Printed from

Ten considerations for projecting self-insured workers compensation losses

Posted On: May. 1, 2017 12:00 AM CST

Ten considerations for projecting self-insured workers compensation losses

Projecting self-insured workers compensation losses is a complex process due to its long-tail nature and the many factors that can significantly impact the final costs of claims. A proper actuarial review can provide valuable insight and tailor estimates to each unique self-insurance program. The following discusses the top 10 considerations used by actuaries for projecting self-insured workers compensation losses and integrating them into the projection process.

1. Collecting claims and exposure data

Claims and exposure data are the key inputs when projecting self-insured losses. It is important to keep clean, organized records of a program’s historical data because inconsistencies can compromise the accuracy of results. Treatment of subrogation and other recoveries should be clearly and consistently applied. Likewise, allocated loss adjustment expenses should be treated consistently if losses are limited to retention, as they may not erode the limit depending on the excess insurance or large deductible policy purchased. In addition, policies may include other limits or deductibles , which need to be captured. Payroll is often chosen as a preferred exposure metric because it directly correlates with the cost of benefits provided to injured employees, but it can be inconsistently measured. Items excluded should remain constant over time, such as bonuses and vacation time. Because exposures are used on a relative basis to measure change in program size, a change in reporting of exposures can signal a false change in program size.

2. Loss development

Loss development serves as the foundation for many estimation methods used by actuaries when selecting ultimate losses. Data is usually organized into so-called loss development triangles, which are constructed using claims data — often organized by accident/policy period — at multiple valuation dates and which present an effective way to analyze how a program’s losses have developed over time. It may be useful to split data into different triangle subsets if it is believed that losses from certain business segments or geographical regions could develop in significantly different ways from others. Losses from different states can show different development patterns due to differences in benefit laws and legal environments. It may also be worthwhile to build triangles by job type or class code.

Loss development factors, or LDFs, are selected for each development period from the triangles. The first step is deciding if a program’s experience is fully credible or if it should be supplemented with industry development factors. The selection of tail development is especially critical, as it affects LDFs for all younger development periods. In addition, LDFs must be tailored to the specific needs and experience of each data set. If LDFs are limited to a retention that is different from the projected losses, it may not be appropriate to use those LDFs in the projection process. LDFs can be adjusted to various retentions, but theoretical increases or decreases may not be appropriate based on a program’s history.

3. Determining trends

In addition to selecting LDFs, frequency and severity trends should also be reviewed. There is a common misconception that they are accounted for when developing losses. This is known as “the overlap fallacy.” Applying trend to claims data adjusts losses to a desired exposure period to account for nominal cost differences between years, while developing losses estimates the final costs of all claims — also referred to as ultimate losses. If a program’s data is considered credible enough to estimate trends and they are significantly different from the industry, it must be demonstrated to auditors that the utilized trends are appropriate. It is important to also recognize what type of trend is needed when dealing with severity, specifically whether inflationary trend should be included. In addition, exposure trends, as measured through payroll, should be accounted for and should consider any shift in class codes of workers.

4. Selection of methods

Actuaries utilize multiple estimation methods when selecting ultimate losses for each accident/policy period. Some methods rely strictly on current claims values and the selected LDFs; others rely on historical claims averages and ignore how claims for a specific year have developed so far; and others weigh actual and theoretical experience together. Understanding the advantages and disadvantages of how the various methods react to data and program changes is essential to selecting appropriate ultimates. There are also times when commonly accepted actuarial methods do not produce intuitive results. It is critical that the actuary step back from the methods and apply business sense, because any movement in ultimate losses affects the bottom line. Additionally, selected ultimate losses for prior years hold an extra level of importance because a prospective year’s loss estimate is typically based on historical experience.

