forecast revisions, one quarter before through two quarters after the meeting), with the residuals – estimated over 263 meetings, R-squared 0.28 – constituting the shock series. Aggregated to monthly frequency and entered with up to 36-48 lags in single-equation regressions for log industrial production and log PPI over January 1970-December 1996 (N=324), a one-percentage-point contractionary shock is estimated to reduce industrial production by a maximum of 4.3 percent after about 22 months (highly significant, t-statistics above 2.5 from month 17 through 27), an effect that becomes imprecisely estimated by month 48 (-1.7 percent, two-standard-error band -6.4 to +3.0); prices are essentially unaffected for roughly the first two years and then fall substantially, reaching -5.9 percent (t=5.5) at month 48, with no statistically significant price puzzle at any horizon – a sharp contrast with regressions using the actual funds rate, which produce a sustained roughly one-percent rise in prices over the same period. A three-variable recursive VAR (industrial production, PPI, cumulated shock) broadly confirms these patterns, with a smaller peak output effect (-2.9 percent) that the authors attribute to the VAR’s shock being transitory rather than the permanent shock assumed in the single-equation exercise. Controlling for world commodity prices changes results only modestly, which the authors read as evidence the measure is not substantially contaminated by responses to supply shocks; correcting for either endogeneity or anticipatory movements alone, by contrast, is shown to produce much smaller effects than correcting for both together. Central caveats: the sample is constrained to end in 1996 by the five-year Greenbook data embargo (excluding the later Greenspan tenure and the entire post-2008 zero-lower-bound period, to which the intended-funds-rate concept does not apply), a positive coefficient on the first lag of the shock in the output regression traces to a single April 1980 outlier that the authors attribute to sampling error, and the authors themselves flag the lag before prices begin to fall as the paper’s “most important uncertainty,” ranging from about six to twenty-two months across specifications.

An alternative explanation of the price puzzle

This 2004 Journal of Monetary Economics paper by Paolo Giordani offers an analytical and empirical alternative to the Sims (1992) forward-looking-information explanation of the “price puzzle” — the finding that prices rise, rather than fall, after a contractionary monetary policy shock in standard VARs — by showing that the puzzle can be generated purely by omitting the output gap from the VAR, even when the resulting misspecified system still produces optimal forecasts of inflation at every horizon. Using a backward-looking Svensson (1997) IS/Phillips-curve/Taylor-rule model as the data-generating process, Giordani shows analytically that a VAR excluding the output gap (retaining only output, inflation, and the policy rate) does not admit a finite-order VAR representation — it is instead a VARMA(2,1) — that the variance attributed to the recovered “monetary policy shock” is strictly positive even when the true policy rule is deterministic, and that the interest rate loads with a positive coefficient in the misspecified inflation equation because movements in the policy rate help retrieve the omitted output gap on which the Taylor rule responds, generating a spurious positive inflation response to what is identified as a contractionary shock. Empirically, using quarterly U.S. data from 1966Q1 to 2001Q3 with a recursively (Cholesky) identified three-variable VAR(3), funds rate ordered last, a system using Federal Reserve capacity utilization as the output-gap proxy (“VARgap”: capacity utilization, CPI inflation, federal funds rate) produces an inflation response to a contractionary shock that is never significantly positive and is significantly negative for several quarters, whereas the otherwise identical system using log real GDP in place of the gap reproduces “a huge price puzzle.” Variance decompositions show monetary policy shocks account for only 2.1% of the 8-quarter-ahead inflation forecast-error variance in VARgap, versus 3.5% at the 2-quarter horizon in the misspecified VAR (where cost-push/supply shocks dominate at 92.1%), and the estimated standard deviation of the monetary policy shock is about 12% lower in VARgap, consistent with the theoretical prediction that omitting the gap inflates the estimated policy-shock variance. The paper further argues that the standard Sims (1992) fix — adding a commodity price index — reduces the puzzle not because it improves inflation forecasts (an F-test shows the commodity index PcomCEE is redundant once capacity utilization is included, p = 0.41, while capacity utilization is not redundant given the commodity index, p ≈ 0.0000) but because commodity prices are themselves correlated with the output gap (r = 0.58 with capacity utilization, 1970Q1–1998Q4) and so act as a cyclical proxy for it. The results are reported as robust to lag lengths 1–8, using CPI in log levels or the GDP deflator, reordering capacity utilization and inflation, adding M2, and using several alternative output-gap measures (linear/quadratic detrending, the CBO gap, unemployment, HP-filtered output), though the fix is less clean in a pre-1979 subsample and in monthly data, and the paper flags remaining measurement-error and gap-construction caveats, including a check that adding leads of GDP to the misspecified VAR does not eliminate the puzzle, addressing the concern that capacity utilization’s construction embeds future information.

Are the Effects of Monetary Policy Shocks Big or Small?

This 2012 American Economic Journal: Macroeconomics paper by Olivier Coibion asks why standard recursive VARs and the Romer-Romer (2004) narrative approach produce such different estimates of the real effects of monetary policy shocks, and argues the discrepancy is driven by three identifiable and largely correctable factors rather than genuine disagreement about the underlying policy shock. Using monthly US data (1970:1-1996:12, following Romer and Romer’s sample and variable choices), Coibion compares a standard recursive (Cholesky) VAR with the funds rate ordered last and 12 lags, the Romer-Romer single-equation approach (a macro variable regressed on 24 autoregressive lags and 36 lags of the R&R narrative shock series), and a “hybrid VAR” that replaces the funds rate in the VAR with the cumulative R&R shock series. At baseline lag specifications the three diverge sharply (Table 2, Panel A): a 100-basis-point shock produces a peak industrial production decline of about 0.7% and a 0.16 percentage-point rise in unemployment in the standard VAR, versus a 4.3% IP decline and 0.93 percentage-point unemployment rise in the R&R single-equation approach, with the hybrid VAR intermediate (1.6% IP decline, 0.40 points). Three mechanisms account for most of this gap: (1) an R&R shock corresponds to a larger, more persistent counterfactual path for the funds rate than a VAR Cholesky shock, and rescaling impulse responses to a common FFR path largely eliminates the difference; (2) R&R shocks during the 1979:10-1982:9 non-borrowed-reserves (NBR) targeting episode are statistically predictable from lagged macro variables (violating the shocks’ assumed exogeneity) and disproportionately drive the large R&R estimates; and (3) the R&R two-step procedure’s results rise nearly monotonically with the number of shock lags included, while AIC- or model-averaging-based lag selection picks shorter lags and shrinks the estimated effects (reducing the peak IP effect from -4.3% to -3.4%), whereas the VAR is comparatively insensitive to lag length. Once the NBR-targeting period is excluded and lag length is chosen by AIC or model averaging, the three approaches converge (Table 2, Panel B) on a “medium-range” peak effect of roughly a 1.8-2.8% industrial production decline and a 0.19-0.70 percentage-point unemployment increase; three additional Taylor-rule-residual shock measures constructed in Section III (GARCH, time-varying-coefficient, and Smets-Wouters DSGE-filtered shocks) corroborate this medium range and are markedly less sensitive to lag length or episode selection than the original R&R series, though the paper is explicit that its evidence covers only the pre-1997 sample and does not speak to the post-1996 or zero-lower-bound period.

Comments on "A New Measure of Monetary Shocks: Derivation and Implications"

These are John Cochrane’s discussant remarks, delivered July 17, 2004 at an NBER Economic Fluctuations and Growth meeting, on Christina Romer and David Romer’s “A New Measure of Monetary Shocks: Derivation and Implications” (published in the American Economic Review later that year). Cochrane opens by declaring he wants to argue the paper is “much better and deeper than the authors think it is,” identifying its two key innovations as extending the Fed’s intended (target) funds rate series back in time using historical “intentions” data, and, more importantly, establishing via narrative evidence that Federal Reserve staff Greenbook forecasts function as a “sufficient statistic” for the Fed’s information set at the time of each policy decision. Cochrane shows this sufficient-statistic property lets both halves of Romer-Romer’s procedure be radically simplified: identifying a valid policy shock requires only that it be orthogonal to the Fed’s own output forecast (not to price, exchange-rate, or other forecasts, since only reverse causality running through output needs to be purged), and once such a shock is in hand, the output or price response can be read directly off a sequence of single regressions rather than a full dynamic simulation with Monte Carlo standard errors. Re-estimating with his simplified shocks, Cochrane obtains essentially the same headline result as Romer-Romer – a 100-basis-point tightening produces a 6-8% decline in output peaking around 30 months out, with statistical significance only modestly above conventional thresholds (t-statistics “just above 2”) – alongside a “price puzzle” that he argues is not legitimately fixable by adding commodity prices to the shock regression, since the historical FOMC record shows no narrative evidence that the Fed actually responded to commodity-price movements in the way that fix presumes. He flags the response’s long delay and the near-total absence of any price response as jointly troubling for monetary theory, since most theories tie the two together. In a preliminary extension, Cochrane recomputes the output effect of a one-time, one-month target increase not followed by the Fed’s usual further tightening, and finds it roughly a tenth the size and considerably shorter-lived than the standard “full path” response – suggesting that whether anticipated policy actions matter is, in his view, at least as consequential a modeling choice as any other in this literature. He closes by suggesting the field’s next step should be estimating the Fed’s systematic policy rule directly (via narrative methods, since Fed statements arguably already summarize the rule) rather than continuing to refine shock-and-response estimates, and offers a conjecture that ordinary regression evidence cannot in principle distinguish a determinate (Taylor-rule-consistent) policy regime from an indeterminate one, because such regressions can only recover the stable root of the model’s dynamics, not the off-equilibrium explosive threats that actually enforce determinacy.

