building value together 6 April 2012 GPV for Takaful Hassan Scott Odierno, FSA Kuala Lumpur www.actuarialpartners.com
Actuarial Work in Takaful Pricing / Product Development Risk Management / Corporate Governance Valuation and Projections 2
8 Step Process for Performing an Actuarial Valuation Data Accuracy Data Consistency Data Completeness Assumptions Methodology Calculations Results Report generation 3
Why do we care about GPV? With fair value accounting being implemented worldwide, our actuarial reserves must be at best estimate plus padding. With the old standard net premium valuation (NPV) basis it is not clear if this is the case. Also, we must strive for fairness with respect to surplus sharing, which can only be achieved if assets and liabilities are using a best estimate approach 4
How do we calculate GPV? In GPV we estimate all cash flows including bonuses and take their present value using best estimates plus padding 5
How do we determine the padding? Padding can be defined by a regulator as a certain increase from base assumptions such as in South Africa or determined at a percentile such as 75 th percentile in Malaysia 6
How do we determine the 75 th percentile reserve? We analyze the experience of each assumption getting the mean, standard deviation, other descriptive statistics and fit to a distribution such as normal or lognormal Scenario 1 Scenario 2 Assets 1000 1000 VRBE 500 500 VRLCC 800 800 VR75 550 650 TCR 250 (800-550) 150 (800 650) TCA 450 (1000-550) 350 (1000-650) CAR 180% (450 / 250) 233% (350 / 150) 7
Expense Analysis Accuracy of the GPV reserves depends on accurate splits of expenses of the operators fund between family and general Takaful, group and individual, first year and renewal and percent of contribution and per policy. Simplified methods such as all expenses assumed to be at inception for single contribution plans results in under reserving Year Actual Expense Pricing 1 100 150 2 10 0 3 10 0 4 10 0 5 10 0 6 10 0 8
Expense Overruns If operations are still in the growth stage then expenses as well as new business can be projected several years into the future to determine expense assumptions. In this case there will be expense overruns in the first few years which must be reserved for 2011 2012 2013 A. Policies 2 4 8 B. Actual Expenses 1,000 1,100 1,200 C. Assumption (B/A) 500 275 150 D. Expense Assumptions 2011 2,000 4,000 E. Overruns / (Underruns) (900) (2,800) F. Long Term Expense Assumptions (50) 200 400 G. Overruns / (Underruns) 900 800 H. Expense Assumptions 2013 600 1,200 I. Overruns / (Underruns) 500 0 9
Expense confidence levels One approach to confidence levels is to compare actual expense levels versus those projected from management the prior year to estimate managements ability to predict expense levels 10
Mortality experience splits Potential splits in experience: Agency versus bancatakaful Level of underwriting Age or age band Males versus females Select versus ultimate period 11
Mortality confidence levels We need credibility (at least 64 deaths) in using subsets of data, but the question is whether volatility is expected to vary by the subsets. Collecting industry data by industry associations or retakaful operators would be useful here, but need to account for higher expected volatility of experience on an individual basis 12
Critical illness experience There is very little critical illness experience available, company or industry wide. As actuaries we are also much less confident of how experience will emerge. Thus confidence interval estimation will be a challenge. This is also where industry associations or retakaful operator could assist 13
Lapse / surrender analysis Lapse experience is at minimum split by duration and single / regular contributions. When calculating confidence levels we need to test both high and low lapsation, which might change over time even for the same policy. At issue is whether volatility varies by duration. 14
Projecting investment returns / discount rates Two methods are used for investment return / discount rate assumptions for the risk fund, best estimates plus padding or the risk free yield curve. When using best estimates it is important to have benchmarks by asset classes to allow comparisons of volatility of benchmark versus actual experience 15
Savings fund investment returns / discount rates In GPV reserving for plans with savings accounts we need to assess the probability of the savings fund to run out of money and don t allow for Tabaru drips in the risk fund once it does run out. Padding would need to be included here 16
Risk free yield curve To calculate the risk free yield curve we obtain the yields to maturity for key durations, interpolate by six month period, then use bootstrapping in excel and goal seek to get the zero coupon yields. This is difficult to obtain in most countries 17
Operators fund GPV methodology This is dependent on the model, but we take the present value of outgo minus income: Outgo Management Expenses Commissions Taxation / Zakat Income Wakalah fees in whatever format Mudharabah investment sharing Risk fund surplus sharing 18
Sterling Reserves GPV is similar to sterling reserves, but sterling reserves are slightly more conservative as some offsetting is not allowed Year Cash Flow 1 (100) 2 50 3 50 Ignore discounting: GPV = 0 Sterling = 100 19
Risk fund GPV methodology Outgo Benefit payments Expenses (some models) Mudharabah investment sharing (some models) Retakaful contributions Income Tabarru drips (some models) Gross contributions net of wakalah fees (some models) Retakaful surplus sharing (some models) 20
Following the bigger picture Where we build conservatism into tabarru rates as a buffer we must also build this into the valuation assumptions. If not the buffer is completely released at inception 21
Illustration of GPV A mortgage reducing term Takaful (MRTT) drip plan is taken as an example to show the changes from NPV to GPV in terms of the reserve itself as well as on first year strain and overall profitability 22
MRTT Under NPV Cash Flow NPV NPV Mort Change NPV Reserving Rate Change A. Wakalah Fee 2,594 2,594 2,594 2,594 B. Mudharabah share Investment income 31 31 31 31 from PIA C. Investment Income of Operators Fund 4 4 4 9 D. Surplus share from Risk Fund 0 44 0 44 E. Management Expenses 210 210 210 210 F. Commission 1,038 1,038 1,038 1,038 G. Increase in expense reserve 0 0 0 139 H. Increase in solvency margin 0 0 0 802 I. Surplus after solvency and corporate tax 1,036 1,069 1,069 367 (A+B+C+D-E-F-G -H) * (1-corporate tax) J. Qard 2,738 0 0 0 K. Total Capital Available (H I) 1,702 1,069 1,069 367 L. Reserves at end of year 1 2,874 0 0 0 GPV 23
MRTT under NPV Reserves and thus Qard under NPV drops tremendously once a best estimate plus padding mortality table is used, but not much of a change due to using the risk free yield curve. The final GPV includes lapsation as well as solvency margin and expense reserve. 24
Conclusion GPV in Takaful follows the same processes and general methodology as conventional insurance, but the actuary must keep in mind both the detail level issues specific to Takaful as well as the big picture of how and where reserves should be held. 25
Questions hassan.odierno@actuarialpartners.com