12/11 IBNR Calculation Sample Company
Table of Contents Executive Summary and Best Estimate IBNR... 1 Analysis of Historical IBNR Claims... 2 Analysis of Claim Seasonality... 3 A Note Regarding Margins of Error... 4 Actuarial Certification and Disclaimers... 5 Appendix Discussion of Methodology... 6 Overview of Methodology... 6 Method 1-A: Completion Factor Method via Geometric Averages of Durational Claim Ratios... 6 Method 1-B: Completion Factor Method via Harmonic Averages of Durational Claim Growth Rates... 6 Method 2: Projected Claims Method by Duration... 7 Method 3: Bornheutter-Ferguson Method... 7 Method 4: Projected Claims Method by Month Incurred... 7 Blending of Methods... 7 Exhibit A Per-Member-Per-Month Medical Claims... 8 Exhibit B Historical Plan Enrollment... 8 Exhibit C Claims Triangulation Data... 9
Executive Summary and Best Estimate IBNR As patients incur medical claims, they are reported to the plan administrator to be paid out. Many payments are made quickly, but some may not be paid until many months after they are incurred. These outstanding claims are known as the incurred but not reported (IBNR) claims. Were the plan to be terminated, the plan would still be liable for all of these costs. Keeping cash reserves on hand guarantees that funds will be available to pay for this liability. The figures below represent my best estimate of the liability as of December 31, 2011. These figures are only estimates, and actual results could vary significantly. See the following page for a better idea of how Sample Company s IBNR liability has varied historically. Incurred and Paid to Date Percentage Complete Total Projected Claims Incurred But Not Reported Prior Months 40,166,801 99.92% 40,199,282 32,481 January 2011 2,532,435 99.54% 2,544,047 11,613 February 2011 2,721,081 99.47% 2,735,568 14,487 March 2011 2,928,099 99.33% 2,947,843 19,744 April 2011 2,870,085 99.13% 2,895,390 25,306 May 2011 3,385,341 98.99% 3,419,785 34,443 June 2011 3,348,713 97.95% 3,418,865 70,151 July 2011 3,420,794 96.04% 3,561,887 141,093 August 2011 3,603,977 94.03% 3,832,725 228,748 September 2011 3,731,000 92.17% 4,048,112 317,112 October 2011 3,844,986 90.25% 4,260,551 415,566 November 2011 3,755,552 80.86% 4,644,297 888,745 December 2011 1,488,131 34.16% 4,356,853 2,868,721 Total Incurred But Not Reported Medical Claims 5,068,210 Page 1
Analysis of Historical IBNR Claims $6,000,000 Historical Incurred But Not Reported Medical Claims $5,000,000 $4,000,000 $3,000,000 $2,000,000 $1,000,000 $0 Over the period of time from January 2011 through December 2011, the magnitude of incurred but not reported medical claims has varied from a minimum of $3,720,087 to a maximum of $5,354,550. The average IBNR liability was $4,368,832. My estimate for the current IBNR is 16% higher than the historical average. If we assume IBNR liabilities are normally distributed, the liability should be within a range of $3,812,126 and $4,925,538 approximately 2/3 of the time. It should be within a range of $3,255,419 and $5,482,245 approximately 95% of the time. Page 2
Analysis of Claim Seasonality $600 Incurred Per-Member-Per-Month Medical Claims by Month 2010 $500 2011 Average $400 $300 $200 $100 $0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec The claims experience for Sample Company displays a significant degree of seasonality. Claims for the later months of the calendar year are likely to be twice the level of claims in the early months of the calendar year. This is typical for a high deductible health plan. As the year progresses, participants reach their deductibles and out-of-pocket maximums, resulting in a greater portion of costs being paid by the plan. The reserve at December 31, 2011 is likely to be higher than at any other point in the calendar year. Page 3
A Note Regarding Margins of Error As claims develop, the incurred but not reported medical claims will be higher or lower than my projections. For this reason, many companies choose to be conservative and hold reserves greater than the expected liability. I have not included any conservatism in my calculations. Excess reserves are often expressed as an intended percentage of the liability or an amount sufficient to cover expected medical claims for a particular period of time. Below are some reserve totals including various margins for excess claims. No calculation below is intended as a recommendation. The reserve Sample Company chooses to hold will depend on your particular financial position and level of risk aversion. Reserve Without Margin 5,068,210 Reserve Plus 5% Margin 5,321,620 Reserve Plus 10% Margin 5,575,030 Reserve Plus 15% Margin 5,828,441 Reserve Plus 20% Margin 6,081,851 Reserve Plus 25% Margin 6,335,262 One Month of Claims 3,555,494 Two Months of Claims 7,110,987 Three Months of Claims 10,666,481 Page 4