Firm-specific capital, nominal rigidities and the business cycle

Macroeconomic data show inertial inflation, and the standard way of accounting for it inside New Keynesian models is to assume firms re-optimise prices only once every six quarters or, without indexation to lagged inflation, once every two years or more – an assumption that clashes directly with micro evidence that firms change prices more often than once every two quarters. This paper formulates and estimates a three-shock US business cycle model that reproduces inflation inertia while firms re-optimise prices on average once every 1.8 quarters, and traces the difference to a single modelling assumption: capital is firm-specific rather than homogeneous and traded in economy-wide rental markets. With a predetermined firm-level capital stock, a firm’s short-run marginal cost curve slopes up in its own output, so a contemplated price rise – which cuts the firm’s demand and output – also cuts its marginal cost, working against the price rise. The authors work with two versions of the Christiano-Eichenbaum-Evans (2005) model that differ only in this respect, and show that the log-linearised equilibrium equations differ only in the mapping from structural parameters to the reduced-form coefficient linking the change in inflation to average real marginal cost. Parameterised in terms of that coefficient the two models are observationally equivalent for aggregate data, which means macro evidence cannot adjudicate between them and the case must be made on micro implications. Estimation follows the CEE limited-information strategy, matching model impulse responses to those from a ten-variable identified VAR on quarterly US data for 1982:1-2008:3, with long-run restrictions identifying neutral and capital-embodied technology shocks and a recursive-timing restriction identifying the monetary policy shock; the three shocks together account for roughly 60 percent of the cyclical variance of aggregate output, with capital-embodied technology the largest single contributor and, notably, about 30 percent of the cyclical variation in the real wage. The point estimate of the inflation-marginal cost coefficient is 0.014, implying that a temporary one percent change in marginal cost moves the aggregate price level by only about 0.02 percent; under homogeneous capital this implies price re-optimisation once every 9.36 quarters, while under firm-specific capital it implies once every 1.8 quarters. Wage contracts are re-optimised on average once every 4.5 quarters, the habit parameter is 0.76, and the estimated cost of varying capital utilisation is higher than in CEE. The decisive micro comparison concerns the cross-firm distribution of production after a monetary policy shock: under homogeneous capital roughly 70 percent of firms produce essentially all of the economy’s output four periods after the shock while the rest effectively shut down, an implication the firm-specific capital model does not share. The authors conclude they “strongly prefer the firm-specific capital model,” while leaving open that other propagation mechanisms – firm-specific labour, sectoral heterogeneity in price-change frequency, intermediate inputs, rational inattention and sticky information – “may be at least as important.”

Forward-looking information in VAR models and the price puzzle

This 2006 Journal of Monetary Economics paper by Sophocles Brissimis and Nicholas Magginas asks whether augmenting a standard monetary VAR with forward-looking information can resolve the “price puzzle” — the counter-theoretical finding that prices rise, rather than fall, after a contractionary monetary policy shock — in a modern US sample dominated by the Greenspan Federal Reserve. Using monthly US data from January 1989 to June 2004 (the period for which federal funds futures data are available, and one the authors characterize as a homogeneous policy regime), they first show that a standard four-variable recursive (Cholesky) VAR — industrial production, CPI, a commodity price index, and the federal funds rate, with 6 lags chosen by AIC and SBC — produces “a strong counter-theoretical response of CPI to the monetary policy shock,” with CPI rising throughout all 48 months of the impulse response. They then replace this baseline with a system ordered {composite leading indicator (LCOM), industrial production, CPI, federal funds rate}, in which the one-month-ahead federal funds futures rate (FFF) — the market’s expectation of the current month’s funds rate — enters as an exogenous variable rather than as part of the endogenous ordering, so that the monetary policy shock is identified as the orthogonalized residual of the funds-rate equation. In this augmented system “the price puzzle is solved”: CPI shows only a small, statistically insignificant positive response for the first 2-3 months, then declines gradually to a trough 30 months after the shock, with the decline becoming statistically significant after 14 months; industrial production declines steadily to a trough 5-6 months out and returns to its pre-shock level roughly two years (24 months) after the shock, tracking LCOM’s response with a 1-2 month lag; and the funds-rate shock itself dies out quickly (4-5 months), so there is “no policy innovation paradox.” The result is robust to replacing FFF with an implied one-month forward rate constructed from two-month Libor and the effective funds rate, both over the same 1989-2004 window and over a longer 1986-2004 sample (the earliest for which Libor data exist), and the authors note that further extending the system to seven variables in a Christiano-Eichenbaum-Evans/Kim-style specification with total and non-borrowed reserves and M2 yields “no significant gains” in resolving the puzzle. The paper’s overall diagnosis is that the price puzzle reflects the VAR’s omission of the richer, forward-looking information set the central bank actually uses to forecast inflation and output; the federal funds futures rate and the composite leading indicator proxy for that omitted information without expanding the VAR’s dimensionality.

Identifying Monetary Policy in a Small Open Economy under Flexible Exchange Rates

This 1997 Journal of Monetary Economics paper by David Cushman and Tao Zha addresses the “exchange rate puzzle” and “price puzzle” that plague conventional recursive (Choleski) VAR studies of monetary policy in small open economies – where a contractionary domestic interest-rate shock perversely depreciates the currency and raises prices – by arguing these puzzles “derive from an identification of monetary policy that is inappropriate for such economies,” specifically the recursive ordering’s implicit assumption that the central bank cannot respond contemporaneously to the exchange rate, which is inconsistent with how a small-open-economy central bank like the Bank of Canada actually behaves. Using monthly Canada-US data from 1974-1993 (chosen to avoid the 1973 oil shock and the pre-float period) with 12 lags, the authors estimate a non-recursive structural VAR with block exogeneity, treating the US block (output, prices, the federal funds rate, and a world commodity price index) as exogenous to Canada while specifying the Canadian block’s contemporaneous structure across three sectors: a money market (a money-demand equation plus a policy reaction function that lets policy respond within the month to the exchange rate, the Canadian money stock, the foreign interest rate, and commodity prices), an information market (an exchange-rate equation responding to all eleven contemporaneous variables), and a recursively-ordered production sector (imports, exports, output, prices) excluded from contemporaneous foreign and financial influences. Estimated by maximum likelihood with Monte Carlo error bands (Zha 1996 method) and passing a model-stability test (chi-square(364) = 346.1), the identification eliminates both puzzles: a contractionary monetary policy shock produces an immediate, persistent decline in M1, an immediate and significant appreciation of the Canadian dollar lasting about 12 months, only a brief and small rise in the nominal interest rate, and a gradual, non-puzzling decline in the price level. Transmission operates primarily through the exchange rate rather than the interest rate – in contrast to the strong interest-rate effects found in US studies such as Gordon and Leeper (1994) – with a J-curve pattern in the trade balance (imports down roughly 0.5% by month 4) and approximate uncovered interest parity (UIP deviations significant for only about 4 months). Forecast-error variance decompositions show monetary policy shocks explain a small share of Canadian output variance – about 1.00% at 12 months, no more than 2.75% at any horizon up to 6 months – with the production and information sectors dominating in the first few months and foreign shocks becoming dominant after 12 months (roughly 72-74% of output variance at 12, 24, and 48 months), a result the authors caution does not imply that endogenous monetary policy itself is ineffective.

