WHITE PAPER. Risk Management Triangles Triangles may be frustrating, but it s not safe to throw them

Transcription

1 frustrating, but WHITE PAPER Authors: Meredith Randall Kelly Barton Riskonnect, Inc.

2 Introduction Those in the risk management industry tend to make many assumptions around triangles. Many have heard of triangles, perhaps even been involved in running triangle reports, but have never had a full understanding of the fundamentals, the nomenclature of triangles and the many purposes they serve. The first section of this paper will give a brief overview or Triangles 101. For the novice, and even those who have been in the industry for years, this explanation may be beneficial as you discuss triangle reports with others beyond the risk management department, including accountants, internal and external auditors, and especially your actuary. For those involved in risk management, the second section of this paper discusses tailoring loss triangles to meet more advanced requirements. Many of you probably have extensive experience producing numerous types of triangles and analyzing data going into and coming out of your triangle reports. We will address the frustrations you may have experienced meeting your actuary s varying triangle requirements and illustrate ways to challenge and satisfy your actuary with more complete data. Through a step-by-step process for generating triangles varying in type, complexity, and requirements you will see how triangle reporting, with technology now available to this niche risk management industry, can be fast and painless. You and your actuary can then focus on the data s interpretation and meaning, not mechanics. Those already well-acquainted with triangles may find that after you deliver your triangles to the actuaries, they manipulate the reports and apply factors, then return a voluminous report. You may question the outcome of the report, or certain assumptions. We will discuss suggestions that will empower you to challenge the report, and techniques to help the actuaries better understand the nuances of your company, its history and future direction. The result: greater confidence in your interactions with actuaries and a report that more accurately projects your firm s expected losses. Page 2

3 Triangles 101 Why do we run triangles? And, why do actuaries want triangles? Let s look at Liability Claims for an example. A claim happens on a given day and is reported to your Third Party Administrator (TPA), hopefully on the same day. Then, an adjuster is responsible for settling that claim. Some claims, like minor property damage, may be settled and closed within 30 days. However, many claims, especially those involving bodily injury or litigation, remain open for months or years. Loss Development describes the tendency of open claims to grow, in dollar terms, over time. While it s not possible to reliably calculate loss development on an individual claim, grouping claims together and evaluating them at successive points in time allows one to estimate loss development factors. These factors can then be applied to a set of new claims to estimate their likely, ultimate value to your organization. The length of time the claim is open impacts the development of the claim. As the claim develops over time, new information comes to light, additional expenses are incurred and law suits may develop that cause even further financial loss. This financial deterioration over time is what we mean by Financial Loss Development. It is the job of the actuary to analyze past development trends to predict future development trends. While the statistical methods used to accomplish this are sometimes complicated, the concept itself is really that simple. When your organization prepares financial statements, outstanding liabilities for claims go beyond the individual case reserves captured in your risk management information system. Known claims may not be adequately reserved, and incidents may have occurred that have not yet been reported as a claim. For the former, looking at historical trends, additional reserves can be estimated. The later scenario is the Incurred But Not Reported or IBNR portion of total claims obligations. IBNR values can be estimated, based on your existing claims data, and then added to outstanding case reserves to estimate the company s total claims liability. Page 3

4 While actuarial methods are beyond the scope of this paper, their techniques estimate the impact of: 1. Late-reported claims 2. Open claims that require additional reserves; and 3. Closed claims that re-open to post additional reserves or payments. Within the actuary s voluminous report provided to your organization you will see that different methodologies and factors have been applied to your data. They may project, for example (see Figure 1.), that the 2004 year for Worker s Compensation will have a total cost to you of $462,000, once all the claims are closed. Triangle reports enable the actuaries to apply actuarial methods resulting in estimates of ultimate loss Figure 1. and an understanding of an organization s liabilities. The estimated liability is reported on the balance sheet, and changes in liability are reported as an expense on the income statement. Why is it called a triangle? Risk management triangles derive their name from the shape formed when the report is properly formatted. The rows of the triangle reflect distinct groups of claims, such as all claims occurring in a given calendar year, or Policy Year as shown in Figure 2. The columns represent the changing value of each set of claims as measured at different points in time, or evaluation periods. Recent claims may only have a single evaluation period, while older claim years can be measured across multiple periods, giving the report its distinctive triangle shape. Figure 2. Page 4

