Research

Job Market Paper

• Harnessing Machine Learning for Real-Time Inflation Nowcasting. [De Nederlandsche Bank Working Paper; SUERF Policy Brief]
  With Aishameriane Schmidt (Erasmus University Rotterdam, Tinbergen Institute and De Nederlandsche Bank) and Guilherme Valle Moura (Federal University of Santa Catarina).

  • Abstract: We analyze the predictive ability of machine learning methods for weekly inflation nowcasts using high-frequency macro-financial indicators and a survey of professional forecasters. Within an unrestricted mixed-frequency framework, we provide guidelines to improve inflation nowcasts upon experts' expectations. First, we find that LASSO-based variable selection is crucial for effective inflation nowcasting, outperforming tree-based methods. Second, timely price data and survey expectations are key to discipline model-based nowcasts, particularly during the post-COVID-19 inflation surge. Third, we show that aligning the model specification with real-time data releases and leveraging the most recent high-frequency signals significantly boosts predictive accuracy at longer nowcast horizons.

  •  Online Appendix with  supplementary results.

  •  R codes are available on this GitHub repository.
     

Work in Progress

• Nowcasting Consumer Price Inflation Using High-Frequency Scanner Data: Evidence from Germany. [Deutsche Bundesbank discussion paper; ECB Working Paper; SUERF Policy Brief]
  With Günter W. Beck (University of Siegen), Kai Carstensen (Kiel University), Jan-Oliver Menz (Deutsche Bundesbank) and Elisabeth Wieland (Deutsche Bundesbank).

  • Abstract: We study how millions of highly granular and weekly household scanner data combined with novel machine learning techniques can help to improve the nowcast of monthly German inflation in real-time. Our nowcasting exercise targets three hierarchy levels of the official consumer price index. First, we construct a large set of weekly scanner-based price indices at the lowest aggregation level underlying official German inflation, such as those of butter and coffee beans. We show that these indices track their official counterparts extremely well. Within a mixed-frequency modeling framework, we also demonstrate that these scanner-based price indices improve inflation nowcasts at this very narrow level, notably after the first seven days of a month. Second, we apply shrinkage estimators to exploit the large set of scanner-based price indices in nowcasting product groups such as processed and unprocessed food. This yields substantial predictive gains compared to a time series benchmark model. Finally, we nowcast headline inflation. Adding high-frequency information on energy and travel services, we construct highly competitive nowcasting models that are on par with, or even outperform, survey-based inflation expectations that are notoriously difficult to beat.

Publications in Research and Policy-Based Columns

•  Forecasting HICP package holidays with forward-looking booking data. Deutsche Bundesbank Technical Paper 04/2024.
  With Elisabeth Wieland and Patrick Schwind.

  • Short summary: Forecasting package holiday prices, a key driver of inflation in Germany, presents challenges due to strong seasonality, volatility, and methodological breaks. We propose a forecasting strategy that leverages a forward-looking price indicator based on high-frequency booking data providing early signals about the underlying trend of the target series. We show that accurate forecasts are obtained with a modeling strategy tailored to the seasonally adjusted target series, alongside precise projections of the future seasonal component. Finally, augmenting the forecasting model with the forward-looking price indicator yields considerable gains that increase with the forecast horizon.

• Wie Haushaltsscannerdaten bei der Inflationsprognose helfen. Deutsche Bundesbank Research Brief, 30 January 2024.
  With Günter W. Beck, Kai Carstensen, Jan-Oliver Menz and Elisabeth Wieland.

  • Short summary: Die Prognose der Inflationsrate für den jeweils laufenden Monat („Nowcasting“) ist für Zentralbanken und Marktteilnehmer von hoher Bedeutung, insbesondere in turbulenten Zeiten. In einer neuen Studie untersuchen Forscherinnen und Forscher, ob sich der Nowcast der monatlichen Inflationsrate in Deutschland mithilfe Millionen granularer, wöchentlicher Scannerdaten von privaten Haushalten und Techniken des Maschinellen Lernens (ML) verbessern lässt.

• Real-time food price inflation in Germany in light of the Russian invasion of Ukraine. VoxEU.org, 24 June 2022.
  With Günter W. Beck, Kai Carstensen, Jan-Oliver Menz and Elisabeth Wieland.

  • Short summary: Unprecedented events, including the Covid-19 pandemic and the Russian invasion of Ukraine, have boosted the demand for real-time measures of consumer price inflation. This column demonstrates the usefulness of high-frequency retail scanner data to document the dynamics of both prices and quantities of German food products in the aftermath of the invasion of Ukraine. Immediate price and quantity responses can be detected and quantified and their subsequent dynamics can be tracked on a continuous base. Both prices and quantities rose for oil and flour due to stockpiling behavior by consumers, whereas only prices rose for other goods. 

Working Papers

•  Bond portfolio optimization in turbulent times: A dynamic Nelson-Siegel approach with Wishart stochastic volatility. 

  • Short summary: Modeling and forecasting the time-varying volatility of bond yields plays a prominent role in many finance applications. However, amid periods of financial turmoil, managing interest rate risk on a daily basis is rather a challenging task due to extreme realizations and sudden changes in bond yields that can easily lead to implausible density forecasts. To reduce forecasting uncertainty and account for structural instability in volatile bond markets, the predictive performance of flexible yield curve models with time-varying VAR parameters and Wishart stochastic volatility is investigated under a Bayesian MCMC scheme. A bond portfolio optimization and Value-at-Risk forecasting application to daily US Treasury yields also highlight the potential gains of modeling frameworks with factor Wishart stochastic volatility. The results clearly indicate that the proposed modeling features are economically motivated due to their outperformance in terms of portfolio allocation and risk management during turbulent times including the Great Recession and the Covid-19 pandemic. 

  • The working paper is available here. Last update: May 2022.