Macroeconomic shocks and their propagation

This 2016 Handbook of Macroeconomics chapter by Valerie Ramey is a survey and methodological stock-taking exercise on how macroeconomists identify structural shocks to monetary policy, fiscal policy (government spending and taxes), and technology, and how those shocks propagate through the economy — organized around the question Cochrane (1994) posed of whether researchers are “destined to remain forever ignorant of the fundamental causes of economic fluctuations,” to which Ramey’s answer is “no.” Rather than introducing a new identification scheme, the chapter sets out a common reduced-form/structural framework (reduced-form residuals e_t = Bepsilon_t, impulse responses C(L) = A(L)^-1B) and classifies the literature into roughly nine identification strategies — recursive/Cholesky orderings (with the policy variable ordered either last, as in Christiano-Eichenbaum-Evans, or first), non-recursive structural restrictions, narrative methods, high-frequency identification, external-instrument/proxy SVARs, long-run restrictions, sign restrictions, and factor-augmented VARs — and then re-estimates or directly compares many of them on overlapping or updated data. For monetary policy, a comparison of eleven leading published methods on their original samples finds trough effects of a 100-basis-point funds-rate increase on output/industrial production ranging from a positive, statistically insignificant response (Uhlig’s 2005 sign-restriction estimate) to declines of several percent (Romer-Romer’s 2004 narrative measure implies -4.3% at 24 months; Barakchian-Crowe’s 2013 high-frequency estimate implies -5% at 23 months), with no consensus on the “true” magnitude. Ramey’s own re-estimation of the Christiano-Eichenbaum-Evans-style recursive VAR shows the effect shrinking from a -0.48% trough (accounting for 6.6% of the output forecast-error variance at 24 months) over 1965-1996 to a -0.20% trough (0.5% of variance) over 1959-2007, and turning expansionary when the sample is restricted to 1983-2007. She then subjects the Gertler-Karadi (2015) high-frequency proxy instrument to validity checks and finds it has a nonzero mean (-0.013), is serially correlated (AR(1) coefficient of 0.31, robust SE 0.11), and is predictable from Federal Reserve Greenbook forecasts (R-squared of 0.21, p=0.027) — evidence that the instrument is contaminated by the Fed’s private information rather than being a clean exogenous surprise — leading her to conclude that what is now identified as a monetary policy shock is “really mostly the effects of superior information on the part of the Fed, foresight by agents, and noise.” For fiscal shocks, government-spending multiplier estimates cluster mostly between 0.6 and 1.5 once computed correctly as the integral of the output response over the integral of the spending response (rather than a peak-to-impact ratio), while tax-multiplier estimates vary more across method, from roughly -0.5 to -0.8 (SVAR-based estimates using outside output elasticities) up to about -3 (Romer-Romer 2010 narrative and Mertens-Ravn 2014 proxy-SVAR estimates). For technology shocks, the chapter documents low correlations between shocks extracted from different identification approaches and finds that investment-specific technology (IST) and marginal-efficiency-of-investment (MEI) shocks — including news about future IST change (Ben Zeev-Khan 2015, estimated to explain 73% of output variance at 8 quarters in one specification) — tend to matter more for business-cycle fluctuations than neutral total-factor-productivity shocks, whose estimated importance ranges from negligible (Galí 1999) to more than 40% of output variance depending on method. A combined VAR that Ramey estimates herself, incorporating leading fiscal, technology, and monetary shock measures together over 1954-2005, finds that the Ben Zeev-Khan IST-news shock is the single most important driver of both output and hours (accounting for up to 40% of the hours forecast-error variance at 8 quarters), that the Justiniano-Primiceri-Tambalotti MEI shock explains 42% of output variance on impact falling to 24% within a year, that the federal funds rate accounts for up to 8% of output and 18% of hours variance, and that all the included shocks together explain 63-79% of the variance of output and hours at horizons of 4-20 quarters — evidence Ramey reads as showing genuine, if incomplete, progress against the identification problem Cochrane posed two decades earlier, even as she stresses that none of the individual identification strategies commands full consensus.

Measuring Monetary Policy

Bernanke and Mihov (1998, QJE) address a measurement problem in the monetary-VAR literature: the era’s dominant methods for extracting a “monetary policy shock” – Cholesky orderings built around the federal funds rate (FFR), nonborrowed reserves (NBR), or the ratio of nonborrowed to total reserves (NBR/TR, following Strongin 1995) – impose ad hoc restrictions rather than an explicit model of how the Fed actually operates in the market for bank reserves. The paper builds a “semi-structural” six-variable VAR (total reserves, nonborrowed reserves, and the federal funds rate in a policy block; real GDP, the GDP deflator, and a commodity price index in a block-exogenous macro block) over monthly U.S. data, 1965:1-1996:12, with lag length chosen by a general-to-specific procedure (13 lags retained for the full sample). Within the policy block they specify structural equations for reserves demand, borrowed-reserves supply, and the Fed’s nonborrowed-reserves supply/policy rule, estimated by two-step efficient GMM using the reduced-form VAR residuals; this yields five nested identifications – FFR, NBR, NBR/TR, borrowed-reserves (BR), and a just-identified (JI) benchmark that restricts only the reserves-demand elasticity to zero – each corresponding to different assumptions about how much the Fed accommodates reserves-demand shocks (phi^d) and borrowing shocks (phi^b). Overidentifying-restriction tests reject the NBR model in every subsample except the 1979:10-1982:10 Volcker NBR-targeting episode, favor NBR/TR as the best overall fit across subsamples, and favor FFR pre-1979 and post-1988; a Hamilton (1989) regime-switching model applied directly to (phi^d, phi^b) independently locates two structural breaks matching the conventional 1979/1982 Volcker-experiment dates, with phi^d falling from roughly 0.86 outside that window to roughly 0.18 within it. Although the five identifications generate qualitatively similar impulse responses (output rising with a 12-18 month peak, prices moving more slowly and persistently), they diverge sharply in magnitude – the cumulative output response under NBR/TR is more than twice as large as under NBR, and at a four-year horizon the price response under NBR is roughly four times that under FFR – because, the authors show algebraically, any overidentified indicator is a contaminated mixture of the true policy shock with reserves-demand and borrowing shocks whenever the Fed accommodates those shocks. This contamination framework offers an explanation for the post-1984 “vanishing liquidity effect” as a measurement artifact of using the NBR indicator rather than a change in the economy: a derived bias formula, evaluated at the JI model’s full-sample GMM estimates, is negative in the full sample (bias term -0.777) and grows more negative in 1988-1996 (-0.904) and 1984-1996 (-1.035, where it flips the estimated liquidity effect’s sign). The paper closes by constructing an overall monetary-policy-stance index from the full vector of structural innovations, normalized by a 36-month moving average, which correlates 0.71 (monthly) with the Boschen-Mills (1991) index; the authors caution that the empirical illustrations are meant only to be illustrative, since the Figure II impulse responses pool full-sample JI parameter estimates across regimes that the paper’s own regime-switching results show are unstable.

Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) Approach

This 2005 Quarterly Journal of Economics paper by Ben Bernanke, Jean Boivin, and Piotr Eliasz proposes the Factor-Augmented VAR (FAVAR) to address a specific problem with standard small monetary VARs: because Fed policymakers actually condition on a wide range of economic and financial data, a VAR built from only a handful of observable series omits conditioning information, which biases the identified policy shock (producing the “price puzzle,” a spurious rise in prices after a tightening), makes the choice of which few variables to include arbitrary, and limits impulse responses to only the included series. The FAVAR augments a standard VAR in observable variables (in the benchmark specification, just the federal funds rate) with a small number of latent factors extracted by principal components from a large panel of 120 monthly U.S. macroeconomic series (DRI/McGraw Hill Basic Economics Database, January 1959-August 2001); the factors and the observable are then modeled jointly in a VAR (13 lags, 3 factors in the benchmark), with the policy shock identified recursively (Cholesky) by ordering the federal funds rate last, and with the factors themselves estimated in a first step using only “slow-moving” variables that do not respond contemporaneously to policy shocks, so as not to contaminate the factor estimates with the policy shock itself. Two estimation approaches are compared: a computationally simple two-step principal-components procedure (following Stock and Watson 2002), which the authors prefer for its “greater plausibility,” and a one-step Bayesian likelihood approach via Gibbs sampling (10,000 draws, 2,000 discarded as burn-in) that is more rigorous but yields qualitatively similar, quantitatively more imprecise results. Moving from a standard three-variable VAR (industrial production, CPI, funds rate), which produces a strong and persistent price puzzle, to a FAVAR with even a single additional factor considerably reduces the puzzle; the preferred three-factor, two-step-PC FAVAR generates impulse responses for many more series that are broadly consistent with standard theory (real activity and prices eventually decline, money aggregates fall, the dollar appreciates on impact), and a five-factor version leaves these qualitative conclusions unchanged. At a 60-month horizon the identified policy shock accounts for a sizeable share of the variance of interest rates (45.4% of the funds rate itself, 43.3% of the 3-month T-bill, 40.3% of the 5-year bond) but a much smaller share of real activity and prices (5.4% of industrial production, 3.8% of CPI) and of monetary aggregates (0.5% of both the monetary base and M2); the authors note that IRFs for the monetary aggregates should be interpreted with caution given the low R² of their common components (10.4% and 5.2%, respectively), and they hedge their central claim, describing the price-puzzle result as offering only “some support” for the view that the puzzle stems from omitted conditioning information rather than a definitive resolution. The sample ends in August 2001, before the financial crisis and the zero lower bound, so the paper does not speak to whether the FAVAR’s resolution of the price puzzle extends to that later, constrained-policy environment.

