Monetary Policy Shocks

We have a new paper (https://lnkd.in/eQHt8att) on the effects of conventional monetary policy on output and prices, based on a multi-year data collection effort. We collect and analyse 146,463 point estimates and confidence bands from 4,871 impulse-response functions reported in 409 primary studies (joint work with Matthias Enzinger, Sebastian Gechert, Franz Prante and Daniel Romero). Our main finding: We show that the results reported in the literature on how output and prices respond to conventional monetary policy shocks are plagued by p-hacking and publication bias, leading to inflated effect sizes. p-hacking is the preference for statistically significant results. Publication bias includes p-hacking but also a tendency to prefer large effect sizes or to conform to theoretical expectations and seminal publications. We document a robust pattern of selective reporting of statistically significant dampening effects of contractionary monetary policy shocks on output and prices, in particular at the most relevant response horizons. The naïve average of all IRFs points to substantial contractionary effects of a 100 basis points interest rate hike on output (with a peak effect of −1 percent after around 2 years) and the price level (with a peak effect of −0.75 percent after 4 to 5 years). However, such a conclusion would be misleading since the literature suffers from substantial publication bias according to a series of established tests. When we correct for this bias, the resulting range of corrected IRFs points to substantially smaller dampening effects of contractionary monetary policy shocks on the economy. The strongest corrections would be consistent with zero effects and the mildest corrections would imply a peak effect of −0.7 percent for output and −0.5 percent for the price level. The mean IRF beyond publication bias for output peaks at −0.25 percent after 1 to 2 years and for the price level at −0.15 percent after 4 years. Bias corrections reduce effect sizes by half or more. Our findings suggest that the power of conventional monetary policy to steer prices and the business cycle may have been overstated in the past, based on a simple average assessment of the empirical literature, seminal empirical studies in leading journals, the predictions of standard New Keynesian models and a summary given by a leading AI. We also investigate how study and estimation characteristics such as identification strategies, samples, author affiliations, and journal ranking are related to the variation of reported effect sizes. Shock identification choices and publication characteristics correlate with effect sizes but are quantitatively less important than publication bias. Link to our paper (comments welcome): https://lnkd.in/eQHt8att The documentation and replication files are available via: https://lnkd.in/e-wJG8cj Pre-registration: https://osf.io/cduq4

𝐎𝐩𝐞𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 𝐁𝐥𝐚𝐜𝐤 𝐁𝐨𝐱 𝐨𝐟 𝐋𝐨𝐜𝐚𝐥 𝐏𝐫𝐨𝐣𝐞𝐜𝐭𝐢𝐨𝐧𝐬 is my new working paper with Karin Klieber. (paper: https://lnkd.in/e8Z29EeJ, deck: https://lnkd.in/eGEUCJsr). Local projections are now routinely used to estimate impulse response functions (IRFs) in empirical macroeconomics. Yet in many ways, they are black boxes. The mechanisms behind the curves are often unclear. Perhaps most importantly: do the episodes we think are driving the causal effect estimates actually 𝘥𝘰 𝘴𝘰? Are those numerous enough to be confident about external validity? We introduce a 𝗻𝗲𝘄 𝗱𝗲𝗰𝗼𝗺𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻 of LP estimates that makes their historical foundation transparent. Each estimate is expressed as a sum of contributions from individual events—weights times observed responses—revealing how specific periods shape the average effect. We plot cumulative contributions over time, which converge by construction to the LP estimate at horizon 𝘩 (see figure below). We also visualize the weights as a time series and introduce LP concentration statistics to summarize how broad or narrow the evidence base is. 𝗧𝗵𝗲 𝗺𝗲𝗮𝗻𝗶𝗻𝗴(𝘀) 𝗼𝗳 𝘄𝗲𝗶𝗴𝗵𝘁𝘀. In linear local projections estimated via least squares, the weight series admits two interpretations. First, by a variation of the Frisch-Waugh-Lovell theorem, the weights can be viewed as purified and standardized shocks. Second, they are proximity scores, measuring the similarity between the projected intervention and past interventions in the sample. 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴. The proximity-based interpretation of weights extends beyond linear models, as many machine learning (ML) algorithms generate IRFs that, while nonlinear in regressors, remain linear in the outcome variable. Proximity weights offer a common scale for comparing linear and ML-based LPs. 𝗘𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹𝗹𝘆, we apply the method to various domains: ▶ 𝘔𝘰𝘯𝘦𝘵𝘢𝘳𝘺 𝘱𝘰𝘭𝘪𝘤𝘺: We find that Cholesky VAR shocks produce a price puzzle by misreading stagflation episodes from the 1970s. Romer and Romer (2004) shocks get the IRF sign right by offsetting this misinterpretation with a long stretch of monetary loosening shocks coinciding with the peak of late 1970s inflation. Random Forest refines the portrait by assigning elevated proximity weights only to well-known episodes of political interference with the Fed. ▶ 𝘍𝘪𝘴𝘤𝘢𝘭 𝘱𝘰𝘭𝘪𝘤𝘺: Ramey and Zubairy’s (2018) state-dependent fiscal multipliers in recessions are driven almost entirely by a single event: World War II. ▶ 𝘊𝘭𝘪𝘮𝘢𝘵𝘦 𝘴𝘩𝘰𝘤𝘬𝘴: The long-run GDP impact of global temperature shocks in Bilal and Känzig (2024) appears fragile, primarily driven by the pairing of a single 1960s volcanic eruption and exceptional post-war growth. École des sciences de la gestion (ESG UQAM) Oesterreichische Nationalbank, European Central Bank #econometrics #causalinference #economics #machinelearning

Household savings rose above trend in many developed countries after the onset of COVID-19. But, how these ``excess savings" affect the transmission of monetary policy? Julio Ortiz, Nils Goernemann, and I answer this question empirically and theoretically. Using a panel of euro-area economies and high-frequency monetary shocks, we document that household excess savings dampen the effects of monetary policy on economic activity and inflation, with this effect being heightened during the pandemic period. To rationalize our empirical findings, we build a New Keynesian model in which savings provides insurance against consumption declines due to unemployment. https://lnkd.in/dqDCMmmp

JUST PUBLISHED: This paper investigates the effects of monetary policy on the credit card spending on different sectors. https://lnkd.in/eqYqAXbd The investigation is based on a structural vector autoregression model, where sector-specific real credit card spending data (adjusted for inflation) representing an overall country, Türkiye, are used. The empirical results (in the long run) suggest that a positive shock to the monetary policy rate reduces real credit card spending in cars, health, insurance, and shopping in a statistically significant way, whereas it increases real credit card spending on airlines and travel. Monetary policy shocks contribute to the volatility of credit card spending by up to 36% for insurance, 26% for markets and shopping centers, and 22% for travel sectors, whereas this contribution is only about 3% for contractor services and about 4% for car rentals, jewelry, and casino sectors. It is implied that there are uneven effects of monetary policy across sector-specific credit card spendings. These results are robust to the consideration of changes in unemployment rate, inflation rate, nominal effective exchange rate, and the number of credit card transactions as well as alternative model specifications with different numbers of lags, different variables, and different estimation strategies. Important suggestions follow for monetary, fiscal, and macroprudential policies to mitigate the uneven effects of monetary policy across sectors.