5 Loss Development Factors (LDF) To analyze and challenge the actuaries work, it is important to understand the actuarial methods they use. Another term commonly used in discussions around triangles is "Loss Development Factors," or LDF. A loss development factor represents the ratio of claim values from one period to another period. So in Figure 2, claims from the 2004 Policy Year valued at the end of 12 months (12/31/2004), had an Incurred value of $407,000, but by the end of 24 months (12/31/2005), Incurred increased to $432,000. The ratio of $407,000 to $432,000 is Paid period n +1 / Paid period n = Loss Development Factor (LDF) In the next period, the LDF would be different ($533,000 to $614,000). That ratio would be Actuaries take period-to-period factors like these, and calculate average factors, weighted factors and other statistics to measure your organization s historical experience, and estimate future claims volumes and costs. Are triangles always financial triangles? Frequently those working with triangles think mainly of financial triangles. However, there are other kinds of triangles, called "frequency triangles," which are based on claim counts. These triangle reports help estimate how many claims your organization will actually have. Limited Incurred For example, policy year 2004 may show 314 claims. But were there really 314 claims on record by 12/31/2004? Most likely not. If we run a frequency triangle, you would see that at the end of 2004, there were 306 claims Triangle Loss Paid Triangle Occurrence Triangle Open Count Triangle Paid Triangle reported. We would discover that another six claims trickled in during the 2005 evaluation year, and the last two came in during the 2006 evaluation year or in other words, the remaining claims came in between 12 and 24 months, and between 24 and 36 months. With a claim count triangle, your actuary can estimate how many claims you will likely have for a given year, even before they ve all been filed. The Data Inside There are many types of triangles such as: Claim Count Triangle Closed Count Triangle Expense Paid Triangle Gross Incurred Triangle Incurred Triangle With this foundation of the basic purpose of triangles and some defined terminology, let s now look at the data that goes into a triangle. Developing financial triangles introduces new levels of complexity, as you work with your actuary to determine which loss dollars to include. Beyond distinguishing between Paid and Incurred losses, should your triangle be based on gross or net losses? Individual claims in your triangle should be limited at the appropriate selfinsured retention level, so that insured, large losses do not distort your results. You may wish to examine different types of claims, such as medical-only and lost time for workers compensation, in separate triangles. Working closely with your actuary to define specific data requirements is critical to the accuracy and usefulness of your loss forecast. Page 5

6 Moving on from Triangles 101, we will now walk through a step-by-step process to create triangles. Figure 3 shows a triangle with General Liability claims with paid loss dollars. The report wizard is a part of the Riskonnect RMIS system which allows you to see the information specified for a given triangle and will enable us to illustrate specific techniques. Step 1 Report Type. The wizard allows you to select the type of report. We have Figure 3 selected a matrix report, with columns going across and rows going down, for the creation of a financial triangle. Step 2 Calculations on Numerical Fields. We have incurred valuations and paid valuations in our pick list (see Figure 4). If this was a claim count or frequency triangle, the wizard would not show us options for incurred and paid. We would have counts - count valuation as of 12 months, 24 months, 36 months, and on as far back as needed for your history. You may be on a cycle with your actuaries where you provide triangle reports on a semi-annual basis. If that's the case, then we would have valuations for 12, 18, 24, 30, 36, etc. Step 3 Rows and Columns. The next step (see Figure 5) allows us to specify columns. The columns are based on the policy periods as they are defined for your organization. It would be very nice if all policy periods fell cleanly within a calendar year. However, the reality is that policy periods and calendar year rarely match up. This is a very important point, and one of the challenges that comes with triangles. Figure 4. Step 2 Calculations on Numerical Fields. Figure 5. Step 3 Rows and Columns. Page 6