Micro Jumps, Macro Humps: Monetary Policy and Business Cycles in an Estimated HANK Model

This paper builds and estimates a Heterogeneous-Agent New Keynesian (HANK) model that simultaneously matches two sets of moments the prior literature could not reconcile: “micro jumps” – the large, immediate marginal propensity to consume (MPC) out of a transitory income shock documented in household data (Fagereng, Holm and Natvik 2018) – and “macro humps” – the hump-shaped, delayed peak response of aggregate output to an identified monetary policy shock. Representative-agent models with habit formation can match the macro hump but feature MPCs far too low to match the micro jump; existing heterogeneous-agent models match high MPCs but produce an aggregate response to monetary policy that peaks immediately, failing to display a hump. The paper resolves this tension by adding “sticky expectations” (following Carroll, Crawley, Slacalek, Tokuoka and White 2018) – households update their beliefs about future aggregate variables only infrequently, though they always observe current prices and income – to an otherwise standard incomplete-markets model with long-term debt, illiquid assets, and investment adjustment costs, estimated using the sequence-space impulse-matching methodology of Auclert, Bardóczy, Rognlie and Straub (2019) to jointly fit Christiano-Eichenbaum-Evans-style monetary policy impulse responses and Smets-Wouters-style aggregate time series. Two central findings follow. First, investment plays an outsized role in monetary transmission: eliminating the investment response cuts the cumulative five-year output response by over 80%, far more than investment’s roughly 40% direct accounting share of GDP, because a rise in investment raises labor income which, given high MPCs, is largely spent, generating a consumption-investment feedback loop absent from representative-agent or simple two-agent models (which, even calibrated to the same average MPC, lack the right intertemporal-MPC profile and show almost no such amplification); by the same logic, fiscal policy turns out to be comparatively unimportant for monetary transmission once the model is calibrated with realistic five-year-duration long-term government debt and a gradual, rather than balanced-budget, fiscal adjustment rule. Second, when the model is extended with a full set of seven orthogonal shocks and re-estimated on 1966-2018 US data, investment shocks explain about 65% of business-cycle-frequency output variation and 55% of consumption variation (versus about 15% and near-zero, respectively, in the estimated representative-agent counterpart), because the same high-MPC investment-consumption complementarity that amplifies monetary transmission also lets investment shocks generate the procyclical consumption-investment comovement that representative-agent models have struggled to explain since Barro and King (1984).

Monetary policy matters: Evidence from new shocks data

This 2013 Journal of Monetary Economics paper by S. Mahdi Barakchian and Christopher Crowe argues that the standard toolkit for identifying monetary policy shocks – recursive VARs (Christiano-Eichenbaum-Evans), over-identified VARs (Bernanke-Mihov), non-recursive VARs (Sims-Zha), and the Romer-Romer narrative measure – stops delivering plausible results once the sample extends past 1988, because the Federal Reserve’s increasingly forward-looking behavior violates the identifying assumptions these methods rely on: when a VAR omits the forward-looking variables the Fed actually reacts to, its endogenous response to anticipated conditions gets misread as an exogenous shock, biasing the estimated output effect upward. Using each method’s own original specification, the authors show all four produce the “wrong” sign in the post-1988 period – a contractionary shock followed by rising, not falling, output (for example, the CEE recursive VAR shows output declining in 1960Q1-1992Q4 but increasing significantly in 1988Q4-2007Q3, and the Bernanke-Mihov, Sims-Zha, and Romer-Romer methods show the same reversal over their own comparison samples). To sidestep the problem, they construct a new high-frequency-identification (HFI) shock measure from the innovations in six Fed Funds futures contracts (current month through five months ahead) around each of 157 FOMC meetings between 1988:12 and 2008:06, extracting a two-factor structure by maximum likelihood in which the first factor – explaining 92% of the variance, interpreted as a “level” shift in the expected medium-term policy path – is used as the shock, while a second “slope” factor (9% of variance, associated with forward-guidance content) is set aside. Feeding the cumulated new shock into a three-variable monthly VAR (log industrial production, log CPI, and the cumulated shock, with 36 lags and the shock ordered last) over 1988:12-2008:06, industrial production shows a statistically significant, sustained decline after a contractionary shock, with the maximum impact around a two-year horizon – recovering the sign that the conventional methods lose in this period – while prices exhibit a milder price puzzle (turning significantly negative only after about four years) that the authors cannot fully resolve even after adding a commodity price index or survey-based inflation expectations. Forecast error variance decompositions show the new shock accounts for roughly 40-50% of industrial production’s variance at a three-year-plus horizon, around twice the share attributed to existing shock measures over the same period, though the authors note this partly reflects the historically low volatility of the “Great Moderation” sample. A regression of the new shock on the Fed’s exclusive information (Greenbook-Blue Chip forecast gaps for 17 variables, following Romer-Romer, over 113 meetings in 1988-2002) finds no significant joint explanatory power (R-squared = 0.185, F(17) = 1.50, p = 0.132), suggesting the measure is relatively free of the simultaneity bias that could contaminate a directly observed policy-rate innovation – though two individual coefficients (current-quarter output growth and current-quarter GDP deflator) are significant, pointing to some remaining contamination that the authors argue would bias the estimated effects toward zero rather than away from it.

Monetary policy shocks: What have we learned and to what end?

This chapter in the 1999 Handbook of Macroeconomics (Volume 1) by Lawrence Christiano, Martin Eichenbaum, and Charles Evans surveys and unifies the VAR-based literature on identifying monetary policy shocks and tracing their dynamic effects on the U.S. economy, comparing multiple identification schemes within a common recursive framework. Using quarterly U.S. data (1965:Q3-1995:Q2) and three benchmark recursive identification schemes — treating the federal funds rate, nonborrowed reserves, or the ratio of nonborrowed to total reserves as the policy instrument — the authors find a robust set of qualitative facts following a contractionary monetary policy shock: the federal funds rate rises persistently, monetary aggregates decline (some with a delay), the price level responds very little for roughly a year before declining, and real GDP falls in a hump-shaped pattern with its maximal decline about one to one-and-a-half years after the shock. They also document and help resolve the “price puzzle” — the anomalous finding that a contractionary shock appears to raise prices — showing it arises from omitting commodity prices (a leading indicator of inflation available to the Fed) from the policy reaction function’s information set, and that including them typically eliminates the puzzle. Comparing the recursive approach against a fully simultaneous (Sims-Zha) identification and against narrative-based measures of policy shocks (Romer-Romer episodes), the chapter concludes that “qualitative inference about the effects of a monetary policy shock is quite robust to the different shock measures,” even though quantitative estimates — particularly the fraction of output variance attributable to policy shocks (ranging from about 7% to 44% at the 4-to-8-quarter horizon depending on the shock measure) — are considerably more sensitive to the specific identification chosen.

Monetary policy surprises and interest rates: Evidence from the Fed funds futures market

This 2001 Journal of Monetary Economics paper by Kenneth N. Kuttner addresses a basic errors-in-variables problem in earlier event studies of monetary policy and interest rates (notably Cook and Hahn 1989 and Roley and Sellon 1995): those studies regressed changes in market interest rates on the raw federal funds target change, without separating the portion markets had already priced in from the portion that came as a surprise, which biases the estimated response toward zero. Kuttner instead uses prices from the Chicago Board of Trade’s Fed funds futures market, established in 1989, to decompose each change in the target rate into an anticipated component and an unanticipated (surprise) component: the surprise is the change in the spot-month futures rate around the announcement, rescaled by a factor that converts from the contract’s monthly-average settlement basis into the point-in-time target-rate unit (and that also cancels out the futures risk premium). He then runs event-study OLS regressions of one-day changes in Treasury bill and bond yields, across maturities from 3 months to 30 years, on the anticipated and unanticipated components separately, using the 42 target changes between June 1989 and February 2000 (22 of them at scheduled FOMC meetings). The central finding is that the anticipated component’s coefficient is small and statistically insignificant at every maturity, while the unanticipated component’s coefficient is large and highly significant, ranging from about 79 basis points at the 3-month maturity down to about 19 basis points at 30 years, with the response declining roughly monotonically across the maturity spectrum; a Wald test rejects equality of the two coefficients at conventional levels for every maturity. By contrast, the naive regression on the raw, undecomposed target change yields only a 27-basis-point response at 3 months, illustrating the scale of the errors-in-variables bias the decomposition removes. The results are robust to restricting the sample to FOMC meeting dates only and to using monthly rather than daily observations, and a companion regression of futures rates at one- to five-month horizons on the same decomposition shows unanticipated-component coefficients clustered between 0.55 and 0.64 that cannot be statistically distinguished across horizons, indicating that a surprise target change mainly shifts the perceived level of the near-term target rather than expectations about further future changes. The paper documents only the interest-rate response to policy surprises and does not trace effects through to output or inflation, and its sample is necessarily confined to the post-1989 period since that is when the Fed funds futures market began trading.

