My research is also posted at IDEAS, SSRN, GoogleScholar, ResearcherID, ORCID.
Codes can also be found at GitHub. Below you also find GitHub links to individual projects.
Word cloud generated with Scholar Goggler.
with Todd E. Clark (Johns Hopkins University)
Abstract: Professional forecasts play a crucial role in shaping expectations and guiding policy decisions. Among these, the Survey of Professional Forecasters (SPF) provides a uniquely detailed and long-standing record of macroeconomic expectations. In their recent work, Clark, Ganics, and Mertens (2025, forthcoming REStat; hereafter CGM) developed a state-space framework that reconstructs a rich, SPF-consistent term structure of expectations and forecast uncertainty by integrating the SPF’s fixed-horizon and fixed-event forecasts. This approach allows for the estimation of latent quarterly expectations at long horizons, even beyond the ragged edge of the SPF’s published data, while accounting for potential inefficiencies and time-varying volatility. Importantly, the CGM model also accommodates deviations from full-information rational expectations by allowing for serially correlated forecast updates, in the spirit of Coibion and Gorodnichenko (2015, AER) and Angeletos, Huo, and Sastry (2021, NBER Macro Annual). This feature captures the possibility of imperfect or delayed information processing by forecasters and proves critical for our analysis of how structural shocks are transmitted through expectations.
Building on the CGM framework, this project investigates the sources of variation in the SPF's term structure of expectations. Specifically, we examine the extent to which updates in SPF-consistent forecasts can be explained by structural shocks related to uncertainty, monetary policy, and geopolitical developments. In doing so, we project the shifts in expectations across the term structure onto exogenous shock series, thereby providing a structural interpretation of the dynamics captured in the SPF forecasts.Our empirical strategy involves two key steps. First, we extract the time series of SPF-consistent term structures of expectations and their updates using the CGM model. These updates, which reflect revisions to forecasters’ beliefs about future macroeconomic outcomes, serve as our dependent variables. Second, we regress these updates onto a set of structural shock proxies. These include widely used measures of economic and policy uncertainty (e.g., the Economic Policy Uncertainty Index), high-frequency monetary policy shocks derived from surprise components around FOMC announcements, and indices capturing geopolitical risk (e.g., Caldara and Iacoviello’s Geopolitical Risk Index; 2022, AER).
Our analysis proceeds across multiple dimensions of the SPF: real GDP growth, inflation (both CPI and GDP deflator), and the unemployment rate. For each variable, we estimate the responsiveness of short-, medium-, and long-term expectations to contemporaneous structural shocks. This decomposition allows us to quantify the horizon-specific impact of each shock, revealing, for instance, whether monetary policy surprises primarily influence near-term forecasts or have persistent effects extending further into the future. Overall, our results contribute to a deeper understanding of how structural shocks are embedded into professional forecasters’ expectations, and how these effects vary across horizons. By marrying the granular SPF term structures constructed by CGM with economically meaningful structural shock measures, this work enhances the interpretability of survey-based forecasts and provides insights into the transmission channels of macroeconomic shocks through expectations.
with Todd E. Clark (Johns Hopkins University)
Abstract: This paper develops and applies analytical solutions for entropic tilting of predictive distributions to match the histogram forecasts provided in the U.S. Survey of Professional Forecasters (SPF). We focus on tilting to histogram probabilities directly---rather than to moments of fitted distributions---and derive analytic solutions to the tilting problem. We evaluate the method using updated SPF data and compare forecast accuracy with and without tilting. Our results suggest that direct tilting to histograms improves model calibration and predictive accuracy, particularly since the Global Financial Crisis.
with Giovanni Nicolò (Federal Reserve Board)
Abstract: The post-pandemic increase in long-term Treasury yields raises questions about the level of the natural rate of interest, r*. Policymakers typically assess the stance of monetary policy relative to the natural rate of interest. However, r* is unobservable, and policymakers rely on real-time estimates generated from data on output, inflation, and other indicators. We show that incorporating those estimates into the setting of monetary policy alters the systematic response to endogenous outcomes in ways that can violate the Taylor principle. Even when the policy rule is ``active'' in response to policymakers' real-time estimates of economic conditions, the effective policy rule---which arises from the effects of data on policymakers' real-time estimates---can violate the Taylor principle. As a result, multiple equilibria may exist, and non-fundamental sunspot shocks can influence macroeconomic outcomes. A policy rule that responds only to observed outcomes, rather than estimates of latent variables like r*, would avoid such multiplicity.
We consider a simple New Keynesian model in which policymakers observe aggregate output and inflation but not the underlying shocks to potential output and the supply side of the economy. We model policymakers' estimates of r* as rational expectations that are consistent with equilibrium dynamics. While the true r* is driven only by shocks to potential output, policymakers' estimates of r* depend on observed data, which in turn leads to violations of the Taylor principle and multiple equilibria. In this case, the resulting equilibrium is not unique. In addition, non-fundamental belief shocks distort policymakers’ ability to perfectly infer structural shocks---leading to imperfect inference about r*---and can contribute significantly to fluctuations in macroeconomic outcomes.
