Partial Disability and Graded Benefits

Why Partial Disability Requires Its Own State

A disability income contract may pay different amounts depending on the claimant’s condition. A claimant who can perform some, but not all, occupational duties may qualify for a partial benefit; a claimant unable to perform any covered duties may qualify for a larger total-disability benefit. The valuation must reflect both which benefit state the claimant occupies and how likely the claimant is to enter or leave that state.

For this tutorial, a graded benefit means that the payment rate depends on the current disability state. It does not mean that the benefit automatically increases with the duration of a claim. A benefit that changes over time is a different feature. The benefit amounts and transition assumptions below are hypothetical.

Definition: A partial-disability model assigns mutually exclusive states to the insured, such as active $A$, partially disabled $P$, totally disabled $T$, and dead $D$. A transition specifies a permitted movement between states. Under a continuous-time model, $\mu_{ij}(t)$ is the instantaneous transition intensity from state $i$ to state $j$ at time $t$, conditional on being in state $i$ immediately before $t$. A state-dependent benefit rate $b_j(t)$ is the amount payable per unit time while the insured occupies state $j$.

The state definitions must match the contract. For example, the model must specify which functional limitations qualify for partial disability and whether partial disability can progress to total disability or recover to the active state. States should not overlap: at any instant, the insured occupies one state, not both partial and total disability.

A common transition structure permits $A\to P$, $A\to T$, $P\to A$, $P\to T$, and transitions from disabled states to death. Other contracts may permit additional movements, such as recovery from total disability to partial disability. A transition omitted from the model is assumed not to occur; that is a substantive assumption, not merely a diagramming choice.

Valuing Continuous Benefits by State

Let $p_j(t)$ be the probability that the insured is in state $j$ at time $t$, given the specified initial state. Suppose benefit payments are made continuously at rate $b_j(t)$ in state $j$, and let $\delta$ be the constant force of interest. The expected present value over a term of $n$ years is the sum of the discounted expected payments in each benefit state.

\operatorname{EPV} = \int_0^n e^{-\delta t} \sum_{j\in\{P,T\}} p_j(t)b_j(t)\,dt.

The sum includes only states in which a benefit is payable. For example, if the contract pays nothing while active or dead, those states contribute zero. A death benefit, if present, would be valued as a separate payment at the death transition rather than as a continuous disability income rate.

Justification: Let $I_j(t)$ equal 1 when the insured is in benefit state $j$ at time $t$, and 0 otherwise. The present value of continuous payments is $\int_0^n e^{-\delta t}\sum_j b_j(t)I_j(t)\,dt$. Taking expectations and using $E[I_j(t)]=P(\text{state }j\text{ at }t)=p_j(t)$ gives the EPV formula. Interchanging expectation and integration is valid here when the benefit rates are bounded over the finite term, as in the examples below.

This result shows why the benefit amount alone is not enough. A larger total-disability rate may have a small contribution to the EPV if the probability of occupying the total-disability state is small. Conversely, even a modest partial benefit can have a material value if partial disability is relatively likely or persistent.

Worked Example: A Partial Benefit with Recovery

A hypothetical one-year contract starts with an active insured. The only transitions are $A\to P$ at constant intensity $\lambda=0.2$ per year and $P\to A$ at constant intensity $\mu=0.3$ per year. There is no death or total-disability state in this simplified model. The contract pays continuously at $ \$12{,}000$ per year while partially disabled. The force of interest is $\delta=0.1$. Find the expected present value, rounded to the nearest cent.

Method. First find the probability of partial disability at time $t$, then integrate the discounted benefit rate over the one-year term. Let $p_P(t)$ be this probability. Over a short interval, entry into $P$ occurs from $A$, while exit from $P$ occurs through recovery. Thus

\frac{dp_P(t)}{dt} =\lambda(1-p_P(t))-\mu p_P(t) =0.2-0.5p_P(t),\qquad p_P(0)=0.

Solving this differential equation gives

p_P(t)=\frac{0.2}{0.5}\left(1-e^{-0.5t}\right) =0.4\left(1-e^{-0.5t}\right).

Therefore, the expected present value is

\begin{aligned} \operatorname{EPV} &=12{,}000\int_0^1 e^{-0.1t}\,0.4(1-e^{-0.5t})\,dt\\ &=4{,}800\left[ \frac{1-e^{-0.1}}{0.1} -\frac{1-e^{-0.6}}{0.6} \right]\\ &=4{,}800(0.95162582-0.75198061)\\ &=4{,}800(0.19964521)\\ &=\$958.30. \end{aligned}

The calculation uses the partial-state occupancy probability at each time, not the probability of ever becoming partially disabled. Someone may enter $P$, recover, and later re-enter it; the occupancy probability accounts for the state at time $t$. The result is nonnegative and is less than $ \$12{,}000$, the maximum possible undiscounted payment over one year, which is a useful reasonableness check.

