Measuring the Effectiveness of US Monetary Policy during the COVID-19 Recession
MMeasuring the Effectiveness of US MonetaryPolicy during the COVID-19 Recession
MARTIN FELDKIRCHER
1, 3 , FLORIAN HUBER ∗ , and MICHAELPFARRHOFER Oesterreichische Nationalbank (OeNB) University of Salzburg Vienna School of International Studies (DA)
Abstract . The COVID-19 recession that started in March 2020 led to anunprecedented decline in economic activity across the globe. To fight thisrecession, policy makers in central banks engaged in expansionary monetarypolicy. This paper asks whether the measures adopted by the US FederalReserve (Fed) have been effective in boosting real activity and calmingfinancial markets. To measure these effects at high frequencies, we propose anovel mixed frequency vector autoregressive (MF-VAR) model. This modelallows us to combine weekly and monthly information within an unifiedframework. Our model combines a set of macroeconomic aggregates such asindustrial production, unemployment rates and inflation with high frequencyinformation from financial markets such as stock prices, interest rate spreadsand weekly information on the Feds balance sheet size. The latter set ofhigh frequency time series is used to dynamically interpolate the monthlytime series to obtain weekly macroeconomic measures. We use this setup tosimulate counterfactuals in absence of monetary stimulus. The results showthat the monetary expansion caused higher output growth and stock marketreturns, more favorable long-term financing conditions and a depreciation ofthe US dollar compared to a no-policy benchmark scenario.
JEL : E52, E58, H12
KEYWORDS : Unconventional monetary policy, mixed frequency model,monetary policy effectiveness ∗ Corresponding author: Florian Huber. Salzburg Centre of European Union Studies, University ofSalzburg.
Address : M¨onchsberg 2a, 5020 Salzburg, Austria.
Email : fl[email protected]. FlorianHuber and Michael Pfarrhofer gratefully acknowledge financial support from the Austrian Science Fund(FWF, grant no. ZK 35). a r X i v : . [ ec on . E M ] J u l . INTRODUCTION Worldwide restrictions to contain the spread of the novel Coronavirus (COVID-19) triggered asharp drop in global economic activity, a collapse in trade and a severe rise in unemployment.First estimates for 2020 point at considerable contractions of GDP in most advanced economies(McKibbin and Fernando, 2020). Policymakers responded swiftly, with unprecedented fiscal stim-ulus packages in the magnitude of nearly 15% of global GDP. In the same vein, central banksprovided stimulus by loosening their policy stance considerably. In many emerging economies,central banks successfully introduced forms of quantitative easing for the first time (Arslan et al. ,2020; Hartley and Rebucci, 2020), while in advanced economies with policy space, easings tookmostly the form of rate cuts, which further facilitated the use of fiscal stimulus packages.In the US, the economic effect of the pandemic was felt strongly on labor markets: employmentdropped sharply and wages were cut (Cajner et al. , 2020; Kurmann et al. , 2020). This weakeneddemand and inflation considerably. The negative business climate also deterred financial markets,with equity prices collapsing more strongly than in any previous crises triggered by infectiousdisease outbreaks (Baker et al. , 2020). Relatedly, US Treasury markets experienced a sharp sell-off, leading to spikes in long-term yields (Schrimpf et al. , 2020). The US Federal Reserve (Fed)responded with several measures including the opening of credit facilities to support malfunctioningmarkets and actions aimed at relieving cash-flow stress for small and medium-sized businesses, aswell as municipalities. The most prominent actions, however, were moving the policy rate backtowards the zero lower bound and resuming the monthly purchase of massive amounts of securities.This paper tries to give a first assessment of how successful the monetary easing in the USwas in stabilizing prices and providing stimulus to the economy. One concern when assessing theeffectiveness of policy responses in real-time is the low frequency nature of many macroeconomicaggregates (with most of them available on a monthly or quarterly frequency, at best). Even ifwe rely on monthly data we are left with only very few observations that we can use to infer theeffects of