7 In a realistic use-case, your policy year might run on an April through March basis. There may have been an acquisition a few years ago, perhaps in 2005, where the acquired company ran their policies on a July through June basis. After you merge the data, you might want all the information coterminous. You can see how things begin to get complicated, as in the real world. Continuing our example, for the acquired data in 2005 they might have a nine-month policy period instead of a 12- month policy period. That's just one example of complex data issues risk managers are dealing with today. Whatever system you are using for triangle reports needs to be able to deal with that reality. It s not uncommon for actuaries to request two or three series of triangles: one series of triangles based upon the most common 12-month policy periods, a second series that accounts for non-standard policy periods (i.e. a six-month policy, an 18- month policy, etc.) Figure 6. Step 4 Select Columns. Comparing the series enables them to take out the fluctuations of those odd periods, develop factors to accommodate the odd periods and apply the factors to the odd policy periods. You need to have a system that can handle all of those nuances that are a part of the complicated definition of policy period. With Riskonnect RMIS, the system is flexible enough to easily define policy periods, or policy year-ends, any way needed for your organization. Step 4 Select Columns. In the next step of the wizard, we choose the cells for the interior of our triangle. This example is a paid triangle, so you see in Figure 6 that the boxes are selected for Financial Loss Triangle for Paid Valuation 12, 24, 36, 48, 60 and 72 months. Figure 7. Step 5 Column Order. Step 5 - Column Order. Step 5, shown in Figure 7, allows you to select the order of your columns. In triangle reports, the columns need to be ordered sequentially. Step 6 Filter Criteria. While each step of the process is important to building the triangle and ensuring the data is painting the most accurate picture of your organization, this step may be the most significant in delivering valuable data to Page 7

8 your actuary and enabling them come up with loss development factors that make sense. Using filters, as shown in Figure 8, you can select homogeneous sets or subsets of your data to send to the actuaries. This allows you to segment data according to specific lines of coverage, so they are not comingled. By providing several different types of triangles, not only is the data a better foundation for calculating loss development factors, but you are also better able to illustrate the nuances of your business to your actuary. Filter criteria configured within the wizard allows you Figure 8. Step 6 Filter Criteria to generate separate triangles for Workers Comp coverage, GL coverage and other lines. Triangles that comingle major coverages have mixed together claims that are developing very differently and the resulting factors would not be a good representation of your activities. Because this is such an important part of the process, let s apply it using an example. Your organization may have many different types of businesses perhaps some retail, some manufacturing, and some products along with manufacturing. If you comingle workers comp claims from your manufacturing business with workers comp claims from your retail business, you will have two very different types of exposures. It would be better to come up with factors that are distinct and specific to the experience that your organization has in the manufacturing and retail segments. At Riskonnect, our team of risk management technology professionals realize that actuaries will take a set of triangles, and, after review and analysis, often request another set triangles produced from a different set of parameters. Thus, another very important capability is to quickly produce triangles based on expense dollars, based on limits, historical limits that represent your self-insured retention, loss dollars, net dollars, gross dollars and many other criteria (see Figure 9). You are able to define the criteria and configure it to meet your needs. In this way, you are not limited by canned triangle reports. Figure 9. Page 8

9 Figure 10. So this completes the wizard process for the creation of a triangle report. As with any report, you want to be able to show the final outcome, but also want to be able to tie that directly to the underlying data. In Figure 10, you can see that there were 95 records in the 2004 period. One can click on Show Details or Hide Details button to toggle back and forth between the summary data which forms the triangle report and the individual claim records that are the underlying data for those 95 records at 12 months, at 24 months, at 36 months. While this level of detail is not something the risk manager or the claims analysts would want to see often, it is actually very important to have the information and to be able to provide that information to the actuary, as well as your accountants who may wish to reconcile source data in the completed actuarial report to individual case reserves in your system. This becomes critical when you have an outlier claim or an aberration, which is worth hundreds of thousands or even millions of dollars. This can really skew your results. If you can provide that individual detail to your actuary showing how a particular claim looked at 12 months and again at 24 months, they can make certain allowances when they are determining your loss development factors. This helps prevent a single large claim from skewing the results in a negative way, and provides an opportunity to discuss with your actuary how they plan to address any specific large claims in your loss experience. Don t let the delivery of your data become Deliverance The next challenge is that you need to provide these triangles, and all of the underlying data, to your actuary in a useful format. While the security and consistency of a PDF file is attractive for certain report applications, it s simply not appropriate for providing actuarial data. It is extremely important that you provide the data in individual cells because your actuary will most likely convert your data from spreadsheet form into their own software tools to further manipulate the data and apply their actuarial methodologies. Figure 11. Page 9