Review of Milton Friedman and Anna J. Schwartz's 'A Monetary History of the United States, 1867-1960'

Writing for the 30th anniversary of Milton Friedman and Anna Schwartz’s A Monetary History of the United States, 1867-1960, Lucas argues that the book’s enduring contribution is not merely its “beautiful time series on the money supply and its components” but a coherent normative narrative: nearly a century of U.S. monetary history, organized around two principles – long-run monetary neutrality and a short-run non-neutrality operating through unexplained but transient price rigidities – in which every major depression is traced to an avoidable policy mistake or an unchecked banking panic, so that the whole period “might have evolved, with stable prices and smoothly growing real output” had the monetary authority acted differently. Lucas states he finds this normative argument “wholly convincing,” particularly for the 1929-33 contraction, but presses on what a model-free narrative history cannot do: answer how much smoother money growth would have helped, since the book’s own conclusions are, in his words, not “a verbal summary of tables describing the results of a numerical simulation” but “the simulation” itself. He then surveys three later research programs against this yardstick – 1970s rational-expectations models that reconciled the book’s two neutrality principles but reached opposite normative conclusions about optimal policy depending on how price rigidity is modeled; Christopher Sims’s atheoretical statistical approach and Romer and Romer’s “natural experiments,” each proposing a different, non-equivalent notion of monetary “independence” from Friedman and Schwartz’s own; and real-business-cycle theory, which Lucas judges incapable of explaining the Depression’s actual magnitude – the Solow residuals for 1928-1933 are far too small to map into a 40% decline in output – while nonetheless reshaping how the discipline reads the comparatively small role of money in accounting for postwar fluctuations, not as evidence money is unimportant but as evidence postwar monetary policy has been close to efficient.

Shocks

Cochrane surveys the empirical evidence for the leading candidate shocks behind postwar U.S. business cycles – monetary policy, technology, oil prices, and credit – and concludes that none of them robustly accounts for the bulk of output fluctuations, while unforecastable movements in endogenous variables like consumption and output themselves explain a large and comparatively stable 50-70% of output variation. Working mostly through VARs, he shows for monetary shocks that estimated contributions to output variance range from as high as 82% in simple specifications down to under 10% once more “level” variables (consumption, hours), alternative orderings, and long-run restrictions are imposed, with virtually no explanatory power at horizons under a year; he argues the largest credible estimate is around 15-25% at a two-to-three-year horizon, tenuous even then. For technology shocks, Prescott’s famous calculation that 70% of output variance is technology-driven proves to be extremely sensitive to sampling error, the choice of statistic (variance decomposition versus long-horizon forecastability versus Beveridge-Nelson-detrended variance), and the production-function specification, with several re-calculations – inspired by Blanchard-Quah, Rotemberg-Woodford, and Christiano – pushing the figure down toward a small fraction of a percent; the concept of a “technology shock” is also shown, following Hansen and Prescott’s own broadening of the term, to risk becoming vacuous, standing in for essentially any distortion that lowers measured output given capital and labor. Oil-price and credit shocks receive briefer treatment and are found quantitatively too small (each explaining well under 20% of output variance in Cochrane’s VARs) despite genuine, if modest, supporting descriptive evidence. Faced with this shortfall, Cochrane examines whether unobservable “consumption” or “news” shocks – information individual agents have about their own prospects that, aggregated, forecasts future aggregate activity – can generate genuine business-cycle dynamics; he shows that a standard real-business-cycle model does not naturally produce consumption-led downturns from good news (news of future productivity growth instead triggers an immediate consumption rise and a decline in current output and labor), but that adding an explicit persistent news-shock process, or feeding VAR-based technology forecasts through the model, can reproduce the data’s characteristic transitory-output, forecastable-growth pattern. He closes by noting that if this news-shock view is correct, economists may remain permanently unable to name the true underlying causes of business cycle fluctuations.

Structural Approaches to Vector Autoregressions

This 1992 Federal Reserve Bank of St. Louis Review paper (a policy review, not peer-reviewed) by John W. Keating is an expository survey of structural vector autoregression (VAR) methods that shows how a reduced-form VAR, x_t = β(L)x_{t-1} + e_t, can be derived from an underlying structural simultaneous-equations model, Ax_t = C(L)x_{t-1} + Dz_t, and then develops two families of identifying restrictions needed to recover the structural parameters in A and D from the estimated reduced-form residual covariance matrix Σ_e = A^{-1}DΣ_εD’A^{-1’}: contemporaneous restrictions on the impact matrix A (of which the Choleski, or recursive, decomposition is a special, atheoretical case) and long-run restrictions on the cumulative multiplier matrix θ(1), following the Blanchard-Quah (1989) approach, which let the data determine short-run dynamics while theory constrains only shocks’ permanent effects. Keating illustrates both strategies with a 4-variable quarterly U.S. system — the GNP deflator, real GNP, the 3-month Treasury bill rate, and M1, all first-differenced, over 1959:Q1-1991:Q3 with 4 lags — specifying a contemporaneous model (a predetermined price/aggregate-supply equation, an IS equation, a money-supply reaction function, and a buffer-stock money-demand equation) and a long-run model that restricts aggregate-supply shocks to be the sole source of permanent output movements. For the long-run model only, the paper reports estimated structural shock standard deviations (aggregate supply 0.0144, IS 0.0092, money demand 0.0149, money supply 0.0172, all significant at 5%), finds the long-run money-demand and money-supply coefficients statistically significant at 5% while the long-run IS coefficient on output is not, and shows via impulse responses (with Runkle 1987 Monte Carlo confidence bands) that supply shocks raise output permanently and lower prices, IS shocks raise output only temporarily but raise prices and the interest rate permanently, money-demand shocks have essentially no effect on output, prices, or rates, and money-supply shocks have a “relatively small” effect on output that peaks at 13 percent of its variance about two years out; variance decompositions show supply shocks explaining 90 percent of output variance at a 48-quarter horizon. For the contemporaneous model, by contrast, all ten estimated structural coefficients (Table 3) are statistically insignificant even though several carry theory-consistent signs (money-demand coefficients of roughly one-half on nominal spending and almost -1.0 on the interest rate; a positive coefficient on money in the money-supply/interest-rate equation, read as evidence the Fed “attempts to stabilize money growth by raising interest rates”), and its impulse responses (Figures 5-8) and variance decomposition (Table 4) mix expected patterns — supply shocks dominating output and price variance, IS shocks dominating short-run output movement and long-run interest-rate variance, much as in the long-run model — with patterns Keating calls “inconsistent with most macroeconomic theories,” namely the money-demand shock’s effects concentrated on long-run output and short-run interest-rate variance rather than money itself, and the money-supply shock’s negligible effect on output paired with a large, persistent effect on real money. Keating’s own comparison concludes that “wherever a significant discrepancy exists between the two models, the model with long-run restrictions yields sensible results, while the results from the contemporaneous model are inconsistent with standard economic theories,” and that long-run parameters are estimated more precisely; he attributes this to macro theories often sharing long-run properties while differing in short-run dynamics, and to long-run identification avoiding contemporaneous exclusion restrictions that his 1990 paper argues can be invalid under rational expectations, while flagging that further research is needed to know whether this result generalizes beyond the paper’s illustrative system.

Taper Tantrums: Quantitative Easing, Its Aftermath, and Emerging Market Capital Flows

