Managing Your Risk
Forecasting future liability isn't weird science

Reliability and consistency of several years of loss history is key to making accurate projections and budgeting for right amount of coverage

By Michael Kelly
Property and Casualty Program Specialist

Drawing upon my July column, "Identify exposures at outset of risk management initiative," once your county identifies the risks and exposures facing its operations, it is necessary to analyze the results of that initial discovery phase.

Two major risk management evaluation category types are "qualitative" and "quantitative" analysis.

Qualitative areas of analysis will help you understand the "why" – its purpose is to identify loss exposures that are not easily measured by traditional financial or statistical methods. Often they will reveal a county's general appetite for taking on risk and reflect some internal operational policies.

Other qualitative areas such as a review of contractual exposures, which is an assumption of legal responsibility through written or oral agreements, can reveal a potential for loss without necessarily illuminating the possible level of severity for damages.

Employee safety issues and management's willingness to develop and enforce safety programs should also be considered and fall within the qualitative purview. Specific departmental job process mapping can often help uncover conditions or situations that, if left unchanged, amount to an injury literally waiting to happen.

Finally, a compliance and regulatory review, which tends to be more applicable to the private for-profit sectors that are more subject to stringent governmental regulation, should at least be addressed at the cursory level. Regulations pertaining to wastewater treatment facilities, utility services and employee management compliance practices should be reviewed.

Quantitative analysis tends to come from statistical processes that often utilize loss history over a period of many years. It is critical that the loss data used in the analysis is complete, consistent in nature, reliable in its integrity, and relevant. The old adage "Garbage In, Garbage Out" could never be more true.

In addition, in order to be able to derive a functional level of credible data, it is necessary to accumulate a minimum of six or seven years of detailed loss information. Anything less than six years of experience will widen the margin of potential statistical error in forecasting future losses.

My personal preference for workers' compensation loss projections is 11 years, discounting the most recent 12-month term. This allows a solid 10 years of data to work with. If during the 10-year period your county changed insurance carriers, pay special attention to the types of loss report information, for consistency's sake. For example, if one insurance carrier is reporting only paid loss data and the other is reporting both incurred and paid loss data, it will be paramount to maintain the same type of data used, otherwise it will skew your analysis. Consistency is extremely important.

Using this longer time frame also will likely generate enough actual loss count numbers to allow you to treat the county's average as if it were drawn from a "normal distribution." The rule-of-thumb minimum number of losses or occurrences necessary to make calculations over the time in question is 30, with more being better. Indexing the loss values reported for inflation and variance in the value of money over time is also important so that the loss data reflects today's actual monetary value.

The complete statistical methodology for quantitative analysis is beyond the scope of this column. Suffice to say calculations based on the standard deviation from a county's expected average number of losses per year, as well as its average severity of losses per year, can help measure the level of loss dispersion. Knowing this amount of dispersion from the average is the basis for future loss forecasting.

The normal probability loss distribution – or traditional "bell-curve" if portrayed graphically – makes statistical analysis easier to project because then it is only necessary to determine measures of central tendency (average) and the dispersion (variance) from the central tendency value in order to fully describe the distribution. Through the use of standard deviation calculations, it becomes possible to project the amount of variance from the average number of loss outcomes (maximums and minimums).

Given sufficient credible data, if your calculations are based on adding/subtracting twice the value of the standard deviation from the average loss number/loss amount, it is mathematically possible to project with a confidence level of 95.4 percent the future loss numbers and corresponding loss amounts. The loss numbers and loss amounts will fall between two given values (a high and low number).

This essentially will provide a risk manager with a working projected maximum expected loss number and maximum expected per loss value. It will also provide a corresponding forecasted minimum loss number and minimum per loss value as well, but it is the maximums that are more important to determine.

Once you have reasonably reliable expected future loss data, it becomes easier to establish and make informed decisions for variables such as deductibles/self-insured retention attachment points, annual aggregates, and even probable likely total liability limits.

While serving as the county's risk manager is somewhat a position of having to "project the future," using both qualitative and quantitative analysis tools can make it resemble more of a science than someone reading a crystal ball.

NCACC Property and Casualty Program Specialist Michael Kelly writes a regular column on risk management for CountyLines. With more than 32 years of risk management/insurance experience, he holds the Associate in Risk Management, Associate in Risk Management for Public Entities, Certified Risk Managers and Certified Insurance Counselors professional designations. Archived versions of the column can be found online at www.ncacc.org/managingyourrisk.html.