Model Behavior or Behavior Modeling’ Tools to Put Your Submission on Top
Underwriting risk is a tricky business whether in a hard or soft market. Competence and experience will always be a hard combination to find, but is greatly valued by strong companies. It’s amazing that even the best underwriters base acceptability and pricing decisions with the limited information provided. The questions of an ACORD application touch the surface issues. The producer narratives tend to sound all the same and provide more feelings than fact. Supplemental applications do provide more depth but are hard to use in the field and harder yet to compile with any consistency. Maybe there’s more information on the Web or through third party or in-house resources, which should also be considered.
How does it all add up to an underwriting decision? It seems different in most cases. What can be done to simplify the information used in underwriting liability risks while providing enough data and supporting materials that are credible and verifiable?
Pemco mentions in one of its advertisements how cars don’t crash themselves. People crash them. Most of the industry already realizes the importance of behavior in evaluating risk. Currently many of us look to loss runs and the various ways we can analyze or dissect the information. Doesn’t it seem a bit redundant if we compound our discretionary judgment with calculated experience factors?
Looking forward at a risk, we tend to ask questions about controls, policies, procedures and the people involved. We may also question the credibility of that information as it ties in to past experience on a case by case basis. Yes, we have to dig hard to get at enough information to form a credible opinion, making that file more complex and time consuming to underwrite. If working with a prospecting broker, this could mean a huge potential waste of time. At the same time, we can’t go to the efficient extreme and automate judgment either. As critical as behavior is to an account’s performance the insurance industry has struggled with how to provide more effective and efficient tools that allow for measurement of these factors. Asking the key questions is important. Seeing the answers proven out in some way is equally important.
In the area of property underwriting, the industry has seen some success using automated modeling tools which take into account the agreed upon characteristics that have a bearing on the risk being measured. The track record is well over five years long for these tools and they have proven themselves useful in enhancing results. A delay in developing similar tools for liability underwriting seems to come from a “disconnect” between the technology people and the qualified underwriting people.
This problem has been observed to come from two points of view. Casualty underwriters may not understand the logic requirements and limitations of technology enough to effectively help technology people in developing the needed tools. In fact, such projects may be a source of frustration and seem threatening to their vital role. On the other hand, technology people may be too ambitious, assuming they actually can automate experience, intuition and judgment—resulting in over-reaching and unqualified efforts to do a good thing. In both cases, the issues of role and simplicity aren’t stressed enough. The answers to such modeling tools lie with limiting the role of such a tool to just that, a tool for the underwriter to use in simplifying, verifying and later compiling relevant information.
Simplified, verifiable information measuring behavior needs to be formatted so it can be later compiled and tracked. With such a format, the model can later be adjusted for its presumed impact on losses or losses avoided using a large enough number of similar accounts to provide credibility. In fact, over time, the true impact of behavior as measured practices, procedures, etc., on losses will allow underwriters to back up their hunch or intuition with prior, compiled case history. Key information has to be verifiable upon file submission and later upon loss control or relevant claim incident. With the behavior model gathering and measuring data from the integrated functions of claims, loss control and underwriting—true correlation to loss results with large enough numbers of similar accounts can credibly determine if a specific set of practices or procedures can affect experience.
What might an initial modeling tool look like? As seen from a small part of a “Contractor Scorecard” (see chart above) the presumed relevant practices and procedures are indicated and supporting file information requested. Too many times underwriters are busy with a flood of submissions, especially in today’s hard market for certain industry classes like contracting or transportation. The scorecard developed by Olympic Advisors separates the file from all the others and gets to the key factors affecting acceptability and pricing decisions.
“Initial” is the key word. As numbers of similar accounts becomes credible and loss results can be specifically tied to behavior measured, adjustments can be made—adding or deleting questions or weighting questions or documentation requirements so that the affect scores more or less.
Please note how such a tool does not underwrite. It is merely a tool requesting and assembling relevant information in hopes of simplifying and differentiating risks with some sense of consistency. The model does not price an account. It does not determine discounts or acceptability. It does provide measurable data whose relevance can be adjusted or verified later to loss results.
As a bonus, retailers have found this tool as a selling point. It gives them a point of differentiation and added value. A competing broker using this tool might ask, “Could you show me the score card or practices calculator your current broker uses to separate your file from the larger pile the underwriters are looking at?” Controlling brokers might well use the tool to better position the risk for renewal and obtain more markets willing to look at the file, much less prevent a prospecting broker from using it first. Any tools that help a client understand the importance of practices and procedures can be a selling point, especially if the broker is equipped to help the risk make improvements.
No model will be perfect, especially at first. We could start with something designed to self improve as time goes on. By using these models consistently, the transparency of a file and its ability to be communicated accurately and completely up and down the insurance distribution chain of command is greatly improved. Given large enough numbers of similar accounts using such tools, program managers, underwriting managers and even reinsurers may better be able to look into more of the “how and why” of company loss results and the people affecting them.
Bottom line, if you can’t measure or track the information and its quality, why use it?
Treacy Duerfeldt has been a part of the insurance industry since 1989 as a surplus line broker and owner of his own firm in Olympia, Wash. Contact him at treacy2@comcast.net or visit his Web site at www.oallc.com.