Grading Benefits Across Partial and Total Disability

When a model has several benefit states, each state has its own occupancy probability and benefit rate. Consider a hypothetical progression-only model with states $A$, $P$, and $T$: transitions occur from active to partial disability at intensity $\lambda$, and from partial to total disability at intensity $\alpha$. There is no recovery or death during the valuation term. If the insured starts active, the probability of being active at time $t$ is $e^{-\lambda t}$.

The probability of being in $P$ is found by considering each possible time of entry from $A$. If entry occurs at time $s$, the probability of remaining in $P$ until time $t$ is $e^{-\alpha(t-s)}$. Hence

p_P(t) =\int_0^t \lambda e^{-\lambda s}e^{-\alpha(t-s)}\,ds =\frac{\lambda}{\alpha-\lambda} \left(e^{-\lambda t}-e^{-\alpha t}\right), \qquad \alpha\ne\lambda.

The three state probabilities must sum to 1, so $p_T(t)=1-p_A(t)-p_P(t)$. The formula for $p_P(t)$ has a limiting form $\lambda t e^{-\lambda t}$ when $\alpha=\lambda$; the displayed quotient should not be used with a zero denominator.

Worked Example: Different Rates for Partial and Total Disability

Use the progression-only model with $\lambda=0.2$ and $\alpha=0.4$ per year. A hypothetical contract pays continuously at $ \$6{,}000$ per year in $P$ and $ \$12{,}000$ per year in $T$. The term is one year and $\delta=0.1$. Find the total expected present value, rounded to the nearest cent.

Method. Calculate each state occupancy probability, then value the two state-specific payment streams separately. Since $\alpha-\lambda=0.2$,

p_A(t)=e^{-0.2t},\qquad p_P(t)=\frac{0.2}{0.2}(e^{-0.2t}-e^{-0.4t}) =e^{-0.2t}-e^{-0.4t}.

The total-disability probability is

p_T(t)=1-p_A(t)-p_P(t) =1-2e^{-0.2t}+e^{-0.4t}.

For compact arithmetic, define $J(k)=(1-e^{-k})/k$. The required values are $J(0.1)=0.95162582$, $J(0.3)=0.86393926$, and $J(0.5)=0.78693868$. Discounting the partial benefit gives

\begin{aligned} \operatorname{EPV}_P &=6{,}000\int_0^1 e^{-0.1t}(e^{-0.2t}-e^{-0.4t})\,dt\\ &=6{,}000[J(0.3)-J(0.5)]\\ &=6{,}000(0.86393926-0.78693868)\\ &=\$462.00. \end{aligned}

For total disability,

\begin{aligned} \operatorname{EPV}_T &=12{,}000\int_0^1 e^{-0.1t}(1-2e^{-0.2t}+e^{-0.4t})\,dt\\ &=12{,}000[J(0.1)-2J(0.3)+J(0.5)]\\ &=12{,}000(0.95162582-1.72787852+0.78693868)\\ &=\$128.23. \end{aligned}

Adding the two mutually exclusive benefit-state values gives $\$462.00+\$128.23=\boxed{\$590.23}$. As a check, $p_A(t)+p_P(t)+p_T(t)=1$ for every $t$, and both disability probabilities are nonnegative over the term. The method avoids treating the total-disability payment as if it were payable during partial disability.

Annual Payments and State Probabilities at Payment Dates

Not every contract pays continuously. In a discrete-time model, let $p_j(k)$ denote the probability of being in state $j$ at the $k$-th payment date. If a benefit $B_j(k)$ is paid at that date whenever the insured is in state $j$, then the expected present value is the sum of discounted expected payments. With effective annual interest rate $i$, the discount factor is $v=1/(1+i)$.

\operatorname{EPV} =\sum_{k=1}^{n}v^k\sum_{j\in\{P,T\}}p_j(k)B_j(k).

This formula assumes the payment is determined by the state at the payment date. If a contract instead pays based on the state during the preceding year, or pays a prorated amount after a transition, the payment rule must be represented differently. State probabilities and benefit timing must use the same convention.