5. Large losses

Large losses are inevitable. Two key considerations are how much do current estimates include for possible future large losses, and how much impact do large losses have on future projections. It must be considered whether a large loss is an anomaly or an indicator of possible future large losses. In order to minimize fluctuation in reserve estimates that are due to large losses, selected ultimates should include some provision for potential large losses or additional incurred but not reported, or IBNR, costs. IBNR losses represent unreported losses and can arise from unreported claims, development on known claims, reopening of claims or claims in transit. The majority of workers compensation IBNR costs are for development on known and reopened claims. A conservative approach would hold a healthy amount of IBNR and slowly lower estimates over time if large losses do not occur as a year matures. For older years, the reporting of a large loss may result in a dollar-for-dollar increase in ultimates, because less IBNR is typically held for more mature years. If large losses are not expected to occur every year, estimates for each individual year may not need to hold enough protection against a large loss. Collectively, however, there should be enough IBNR to absorb some expected level of large losses without significantly raising ultimates in total.

6. Stability vs. responsiveness

Selected LDFs, ultimates, severity and frequency trends, and other model assumptions all affect the level of stability when projecting workers compensation losses. Estimates for younger years are typically based on theoretical methods that are more stable compared with methods that rely heavily on actual claims experience. For example, if a large loss is reported within the first six months of a policy year, a method that relies solely on actual claims experience will assume continuing large development in the future and can potentially overestimate ultimate losses. A balance must be achieved between overreacting and reacting too slowly to new information.

7. Program changes

Possible changes to a program may include size of retention, purchase or sale of a division, changes in claims handling and new loss initiatives. Adjustments for retention typically use increased limit factors, or ILFs, which can be based on an industry source or a program’s own data. Similar to limiting LDFs, theoretical increases or decreases based on industry experience may not be appropriate, and actuarial judgment must be exercised when selecting appropriate ILFs. Adjustments for the purchase or sale of a division are more complex and must be tailored for each scenario. Adjustments for changes in claims handling and new loss initiatives must also be tailored for each scenario. For example, a switch from using a third-party administrator to handling claims in-house may result in more focus being placed on gathering claims detail and closing claims more quickly.

8. Special considerations

Changes to reserving methodology and timing of claims payments can have significant effects on the accuracy of LDFs and associated estimation methods. For example, if there is an increase in case reserve adequacy, ultimate losses will likely be overstated unless the necessary adjustments are made. Changes to employee mix — due to layoffs or changes in business — may also have a significant impact on the accuracy of estimation methods. An additional consideration includes any new initiatives to improve safety that will influence workers’ attitudes toward claims.

9. Issues specific to workers compensation

Workers compensation benefits are highly dependent on state and jurisdiction. Prior claims must be adjusted for any new benefit laws put into place when projecting losses. Additions of new locations may also require adjustments to make sure prior claims are level with expected future losses, if the new locations fall within jurisdictions known to be different from current locations.

Workers compensation exposures have very long tails. For example, a minor back injury today may require back surgery five years from now. An experienced actuary will factor in a provision for IBNR to protect against these late claims cost increases.

10. Financial reporting

Although the long-tail nature of workers compensation adds complexity to projecting losses, it offers a great advantage for discounting liabilities from a financial perspective. Discount factors are based on a chosen interest rate and the estimated payment pattern from the selected LDFs. The discount rate is usually provided by the company, but an actuary would be able to provide guidance related to interest rates similar companies are using. An actuary would also be able to provide expertise in estimating liabilities at various probability levels above the actuarial central estimate using loss distributions for programs that are interested in evaluating different degrees of conservatism.


When projecting workers compensation losses, it is beneficial to have a strong understanding of the many pieces that affect results. If carried reserves are understated, accruals will be inadequate. If carried reserves are overstated, capital is tied up and unavailable for other strategic uses. It is always a best practice that management works closely with the actuary to fully understand the analysis and newlyfounded insights.

Carly Rowland is a consulting actuary at Milliman Inc.. She can be reached at or 312-577-2912. Richard Frese is a principal and consulting actuary at Milliman. He can be reached at or 312-499-5648.