Forecast Combination and Interpretability Using Random Subspace
Boris Kozyrev
IWH Discussion Papers,
No. 21,
2024
Abstract
This paper investigates forecast aggregation via the random subspace regressions method (RSM) and explores the potential link between RSM and the Shapley value decomposition (SVD) using the US GDP growth rates. This technique combination enables handling high-dimensional data and reveals the relative importance of each individual forecast. First, it is possible to enhance forecasting performance in certain practical instances by randomly selecting smaller subsets of individual forecasts and obtaining a new set of predictions based on a regression-based weighting scheme. The optimal value of selected individual forecasts is also empirically studied. Then, a connection between RSM and SVD is proposed, enabling the examination of each individual forecast’s contribution to the final prediction, even when there is a large number of forecasts. This approach is model-agnostic (can be applied to any set of predictions) and facilitates understanding of how the aggregated prediction is obtained based on individual forecasts, which is crucial for decision-makers.
Read article
Forecasting Economic Activity Using a Neural Network in Uncertain Times: Monte Carlo Evidence and Application to the
German GDP
Oliver Holtemöller, Boris Kozyrev
IWH Discussion Papers,
No. 6,
2024
Abstract
In this study, we analyzed the forecasting and nowcasting performance of a generalized regression neural network (GRNN). We provide evidence from Monte Carlo simulations for the relative forecast performance of GRNN depending on the data-generating process. We show that GRNN outperforms an autoregressive benchmark model in many practically relevant cases. Then, we applied GRNN to forecast quarterly German GDP growth by extending univariate GRNN to multivariate and mixed-frequency settings. We could distinguish between “normal” times and situations where the time-series behavior is very different from “normal” times such as during the COVID-19 recession and recovery. GRNN was superior in terms of root mean forecast errors compared to an autoregressive model and to more sophisticated approaches such as dynamic factor models if applied appropriately.
Read article
Media Response
Media Response November 2024 IWH: Manchmal wäre der Schlussstrich die angemessenere Lösung in: TextilWirtschaft, 21.11.2024 IWH: Existenzgefahr Nun droht eine Pleitewelle in: DVZ…
See page
On Modeling IPO Failure Risk
Gonul Colak, Mengchuan Fu, Iftekhar Hasan
Economic Modelling,
April
2022
Abstract
This paper offers a novel framework, combining firm operational risk, IPO pricing risk, and market risk, to model IPO failure risk. By analyzing nearly a thousand variables, we observe that prior IPO failure risk models have suffered from a major missing-variable problem. Evidence reveals several key new firm-level determinants, e.g., the volatility operating performance, the size of its accounts payable, pretax income to common equity, total short-term debt, and a few macroeconomic variables such as treasury bill rate, and book-to-market of the DJIA index. These findings have major economic implications. The total value loss from not predicting the imminent failure of an IPO is significantly lower with this proposed model compared to other established models. The IPO investors could have saved around $18billion over the period between 1994 and 2016 by using this model.
Read article
An Evaluation of Early Warning Models for Systemic Banking Crises: Does Machine Learning Improve Predictions?
Johannes Beutel, Sophia List, Gregor von Schweinitz
Abstract
This paper compares the out-of-sample predictive performance of different early warning models for systemic banking crises using a sample of advanced economies covering the past 45 years. We compare a benchmark logit approach to several machine learning approaches recently proposed in the literature. We find that while machine learning methods often attain a very high in-sample fit, they are outperformed by the logit approach in recursive out-of-sample evaluations. This result is robust to the choice of performance measure, crisis definition, preference parameter, and sample length, as well as to using different sets of variables and data transformations. Thus, our paper suggests that further enhancements to machine learning early warning models are needed before they are able to offer a substantial value-added for predicting systemic banking crises. Conventional logit models appear to use the available information already fairly effciently, and would for instance have been able to predict the 2007/2008 financial crisis out-of-sample for many countries. In line with economic intuition, these models identify credit expansions, asset price booms and external imbalances as key predictors of systemic banking crises.
Read article
Too connected to fail? Wie die Vernetzung der Banken staatliche
Rettungsmaßnahmen vorhersagen kann
Friederike Altgelt, Michael Koetter
Wirtschaft im Wandel,
No. 4,
2017
Abstract
Seit der globalen Finanzkrise 2007/2008 liegt aufgrund ihrer Schlüsselrolle für ein funktionierendes Finanzsystem ein besonderer Fokus auf den so genannten systemrelevanten Finanzinstitutionen (systemically important financial institutions, SIFIs). Neben der Größe von Finanzinstitutionen ist auch das Ausmaß ihrer Vernetzung im internationalen Finanzsystem entscheidend für die Klassifikation als systemrelevant. Obwohl die Vernetzung von Banken untereinander in der Regel schwer zu messen ist, kann sie aus der Entwicklung von Prämien von Kreditausfallversicherungen (den so genannten Credit Default Swap (CDS) Spreads) und Aktienrenditen abgeleitet werden. Dieser Beitrag untersucht, inwieweit sich mit Hilfe der sich daraus ergebenden Co-Crash-Probability vor der Finanzkrise vorhersagen lässt, welche Finanzinstitutionen während der Krise Teil von staatlichen Rettungsprogrammen (bailout programmes) wurden.
