Disability Income Insurance and Multi-State Models
Price disability income insurance using multi-state Markov models for Exam LTAM.
Disability Income (DI) Model
Disability income insurance replaces a portion of the insured's income during periods of disability, typically paying 60% to 70% of pre-disability earnings. The standard modeling framework is a three-state continuous-time Markov model with states: active (state 0), disabled (state 1), and dead (state 2). Unlike life insurance where the only decrement is death, DI insurance features transitions in both directions between the active and disabled states, making it a genuinely multi-state problem.
The transition intensities governing this model are: sigma_x (active to disabled), which represents the disability incidence rate; rho_x (disabled to active), the recovery rate; mu_x^a (active to dead), mortality for healthy individuals; and mu_x^d (disabled to dead), mortality for disabled individuals. Mortality among disabled lives is typically higher than among active lives of the same age, reflecting the health conditions underlying the disability.
Elimination Period and Benefit Period
The elimination period (also called the waiting period) is the duration of continuous disability that must elapse before benefits begin. Common elimination periods are 30, 60, 90, or 180 days. A longer elimination period reduces premiums because it eliminates short-duration claims and shifts more of the initial disability cost to the policyholder. The elimination period functions like a deductible in property insurance, screening out minor claims.
The benefit period specifies how long benefits continue once the elimination period is satisfied. Benefit periods range from 2 years to "to age 65" or even lifetime. A 2-year benefit period covers short-term disabilities, while a "to age 65" benefit period protects against long-term career-ending disabilities. The benefit period directly affects the expected claim duration and therefore the premium. Longer benefit periods increase the premium substantially because they expose the insurer to extended claims with lower recovery probabilities.
Computing Transition Probabilities
The transition probabilities t_p_x^{ij} give the probability of being in state j at age x+t, given the individual is in state i at age x. For the three-state DI model, there are four relevant transition probabilities: t_p_x^{00} (active to active), t_p_x^{01} (active to disabled), t_p_x^{10} (disabled to active), and t_p_x^{11} (disabled to disabled). The probabilities of being in state 2 (dead) from either starting state are determined by the complement.
These probabilities satisfy Kolmogorov's forward equations, a system of differential equations involving the transition intensities. For a Markov model with constant intensities within each year, the equations can be solved analytically or numerically. The solution involves matrix exponentials: the matrix of transition probabilities over time t is P(t) = exp(Q * t), where Q is the generator matrix with off-diagonal entries equal to the transition intensities and diagonal entries equal to the negative sum of outgoing intensities.
Premium Calculation
The net annual premium P satisfies the equivalence principle: P * a-bar_x^{00} = b * a-bar_x^{01}. Here a-bar_x^{00} = integral from 0 to n of v^t * t_p_x^{00} dt is the expected present value of a continuous annuity of 1 per year payable while in the active state, and a-bar_x^{01} = integral from 0 to n of v^t * t_p_x^{01} dt is the expected present value of a continuous annuity of 1 per year payable while in the disabled state (starting from the active state). The benefit rate b represents the annual disability income benefit.
Solving for the premium gives P = b * a-bar_x^{01} / a-bar_x^{00}. The ratio a-bar_x^{01} / a-bar_x^{00} captures both the probability of becoming disabled and the expected duration of disability, relative to the expected premium-paying period. Computing these annuity values requires the transition probabilities obtained from solving Kolmogorov's equations.
Reserves by Current State
Reserves for DI insurance depend on the insured's current state because the future expected cash flows differ dramatically depending on whether the individual is currently active or disabled. The active reserve (state 0) at time t is: tV^{(0)} = b * a-bar_{x+t}^{01} - P * a-bar_{x+t}^{00}. This is the prospective reserve given that the insured is healthy at time t.
The disabled reserve (state 1) at time t is: tV^{(1)} = b * a-bar_{x+t}^{11} - P * a-bar_{x+t}^{10}. Here a-bar_{x+t}^{11} is the expected present value of future benefits while remaining disabled, and a-bar_{x+t}^{10} is the expected present value of future premiums if the insured recovers. The disabled reserve is typically much larger than the active reserve because the insured is currently receiving benefits and the probability of continued disability is high.
Thiele's Equations for DI
Thiele's differential equations describe how reserves change continuously over time. For the three-state DI model, there is one equation for each state. For state 0 (active): d/dt tV^{(0)} = delta * tV^{(0)} + P - sigma_{x+t} * (tV^{(1)} - tV^{(0)}) - mu_{x+t}^a * (0 - tV^{(0)}). For state 1 (disabled): d/dt tV^{(1)} = delta * tV^{(1)} - b + rho_{x+t} * (tV^{(0)} - tV^{(1)}) - mu_{x+t}^d * (0 - tV^{(1)}). These equations show how the reserve changes due to interest accumulation, premium or benefit flows, and transitions between states.
Common Pitfalls
A frequent mistake is forgetting that reserves depend on the current state. A DI policy has two reserves at each duration, one for each living state. The problem must specify the current state, or you must compute both. Another common error is omitting the recovery transition when computing disabled life annuity values. The possibility of recovery reduces the expected claim duration and therefore the disabled reserve.
Students also sometimes confuse the elimination period with the benefit period. The elimination period is a waiting time before benefits start; the benefit period is how long benefits continue once started. Both affect pricing, but in different ways.
Study Tips
Exam LTAM tests the three-state disability model extensively. Practice setting up the generator matrix Q from given transition intensities, computing transition probabilities by solving Kolmogorov's equations, and using those probabilities to find multi-state annuity values. Be comfortable computing both active and disabled reserves and interpreting their relative magnitudes. Understand Thiele's equations for each state and how they relate to the reserve recursion. Work through complete pricing examples that go from transition intensities to final premium determination.