Actuarial Certification and Disclaimers I, Michael Taylor, am an Associate of the Society of Actuaries, an Associate of the Conference of Consulting Actuaries, and a Member of the American Academy of Actuaries. I meet the education and experience requirements of the U.S. Qualification Standards, as issued by the American Academy of Actuaries, necessary to perform these calculations. My analysis was conducted in accordance with all applicable Actuarial Standards of Practice, as issued by the Actuarial Standards Board, specifically ASOP 5 Incurred Health and Disability Claims This report summarizes the results of an estimate of incurred-but-not-reported claims prepared by Liability Analytics, LLC for Sample Company. The figures contained in this report are intended to be used in Sample Company s plan monitoring and financial reporting. These figures are not intended for any other purpose. The intended audience for this report is the management of Sample Company and its auditors. This report contains confidential information and should not be distributed to any other party without permission from Sample Company. If this report is provided to any other party, it should be provided in its entirety. All figures provided in this report, including reserve recommendations, are my best estimates based on the data available. Actual claims will differ from these figures. I make no guarantee that the recommended reserve will be adequate to cover all outstanding incurred claims. I have not included a reserve for any outstanding catastrophic claims. If the plan were terminated, Sample Company should work with its third party administrator to determine if any such claims exist and analyze their potential financial impact on a case-by-case basis. I have not included any reserve for administrative costs that would be required were the plan to be terminated. This report was created using claims triangulation and enrollment data as provided by Sample Company. Data was checked for reasonableness and adjustments were made for data considered to be outliers. I have made no attempt to audit the data or verify its accuracy. This report supersedes any estimates or drafts previously provided. Nothing within this report should be considered as accounting, tax, investment, or legal advice. There is no conflict of interest which would interfere with the objectivity of my work. February 1, 2012 Date Michael S. Taylor, ASA, ACA, MAAA Page 5
Appendix Discussion of Methodology Overview of Methodology Claims triangulation data was provided for the last 24 months, split by month incurred and month paid. The data was checked for reasonability and outlier data points were adjusted or removed. Projections were made for ultimate claims based on four core methodologies. For the first, projections are based on historical claim runout patterns. For the second, projections are based on historical claims by duration. For the third, projections are based on payment percentages by duration. For the fourth, claims were projected based on historical ultimate claims. The results of these four methods were blended to determine final claims projections. Method 1-A: Completion Factor Method via Geometric Averages of Durational Claim Ratios For each value in the claims triangle, ratios were calculated for the aggregate claims paid in the following month to aggregate claims already paid. The average of these values was calculated for each period of duration. The arithmetic average can give inappropriate results in this scenario. Instead, a geometric average was used. The geometric average is found by multiplying the N number of items in the group together and taking the Nth root of the product. For a set consisting of all positive numbers, the geometric average will always be less than the arithmetic average. It is most commonly used when dealing with normalized data sets. The minimum and maximum claim ratios were excluded from each calculation. The resulting averages were applied to the triangle to project future payments. Results for this method were calculated three times: using data for the most recent 6, 12, and 24 months. Results were blended with more weight assigned to the most recent months. These results were multiplied by the amounts incurred and paid to date, yielding the projected ultimate claims for each month incurred. Method 1-B: Completion Factor Method via Harmonic Averages of Durational Claim Growth Rates For each cell in the claims triangle, growth rates were calculated for aggregate claims to the next duration. The average of these values was calculated for each period of duration. The arithmetic average can give inappropriate results in this scenario. Instead, a harmonic average is used. The harmonic average is found by taking the reciprocals for each item in the data set, taking the arithmetic average of the reciprocals, and taking the reciprocal of the average. For a set consisting of all positive numbers, the harmonic average will always be less than both the arithmetic and geometric averages. It is most commonly used when dealing with percentage rates. The minimum and maximum claim growth rates were excluded from each calculation. As a reciprocal cannot be taken for the number zero, an arithmetic average was used when any cell in the calculation had a claim growth rate of 0%. This only occurred in later durations which are immaterial to the overall results. The resulting averages were applied to the triangle to project future payments. Results for this method were calculated three times: using data for the most recent 6, 12, and 24 months. Results were blended with more weight assigned to the most recent months. Growth rates were applied to the amounts incurred and paid to date, yielding the projected ultimate claims for each month incurred. Page 6