Identifying US monetary policy shocks from daily moves in five-year Treasury futures around FOMC announcements, this paper shows that during the unconventional-policy years those shocks largely represent revisions to required risk compensation rather than to the expected short-rate path, and that their effects on emerging market portfolio positions run mainly through valuations rather than physical flows – with by far the largest effects during the taper period. The shock measure follows Rogers, Scotti and Wright (2014): the daily change in the implied yield of the five-year Treasury futures contract on FOMC announcement dates, plus the additional policy events in Gagnon et al. (2011) and the taper-tantrum date of 22 May 2013. Its average value is a fall of 2.0 basis points during the QE period and a rise of 1.6 basis points during the taper period, against minus 0.6 for the full sample and minus 0.5 pre-crisis, with the period differences statistically significant. Feeding the shock through the Kim and Wright (2005) affine term structure decomposition shows it moves both the expected short rate and the term premium in the conventional period, but that in the unconventional periods the largest effects are on term premia and those effects rise monotonically with maturity – a one-standard-deviation shock raises the ten-year yield by 4.7 basis points pre-crisis but 12.2 basis points during QE, against unconditional daily ten-year standard deviations of 5.8 and 7.1 basis points respectively. The capital-flow analysis uses Bertaut-Tryon and Bertaut-Judson monthly estimates built from US Treasury International Capital data, covering 15 emerging markets monthly from 1994 to 2014, with positions, flows and valuation changes for debt and equity separately scaled by annual GDP, estimated in a random-effects panel with lagged dependent variables, an extensive set of lagged push and pull controls, and country-clustered robust standard errors. Three kinds of heterogeneity emerge. Flows versus prices: “in nearly every specification, the effect of monetary policy shocks on asset returns is larger than that for physical flows,” which the authors read as consistent with the shocks capturing revisions in required risk compensation. Debt versus equity: during QE the coefficient on equity valuations is ten times that on debt valuations, and during the taper period equity effects are double or triple debt effects. QE versus tapering: during QE the significant responses are confined to debt and equity valuations and equity positions, whereas after tapering was first mentioned the coefficients are inversely signed and significant at the 1 percent level across essentially every variable, and an order of magnitude larger than pre-crisis for debt positions, debt valuations and equity flows. Because the shock has a magnitude, the paper can price these effects: a mean-sized QE shock corresponds to roughly a $153.5 million monthly increase in US emerging market equity positions per country, a mean-sized taper shock to a $144.1 million monthly outflow, with one-standard-deviation ranges of roughly minus $672 million to plus $979 million during QE. The paper is explicit that its estimates are associations from a controlled panel regression rather than structural effects: coefficients are described throughout as correlations, the shock’s channel is inferred from coefficient signs rather than separately identified, and the exchange rate results, while statistically significant, are reported as economically modest against an unconditional monthly bilateral exchange rate standard deviation of 3.57 percent.

The Dynamic Impacts of Monetary Policy: An Exercise in Tentative Identification

This 1994 Journal of Political Economy paper by David B. Gordon and Eric M. Leeper proposes a structural vector autoregression (SVAR) that identifies monetary policy shocks by modeling supply and demand simultaneously in the reserves market and, separately, in the M2 market, rather than treating a single monetary aggregate or interest rate as predetermined the way standard Cholesky, nonborrowed-reserves, or federal-funds-rate identification schemes do. In each market a demand equation (the monetary aggregate as a function of its own interest rate, the price level, and output, with the interest elasticity a_1 restricted negative) and a supply equation (the interest rate as a function of the aggregate, the 10-year Treasury yield, and a commodity price index, with elasticity a_4 restricted positive) are estimated jointly by maximum likelihood, identified by two restrictions: private demanders of money observe only the current interest rate, prices, and output within the month (the long-term rate is excluded from demand), while the Federal Reserve observes current financial-market variables (funds rate, 10-year yield, commodity prices) but not current goods-market data (price level, output, unemployment) because of a roughly one-month reporting lag, so those goods-market variables are excluded from the supply/Fed-behavior equation; the goods-market block itself is treated as block-recursive, unaffected contemporaneously by either money market. Using monthly U.S. data over 1982:12-1992:4 with six lags, the estimated interest elasticity of reserves demand is -0.028 (se 0.014) and the reserves supply elasticity is +30.28 (se 10.95); for M2 the demand elasticity is -0.0082 (se 0.0007) and the supply elasticity is +253.84 (se 10.79); the reserves-market model narrowly fails an overidentification test (chi-squared(5)=11.864, p=.04, which the authors describe as “not… a strong statistical rejection”), while the M2-market model comfortably passes (chi-squared(5)=2.715, p=.74). A one-unit expansionary supply shock in the reserves market produces a liquidity effect – the funds rate falls sharply and reserves rise – that lasts about 8 months before the rate climbs back above baseline after roughly two years, with the price level and industrial production rising significantly within about six months and unemployment falling with roughly a three-month lag, responses the authors call “fully consistent with traditional analyses,” in contrast to the liquidity-puzzle and price-puzzle results the paper documents earlier from reduced-form and single-variable-innovation identifications. The analogous M2 supply shock produces qualitatively similar but “less significant” responses, and a regression linking the two shows the M2 supply shock loading on current and lagged reserves supply shocks with R-bar-squared only 0.19, implying that a substantial share of M2 supply-shock variation reflects nonpolicy behavior of financial institutions rather than monetary policy, so the authors caution that inference from M2-based studies “should be drawn cautiously.” Variance decompositions show policy shocks explaining a large share of reserves and funds-rate forecast-error variance on impact but declining quickly, while their share of price-level and output variance rises to roughly 30 percent by three years out. When the same reserves-market model is re-estimated on 1971:1-1979:9 data, the identifying restrictions “fail to identify demand and supply relationships for reserves in the 1970s”: the demand elasticity flips to the wrong sign (+0.0069) and neither elasticity differs significantly from zero, which the authors attribute to a different Federal Reserve operating procedure in that earlier period rather than to a flaw in the identification strategy itself, whereas the M2-market model still recovers sensible elasticities in the 1970s but now strongly rejects overidentification (chi-squared(5)=14.958, p=.01).

The Fed and Interest Rates—A High-Frequency Identification

This 2002 American Economic Review Papers and Proceedings essay by John Cochrane and Monika Piazzesi proposes a high-frequency alternative to monthly-VAR identification of monetary policy shocks, built from daily interest-rate movements around actual Federal Open Market Committee target-change dates rather than from orthogonalized VAR residuals. Following Piazzesi (2001), they construct two shock measures that are zero in any month without a target change: a “target shock” (the change in a given yield from two days before to one day after a target change, regressed on the target-rate change itself) and a “Eurodollar shock” (the change in the one-month Eurodollar rate over the same window). Using daily U.S. interest-rate data and monthly nonfarm employment, CPI, and commodity-price data over 1984-2001, they estimate two regressions on actual target-change dates (Table 1): the target-shock regression (Panel A) finds coefficients ranging from 0.52 (t=9.1) at the one-month Eurodollar rate down to just 0.19 — 19 basis points (t=3.5) — at ten years; the separate Eurodollar-shock regression (Panel B) fits much better (R-squared as high as 0.87 at three months) and is the source of the paper’s “startling” headline figures — a 1-percent unexpected target change moving Treasury yields by 60-70 basis points from three months to three years and by 52 basis points even at ten years. This produces a “level” effect on the yield curve — all maturities moving together — in sharp contrast to the “slope” effect (short rates moving, long rates barely responding) found using monthly Christiano-Eichenbaum-Evans (CEE, 1996) VAR shocks. A separate forecasting regression (Table 2, 1984-2001, R-squared 0.64) shows Fed target changes are much better predicted by long-term rates than short rates — the two-year rate (b=0.87, t=6.7) and five-year rate (b=-0.87, t=-3.5) dominate — while an earlier, broader version of that regression that included all yields (R-squared 0.66, before variables with small t-statistics were dropped to reach the final Table 2 specification) found the one-month rate contributing almost nothing (b=-0.06, t=-0.8); this pattern implies the Fed reacts to market-embedded inflation expectations and to the yield-curve slope’s real-activity signal, consistent with (but outperforming) a Taylor rule. Two further findings complicate the identification: employment rises, rather than falls, following a high-frequency contractionary shock (versus a slow decline under CEE shocks), a difference the authors trace to how each measure classifies the 1979-1982 episode; and neither shock measure shows a statistically significant inflation decline, with the target-shock measure instead showing a large but “dubiously significant” price-puzzle-like increase. Dynamically, all interest rates keep rising for two years after a shock under the high-frequency measures — the ten-year rate rises 0.8 percentage points versus only 0.2 for the CEE shock — a pattern the authors call “troubling” because it runs against the standard intuition that tight policy lowers long-run inflation expectations and hence long rates. The paper’s own tentative conclusion is that because the Fed’s actions are so often forecastable responses to market information, “perhaps there are no true shocks,” and results throughout are described as preliminary, based on a small number of genuine surprises, and not resting on formal statistical inference beyond the reported coefficients and standard errors.

