Two Different Changes in Reserve
A multi-state reserve calculation involves two kinds of change that are easy to confuse. While a policy remains in one state, its conditional reserve generally moves continuously with time. If the policy changes state, the reserve appropriate to its new state can be different immediately, producing a jump in the state-conditional reserve. A benefit paid at that transition is a separate cash flow; it is not the reserve jump.
For example, a policy may have a small reserve while the insured is active because future care costs are uncertain. If the insured enters a long-term care state, the relevant reserve is then the reserve conditional on being in care. The policy has not gradually moved through the care-state reserve curve before entering care: at the transition, the applicable reserve changes from the active-state value to the care-state value.
The transition strain is a useful accounting measure. A positive value means that paying the transition benefit and establishing the destination-state reserve together increase the insurer’s outgo-plus-reserve position relative to the source-state reserve released. A negative value means the combined position decreases. It is not itself the reserve jump: the reserve jump is only \(V_k(t)-V_j(t)\).
Continuous Evolution Within a State
Use the following convention. The reserve is the expected present value of future benefits and expenses minus future premiums, conditional on the current state. In state \(j\), let \(p_j(t)\) be the premium income rate and \(c_j(t)\) the continuous benefit and expense outgo rate. Let \(\delta\) be the force of interest, and let \(\mu_{jk}(t)\) be the transition intensity from state \(j\) to state \(k\). A transition benefit \(B_{jk}(t)\) is an insurer outgo. The quantities \(p_j\) and \(c_j\) are amounts per unit time; \(B_{jk}\) is an amount per transition.
With these definitions, the state-specific Thiele equation is
The derivative describes the slope of the reserve function for a policy currently in state \(j\), as time advances while conditioning on that state. It does not describe the size of a realized jump at a state transition. The transition term uses the strain \(B_{jk}+V_k-V_j\), with a minus sign: in this backward reserve equation, the expected future transition outgo reduces the derivative relative to the other terms.
Why the Transition Term Has This Sign
The sign can be checked directly over a short interval of length \(dt\). Suppose the policy is in state \(j\) at time \(t\), with transition intensities and cash-flow rates effectively constant over that interval. The probability of a transition \(j\to k\) during the interval is approximately \(\mu_{jk}(t)dt\); the probability of remaining in \(j\) is approximately \(1-\sum_{k\ne j}\mu_{jk}(t)dt\). Discounting one interval at force \(\delta\), the reserve is the immediate continuous net outgo plus the expected discounted reserve and transition outgo at the interval’s end.
Use \(V_j(t+dt)=V_j(t)+V_j'(t)dt+o(dt)\), and the corresponding expansion for \(V_k\). Also use \(e^{-\delta dt}=1-\delta dt+o(dt)\). Subtract \(V_j(t)\), divide by \(dt\), and let \(dt\) tend to zero. The resulting equation is
Rearranging gives the stated Thiele equation. In particular, the transition contribution is \(-\mu_{jk}(B_{jk}+V_k-V_j)\), which is the negative of the transition strain multiplied by its intensity.
Worked Examples: Following a Reserve Across a Transition
Worked Example: A Care-State Entry Changes the Applicable Reserve
Consider a hypothetical policy with two states: active \(A\) and disabled \(D\). The policy can move from \(A\) to \(D\), but cannot recover. The transition intensity is constant at \(1\) per year. Interest is zero, there are no premiums or transition benefits, and the disabled-state benefit is paid continuously at \(4\) dollars per year. The term is \(T=1\) year, and both state reserves are zero at the term. Calculate the conditional reserves at \(t=0.5\), then the reserve jump if an observed policy enters disability at that time.
Method. First calculate the reserve conditional on being disabled. Since there are no future transitions out of \(D\), it is the remaining certain continuous benefit:
Thus \(V_D(0.5)=4(0.5)=2\) dollars. In state \(A\), the Thiele equation is \(V_A'=V_A-V_D\), with \(V_A(1)=0\). Equivalently, a transition occurring \(u\) years after time \(t\) has density \(e^{-u}\) under the constant intensity of \(1\), and the remaining disabled benefit at that time has value \(4(1-t-u)\). Therefore, writing \(x=1-t\),
At \(t=0.5\), \(x=0.5\), so
To the nearest cent, the active-state reserve is \(0.43\) dollars and the disabled-state reserve is \(2.00\) dollars. If the policy transitions from \(A\) to \(D\) at that time, the reserve jump is \(2.00-0.42612264=1.57387736\), or \(1.57\) dollars to the nearest cent. There is no transition benefit in this example, so the combined transition strain is also \(1.57\) dollars.
Check. The source-state reserve is below the destination-state reserve, so the jump is positive, as the calculation shows. A common mistake is to call \(V_A'(0.5)\) the jump. In fact \(V_A'(0.5)=V_A(0.5)-V_D(0.5)=-1.57387736\) dollars per year: that is the instantaneous slope of the active-state reserve function, not the positive \(1.57\)-dollar jump upon entering disability.
Worked Example: Separating the Reserve Jump from a Lump-Sum Benefit
At a specified time, a hypothetical policy moves from state \(R\) to state \(C\), with no other transition possible at that instant. The prospective reserves calculated on the valuation basis are \(V_R(t)=60\) dollars and \(V_C(t)=90\) dollars. A transition benefit of \(12\) dollars is payable on entry to \(C\). The effective annual interest rate is \(3\%\), and the transition intensity is \(0.20\) per year. Find the reserve jump and the combined transition strain.
