Latest draft: [ coming soon ] 

Coauthor(s): Andrzej Kociecki and Christian Matthes 

Executive summary: We extend our work on SVAR models with t-distributed shocks to factor models

Latest draft: [ coming soon ] 

Coauthor(s): Annika Camehl and Malte Rieth 

Executive summary: We study how vaccination rates affected the evolution of the Covid-19 pandemic, using a large Panel Threshold VAR model with shocks identified via sign restrictions

Latest draft: [ coming soon ] 

Coauthor(s): Anastasia Allayioti and Francesca Monti 

Executive summary: We study how the effects of monetary policy change depending on how much agents fear high levels of inflation or deflation

Latest draft: paper and online appendix

Coauthor(s): Davide Brignone 

Executive summary: This paper shows how to use the structural form of a SVAR to form a narrative/interpretation of forecast errors and forecast revisions, hence showing a link between reduced form and structural VARs

Abstract: This paper shows how the structural representation of a vector autoregressive model can support forecasting. We offer a unified framework between reducedform forecast and structural analysis, and describe how the use of the latter can help form a narrative of two reduced-form objects of real-time forecasting: the forecast errors made relative to the outturn of the data, and the revisions of the full forecast made when new data become available. We conduct a real-time exercise on the UK focusing on the inflation surge that followed the pandemic. We show that the inflation forecast was revised up not only due to contractionary supply-side shocks, but also due to a mix of expansionary demand-side shocks and a change in the underlying unconditional mean

Latest draft: paper, online appendix

Coauthor(s): Andrzej Kociecki and Christian Matthes 

Executive summary: This paper shows that a simple reparameterization allows for the development of the first Gibbs sampler for SVAR models with t-distributed shocks. This makes the analysis computationally more tractable compared to existing models, and  paves the way for a larger model specificaiton within this class of models. We apply our method to  US data and show that there is no such thing as a Main Business Cycle shock that explains most of the US business cycle

Abstract: Non-Gaussian shocks can identify causal effects in structural VARs, but existing approaches face both conceptual and computational challenges. Conceptually, shocks are only identified up to sign and permutation, requiring careful normalization. Computationally, standard algorithms scale poorly, limiting applications to small models. We develop a Bayesian framework that resolves both issues and enables estimation in high-dimensional settings. Our approach applies to all common departures from Gaussianity - such as heavy tails and stochastic volatility - and accommodates external instruments to sharpen identification. This expands the scope of non-Gaussian SVARs to the large-scale models increasingly used in economics

Additional info: the paper previously circulated with the title "A Non-Gaussian GDP Anatomy"

Latest draft: paper, online appendix

Coauthor(s): Fabio Canova and Andrzej Kociecki

Executive summary: The paper develops a new prior for SVARs that allows introducing priors on the shape of the impulse  horizons, while still working with the convenient Normal distribution. The method is compatible with standard identification strategies and is an alternative to the Minnesota prior

Abstract: We design a prior for VAR coefficients that allows for flexible, non-dogmatic beliefs on the shape and the timing of the structural impulse responses. We achieve this with a particular setting of the moments of a Normal distribution. Posterior computations are no more demanding than with existing specifications; yet, the methodology provides shrinkage on impulse responses. We study the transmission of monetary policy shocks. Replacing standard priors with a prior that assumes that monetary policy shocks generate temporary but persistent effects leads to a hump-shaped posterior response of industrial production. The trough occurs eight months after the shock

Final draft: paper, online appendix

Journal version: [ coming soon ] 

Coauthor(s): Haroon Mumtaz 

Executive summary: The paper proposes a nonparametric version of Local Projections by applying Bayesian Additive Regression Trees model to LPs. Develops the model, shows 3 simulation exercises in Monte Carlo. Applies the method to fiscal shocks and to financial shocks

Abstract: This paper introduces a flexible local projection that generalises the model by Jorda (2005) to a non-parametric setting using Bayesian Additive Regression Trees. Monte Carlo experiments show that the BART-LP model is able to capture multiple types of non-linearities in the impulse responses within the same framework. Our first application shows that adverse financial shocks generate effects on the economy that can increase more than proportionately in the size of the shock, potentially explaining the large economic effects of strong financial disruptions. Our second application shows that before Covid-19 monetary policy shocks were more effective during economic expansions if they were contractionary shocks, while expansionary shocks generate symmetric effects over the business cycle.

Codes: [ coming soon ] 

Final draft: paper, online appendix, some derivations

Journal version: The Econometrics Journal  (2024), 27(3), pp. 343-361. Link 

Coauthor(s): Martin Bruns 

Executive summary: The paper proposes a way of estimating the parameters gamma and c in Smooth Transition SVAR models. It hence offers an alternative to calibrating such parameters, since we show that they have a strong effect on results, hence they are best estimated than calibrated

Abstract: We develop a tractable way of estimating the parameters ruling the nonlinearity in the popular smooth transition VAR model, and identify structural shocks using external instruments. This jointly offers an alternative to the option of identifying shocks recursively and calibrating key parameters. In an illustration, we show that monetary policy shocks generate larger effects on economic activity during economic expansions compared to economic recessions. We then document that calibrating rather than estimating the parameters ruling the nonlinearity of the model can lead to values for which the key results are lost. This suggests caution in the calibration of these parameters

Codes: See dedicated page

Final draft: paper, online appendix

Journal version: Quantitative Economics  (2023), 14(4), pp. 1221-1250. Link 

Coauthor(s): Martin Bruns 

Executive summary: We propose a methodology that can handle a wide range of prior beliefs on contemporaneous impulse responses, while ensuring a tractable posterior distribution. We combine sign restrictions with information on the scaling of the variables and show that this delivers sharper inference. We show an application to the oil market and find that oil supply shocks do matter for oil price dynamics

