On December 27, we released significant revisions to the liability catastrophe model available to Oortfolio users. The modeling and data enhancements in this release include:
- Addition of more than 35,000 named companies with modeled liability catastrophe losses
- Deployment of Praedicat’s stochastic loss allocation model with improved estimates of company-level business activity revenue and market share
- Models of limit impairment under the occurrence form and integrated occurrence declaration contagion under the Bermuda form
Brief explanations of these and other modeling, data, and application-level enhancements follow. Please contact your account manager for more detailed explanations.
39,248 companies now available for modeling
This release extends Oortfolio’s modeling capabilities to more than 39,000 companies representing approximately 85 percent of U.S. gross domestic product. The Oortfolio company library now includes every company with U.S. revenue of at least $100 million. Both parent (~27,000) and subsidiary (~12,000) companies are represented in the data. Look here for more information on how Praedicat scaled its company profiling process to support this major expansion in company-level liability catastrophe loss information.
Stochastic loss allocation model (SLAM!) 1.0
Stochastic loss allocation at the company-level was first introduced in Oortfolio in March 2018. This release significantly advances that loss allocation framework. For a given event, economy-wide losses are first spread between distinct business activities according to the ease with which plaintiffs can attribute exposure to a company engaged in that activity and then by market share within the business activity. Both these steps in loss allocation are now stochastic in nature reflecting the uncertainty in how a given mass litigation event will be distributed among potentially culpable business activities, whether a given company is engaged in those business activities, a company’s share of a given business activity market, and whether the company’s ultimate losses will correspond to its market share.
Business activity market share estimates are derived from detailed revenue data at all levels of the corporate tree structure. The stochastic loss allocation model resolves inherent market share uncertainties simulation-by-simulation across all potentially responsible parties, regardless of whether they are among Oortfolio’s 39,248 named companies, abiding by industry-level constraints on revenue. The result is a high-resolution market share event set that maps directly to Praedicat’s economy-wide liability catastrophe event set.
The deployment of SLAM results in changes to both the mean and variance of loss allocation shares at the company-level. While there is significant variability in the direction and extent of change in loss share allocation across companies, SLAM, on average, increases loss shares for large account companies. The 3,849 companies available in the previous Oortfolio release, which includes most U.S. publicly-traded companies, now account for 68 percent of modeled economy-wide losses, up from 41 percent previously. All else equal, expected losses in large account portfolios will increase as a result as will tail losses relative to expected losses.
To a lesser extent SLAM also impacts both the mean and variance of industry-level loss shares, the effects of which will vary portfolio-by-portfolio according to the portfolio’s industrial composition.
Limit impairment and IO declaration contagion
Under common interpretations of the occurrence form, multiple policy years are available to respond to a single occurrence. The aggregate limit available in those policy years, however, can erode over time due to attritional losses, insolvency, and other factors. In addition, aggregate nominal limits tend to increase over time. With this release, our model now accounts for this so-called limit impairment, which results in a greater fraction of overall insured losses being allocated to the current policy year. On average the current policy year now accounts for eight percent of insured losses, up from 5.5 percent in the previous model release.
The March 2018 Oortfolio release introduced new estimates of integrated occurrence (IO) declarations modeled as a function of the likelihood an insured’s self-insured retention will be breached. The economy-wide frequency of IO declarations changes little with this release, but the deployment of SLAM results in company-level variability relative to the previous model release. In accordance with historical observation, we also now model IO declaration “contagion,” whereby one company’s declaration by itself increases the likelihood of IO declaration by other companies affected by the same underlying event.
All else equal, the impact of limit impairment is to increase significantly modeled insured losses in the current policy year under the occurrence form. IO contagion results in thicker insured loss tails for Bermuda portfolios.
Two data and model changes result in a five percent increase in economy-wide expected losses attributable to modeled risks (from $82.5 billion to $86.6 billion). First, we have advanced the policy year to 2019. Because losses are expressed in nominal terms, this change increases economy-wide expected losses by about three percent. The balance of the increase in economy-wide expected losses is attributable to the addition of approximately 700 newly-modeled mass litigation events (latent mass actions) for existing Litagion agents.
You can now upload to Oortfolio portfolio files consisting of as many as 50,000 data rows (the total number of policies can be increased from 50,000 by adjusting the number of policies represented by each data row in the Oortfolio portfolio upload template).
Company and industry list views
Oortfolio users will see redesigned company and industry list views that accommodate new company and industry data.
Oortfolio now employs the 2017 version of the North American Industry Classification System (NAICS). All existing portfolios have been updated to reflect the small number of changes between the 2012 and 2017 NAICS taxonomies. Oortfolio will assume 2017 NAICS codes for any newly uploaded portfolio. Those codes can be found in the revised Oortfolio portfolio upload template.
Removal of occurrence (low) and occurrence (high) forms
In previous model releases have maintained “low” and “high” estimates of losses under the occurrence form that reflected relatively extreme interpretations of how occurrence losses could be allocated across multiple policy years. With this release, we offer a single estimate of losses under the occurrence form, first released to Oortfolio in March 2018, that accounts for the most common interpretations of the occurrence form in U.S. courts.
All existing portfolios employing either occurrence (low) or occurrence (high) forms have been updated to reflect our central estimate of losses under the occurrence form. All else equal, estimated insured losses will increase for portfolios previously modeled as occurrence (low) and decrease for portfolios previously modeled as occurrence (high). Occurrence (low) and occurrence (high) forms are no longer available for selection within the Oortfolio portfolio upload template.
New company IDs
The release of more than 35,000 new companies necessitated the creation of a new set of Praedicat company IDs. All existing portfolios have been updated to reflect the revised company IDs. Oortfolio will assume new company IDs for any newly uploaded portfolio. These new company IDs can be found in the revised Oortfolio portfolio upload template.
To improve Oortfolio performance and eliminate redundancies, this release eliminates the exportable “A.E. Policy Detail” file. Users desiring access to the output contained in that file can use the exportable Event Loss Table (ELT) report instead. Please contact your account manager to enable ELT export functionality. Also with this release, the “Litagion/Aggregate Exposure” file will only be generated when a portfolio file is uploaded to Oortfolio.