Method. The rate assumptions govern the prospective valuation but do not change how the already-calculated conditional reserve values are compared at the event. The reserve jump is destination minus source:
The benefit is a separate outgo. The combined transition strain is
Thus the reserve itself increases by \(30.00\) dollars, while the transition benefit plus the reserve increase is \(42.00\) dollars. The intensity-weighted strain appearing in the Thiele equation is \(0.20(42)=8.40\) dollars per year, with a negative sign in the derivative equation. The intensity is not multiplied into the realized \(30\)-dollar jump: it weights the expected transition strain in the continuous-time reserve equation.
Check. The insurer pays \(12\) dollars and the destination reserve is \(30\) dollars greater than the source reserve, giving a total increase of \(12+30=42\) dollars. A common mistake is to report \(42\) dollars as the reserve jump; that incorrectly combines the cash benefit with the reserve change.
Worked Example: The Reserve Slope and the Transition Jump Can Have Different Signs
A hypothetical policy is currently active in state \(A\), with one possible transition to state \(B\). At time \(t\), its conditional reserves are \(V_A(t)=20\) dollars and \(V_B(t)=50\) dollars. The transition benefit is \(5\) dollars. The force of interest is \(0.04\), the transition intensity is \(0.10\) per year, active-state premium income is \(3\) dollars per year, and active-state continuous outgo is \(2\) dollars per year. Find the slope of \(V_A\) at \(t\), the reserve jump if the transition occurs, and the combined transition strain.
Method. Apply the state-\(A\) Thiele equation. The interest and net continuous cash-flow contribution is
The transition strain is
Consequently, the transition contribution to the derivative is \(-0.10(35)=-3.50\) dollars per year, and
If the transition actually occurs, the reserve jump is \(V_B-V_A=50-20=30\) dollars. Including the \(5\)-dollar benefit, the combined transition strain is \(35\) dollars. Thus the reserve function slopes downward at \(1.70\) dollars per year while the observed transition produces an upward reserve jump of \(30\) dollars.
Check. The derivative has units of dollars per year, whereas the jump and benefit have units of dollars. The negative derivative is consistent with a positive expected transition strain in the Thiele equation. Adding the transition strain with a positive sign would incorrectly give \(1.80+3.50=5.30\) dollars per year.
Reading a Realized Reserve Path
A realized policy path consists of intervals spent in particular states, separated by transition times. On each interval, use the reserve function for the state occupied and its derivative from that state’s Thiele equation. At a transition, switch to the destination-state reserve at the same time and record any transition benefit separately.
Use that state’s reserve function \(V_j(t)\) to describe the conditional reserve immediately before a transition.
Between events, the reserve changes with time according to the Thiele derivative for the state occupied. This is a rate of change, not a transition payment.
For a transition \(j\to k\) at time \(t\), calculate the reserve jump \(V_k(t)-V_j(t)\).
Add \(B_{jk}(t)\) to the reserve jump only when calculating the combined transition strain \(B_{jk}(t)+V_k(t)-V_j(t)\).
This distinction matters when interpreting reserve graphs, checking a state-by-state calculation, or explaining a transition term in Thiele’s equation. The functions \(V_j(t)\) describe conditional liabilities, not a single smooth curve followed by every policy regardless of its state. An actual policy’s reserve path can therefore have a discrete jump even when each state-specific reserve function is smooth.
Common Exam Traps
- Calling the reserve jump a derivative. The difference \(V_k(t)-V_j(t)\) is an amount at an event; \(V_j'(t)\) is an amount per unit time along a state-specific curve.
- Including the transition benefit in the reserve jump. The benefit is a cash flow. The reserve jump is destination reserve minus source reserve.
- Ignoring the source reserve. The transition does not create the full destination reserve from nothing; the source-state reserve is replaced. The reserve change is \(V_k-V_j\), not \(V_k\).
- Using the wrong sign in the differential equation. Under the convention here, the transition term is \(-\mu_{jk}(B_{jk}+V_k-V_j)\). Check the cash-flow convention before using a formula from another source.
- Multiplying an observed jump by an intensity. The jump conditional on an actual transition is not intensity-weighted. Intensity weights the expected transition strain per unit time in the differential equation.
Check Your Understanding
For each problem, distinguish amounts at an event from rates of change within a state.
- At a transition \(A\to B\), \(V_A=30\), \(V_B=45\), and the transition benefit is \(4\). Find the reserve jump and combined transition strain. (Answer: reserve jump \(=15\) dollars; strain \(=19\) dollars.)
- In state \(D\), a policy has no further transitions or premiums, pays a continuous benefit of \(6\) dollars per year, and has a term at \(T=3\). Interest is zero and the terminal reserve is zero. Find \(V_D(2)\). (Answer: \(6.00\) dollars.)
- For a transition \(A\to B\), let \(V_A=30\), \(V_B=45\), \(B_{AB}=4\), \(\delta=0.02\), \(\mu_{AB}=0.10\), \(p_A=2\), and \(c_A=1\). Find \(V_A'\) using the stated Thiele convention. (Answer: \(-0.30\) dollars per year.)
- A policy can move from active state \(A\) to care state \(C\), with constant transition intensity \(0.5\) per year. There is no interest, no transition benefit, and no active-state cash flow. In state \(C\), a continuous benefit of \(5\) dollars per year is paid until \(T=2\); there are no transitions out of \(C\). Find \(V_A(1)\) and \(V_C(1)\). (Answer: \(V_A(1)=1.07\) dollars and \(V_C(1)=5.00\) dollars, to the nearest cent.)
- At a transition \(R\to S\), the reserves are \(V_R=100\), \(V_S=70\), and the transition benefit is \(10\). Given \(\delta=0.05\), \(\mu_{RS}=0.20\), \(p_R=4\), and \(c_R=3\), calculate \(V_R'\) and the actual reserve jump. (Answer: \(V_R'=10.00\) dollars per year; reserve jump \(=-30.00\) dollars.)