GMWB Rider Pricing With Thin Lapse Experience
A life insurer is preparing to file a single-premium variable annuity with a GMWB rider. The base contract has a 1.00% M&E charge and the rider adds a 0.90% guarantee fee. The reference portfolio is 60% global equity/40% investment-grade bonds. The insurer has four years of lapse experience from a similar but older VA product without a living benefit; observed annual lapse rates by policy year were 6%, 9%, 12%, and 15%, but the sample is thin for the rider. The pricing actuary used a flat 5% lapse assumption and a 2.5% long-term equity return. The risk committee notes that guarantee costs are highly sensitive to policyholder withdrawals and fund performance, and that the hedge program has not been tested through a prolonged equity drawdown. The company must decide whether to launch, change the design, or seek reinsurance.
1. Evaluate the credibility of the flat 5% lapse assumption and describe the assumption-setting process you would use for the GMWB rider.
Read the worked answer
The flat 5% is not credible. The observed lapse rates on the older VA were 6%, 9%, 12% and 15% by policy year, already above 5% and rising, and a living-benefit rider changes behavior: when the guarantee is in the money, policyholders tend to persist or withdraw only the maximum; when out of the money, lapses rise. A 5% flat rate understates duration risk and overstates fee income.
I would build a dynamic lapse function conditioned on the ratio of account value to benefit base, surrender charge, age, distribution channel and fund volatility, calibrated with credibility weighting between company data and reinsurer/industry data. Given the thin rider sample, I would start with a base lapse of 7–8% in years 2–5 and apply dynamic multipliers, document expert judgment, and set a monitoring trigger. The model should also differentiate full surrenders from partial withdrawals, because GMWB costs depend heavily on withdrawal timing.
2. What profitability and capital stresses would you require before launch, and what design or risk-transfer changes might follow?
Read the worked answer
I would require stochastic projections under real-world and risk-neutral scenarios. Real-world: 10,000 scenarios with equity returns, interest rates, volatility and lapse; risk-neutral for guarantee valuation. Stress: -40% equity in year 1, prolonged low returns, +200bp/-200bp rates, 50% higher withdrawals, 50% lower lapses in stress (persistency risk), and elevated volatility. Measure CTE95/99 and 99.5% one-year VaR, plus liquidity for hedge collateral. The hedge program must be tested through a drawdown, including basis risk and transaction costs.
If capital or hedge cost is excessive, design changes include a lower roll-up rate, a higher guarantee fee, a cap on withdrawals, a waiting period, or a GMDB only with an explicit charge. Reinsurance of the GMWB tail could be used, but compare cost and counterparty risk. Launch only if pricing covers hedge cost, capital and profit margin.
3. Identify the data limitations and propose a post-launch experience study and governance plan.
Read the worked answer
Limitations: no rider-specific lapse/withdrawal experience, limited fund mix history, no stress-period behavior, small sample, and possible distribution bias. The older product's data may not transfer because living benefits alter behavior. Plan: collect policy-level monthly data on account value, benefit base, withdrawals, surrenders, fund allocation, channel and surrender charges.
At 12 and 24 months, compare actual lapse/withdrawal to assumptions by duration and moneyness; use credibility weighting and statistical tests. Establish a trigger: if actual lapse differs by more than 20% or withdrawal utilization exceeds 110% of expected for two quarters, escalate to the product committee. Governance: annual assumption review, model risk validation, and feedback to pricing and hedging. Also track hedge P&L versus expected.