Abstract: We develop an importance sampler for sign restricted Bayesian structural vector autoregressive models. The algorithm nests as a special case the sampler associated with the popular Normal inverse Wishart Uniform prior, while allowing to move beyond such prior in medium sized models. We then propose a prior on contemporaneous impulse responses that provides flexibility on the magnitude and shape of the impact responses. We illustrate the quantitative relevance of the choice of the prior in an application to US monetary policy shocks. We find that the real effects of monetary policy shocks are stronger under our proposed prior than in the Normal inverse Wishart Uniform setup

Codes: replication package

Additional info: 

• This project received  financial support from Michele's Marie Skłodowska-Curie grant agreement number 744010, which is part of the European Union's Horizon 2020 research and innovation program

• The paper previously circulated under the title "Bayesian Structural VAR models: a new approach for prior beliefs on impulse responses"

• List of papers using SVAR models

Final draft: paper, online appendix

Journal version: Journal of the European Economic Association (2020), 18(1), pp. 202–231. Link 

Coauthor(s): Michael Hachula and Malte Rieth

Executive summary: Uses a VAR for the euro area to study the effects and side effects of an exogenous ECB's unconventional monetary expansion. Studies in details how the responses differ across Euro Area members

Abstract: We study the macroeconomic effects of unconventional monetary policy in the euro area using structural vector autoregressions, identified with external instruments. The instruments are based on the common unexpected variation in euro area sovereign yields for different maturities on policy announcement days. We first show that expansionary monetary surprises are effective at lowering public and private interest rates and increasing economic activity, consumer prices, and inflation expectations. We then document that the shocks lead to a rise in primary public expenditures and a widening of internal trade balances

Codes: replication package, instruments for monetary policy shocks

Final draft: paper (journal, open source), online appendix

Journal version: International Journal of Central Banking   (2018), 14(4), pp. 327-358

Coauthor(s): [ none ]

Executive summary: Studies empirically and theoretically the general equilibrium implications of the risk-taking channel of monetary policy  

Abstract: This paper uses a structural VAR model to study the effect of monetary policy on the delinquency rate of business loans and consumer credit. The VAR is identified using, jointly, several external instruments that reflect different approaches from the literature. Delinquency rates, defined as the rate of loans with overdue repayments relative to total loans, are found to decrease in response to an exogenous monetary expansion. The results are consistent with a general equilibrium effect formalized in the paper using a standard model of optimal defaults. According to both the theoretical model and the reported empirical evidence, the decrease in defaults is driven by the fact that monetary expansions increase aggregate demand and push up profits and income, thereby improving the repayment possibility of borrowers

Final draft: paper, online appendix

Journal version: The Economic Journal   (2018), 128(616), pp. 3266-3284. Link 

Coauthor(s): Maximilian Podstawski

Executive summary: Uses variations in the price of gold around uncertainty-related events to compute a proxy for uncertainty shocks. Then identifies uncertainty and news shocks in a proxy SVAR and compares results to the recursive identification. Argues that uncertainty shocks identified recursively look more like news shocks

Abstract: We propose an instrument to identify uncertainty shocks in a proxy structural vector autoregressive model (SVAR). The instrument equals the variations in the price of gold around events associated with unexpected changes in uncertainty. These variations correlate with uncertainty shocks because gold is perceived as a safe haven asset. To control for news‐related effects associated with the events we identify uncertainty and news shocks jointly, developing a set‐identified proxy SVAR. We find that the popular recursive approach underestimates the effects of uncertainty shocks and delivers responses for economic activity and monetary policy that have more in common with news shocks than with uncertainty shocks

Codes: See dedicated page for updated instrument, replication package of published version

Journal version: American Journal of Tropical Medicine and Hygiene (2013), 88(6), pp. 1070-78. Link 

Coauthor(s): Alexandra Hiscox, Angela Kaye, Khamsing Vongphayloth, Ian Banks, Phasouk Khammanithong, Pany Sananikhom, Surinder Kaul, Nigel Hill, Steven Lindsay and Paul Brey

Abstract: We assessed risk factors for vectors of dengue and chikungunya viruses near a new hydroelectric project, Nam Theun 2, in Laos. Immature stages of Aedes aegypti were found only in sites within 40 km of the urban provincial capital, but Aedes albopictus was found throughout. Aedes aegypti pupae were most common in water storage jars (odds ratio [OR] = 4.72) and tires (OR = 2.99), and Ae. albopictus pupae were associated with tires in 2009 (OR = 10.87) and drums, tires, and jars in 2010 (drums OR = 3.05; tires OR = 3.45, jars OR = 6.59). Compared with water storage vessels, containers used for hygiene, cooking, and drinking were 80%less likely to harbor Ae. albopictus pupae in 2010 (OR = 0.20), and discarded waste was associated with a 3.64 increased odds of infestation. Vector control efforts should focus on source reduction of water storage containers, particularly concrete jars and tires.

Quantifying the macroeconomic effects of the uncertainty shock of the Brexit vote

with Malte Rieth

Bayesian assessment of sign restrictions in VAR models

pdf | appendix. Uses Bayesian estimation to study how strongly the data support sign restrictions for a VAR as candidate identifying restrictions. First version July 2015, this version July 2017. This paper initially circulated under the titles "Assessing identifying restrictions in SVAR models" and "Bayesian model comparison for sign restrictions in SVAR models".

Monetary policy, leverage, and default  

pdf | awards. Uses the BGG (1999) model to study leverage and revenue effects after a monetary expansion. 

Counter-cyclical defaults in costly state verification models   

pdf. Proposes a way of generating counter-cyclical defaults in CSV model + calibrates the non-linear model numerically.