The global capital flows cycle: structural drivers and transmission channels

Decomposing a global risk factor built from stock returns in 63 economies into structural shocks, this paper finds that exogenous shifts in the financial sector’s risk-bearing capacity matter more than US monetary policy for driving global risk, and that the transmission of risk to capital flows follows a classical trilemma – countries that are both financially open and pegged are markedly more exposed – driven almost entirely by cross-border bank loans. The data are quarterly, 1990Q1 to 2017Q4, for 50 countries (18 advanced, 32 emerging), with gross capital inflows split into direct investment, portfolio equity, portfolio debt and other investment from the IMF’s Balance of Payments Statistics; several financial centres are excluded outright because their flows “record extremely large values with respect to GDP and are very volatile,” and dependent variables are winsorised at 1 percent. Global risk is proxied by a Global Stock Market Factor, the first principal component of country-average stock returns in 63 economies. Three findings follow. First, in a seven-variable Bayesian structural VAR identifying a US monetary policy shock by external instrument and US demand, global financial and geopolitical risk shocks by sign restrictions, the forecast error variance decomposition of global risk at a twelve-month horizon attributes about 19 percent to US monetary policy against 23 percent to financial shocks, 13 percent to geopolitical risk and under 10 percent to US demand – with the gap widening at higher percentiles, where financial shocks reach roughly 70 percent and US monetary policy no more than about 30 percent. The authors are careful that this does not demote monetary policy: “not only monetary policy is indeed relevant for global risk as the proponents of the global financial cycle have stressed, but its quantitative role is all but negligible.” Second, in country panel regressions with country fixed effects, four lags of the dependent variable and Driscoll-Kraay standard errors, the Global Stock Market Factor is negative and statistically significant for every category of flows, where the VIX is significant only for portfolio flows – but the average magnitude is modest, a one-standard-deviation risk shock cutting gross inflows by between 0.1 percent of GDP for equity and 0.8 percent for other investment, and total inflows by 1.7 percent of GDP against a flow volatility of 14 percent. US monetary policy surprises are significant for portfolio flows but not for other investment or total flows, which the paper reconciles with Bruno and Shin by noting that US policy “can affect these flows only to the extent that it induces significant shifts in global risk.” Third, interacting the risk factor with capital account openness and exchange rate regime dummies yields a classical trilemma: both policy variables matter for the transmission of global risk but neither matters for US monetary policy surprises, strict pegs transmit risk shocks more strongly to other investment and direct investment but “not necessarily to portfolio flows,” and the effect is concentrated enough that the paper concludes the trilemma “is largely driven by one category of capital flows: other investment.” Conditioning on policy raises the economic significance considerably – for open economies with an exchange rate target the impact on total inflows reaches 4.0 percent of GDP against a 1.7 percent unconditional average, and for open, pegging emerging markets it exceeds 4 percent of GDP, four times the 1 percent average across all emerging markets and large enough, against a typical inflow of almost 7 percent of GDP, to constitute a sudden stop.

The Response of Term Rates to Fed Announcements

Demiralp and Jorda argue that since the Fed began publicly announcing its Federal funds rate target in 1994, it has manipulated short-term rates through a Pavlovian “announcement effect” – moral suasion backed by a credible threat – requiring far smaller open-market operations than the conventional liquidity effect would predict, and they show, using a new hazard-model decomposition of target changes into anticipated and surprise components, that unanticipated target changes and surprise Fed inaction both move the term structure in ways consistent with the rational expectations hypothesis. The paper’s starting observation is that the February 3-4, 1994 FOMC meeting inaugurated the practice of publicly disclosing the new Federal funds rate target immediately after each meeting, and the authors argue this gave the Fed a genuinely new policy tool distinct from the textbook liquidity effect, in which the Fed moves rates only by manipulating the supply of nonborrowed reserves. Exploiting institutional developments in the reserves market (declining required reserves, retail sweep programs, minimal use of the discount window), the authors show with structural VARs that the estimated liquidity effect is artificially inflated when target-change dates are included in the sample, and that this inflation is much smaller after a 1989 “Thanksgiving effect” episode after which banks began reading Fed intentions with near-certainty – evidence that an announcement/credibility channel was already emerging before the formal 1994 disclosure policy. To measure the effect of target changes on the term structure, the authors build an autoregressive conditional hazard (ACH) model, paired with an ordered-probit model for the size of changes, that decomposes each Federal funds rate target change into its anticipated and unanticipated components without relying solely on the federal funds futures market (which only began in 1989). Regressing Treasury rates of various maturities on these components, they find that anticipated target changes have no significant effect on term rates (as the rational-expectations hypothesis predicts), while unanticipated target changes move rates significantly, an effect that becomes stronger and extends to longer maturities after 1989; strikingly, after 1989 markets also react significantly to surprising Fed inaction – an expected target change that fails to occur – at medium and long maturities, evidence that the market had come to understand and price the Fed’s own reaction function closely enough to be surprised when it did not act as expected.

U.S. Monetary Policy and the Global Financial Cycle

A single global factor extracted from a large panel of risky asset prices traded around the world falls sharply after a US monetary contraction, alongside deleveraging by global banks, a rise in aggregate risk aversion, contracting credit provision and retrenching international credit flows – and countries with floating exchange rates are subject to financial spillovers of similar magnitude. The paper proceeds in two empirical steps plus a model. First, a dynamic factor model fitted to 858 monthly price series for 1990-2012 – equities from North America, Latin America, Europe, Asia Pacific and Australia, corporate bond indices, and commodities excluding precious metals – supports a unique common global factor that accounts for over 20 percent of the common variation; on a narrower 303-series panel covering only the US, Europe, Japan and commodities but reaching back to 1975, one factor accounts for about 60 percent. Second, monthly Bayesian VARs with 12 lags estimated over 1980-2010 – a 13-variable closed-economy version and 15-variable global versions – are identified with an external instrument built from 30-minute price revisions in the fourth federal funds futures contract around FOMC announcements. For a shock normalised to raise the one-year Treasury rate by 1 percent on impact, the domestic responses are conventional: production, capacity utilisation and housing starts fall, unemployment rises, prices decline without a price puzzle, the excess bond premium and mortgage spreads widen, house prices and the S&P 500 fall, and the dollar appreciates. The global responses are the paper’s point. The global factor drops about 40 percent on impact, which the authors translate – under the explicit assumption that all asset prices load equally on the factor – into roughly an 8 percent fall in a local stock market, a figure consistent with the estimated US, UK and euro-area equity responses. Measured aggregate risk aversion rises by over 50 percent above its average trend. Global domestic credit and cross-border credit inflows to both banks and non-banks contract by several percentage points, with global real activity outside the US left unchanged on impact as a control, and the credit contraction is not driven by the US component. Leverage falls strongly and quickly for US security brokers and dealers and for European global systemically important banks, while broader banking aggregates react later and more weakly – no appreciable change for US banks and a European trough about a year out. Restricting the global aggregates to the 32 countries the IMF classified as independently floating leaves the credit and inflow contractions “very similar to that obtained over the full sample.” For the UK and euro area specifically, equity indices plummet, the dollar appreciates in a reversal-prone way over one to four quarters, corporate spreads widen on impact, and policy rates ease endogenously by about 30 basis points – so the local tightening of financial conditions “cannot be ascribed to a domestic monetary policy tightening.” The authors state the interpretive limit carefully: the floater result “challenges the degree of monetary policy sovereignty of open economies” and echoes Rey’s (2013) trilemma-to-dilemma claim, but “does not mean that exchange rate regimes do not matter,” and whether open-economy models with Value-at-Risk-type frictions can actually reproduce these regularities “still remains to be seen.”

Unconventional but Different After All? A Unified Series of Narrative Monetary Policy Shocks

Asking whether unconventional monetary policy works differently from an interest-rate cut has been hard to answer because the two kinds of policy have been measured with different tools, so any difference in the estimated effects could be a difference in the measuring rather than in the policy. This paper builds one shock series for both, running Romer and Romer’s (2004) narrative regression — policy-rate changes on the Federal Reserve’s own Greenbook/Tealbook forecasts — but substituting Wu and Xia’s (2016) shadow rate for the federal funds rate over the 2009–16 zero-lower-bound period, which yields shocks for 1969–2008 and 2009–16 built the same way. Estimating local projections on the pooled sample with a structural break at the zero lower bound, the authors cannot reject equality of the peak responses of the interest rate, industrial production, unemployment and the consumer price index across regimes (p-values from 0.09 to 0.56), which they read as unconventional policy being about as effective as rate cuts for aggregate activity. Joint Wald tests do reject equality for wealth shares and for the stock-to-house price ratio, and the sign flips: an expansionary conventional shock normalised to 25 basis points lowers the top 10% wealth share by about 0.3% at the trough, while the equivalent unconventional shock raises it by up to about 0.36% and lowers the middle 40% share by about 0.7%. The authors attribute the difference to asset prices — unconventional easing raises the stock-to-house price ratio by over 3% where conventional easing lowers it by about 3.13% — and a mechanical revaluation of Survey of Consumer Finances portfolios reproduces the pattern, with the equity-heavy top 10% capturing most of the gains and households in the bottom groups, who hold little equity and often no housing, receiving comparatively little. The distributional evidence rests on short samples — 1989–2008 for conventional and 2009–15 for unconventional shocks, at quarterly frequency — and the authors do not address the Fed-information-effect and Fed-reaction-to-news concerns directly, assessing them only indirectly by comparison with shock series that are robust to them.

