Interest Rate Models: Vasicek, CIR, and Hull-White
An overview of the major interest rate models used in actuarial valuation and risk management.
Why Interest Rate Models Matter for Actuaries
Interest rate models are essential tools for actuaries valuing insurance liabilities with interest-sensitive cash flows. Life insurance reserves, pension obligations, and annuity guarantees all depend critically on future interest rate paths. A single deterministic interest rate assumption is insufficient when the relationship between rates and cash flows is non-linear, as is the case with products containing embedded options or guarantees. Stochastic interest rate models provide the framework for generating realistic scenarios that capture the full range of possible rate environments.
Actuaries encounter interest rate models in several contexts: economic scenario generators (ESGs) for cash flow testing and principle-based reserving, asset-liability management studies, valuation of embedded options and guarantees in insurance products, and enterprise risk management analyses. Understanding the mathematical foundations, practical strengths, and limitations of the major models enables actuaries to select appropriate tools and interpret results with appropriate skepticism.
The Vasicek Model
The Vasicek model describes the short rate as a mean-reverting Ornstein-Uhlenbeck process: dr = a(b - r)dt + sigma dW, where a is the speed of mean reversion, b is the long-run mean rate, sigma is the volatility, and dW is a Wiener process increment. The mean-reverting property is economically intuitive: when rates are above the long-run mean, the drift is negative (pushing rates down), and when rates are below the long-run mean, the drift is positive (pushing rates up). This captures the observed tendency of interest rates to cycle around a central level rather than trending indefinitely in one direction.
The Vasicek model has several analytical advantages. Bond prices, yields, and various option prices can be computed in closed form, making it computationally efficient. The distribution of the short rate at any future time is normal, with mean b + (r_0 - b)e^(-at) and variance (sigma^2 / 2a)(1 - e^(-2at)). However, the normal distribution permits negative interest rates, which was historically considered a weakness but became more realistic after several countries experienced negative rates in the 2010s and 2020s.
The Cox-Ingersoll-Ross (CIR) Model
The CIR model modifies the Vasicek framework by making the volatility proportional to the square root of the current rate: dr = a(b - r)dt + sigma * sqrt(r) * dW. This specification ensures that as the rate approaches zero, the volatility shrinks, preventing the rate from becoming negative (provided the Feller condition 2ab >= sigma^2 is satisfied). The short rate under the CIR model follows a non-central chi-squared distribution, which has a longer right tail than the normal distribution and is bounded below by zero.
The CIR model shares the mean-reverting property of the Vasicek model but produces rate distributions with positive skewness, consistent with the empirical observation that interest rates tend to have longer right tails. Bond prices under the CIR model also have closed-form expressions, though they are more complex than the Vasicek formulas. The model is well-suited for applications where negative rates are inappropriate, such as modeling discount rates for developing-country currencies or historical periods with consistently positive rates.
A practical consideration is parameter estimation. The CIR model's square-root volatility structure means that the model's parameters interact in ways that can make estimation challenging, particularly when historical data spans periods with very different rate environments. Maximum likelihood estimation, method of moments, and Kalman filter approaches are all used in practice, with each method having distinct strengths and data requirements.
The Hull-White Model
The Hull-White model extends the Vasicek framework by replacing the constant long-run mean b with a time-dependent function theta(t): dr = [theta(t) - ar]dt + sigma dW. The function theta(t) is chosen to ensure that the model exactly reproduces the currently observed yield curve, a property known as "fitting the term structure." This calibration capability makes the Hull-White model particularly practical for actuarial applications where consistency with current market prices is required.
The term structure fitting property is important because it ensures that the model's initial output is consistent with observable market data. Without this feature (as in the basic Vasicek or CIR models), the model's implied yield curve may differ from the actual yield curve, creating inconsistencies in valuations that compare model outputs to market prices. The Hull-White model eliminates this inconsistency by construction.
The Hull-White model retains analytical tractability, with closed-form expressions for bond prices and many interest rate derivatives. It can also be extended to include time-dependent volatility, though this reduces analytical tractability and requires numerical methods. The model's flexibility and calibration properties have made it one of the most widely used models in both financial engineering and actuarial practice.
Economic Scenario Generators
Actuaries use interest rate models as components of economic scenario generators (ESGs), which produce correlated scenarios of multiple financial variables including short-term and long-term interest rates, equity returns, credit spreads, and inflation. The ESG generates thousands of scenarios, each representing a possible path for the economy over the projection horizon. These scenarios are used as inputs to cash flow projection models that compute reserves, capital requirements, and other actuarial quantities.
For regulatory applications such as principle-based reserving, the American Academy of Actuaries prescribes specific ESG calibration criteria. These criteria ensure that the generated scenarios are realistic and span a sufficient range of adverse conditions. The Academy's interest rate generator uses a model structure similar to Hull-White but with specific parameterization and calibration requirements. Actuaries must understand these requirements and verify that their ESG output satisfies the prescribed calibration criteria.
For internal risk management applications, companies may use more complex models that incorporate features like stochastic volatility, regime switching, or jump processes. These additional features capture aspects of rate behavior (such as the tendency for volatility to cluster or for rates to shift abruptly during crises) that simpler models may not reproduce. The trade-off is increased computational cost and the need for more sophisticated calibration procedures.
Calibration and Parameter Estimation
Calibrating an interest rate model involves selecting parameters that produce scenarios consistent with either historical experience, current market prices, or both. Historical calibration uses time series of observed rates to estimate the mean reversion speed, long-run mean, and volatility. Market-consistent calibration uses prices of traded instruments (bonds, caps, floors, swaptions) to infer the parameters that reproduce those prices. Risk-neutral calibration, used for pricing and market-consistent valuation, differs from real-world calibration, used for risk assessment and capital modeling.
The distinction between risk-neutral and real-world calibration is fundamental. Risk-neutral parameters are derived from market prices and are used for pricing embedded options and computing market-consistent values. Real-world parameters reflect the actual expected behavior of rates and are used for projecting future financial outcomes, stress testing, and capital modeling. Using the wrong calibration for a given purpose produces meaningless results, making this distinction one of the most important conceptual points for actuaries working with interest rate models.
Study Tips and Exam Connections
Interest rate models appear on SOA exams in the Quantitative Finance and Investment track, the Corporate Finance and ERM track, and on CAS exams covering enterprise risk management. Candidates should be able to write down the stochastic differential equations for the Vasicek, CIR, and Hull-White models; explain the mean-reversion property and its economic motivation; derive or state the distribution of the short rate under each model; and discuss the advantages and limitations of each model for specific actuarial applications. Practice computing bond prices under the Vasicek model using the closed-form formula, and understand how the Hull-White model's term structure fitting property works conceptually. Exam questions often test the ability to select an appropriate model for a given application and to explain the reasoning behind that choice.