Dr. Jantje Sönksen
Phone: +49 (0)7071 - 29 - 7 25 70
Fax: +49 (0)7071 - 29 - 5546
- Empirical Asset Pricing with Multi-Period Disaster Risk: A Simulation-Based Approach
with Joachim Grammig (2021)
Journal of Econometrics, 222 (1, Part C), 805-832.
Abstract: We propose a simulation-based strategy to estimate and empirically assess a class of asset pricing models that account for rare but severe consumption contractions that can extend over multiple periods. Our approach expands the scope of prevalent calibration studies and tackles the inherent sample selection problem associated with measuring the effect of rare disaster risk on asset prices. An analysis based on postwar U.S. and historical multi-country panel data yields estimates of investor preference parameters that are economically plausible and robust with respect to alternative specifications. The estimated model withstands tests of validity; the model-implied key financial indicators and timing premium all have reasonable magnitudes. These findings suggest that the rare disaster hypothesis can help restore the nexus between the real economy and financial markets when allowing for multi-period disaster events. Our methodological contribution is a new econometric framework for empirical asset pricing with rare disaster risk.
- Estimating the SARS-CoV-2 Infection Fatality Rate by Data Combination: The Case of Germany’s First Wave
with Ingo Bechmann, Thomas Dimpfl, and Joachim Grammig (2022)
The Econometrics Journal, 25 (2), 515-530.
Abstract: Assessing the infection fatality rate (IFR) of SARS-CoV-2 in a population is a controversial issue. Due to asymptomatic courses of COVID-19, many infections remain undetected. Reported case fatality rates are therefore poor estimates of the IFR. We propose a strategy to estimate the IFR that combines official data on cases and fatalities with data from seroepidemiological studies in infection hotspots. The application of the method yields an estimate of the IFR of wild-type SARS-CoV-2 in Germany during the first wave of the pandemic of 0.83% (95% CI: [0.69%; 0.98%]), notably higher than the estimate reported in the prominent study by Streeck et al. (2020) (0.36% [0.17%; 0.77%]) and closer to that obtained from a world-wide meta-analysis (0.68% [0.53%; 0.82%]), where the difference can be explained by Germany’s disadvantageous age structure. Provided that suitable data are available, the proposed method can be applied to estimate the IFR of virus variants and other regions.
Machine Learning for Asset Pricing
book chapter in Econometrics with Machine Learning (editors: Felix Chan and László Mátyás) published in Springer’s Advanced Studies in Theoretical and Applied Econometrics series (2022)
Abstract: This chapter reviews the growing literature that describes machine learning applications in the field of asset pricing. In doing so, it focuses on the additional benefits that machine learning in addition to, or in combination with, standard econometric approaches can bring to the table. This issue is of particular importance because in recent years, improved data availability and increased computational facilities have had huge effects on finance literature. For example, machine learning techniques inform analyses of conditional factor models; they have been applied to identify the stochastic discount factor and purposefully to test and evaluate existing asset pricing models. Beyond those pertinent applications, machine learning techniques also lend themselves to prediction problems in the domain of empirical asset pricing.
- Diverging Roads: Theory-Based vs. Machine Learning-Implied Stock Risk Premia
with Constantin Hanenberg, Joachim Grammig, and Christian Schlag (2021)
presented at the Econometric Society World Congress (2020) and annual meetings of the Society for Financial Econometrics (SoFiE; 2021) and the European Finance Association (EFA; 2021)
Abstract: We assess financial theory-based and machine learning methods to quantify stock risk premia and investigate the potential of hybrid strategies by comparing the quality of the respective excess return forecasts. In the low signal-to-noise environment of a one-month investment horizon, we recommend to rely on a theory-based strategy that exploits the information in current option prices, especially if the risk premium estimate is to be updated at a high frequency. At the one-year horizon, a random forest can improve on the theory-based method, provided that a sufficiently long training period is used. In an effort to connect the opposing philosophies, we identify the use of a random forest to account for the approximation errors of the theory-based approach towards measuring stock risk premia as a promising hybrid strategy. It combines the advantages of two diverging roads in the finance world.
- Non-Standard Errors
with Albert Menkveld, Anna Dreber, Felix Holzmeister, Jürgen Huber, Magnus Johannesson, Michael Kirchler, Michael Razen, Utz Weitzel et al. (2023)
forthcoming in the Journal of Finance
Abstract: In statistics, samples are drawn from a population in a data-generating process (DGP). Standard errors measure the uncertainty in sample estimates of population parameters. In science, evidence is generated to test hypotheses in an evidence-generating process (EGP). We claim that EGP variation across researchers adds uncertainty: non-standard errors. To study them, we let 164 teams test six hypotheses on the same sample. We find that non-standard errors are sizeable, on par with standard errors. Their size (i) co-varies only weakly with team merits, reproducibility, or peer rating, (ii) declines significantly after peer-feedback, and (iii) is underestimated by participants.