Comparing logit-based early warning systems: Does the duration of systemic banking crises matter?

Abstract

Highlights1. This paper examines systemic banking crises.2. We focus on the crisis duration bias 3. We compare multinomial and binomial logit models in correctly predicting crises.4. We consider a large and heterogeneous dataset. We find the multinomial logit model to outperform binomial models AbstractThis paper compares the performance of binomial and multinomial logit models in the context of building early warning systems (EWS) for systemic banking crises. We test the hypothesis that the predictive performance of binomial logit models is hampered by what we define as the crisis duration bias, arising from the decision to either treat crisis years after the onset of a crisis as non-crisis years or remove them altogether from the sample. In line with our hypothesis, results from a large sample of world economies suggest that i) the multinomial logit outperforms the binomial logit model in predicting systemic banking crises, and ii) the longer the average duration of the crisis in the sample, the larger the improvement.

Determinants of financial crises—An early warning system based on panel logit regression

“…The bias arises from the decision to treat crisis years after the onset of a crisis as noncrisis years or simply remove them from the dataset. A valuable approach to deal with the crises that normally last for more than a year is to use multinomial approach, which, along with the so‐called tranquil years, recognizes a year when a crisis erupts and subsequent crisis years (Bussiere & Fratzscher, 2006; Caggiano, Calice, & Leonida, 2014; Caggiano, Calice, Leonida, & Kapetanios, 2016; Hamdaoui, 2016). Such an approach also recognizes tranquil phase, when fundamentals are largely sound, but, opposite to binomial models, it is able to distinguish between a precrisis regime and postcrisis or recovery periods, when economic fundamentals go through an adjustment before reaching a sustainable level.…”

Modeling human behavior in economics and social science

“…In his letter to the G-20, Dominique Strauss-Kahn, Managing Director of the IMF (International Monetary Fund), asks for "developing a reliable early warning and response system [130], able to signal the policy maker the potential arrival of systemic and therefore costly banking crises". In the aftermath of the 2008, this literature has been growing (for a short review see [62]). However, none of the predictive models allow for the possibility that banks, acting as a network, interact with each other and, in turn, this interaction shapes the underlying and unknown distribution of systemic risk endogenously.…”

Network‐based early warning system to predict financial crisis

“…In another study, Li, Chen, and French (2015) considered the data on future contracts and options to develop the logit model. There are other similar studies on the development of early warning systems based on econometric models (e.g., see Giovanni, Calice, & Leonida, 2014; Caggiano, Calice, Leonida, & Kapetanios, 2016; Khallouli & SamiNabi, 2013; Gresnigt, Kole, & Franses, 2015).…”