monetary policy during the COVID-19 crisis on several key quantities of interest forpolicy makers.For that purpose, we borrow strength from data which is available at higher frequencies. Thesetime series are often sampled at daily or weekly frequency and allow us to construct weekly meas-ures of industrial production, inflation and unemployment. This is achieved within a coherentmultivariate framework that allows for dynamic interactions between the macroeconomic and fin-ancial quantities considered. For an overview of the enacted policy measures, see ur proposed econometric framework is a mixed frequency vector autoregression (MF-VAR)which models all variables on a weekly frequency. Using a state space representation of the mul-tivariate system, we recast the lower frequency quantities in terms of a weekly component withmissings between monthly observed values. These missing observations are subsequently estimatedby taking into account the properties of the model and using the higher frequency time series dy-namically. Our model is then used to simulate the effects of monetary policy shocks. Using theseshocks we can compute weekly historical decompositions and perform counterfactual scenarios toinvestigate the effects the monetary policy measures had on the US economy.Our results indicate that without a monetary expansion, US economic activity would havebeen significantly lower. In other words, the US Fed, so far, has been successful in cushioning theeconomic consequences of the COVID-19 crisis. Positive effects on output growth are underpinnedby a rise in stock market returns, an easing of long-term financing conditions and a depreciationof the US dollar. By contrast, effects on inflation and the unemployment rate are statisticallyinsignificant.The remainder of this paper is structured as follows. Section 2 briefly describes the datasetand econometric model used while Section 3 shows the main results. In this section, we discuss thedynamic reactions to a monetary policy shock and discuss the historical decompositions. Finally,the last section briefly summarizes and concludes the paper.
2. EMPIRICAL FRAMEWORK2.1.
A Mixed Frequency VAR Model
As stated in the introductory section, one key issue with adequately assessing the impacts ofCOVID-19 related monetary policy measures is the extremely short time span of available data.To provide a timely estimate, one could focus on high frequency variables such as interest ratespreads or stock prices. But these are typically not of direct interest for policy makers. In policymaking circles, assessing the effects of monetary policy interventions on output, inflation and labormarkets is pertinent. Unfortunately, for all these variables we only have a handful of observations,rendering an adequate assessment of policy effectiveness difficult.As a solution, we propose pairing a panel of weekly indicators, contained in an M H -dimensionalvector y t,H , with monthly indicators stored in an M L -dimensional vector y t,L in a MF-VAR. Thesevectors run from t = 1 , . . . , T , with T denoting the number of weeks in our sample. FollowingSchorfheide and Song (2015), we assume that y t,H is a latent weekly measure of the low frequencyindicator. ne key objective is to infer y H,t to obtain weekly measures of the low frequency variables.This is achieved by defining y t = ( y t,L , y t,H ) (cid:48) , which is an M (= M H + M L )-dimensional vector,and assuming that it follows a VAR( P ) process: y t = A y t − + · · · + A p y t − p + ε t , ε t ∼ N ( M , Σ t ) (1)where A j ( j = 1 , . . . , M ) are M × M coefficient matrices associated with lags j = 1 , . . . , P . ε t is awhite noise Gaussian process with variance-covariance matrix Σ t that varies over time. To speedup computation and assume that the Covid-19 shock led to a sharp increase in the conditionalvariance of all elements in y t , we introduce a common stochastic volatility (CSV) model originallyproposed in Carriero et al. (2016). This implies that Σ t is driven by a scalar factor such that: Σ t = e h t × Σ . We assume that h t evolves according to an AR(1) process: h t = µ h + ρ h ( h t − − µ h ) + σ h v t , v t ∼ N (0 , . Here, µ h denotes the unconditional mean, ρ h the autoregressive parameter and σ h the error vari-ance. h t simply scales the time-invariant variance-covariance matrix Σ . This allows us to capturesudden common shifts in variances while leaving the contemporaneous relations unchanged overtime.Equation (1) can be cast in its companion form: z t = F z t − + η t , (2)with z t = ( y (cid:48) t , . . . , y (cid:48) t − P +1 ) (cid:48) and F being the K × K companion matrix (for K = P M ) with thefirst M rows given by ( A , . . . , A p ). The remaining rows are defined to return an identity such that y t − j = y t − j for j = 1 , . . . , P −
