The Financial Accounting Standards Board (FASB) issued new accounting standards under the Long Duration Targeted Improvements (LDTI) for long-duration insurance contracts, requiring liabilities to be discounted using an "upper-medium grade (low credit-risk)" yield, typically based on A-rated bonds. This paper examines the process of constructing discount curves under LDTI, focusing on criteria and the quality of algorithms for interpolation, extrapolation, and smoothing.
The paper evaluates three models for constructing discount curves: the Nelson-Siegel model, the Cairns model, and the Treasury High-Quality Market (HQM) methodology. Data from ICE's fixed income index service was used, including A1, A2, or A3-rated corporate bonds with outstanding par greater than $250 million and maturities between three months and 30 years. The results show that all four parameterizations produce similar par and zero curves, with the Treasury HQM methodology exhibiting a hump in the 10-15 year range and the Cairns model prone to wiggly short-term rates.
The paper then compares the impact of these curve approaches on GAAP earnings and balance sheet for a cohort of annuity business, using a simplified Treasury forward plus Option-Adjusted Spread (OAS) approach as a benchmark. Under locked-in forward curves, the fitted approach shows a smoother profitability pattern compared to the Treasury plus OAS method, leading to a bias toward negative Accumulated Other Comprehensive Income (AOCI). However, when using a single equivalent rate, the differences between the two approaches become less pronounced, with a more subdued impact on operating income and AOCI.
The paper concludes that the decision to lock in a curve or an equivalent flat rate significantly affects profitability patterns. Further investigations could address the use of internal asset data, inclusion of mixed credit quality in curve construction, and the creation of liability-specific curves.
