This release of CoMeta greatly expands the number of profiled companies to more than 39,000. The expansion in company content is supported by newly-designed algorithms that probabilistically connect companies with the business activities that could draw them into future mass litigation events.
CoMeta employs detailed corporate tree structures to define business groups that could be held liable for the harmful effects of the business activities engaged in by the business entities they own. The 39,147 business groups profiled in CoMeta consist of all business groups generating at least $100 million in annual revenue in the United States. About 27,000 of these business groups are ultimate parents; the remaining roughly 12,000 business groups represent subsidiaries and other business entities owned by those ultimate parents. CoMeta refers to any business group regardless of its position within the corporate tree as simply a “company.”
Algorithmic matching process
CoMeta provides users with detailed information regarding the business activities that connect a given company to a potential mass litigation event. Some companies are connected to potential mass litigation events because they produce the Litagion agent at the center of the event. For example, Eastman Chemical Company is connected to the Litagion agent di(2-exthylhexyl) phthalate because it manufactures the chemical. Johnson & Johnson is swept up in current litigation (and, possibly, future litigation) centered on the Litagion agent talc because it produces feminine hygiene products that contain talc.
Over the years, Praedicat company analysts have reviewed thousands of possible connections between companies and the business activities that could expose them to future mass litigation events. Our data scientists have incorporated these “training data” along with externally-available data on companies in machine-learning and natural language processing models that generate a probability that any one company is connected to any one of the tens of thousands of potential mass litigation events that comprise CoMeta’s liability catastrophe event set.
The probabilistic nature of the model output is reflected in a categorical “match-quality” variable (1) that gives users an indication of the degree of certainty with which companies are connected to Litagion agents through specific business activities.
There are hundreds of thousands of connections between profiled companies and the tens of thousands of hypothetical mass litigation events found in CoMeta. A relatively small number of these connections have been reviewed by analysts. All else equal, we focus analyst review upon connections that are financially consequential and that the model indicates have greater uncertainty of being true. The revised training data generated by ongoing review is incorporated in subsequent runs of the model both improving our assessment of “ground-truth” and the accuracy and precision of the model itself.
Only analyst-reviewed company connections can receive a match-quality of “high.” Analyst-reviewed connections, though, can be assigned a match quality of less than “high” if the analysts themselves are uncertain of the connection.
Availability of projected loss and company clash data
Quantitative liability catastrophe loss data are generated less frequently than the monthly CoMeta release schedule. As a result, more recently added companies may not have company loss data available in CoMeta (2). This is also true of company clash data (3). You can expect to see liability catastrophe loss data for the large number of companies added this release by the time of the January 2019 CoMeta release.
Please also note that a large number of profiled companies have no connections at all to CoMeta’s liability catastrophe event set today.