1. The first M elements of η t are equal to ε t while the remainingelements are equal to zero.The missing values in y t can be obtained by interpreting (2) as a state evolution equation thatprovides information on how the elements in z t (and thus y t ) are related over time. Followingmuch of the recent literature (Koop et al. , 2020a;b; Gefang et al. , 2020), we assume that thefour-week-average of y L,t which we denote by ˜ x L,t , is related to y L,t as follows:˜ x t,L = ( y L,t + y L,t − + y L,t − + y L,t − ) / . his equation states that we view ˜ x t,L as the (observed) average of the weekly latent indicators.Notice that this assumption implies that each month features exactly four weeks (and thus we dropfour weeks per year to arrive at 48 weeks). Define a selection matrix S Lt that equals an identitymatrix in time t only in the last week of a month while being equal to a zero matrix for the initialthree weeks, and Λ L is a matrix such that: x t,L = S L,t ˜ x t,L = S L,t Λ L z t . For the monthly indicators, we assume that the identity x t,H = y t,H holds if the dataset isbalanced. If some monthly values are missing, we introduce a separate selection matrix S M,t with x M,t = S M,t y M,t .Following Schorfheide and Song (2015), the observation equation that relates the observed tothe latent quantities is: x t = M t Λ z t . (3)Here, x t = ( x (cid:48) t,L , x (cid:48) t,H ) (cid:48) , M t is a selection matrix and Λ is composed of Λ L and appropriateselection vectors to single out the high frequency quantities in z t .We estimate the MF-VAR using Bayesian techniques. This implies that we need to specifysuitable priors on all parameters of the model. In this paper, we use the conjugate Minnesotaprior on the VAR coefficients which has also been used by Schorfheide and Song (2015). On theremaining model parameters (which comprise of the parameters of the state equation of h t and Σ ), we use a Beta prior on the autoregressive coefficient ρ h , a normally distributed prior on theunconditional mean µ h and a Gamma prior on σ h . Finally, we use an weakly informative inverseWishart prior on Σ . Estimation is carried out using the Markov chain Monte Carlo (MCMC)algorithm discussed in Schorfheide and Song (2015) and efficiently implemented in the R package mfbvar (Ankargren and Yang, 2019). Data
Our analysis focuses on the reaction of the consumer price index (
CPIAUCSL ), the unemploymentrate (
UNRATE ) and industrial production (
INDPRO ) to a monetary policy easing. All of these focalvariables are on a monthly frequency. Higher-frequency variables consist mainly of financial indic-ators. In particular, we include the money supply ( M2 ) as the policy variable, the five-year forwardinflation expectation rate ( T5YIFR ) to gauge market-based inflation expectations, the NASDAQcomposite indicator (
NASDAQCOM ), the
US dollar/euro foreign exchange rate ( DEXUSEU ) and the ASDAQCOM T5YIFR VIXCLS WGS10YRCPIAUCSL UNRATE INDPRO M2 DEXUSEU0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50−0.250.000.250.500.00.10.20.3−0.04−0.020.000.00.10.2−2−1012−0.075−0.050−0.0250.0000.025−0.015−0.010−0.0050.0000.0050.010−0.075−0.050−0.0250.0000.0250.00.51.0
Horizon I R F Fig. 1:
Impulse response functions to a one-standard deviation shock to M2 . Notes : Median response alongside the 90 percent posterior credible set. The red line marks zero. ten-year treasury constant maturity rate ( WGS10YR ). As measures of financial stress we rely on the
CBOE volatility index (VIX,
VIXCLS ).The sample period runs from the first week of 2011 to week 24 of 2020 (end of week: June 8,2020) and is taken from the FRED database of the Federal Reserve Bank of St. Louis (fred.stlouisfed.org).If the raw data for financial variables is on a higher frequency than weekly (that is, daily for
T5YIFR , NASDAQCOM , DEXUSEU , VIXCLS ), we take the arithmetic average over the respective weekdays. Allvariables enter the model as year-on-year differences.