Method 2: Projected Claims Method by Duration Cells in the claims triangle were adjusted to remove the effects of medical trend and seasonality. Claims were also adjusted to account for the number of days in each month, including additional adjustments for the number of weekend days and federal holidays. Claims were put into a per-member-per-month format by dividing by monthly enrollment. The arithmetic average of the adjusted values was taken for each period of duration. These averages were then adjusted for each period of duration to add back in the effects of trend, seasonality, and enrollment changes. Empty cells in the claims triangle were filled in assuming claims would equal the adjusted average for that duration. These incurred but not reported projections were added to claims incurred and paid for projected ultimate claims. Method 3: Bornheutter-Ferguson Method The claims triangle was completed using Method 1-B Completion Factor Method via Harmonic Averages of Durational Claim Growth Rates, using 24 months of data. This completed triangle was used to determine the percentage of payments that are made at each duration. Interim ultimate claims were assumed to be a blend equal to 40% of method 1-A, 40% of method 1-B, and 20% of method 2. For each future cell of the original claims triangle, payment percentages were multiplied by these interim ultimate claims to estimate claims incurred but not reported. These were added to the known claims to determine the projected ultimate claims. Method 4: Projected Claims Method by Month Incurred Ultimate claims were first projected using Methods 1-A and 1-B. The arithmetic average of these 2 methods was taken to provide interim ultimate claims for each month. Claims were adjusted to remove the effects of medical trend and seasonality. Claims were also adjusted to account for the number of days in each month, including additional adjustments for the number of weekend days and federal holidays. Claims were put into a per-member-per-month format by dividing by monthly enrollment. The median of the adjusted values was taken and adjusted for each month incurred to add back in the effects of trend, seasonality, day counts, and enrollment changes. If this figure was less than the claims already paid for that month, this figure was overridden by an amount halfway between the amount incurred and paid to date and the ultimate claims projected via Method 2. This yielded the projected ultimate claims for each month incurred. Blending of Methods The ultimate claims by month incurred were calculated using the first three methods. These results were blended together using 25% of method 1-A, 25% of method 1-B, 10% of method 2, and 40% of method 3. This resulted in the interim ultimate claims. These were blended with the results from Method 4. The weighted average was taken for each month incurred. In each month, the weight assigned to the interim results was set equal to the square root of the interim results percentage complete, calculated as the incurred claims divided by the projected ultimate claims. Incurred claims were subtracted from ultimate claims, resulting in the final projection of claims incurred but not reported. Page 7
Exhibit A Per-Member-Per-Month Medical Claims $500 $450 $400 $350 $300 $250 $200 $150 $100 $50 $0 01/01/10 04/01/10 07/01/10 10/01/10 01/01/11 04/01/11 07/01/11 10/01/11 Exhibit B Historical Plan Enrollment 10,100 10,050 10,000 9,950 9,900 9,850 9,800 01/01/10 04/01/10 07/01/10 10/01/10 01/01/11 04/01/11 07/01/11 10/01/11 Page 8
Exhibit C Claims Triangulation Data Duration 1 2 3 4 5 6 7 8 9 10 11 12 13 + Prior Months 12,007,091 20,114,122 4,030,354 814,722 812,903 798,687 813,248 406,528 80,001 80,963 40,403 40,165 127,615 Jan '11 770,578 1,276,324 247,414 51,269 47,551 49,268 50,920 23,967 5,094 4,929 2,542 2,580 Feb '11 795,530 1,381,800 283,855 53,004 53,723 56,096 55,629 27,790 5,548 5,359 2,747 Mar '11 895,773 1,473,357 290,273 55,252 57,701 58,641 55,748 29,682 5,962 5,710 Apr '11 876,967 1,425,280 293,926 60,962 58,063 60,936 57,790 30,234 5,929 May '11 1,027,834 1,729,466 328,208 68,727 63,768 66,227 67,431 33,681 Jun '11 1,053,390 1,681,298 347,509 64,997 68,571 68,327 64,621 Jul '11 1,104,019 1,749,895 353,059 71,638 72,803 69,380 Aug '11 1,120,393 1,949,686 384,950 75,064 73,885 Sep '11 1,174,769 2,064,140 411,450 80,641 Oct '11 1,279,750 2,134,415 430,821 Nov '11 1,386,941 2,368,611 Dec '11 1,488,131 Page 9