Read article
On the Trail of Core–periphery Patterns in Innovation Networks: Measurements and New Empirical Findings from the German Laser Industry
Wilfried Ehrenfeld, Toralf Pusch, Muhamed Kudic
Annals of Regional Science,
No. 1,
2015
Abstract
It has been frequently argued that a firm’s location in the core of an industry’s innovation network improves its ability to access information and absorb technological knowledge. The literature has still widely neglected the role of peripheral network positions for innovation processes. In addition to this, little is known about the determinants affecting a peripheral actors’ ability to reach the core. To shed some light on these issues, we have employed a unique longitudinal dataset encompassing the entire population of German laser source manufacturers (LSMs) and laser-related public research organizations (PROs) over a period of more than two decades. The aim of our paper is threefold. First, we analyze the emergence of core–periphery (CP) patterns in the German laser industry. Then, we explore the paths on which LSMs and PROs move from isolated positions toward the core. Finally, we employ non-parametric event history techniques to analyze the extent to which organizational and geographical determinates affect the propensity and timing of network core entries. Our results indicate the emergence and solidification of CP patterns at the overall network level. We also found that the paths on which organizations traverse through the network are characterized by high levels of heterogeneity and volatility. The transition from peripheral to core positions is impacted by organizational characteristics, while an organization’s geographical location does not play a significant role.
Read article
Isolation and Innovation – Two Contradictory Concepts? Explorative Findings from the German Laser Industry
Wilfried Ehrenfeld, T. Pusch, Muhamed Kudic
IWH Discussion Papers,
No. 1,
2015
Abstract
We apply a network perspective and study the emergence of core-periphery (CP) structures in innovation networks to shed some light on the relationship between isolation and innovation. It has been frequently argued that a firm’s location in a densely interconnected network area improves its ability to access information and absorb technological knowledge. This, in turn, enables a firm to generate new products and services at a higher rate compared to less integrated competitors. However, the importance of peripheral positions for innovation processes is still a widely neglected issue in literature. Isolation may provide unique conditions that induce innovations which otherwise may never have been invented. Such innovations have the potential to lay the ground for a firm’s pathway towards the network core, where the industry’s established technological knowledge is assumed to be located.
The aim of our paper is twofold. Firstly, we propose a new CP indicator and apply it to analyze the emergence of CP patterns in the German laser industry. We employ publicly funded Research and Development (R&D) cooperation project data over a period of more than two decades. Secondly, we explore the paths on which firms move from isolated positions towards the core (and vice versa). Our exploratory results open up a number of new research questions at the intersection between geography, economics and network research.
Read article
Glaskugel Prognose – Warum werden ökonomische Prognosen nicht besser?
Oliver Holtemöller
Wirtschaft im Wandel,
No. 2,
2014
Abstract
Während Prognosefehler bei kurzfristigen Wetterprognosen in den vergangenen Jahrzehnten deutlich reduziert werden konnten, hat sich der durchschnittliche absolute Prognosefehler ökonomischer Prognosen für die jährliche Veränderung des Bruttoinlandsprodukts in den vergangenen 45 Jahren kaum geändert. Dies liegt vor allem daran, dass sowohl in Bezug auf die tatsächlichen ökonomischen Wirkungszusammenhänge als auch in Bezug auf die relevanten zukünftigen ökonomischen Schocks eine fundamentale Unsicherheit besteht, die auch mit besseren Methoden und mehr Daten grundsätzlich nicht aufgehoben werden kann. Die Prognosen der Wirtschaftsforschungsinstitute sind allerdings unverzerrt und stellen daher eine solide Grundlage für die wirtschaftspolitische Planung dar. Trotz ihrer Unvollkommenheit sind wissenschaftlich fundierte Prognosen eine wichtige Grundlage rationaler Wirtschaftspolitik.
Read article
Testing for Structural Breaks at Unknown Time: A Steeplechase
Makram El-Shagi, Sebastian Giesen
Computational Economics,
No. 1,
2013
Abstract
This paper analyzes the role of common data problems when identifying structural breaks in small samples. Most notably, we survey small sample properties of the most commonly applied endogenous break tests developed by Brown et al. (J R Stat Soc B 37:149–163, 1975) and Zeileis (Stat Pap 45(1):123–131, 2004), Nyblom (J Am Stat Assoc 84(405):223–230, 1989) and Hansen (J Policy Model 14(4):517–533, 1992), and Andrews et al. (J Econ 70(1):9–38, 1996). Power and size properties are derived using Monte Carlo simulations. We find that the Nyblom test is on par with the commonly used F type tests in a small sample in terms of power. While the Nyblom test’s power decreases if the structural break occurs close to the margin of the sample, it proves far more robust to nonnormal distributions of the error term that are found to matter strongly in small samples although being irrelevant asymptotically for all tests that are analyzed in this paper.
Read article