3. SCENARIO AND COUNTERFACTUAL ANALYSIS
In this section we examine the effects of an expansion of the US money supply on output, inflation,the unemployment rate and several financial indicators. In what follows, we proceed in two steps.First, we look at the overall plausibility of our model by examining impulse response functions.For that purpose, we rely on a simple recursive identification scheme with ordering the monthlyvariables first, followed by M2 . Last, we put all other weekly indicators. Note that this simplerecursive scheme implies zero restrictions on the low-frequency variables. In particular, in ourapplication the Cholesky decomposition implies that there are no contemporaneous effects of thehigh-frequency indicators on inflation, output and the unemployment rate, an assumption withwhich most economist would agree upon.The results are depicted in Fig. 1 which shows the posterior median (solid line) along with90% credible intervals. The figure demonstrates that the expansionary shock to the money supply( M2 ) significantly drives up output growth and lowers the unemployment rate. These effects arerather persistent and take place with a lag. We do not find a significant upward effect on inflation,although we have included inflation expectations which in general should help mitigating the price uzzle (Castelnuovo and Surico, 2010) often encountered in empirical studies. This finding canbe explained by the time period under consideration, which was characterized by low interest andinflation rates. As regards financial variables, we see a a significant and persistent upward effect onequity returns, a front-loaded depreciation of the US dollar and a decrease of long-term yields. Alsothe VIX increases immediately, which could be related to the positive and pronounced shoot-up ofequity returns. Summing up, the mixed-frequency approach generates impulse response functionsthat are in line with predictions of the bulk of empirical studies on the effects of monetary policy.Next, we generate counterfactual scenarios. For that purpose, we construct historical decom-positions that explain deviations of time series from their trend by shocks to the equations in thesystem. Neutralizing shocks to money supply after the onset of the COVID-19 crisis thus yieldsa counterfactual scenario to answer the question how output growth, unemployment and inflationwould have evolved without the Fed having provided monetary stimulus.The results are depicted in Fig. 2. In the upper panels, we show the evolution of actual series(black thick lines) and responses under the counterfactual scenario (grey shaded area, dashed line)along with 90% credible intervals. Since high-frequency movements of low-frequency variables areestimated within the MF-VAR framework, we also depict credible intervals for the historical weeklyevolution of inflation, the unemployment rate and output growth (black thin lines).The results indicate that output growth would have been weaker without monetary policystimulus provided by the US Fed. This finding could be driven by the strong effect monetarypolicy exerted on financial variables: equity returns would have been considerably lower and long-term yields higher under the no-policy scenario. The analysis also suggests that monetary policytriggered a stronger depreciation of the exchange rate and hence a boost to external competitivenessof the US economy. By contrast, the counterfactuals show no significant effect on unemploymentand inflation. Considering the delayed response of unemployment discussed in the context of theimpulse response functions, this might be an artefact of the considered counterfactual period beingto short to detect effects of the expansion yet.To investigate the significance more systematically, the bottom panel of Fig. 2 presents thedifferences of the responses under the no-policy and the policy scenario along with 90% credibleintervals. That analysis corroborates the findings from above that monetary policy led to higheroutput growth, a pick up in equity returns and an easing in long-term financing conditions. It alsoled to a significantly lower value of the US dollar. ASDAQCOM T5YIFR VIXCLS WGS10YRCPIAUCSL UNRATE INDPRO M2 DEXUSEUW01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24−20−1000510152025−2.0−1.5−1.0−0.50.0−25−20−15−10−500100200300400051015−1.00−0.75−0.50−0.250.00−202−20020