10 In the Riskonnect system, one can click on a printable view button which will instantly take the results from this triangle and open it in Excel (see Figures 11 and 12). In addition to the summary data, it also brings in other important information (see Figure 12). The headers tell you the report was produced in Riskonnect; there is a date and time stamp; the filter criteria is displayed and the data is placed into an industry standard matrix. The Data Requirements Figure 12. With whatever system you are using to produce triangles, it is essential that when the data entered or loaded into the system is with the awareness that triangles will be generated from that data. A good understanding of the data from a risk management perspective (i.e. the difference between gross and net paid, or an outstanding reserve and a case reserve) is critical. The requirements for the data include: 1. Snapshot valuations or individual transactions by period-end amount to build the development history 2. Transactions for paid and incurred 3. Detail of each transaction broken out by reserve category or payment bucket to distinguish between medical, indemnity, bodily injury, property damage, expense (including legal) and recovery transactions 4. Coverage fields used to create homogeneous subsets of the data Claim Count Triangles Help Uncover Discrepancies Actuaries will often come back to the risk management team with a discrepancy from year to year or period to period. It can be very difficult to dig through the data to uncover the reason(s) for the discrepancy. A powerful technology system can help. In Figure 13, let s uncover why the GL claims for the 2006, 36-month valuation total $183,000 at 12/31/08. Yet, when you produce the triangles for 2010, that number jumps to $200,000. This task becomes simple if Figure 13. the data is collected in a powerful RMIS capable of analytics. A first step is to look at the claim counts. If the claim counts remain at 139 for both periods, next, you may look at the transaction for an accounting correction. Check for an error, such as something that should have been entered as a 2010 transaction keyed instead as a 2008 transaction. Page 10

11 Another possibility is that when you look at the claim count, you will find that in the 2010 report, the claim record count has increased to 140 records, from 139. This would mean that some record met the filter criteria, when it did not previously meet that filter criteria. This is fairly common in the GL area. Often clients and actuaries separate completed operations and product liability from their GL. Perhaps while the adjuster was busy trying to close the 2006 claim, he may have realized that the claim was coded as completed operations, but should have been coded to GL. The adjuster changes it. Now in your system, it is reflected as a GL claim, but previously it was completed operations. By also checking the triangle with the completed operations data, you will find that it dropped off that triangle when it started to appear on this GL triangle. Have you ever seen an excited Actuary? Show them a Net Change Triangle. A Net Change Triangle should be able to be produced instantaneously from your system. This report enables you to see at a glance any net change in triangles from the prior edition to the current edition. By overlaying the data and subtracting, the majority of cells will be zeros. The only difference should be the last diagonal, which is the evaluation on the current triangle, obviously not there on prior version. However, in the example in Figure 13 of the completed operations claim, you can clearly see $17,000 in an earlier evaluation period, which would indicate the need for investigation. This type of proactive reporting and analysis will excite your actuaries. Instead of having your actuary coming to you questioning this discrepancy, you are able to proactively point out that the difference between the $200,000 in the current edition and the $183,000 in the prior edition is because of a claim that was really GL but was coded as completed operations. The history changed. Your experience with your actuaries will change as a result of this type of proactive reporting and analysis. And, the better news is that higher levels of confidence in the underlying data often leads to lower forecasts as actuaries remove any I don t trust the data assumptions, as well as lower premiums. Dealing with Takeovers One of the other challenge areas with triangles is takeovers. Takeovers, in the risk management industry, occur when you switch to a new TPA and transfer the open claims from the current adjuster to the new. Earlier in our discussion around Figure 8, we reviewed filter criteria. Within the Riskonnect system, to effectively handle takeovers, we set our filter criteria to include the Status Not Equal to V, or void, or deleted records (see Figure 14). In this scenario, if you switched TPAs in the beginning of your 2006 year, the open records taken over by the new TPA are going to reflect new reserves and payments that are not connected to the previous transactions on the Figure 14. Page 11