Understanding the Price Puzzle

This 1994 Federal Reserve Bank of Dallas Economic Review paper by Nathan S. Balke and Kenneth M. Emery investigates the “price puzzle” – the empirical regularity that a positive innovation to the federal funds rate in a standard reduced-form VAR is followed by a rise, rather than a fall, in prices – using recursive (Cholesky) VARs on quarterly U.S. data spanning 1960:1-1993:4, with subsample splits at 1960:1-1979:3 and 1982:4-1993:4 (the 1979:4-1982:3 non-borrowed-reserves period is excluded from the subsamples because the Fed did not target the funds rate then). In a base three-variable VAR (output, prices, funds rate, Cholesky-ordered with the funds rate last), the price puzzle appears in the full sample and is much stronger in the pre-1980 subsample, where prices rise substantially for several years, than in the post-1982 subsample, where the funds rate’s effect on prices, though negative, is small and not statistically different from zero – a pattern the paper also confirms with a Romer and Romer (1989) narrative-shock distributed-lag approach. Adding commodity prices to the VAR (replicating Christiano, Eichenbaum, and Evans) eliminates the puzzle for the full sample and the post-1982 subsample but not the pre-1980 subsample, where prices remain above their original level for nearly three years; among the other candidate variables tested (oil prices, stock prices, unit labor costs, an index of leading indicators, capacity utilization, and individual short- and long-term rates), only the term spread (10-year Treasury bond rate minus 3-month Treasury bill rate) systematically helps resolve the puzzle, eliminating it in both subsamples when included alone – though not in the full sample, possibly because of extreme interest-rate volatility during 1979-82 – and eliminating it across all three samples when combined with commodity prices. Because the federal funds rate responds negatively, not positively, to a positive spread shock, the authors argue the data are inconsistent with Sims’s (1992) inflation-expectations explanation of the puzzle (which implies the Fed should tighten when the spread signals higher expected inflation), and instead favor a supply-shock explanation in which the Fed incompletely offsets the inflationary consequences of negative supply shocks that raise both commodity prices and the general price level while lowering output; the muted post-1982 puzzle is attributed to either the Fed placing more weight on price stability after Volcker or fewer severe supply shocks hitting the economy in the 1980s. Throughout, the paper reports one-standard-error rather than the more conventional two-standard-error confidence bands, understating the statistical uncertainty of its findings, and it is published as a Federal Reserve Bank policy review rather than a peer-reviewed journal article.

What are the effects of monetary policy on output? Results from an agnostic identification procedure

This 2005 Journal of Monetary Economics paper by Harald Uhlig proposes an “agnostic” sign-restriction procedure for identifying monetary policy shocks in a VAR, designed to test rather than assume the conventional view that a contractionary shock lowers real output. Using a monthly six-variable VAR for the United States (real GDP, GDP deflator, a commodity price index, total reserves, nonborrowed reserves, and the federal funds rate, January 1965-December 2003, 12 lags as in Bernanke and Mihov), Uhlig defines a contractionary policy shock as an impulse vector for which, over a benchmark horizon of K=5 months (six months after the shock), the GDP deflator, the commodity price index, and nonborrowed reserves do not rise and the federal funds rate does not fall – while leaving the response of real GDP completely unrestricted. Any impulse vector consistent with a given reduced-form VAR’s covariance matrix can be written as a linear combination of the Cholesky factor’s columns (Appendix A), so the sign restrictions are imposed via a Bayesian procedure that draws candidate impulse vectors from a Normal-Wishart posterior and keeps only those satisfying the restrictions (the “pure-sign-restriction” approach), with a “penalty-function” alternative that instead selects, for every posterior draw, the vector minimizing a penalty for sign violations (Appendix B). The central finding is that once GDP is left free, contractionary monetary policy shocks have an ambiguous effect on real output – moving it up or down by up to about 0.2% with two-thirds probability – and account for perhaps 5-10% (and possibly under 3%) of real GDP’s forecast-error variance at any horizon, versus the roughly 50% implied by a standard Cholesky ordering, which also generates a price puzzle that the sign-restriction approach avoids by construction. Uhlig traces the conventional finding that output falls to a single additional, and in his view spurious, restriction implicit in recursive (Cholesky) orderings – that GDP not respond on impact – and shows that adding that one restriction back in (while the agnostic restrictions alone leave the deflator- or GDP-widened bands compatible with no output effect at all) is what manufactures the textbook contractionary-output result. A penalty-function version of the exercise gives qualitatively similar but somewhat sharper (more precisely estimated, slightly larger-magnitude) results, and is described as more sensitive to the choice of restriction horizon K than the pure-sign-restriction approach.

What do the VARs mean? Measuring the output effects of monetary policy

Cochrane argues that even once a VAR-based monetary policy shock is “correctly identified” – the focus of the extensive VAR literature on variable selection and shock orthogonalization – a further, purely theoretical identifying assumption is still required before the estimated impulse-response function can be read as the causal effect of that shock: whether anticipated monetary policy actions, not just unanticipated shocks, can themselves affect output. Using both a flexible linear model that weights anticipated and unanticipated money by a parameter lambda, and an explicit Rotemberg-style sticky-price model parameterized by a price-adjustment-cost parameter alpha, Cochrane shows that this single identifying choice moves the estimated output effects of monetary policy at least as much as, and typically more than, the variable-selection and orthogonalization assumptions the VAR literature usually debates. Applying both models to a standard M2 VAR and a federal-funds-rate VAR on 1959-1992 U.S. quarterly data, he finds that if anticipated money is assumed to have no effect on output – the implicit assumption behind treating the impulse-response function itself as the policy-invariant effect of a shock – the estimated output response to a shock not followed by the customary further monetary expansion is large, hump-shaped, and takes roughly five years to die out; but allowing anticipated money to matter even slightly shrinks this same estimated response dramatically, because most of the impulse-response function’s persistence then reflects the systematic further money growth that has historically followed such shocks, not the shock’s own lagged causal effect. The sticky-price model reproduces this result through a different mechanism and adds the counterintuitive finding that assuming stickier prices implies a smaller and shorter estimated output response to an unanticipated shock, since more of the VAR’s observed dynamics are then attributed to slow price adjustment. Cochrane concludes that data from a single policy regime cannot settle which identifying assumption is correct, but that circumstantial considerations – the otherwise-coincidental similarity in shape between the money and output impulse responses, and the alternative’s reliance on ad hoc multi-year delay and propagation mechanisms – favor giving at least some role to anticipated, systematic monetary policy, with direct implications for how VAR evidence is used to argue about the length and size of monetary policy’s real effects.

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.

Why Are Target Interest Rate Changes so Persistent?

This 2012 American Economic Journal: Macroeconomics paper by Olivier Coibion and Yuriy Gorodnichenko adjudicates between two observationally similar explanations for the federal funds rate target’s very high persistence (an estimated AR(1) coefficient of roughly 0.9-0.99 in typical Taylor-rule regressions): deliberate interest-rate smoothing (policy inertia, modeled as partial adjustment toward a Taylor-rule target) versus serially correlated monetary policy shocks (unexplained deviations from the Taylor rule that themselves carry over from meeting to meeting). Using real-time Federal Reserve Greenbook forecasts of inflation and the output gap – treated as the best available measure of the Fed’s actual information set at each decision, released only with a five-year publication lag – over a 1987:IV-2006:IV quarterly sample bounded by Greenbook availability and ending before the zero lower bound, the paper estimates a sequence of nested Taylor-rule specifications. A simple specification nesting a smoothing parameter and a persistent-shock parameter in a single AR(1) finds both statistically significant (smoothing coefficient about 0.81, shock persistence about 0.46), and the paper concludes that this simple nested specification cannot overwhelmingly differentiate between the two explanations. But once the paper generalizes to higher-order AR(p) smoothing and AR(q) shock lags and lets the Bayesian information criterion (BIC) select among the resulting grid of specifications, the selected model is robustly IS(2), AR(0) – two lags of interest-rate smoothing and no persistent-shock lags – across the full sample and pre-/post-2000 subsamples; under a richer AR(2)-smoothing-plus-AR(2)-shock specification the smoothing coefficients remain large and significant (summing to roughly 0.75-0.95) while the shock-persistence terms turn insignificant at the first lag and negative and significant at the second. The paper corroborates this with an instrumental-variables strategy that instruments the lagged rate with non-monetary shocks (oil prices, technology, fiscal policy) to address possible endogeneity, yielding smoothing estimates of 0.70-0.87; with a Rudebusch (2002)-style predictability test showing Greenbook forecasts (rather than Eurodollar futures) predict future target changes at 2- and 3-quarter horizons with R-squared of 20% and 12%, consistent with genuine smoothing; with checks showing financial conditions, forecast revisions, and time-varying-inflation-target measures cannot fully account for the estimated smoothing; and with narrative evidence – Greenspan’s 1994 stated preference for 25-basis-point increments and Bernanke’s 2004 “Gradualism” speech – that Fed officials themselves describe deliberate gradual adjustment. A further implication is that the interest-smoothing interpretation implies a long-run inflation response exceeding one (satisfying the Taylor principle), whereas the no-smoothing persistent-shock interpretation implies a contemporaneous inflation response that can fall below one, a determinacy-violating result the smoothing interpretation avoids.