Worked Example: Updating State Probabilities Before Each Payment

A hypothetical annual-payment contract has states $A,P,T,D$, with annual transition probabilities given by the following matrix. Rows are the state at the start of a year and columns are the state at its end. State $D$ is absorbing.

Q= \begin{pmatrix} 0.80&0.20&0&0\\ 0.25&0.50&0.25&0\\ 0.10&0&0.80&0.10\\ 0&0&0&1 \end{pmatrix}.

The insured starts active. At each year-end, the contract pays $ \$4{,}000$ if the insured is in $P$, $ \$10{,}000$ if in $T$, and nothing in $A$ or $D$. The effective annual interest rate is $i=0.05$. Find the expected present value of the first two payments, rounded to the nearest cent.

Method. Multiply the initial state-probability row vector by $Q$ once for the first payment date and twice for the second. At year 1,

$$ (1,0,0,0)Q=(0.80,0.20,0,0). $$

The expected year-1 payment is $0.20(4{,}000)=\$800$, so its present value is $800/1.05=\$761.90$. At year 2, the state probabilities are

$$ (0.80,0.20,0,0)Q =(0.69,0.26,0.05,0). $$

These probabilities sum to 1. The expected year-2 payment is $0.26(4{,}000)+0.05(10{,}000)=\$1{,}540$. Its present value is $1{,}540/(1.05)^2=\$1{,}396.83$. Thus

\operatorname{EPV}=\frac{800}{1.05}+\frac{1{,}540}{1.05^2} =\$761.90+\$1{,}396.83 =\boxed{\$2{,}158.73}.

The transition matrix is applied again before valuing the second payment because the state distribution changes from year 1 to year 2. Reusing the year-1 probabilities for both payments would understate the chance of progressing from partial to total disability.

Common Errors and Practical Checks

State-based valuation is useful because it separates two sources of value: the benefit payable in each state and the probability of occupying that state. It also makes assumptions visible. A model may assume recovery is possible from partial disability but not total disability, or it may allow progression between those states. Such choices can affect both the expected benefit and its timing.

A frequent exam error is to apply the total-disability benefit to every disabled state. In the annual example, that mistake would value the first payment as $0.20(10{,}000)/1.05=\$1{,}904.76$, rather than $0.20(4{,}000)/1.05=\$761.90$. The correction is to match each state probability with its own contractual amount. Another error is to use the first-year state probabilities at every later payment date; update the state distribution through the transition model before each payment.

Check the state definitions. Confirm that partial and total disability are mutually exclusive and reflect the hypothetical contract’s eligibility rules.
Check transition assumptions. A missing recovery, progression, or death transition may materially change occupancy probabilities.
Check payment timing. Continuous payments require integration; annual payments require discounting each payment date.
Check the benefit grade. Apply the partial rate in $P$ and the total rate in $T$, not one rate to all disabled states.
Check probabilities. State probabilities should be nonnegative and sum to 1, including any dead or other absorbing state.

Key takeaway: Value a graded disability income benefit by pairing each payment amount with the probability of occupying its corresponding state at the time the payment is made. The transition model determines those probabilities; the contract determines the state-specific benefits and payment timing.

Check Your Understanding

For each problem, state the key probability or payment calculation before checking the final answer.

A one-year continuous model starts active. The only transitions are $A\to P$ with intensity $0.1$ and $P\to A$ with intensity $0.2$, both per year. The partial benefit is $ \$9{,}000$ per year and $\delta=0.05$. Find the EPV to the nearest cent. (Answer: $\$394.99$)
At the end of one year, a hypothetical annual model has probabilities $0.80$ active, $0.15$ partially disabled, and $0.05$ totally disabled. The respective year-end benefits are $ \$0$, $ \$3{,}000$, and $ \$8{,}000$. At $i=0.04$, find the EPV of that payment to the nearest cent. (Answer: $\$817.31$)
A one-year continuous model starts active, with $A\to P$ intensity $0.25$ and $P\to A$ intensity $0.25$. There is no interest and the benefit rate in $P$ is $ \$8{,}000$ per year. Find the EPV to the nearest cent. (Answer: $\$852.25$)
A two-year annual model has states $A,P,T,D$. Its transition rows are $A:(0.70,0.30,0,0)$, $P:(0.20,0.50,0.30,0)$, $T:(0.10,0,0.80,0.10)$, and $D:(0,0,0,1)$. It starts in $A$, pays $ \$5{,}000$ in $P$ and $ \$9{,}000$ in $T$ at each year-end, and has $i=0.04$. Find the EPV of the first two payments to the nearest cent. (Answer: $\$3{,}855.40$)

What You'll Learn