12 original claim. However, from an actuarial perspective, you need all of those claim records to show their loss development over time as one continuum from the original TPA to the new TPA. It is important that whatever system you are using to generate triangles is capable of merging the new TPA records and the original TPA's history. Often the original record is still in the system and becomes deleted or screened from view. All parties involved in generating the triangles, from the risk management team, to the actuaries, to any vendors or consultants, need to be certain you are consistent with how you are handling your takeover data. Concluding Remarks If triangles are a new concept for you, now you should have greater confidence as you discuss triangles and find ways to generate and analyze the data. For those of you who are intimately familiar with triangles and run triangle reports on an annual or more frequent basis, you should have greater insights in how to eliminate frustrations around the data and the generation of the triangle reports. It is important for those who present the conclusions and recommendations of the actuarial reports up to your internal management, that you have the confidence and the assurance that you have challenged the data, analyzed the data to its fullest and that you are able to get a good estimate of ultimate losses that accurately reflects your company and your company's experience. With Riskonnect s enterprise-class cloud computing risk management technology and our experienced team, we continue to find innovative techniques and solutions for our clients. For more information about Riskonnect and what to look for in a risk management technology solution please visit Read Additional Whitepapers Add Spreadsheets to your Risk Inventory By: Chris Duncan The Seven Standards of Cloud Computing Service Delivery By: Salesforce.com Page 12

13 About Riskonnect, Inc. Riskonnect, Inc. is the provider of a premier, enterprise-class technology platform for the risk management industry. As an independent innovator in risk management software, Riskonnect develops and markets a growing suite of software solutions on a world-class cloud computing model, helping clients elevate their risk management programs, safety solutions and programs for management of risks across the enterprise. Through its strategic, operational and insurable risk software applications, Riskonnect provides the risk management industry with the specific, configurable solutions needed to reduce losses, control risk and affect shareholder value. For more information about Riskonnect, contact us at to info@riskonnect.com or call About the authors Meredith Randall, Riskonnect, Director of Product Management Meredith Randall is responsible for direction, design, configuration and quality assurance for the Riskonnect product offerings. She leverages involvement in client implementations and support to improve upon the Riskonnect Enterprise, Information System and Incident Management products. With more than 25 years of experience in the risk management industry building and delivering technology to lower client overall costs of risk, she is driven to continuously improve the Riskonnect data integration methodology by minimizing the time to implement and maximize consistency and quality. She joined Riskonnect in 2010 after 27 years at Crawford & Company/Risk Sciences Group. (Risk Sciences Group is a wholly owned subsidiary of Crawford & Company.) As Director of Product Development and Support, she was responsible for a multi-million dollar annual budget for the RSG infrastructure including the network, in-house web-enabled, application and database servers, as well as security and SAS 70 audits. She was also responsible for design, development and support of the RSG Sigma Encore risk management information system serving hundreds of client customers. Randall also held positions as diverse as account manager, director of technical operations and branch manager. With a strong actuarial background, Randall was also responsible for multi-client data integration from external sources such as insurers and TPAs. She brings this expertise and attentiveness to the quality of risk management data to her role at Riskonnect. Randall holds an MA in Mathematics from the University of California, Berkeley, and is an Associate in Risk Management. Kelly Barton, Riskonnect, Vice President of Marketing Kelly Barton leads Riskonnect's marketing and communications efforts. She has ten years experience building and managing marketing and public relations for rapidly growing entrepreneurial and publicly traded companies. Previously, Barton was vice president of communications and product marketing at Metro Networks, Inc., a wholly-owned subsidiary of Westwood One, Inc. Westwood One is the largest provider of audio content in the United States. Barton has also advised other entrepreneurial start ups including Nihon Traffic Network in Tokyo, Japan and Australian Traffic Network in Sydney, Australia. Barton received a Bachelors of Arts in Communications and Public Relations from Berry College. Page 13