We provide a unified framework in which policymakers believe that monetary policy is `active,'' while the policy response to actual data (instead of policymakers' real-time estimates) is consistent with multiple equilibrium outcomes. In this case, sunspot shocks emerge as important drivers of macroeconomic dynamics. Our framework integrates empirical results that find an active reaction function of monetary policy in terms of real-time estimates (Orphanides, 2001, AER), while also recognizing a role for non-fundamental sunspot shocks as drivers of inflation and output due to an inactive policy in terms of actual data (Clarida, Galì, and Gertler, 2000, QJE; Lubik and Schorfheide, 2004, AER). Extending earlier work by Lubik, Matthes, and Mertens (2022, RED), we derive analytic solutions for models in which imperfect information about r* faced by policymakers leads to equilibrium outcomes featuring sunspot-driven fluctuations. Our solution framework facilitates the estimation of such models.
Abstract: We present trend estimates of the natural rate of interest in the US and Europe that reflect time-varying uncertainty about economic shocks. Time-varying uncertainty about the relative importance of different shocks crucially affects natural-rate estimates. The natural rate of interest is not directly observed and can only be inferred from macroeconomic aggregates. When the role of different shocks changes over time, natural-rate estimates should also react differently to new data. A model that assumes constant uncertainty about the relative importance of different shocks will instead assign a constant signal-to-noise ratio to incoming data, which can lead to substantial revisions in natural-rate estimates. Moreover, while nominal interest rate data is a key input for estimating the natural rate, their dynamics have been notably affected over the last decade by a binding effective lower bound in many economies.
To address these issues, we extend earlier work by Johannsen and Mertens (2021, JMCB), which provided natural-rate estimates for the US that also reflected the effective lower bound. We adapt the model to euro-area data and generalize the treatment of time-varying volatility in shocks to trend and cyclical components. We also introduce a novel treatment of extreme data points, such as those observed during the COVID-19 pandemic, as stochastic outliers with random timing and magnitude. Our model filters the output gap from real GDP instead of relying on external estimates and implements a unified factor for cyclical volatility, which is robust to variable reordering. We highlight the importance of using extended data samples (including cross-country information) for more reliable filtering of natural rates from the data. A joint model of natural rates in the Euro area and the US is the subject of ongoing work.
with Todd E. Clark (FRB Cleveland), and Gergely Ganics (Banco de España)
Federal Reserve Bank of Cleveland WP: https://doi.org/10.26509/frbc-wp-202237
online appendix: pdf
replication code: https://github.com/elmarmertens/ClarkGanicsMertensSPFfancharts
Abstract: This paper presents a new approach to combining the information in point and density forecasts from the Survey of Professional Forecasters (SPF) and assesses the incremental value of the density forecasts. Our starting point is a model, developed in companion work, that constructs quarterly term structures of expectations and uncertainty from SPF point forecasts for quarterly fixed horizons and annual fixed events. We then employ entropic tilting to bring the density forecast information contained in the SPF's probability bins to bear on the model estimates. In a novel application of entropic tilting, we let the resulting predictive densities exactly replicate the SPF's probability bins. Our empirical analysis of SPF forecasts of GDP growth shows that tilting to the SPF's probability bins can visibly affect our model-based predictive distributions. Yet in historical evaluations, tilting does not offer consistent benefits to forecast accuracy relative to the model-based densities that are centered on the SPF's point forecasts and reflect the historical behavior of SPF forecast errors. That said, there can be periods in which tilting to the bin information helps forecast accuracy.
with Andrea Carriero (Queen Mary University of London), Todd Clark (Federal Reserve Bank of Cleveland), Massimiliano Marcellino (Bocconi, IGIER and CEPR)
Cleveland Fed WP (Revised Jan 2022)
Abstract:
We measure the effects of the COVID-19 outbreak on macroeconomic and financial uncertainty, and we assess the consequences of the latter for key economic variables. We use a large, heteroskedastic vector autoregression (VAR) in which the error volatilities share two common factors, interpreted as macro and financial uncertainty, in addition to idiosyncratic components. Macro and financial uncertainty are allowed to contemporaneously affect the macroeconomy and financial conditions, with changes in the common component of the volatilities providing contemporaneous identifying information on uncertainty. We also consider an extended version of the model, based on a latent state approach to accommodating outliers in volatility, to reduce the influence of extreme observations from the COVID period. The estimates we obtain yield very large increases in macroeconomic and financial uncertainty over the course of the COVID-19 period. These increases have contributed to the downturn in economic and financial conditions, but with both models, the contributions of uncertainty are small compared to the overall movements in many macroeconomic and financial indicators. That implies that the downturn is driven more by other dimensions of the COVID crisis than shocks to aggregate uncertainty (as measured by our method).
Abstract: Real-time estimates of the Output Gap — defined as the cyclical component of GDP — have previously been shown to be unreliable, since they are subject to large revisions when new data comes in. However, this result has so far only been derived for constant parameter models. This paper uses statistical models where the volatility of shocks to trend and cycle can vary over time. In this case, output gap estimates derived from data vintages going back to the 1970s are much closer to “final” estimates derived from all available sample data. The final estimates not only fall mostly within the credible intervals generated by the real-time data. When generated from a model with stochastic volatility, these credible sets are also tighter, at least over low-volatility periods.
Abstract:
This papers presents general methods to compute optimal commitment and discretion policies, when a policymaker is better informed about the realization of some shocks than the public. In this situation, public beliefs about the hidden information emerge as additional state variables, managed by the policymaker.
Under commitment, policy is additive in two components: The optimal policy, as if the government shared the public's information set and the systematic manipulation of that information set. Even under discretion, belief management imparts history dependence.
Illustrated in a New Keynesian economy with time-varying output targets of the policymaker, belief management improves outcomes compared to symmetric information. At the margin, the policymaker tries to be intransparent about policy objectives by engineering disturbances which lower public beliefs about the persistence of output targets.