Probable maximum cyber loss: Empirical estimation and reinsurance design with private-public partnership
Type
conference paper
Date Issued
2019-03-28
Author(s)
Abstract (De)
This study defines the probable maximum cyber loss (PMCL), which stands for the worst cyber loss likely to occur, with an alternative approach to estimating the potential size of a next worst cyber risk event. We identify that series of cyber loss maxima on weekly, bi-weekly and monthly bases are stationary, short-range temporal dependency exists, Fréchet type of generalized extreme value distribution (GEV) well describes fat-tailedness of cyber loss maxima and extreme dependency is observed. We find that the predicted cyber loss likely to occur in the next five years is almost seven times larger than the estimate by the recent literature with a widely used Pareto-based model. Particularly, the comparison between the estimates from the data for the more recent period (after 2014) and the ones for the older period (before 2014) shows a dramatic increase with a significant structural break. Applying PMCL estimates, we further provide empirical benchmark of premiums under reinsurance design with private-public partnership. Our findings are important for risk managers, actuaries and policymakers, who are concerned about a (social) cost by a next extreme cyber loss as a New Normal in the digital era.
Language
English
Publisher
German Insurance Science Association (DVfVW)
Event Title
2019 German Insurance Science Association (DVfVW) annual meeting
Event Location
Berlin
Event Date
March 27-28
Division(s)
Eprints ID
256973
File(s)![Thumbnail Image]()
Name
Executive summary_Probable maximum cyber loss (Jung).pdf
Size
155.82 KB
Format
Adobe PDF
Checksum (MD5)
96596ceae28ef43627e07ea69bd1f481