Series (solid, black) and counterfactual (dashed, grey)
NASDAQCOM T5YIFR VIXCLS WGS10YRCPIAUCSL UNRATE INDPRO M2 DEXUSEUW01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24 W01 W08 W16 W24−25−20−15−10−50−15−10−50−0.50.00.51.0−6−4−20−50050−10−0.250.000.25−2−101−50−40−30−20−100
Differences
Fig. 2:
Counterfactual analysis based on setting identified shocks to M2 after the onset ofthe COVID-19 crisis to zero. Notes : Upper panel: The black solid lines depict the actual evolution of the series (alongside the 90 percentposterior credible set for monthly variables), the dashed line alongside the grey shaded area (90 percentposterior credible set) shows the counterfactual. Lower panel: Posterior of the differences between theactual and counterfactual scenario. The red line marks zero. . CLOSING REMARKS In this note, we gave a first empirical investigation of the effects of US monetary policy to stimulategrowth in response to COVID-19. For that purpose, we have estimated a MF-VAR on monthly andweekly data. This model allows us to estimate weekly measures of industrial production, inflationand the unemployment rate. We then simulate the effects of expansionary monetary policy andassess its effects on the endogenous variables in the model.The results suggest that the US Fed was successful in stimulating growth on the back of higherequity prices and more favorable long-term financing conditions. Also, monetary policy triggered adepreciation of the US dollar supporting external competitiveness of the US economy. By contrast,we do not find significant effects on unemployment and inflation, both variables that typically reactmore sluggishly to economic stimulus. EFERENCES
Ankargren S, and Yang Y (2019), “Mixed-Frequency Bayesian VAR Models in R: the mfbvarpackage,” .
Arslan Y, Drehmann M, and Hofmann B (2020), “Central bank bond purchases in emergingmarket economies,” Technical Report 20, BIS Bulletin.
Baker S, Bloom N, Davis S, Kost K, Sammon M, and Viratyosin T (2020), “The unpre-cedented stock market impact of COVID-19,” Technical Report 26945, NBER working paperseries.
Cajner T, Crane L, Decker RA, Grigsby J, Hamins-Puertolas A, Hurst E, Kurz C,and Yildirmaz A (2020), “The U.S. labor market during the beginning of the pandemic,”Technical Report 27159, NBER working paper series.
Carriero A, Clark TE, and Marcellino M (2016), “Common drifting volatility in largeBayesian VARs,”
Journal of Business & Economic Statistics (3), 375–390. Castelnuovo E, and Surico P (2010), “Monetary Policy, Inflation Expectations and The PricePuzzle*,”
The Economic Journal (549), 1262–1283.
Gefang D, Koop G, and Poon A (2020), “Computationally efficient inference in large Bayesianmixed frequency VARs,”
Economics Letters
Hartley J, and Rebucci A (2020), “An event study of COVID-19 central bank quantitativeeasing in advanced and emerging economies,” Technical Report 27339, NBER working paperseries.
Koop G, McIntyre S, and Mitchell J (2020a), “UK regional nowcasting using a mixedfrequency vector auto-regressive model with entropic tilting,”
Journal of the Royal StatisticalSociety: Series A (Statistics in Society) (1), 91–119.
Koop G, McIntyre S, Mitchell J, and Poon A (2020b), “Regional output growth in theUnited Kingdom: More timely and higher frequency estimates from 1970,”
Journal of AppliedEconometrics (2), 176–197. Kurmann A, Lale E, and Ta L (2020), “The Impact of COVID-19 on U.S. Employment andHours: Real-Time Estimates With Homebase Data,” Technical Report mimeo, LeBow Collegeof Business, Drexel University.
McKibbin W, and Fernando R (2020), “The Global Macroeconomic Impacts of COVID-19:Seven Scenarios,” Technical Report 19/2020, CAMA working paper series.
Schorfheide F, and Song D (2015), “Real-time forecasting with a mixed-frequency VAR,”
Journal of Business & Economic Statistics (3), 366–380. Schrimpf A, Shin H, and Sushko V (2020), “Leverage and margin spirals in fixed incomemarkets during the Covid-19 crisis,” Technical Report 2, BIS Bulletin.(2020), “Leverage and margin spirals in fixed incomemarkets during the Covid-19 crisis,” Technical Report 2, BIS Bulletin.