Ruin Time and Ultimate Ruin Probability

Ruin as a First-Passage Event

An insurer’s surplus changes as premiums arrive and claims are paid. Ruin is not simply the event that a claim is large, nor does it mean that the surplus has reached zero. Under the convention used here, ruin occurs when the surplus first becomes strictly negative. This distinction determines both the ruin time and the ultimate ruin probability.

In the classical surplus model, let $u\geq 0$ be initial surplus, $c>0$ the constant premium income per unit time, and $S(t)$ the aggregate claims paid by time $t$. With no investment income, the surplus at time $t$ is

$$ U(t)=u+ct-S(t). $$

In the compound Poisson version, claim arrivals follow a Poisson process and claim amounts are independent, identically distributed, nonnegative random variables, independent of the arrival process. The definitions below apply to each realized surplus path; the ultimate ruin probability averages over the randomness of that path.

Definition: The ruin time is $$ \tau=\inf\{t\geq 0:U(t)<0\}, $$ where the infimum of the empty set is defined to be $+\infty$. For initial surplus $u$, the ultimate ruin probability is $$ \psi(u)=P(\tau<\infty\mid U(0)=u). $$ Thus $\tau$ is a random time, possibly infinite, while $\psi(u)$ is a probability between 0 and 1.

The strict inequality matters. If $U(t)=0$, the insurer has exhausted its surplus but has not yet entered ruin under this definition. A later claim can make the surplus negative; absent such a later deficit, reaching zero alone does not make $\tau$ finite. If the surplus is never negative, the set in the definition is empty and $\tau=\infty$.

Reading a Surplus Path

Between claims, $S(t)$ is constant and surplus increases linearly at rate $c$. At a claim, surplus drops by the amount paid. To find $\tau$ on a particular path, track the surplus immediately before and after each claim, taking account of all premium income since the preceding time. If no claim or other event makes surplus negative, the ruin time is infinite.

Start from the stated initial surplus.
Set $U(0)=u$, and use the specified premium rate and claim times.

Accumulate premiums between claims.
Over an interval of length $\Delta t$ without claims, surplus increases by $c\Delta t$.

Subtract each claim payment.
A claim of amount $x$ reduces the current surplus by $x$. Record whether the post-claim value is strictly below zero.

Identify the first crossing.
The first time the path becomes negative is $\tau$. If it never does, record $\tau=\infty$.

Worked Example: Finding Ruin Time From Two Claims

A hypothetical fund starts with $\$7$, receives premiums continuously at $\$3$ per year, and pays claims of $\$8$ at $t=1$ and $\$5$ at $t=2.5$. There are no other claims. Find the post-claim surpluses and the ruin time. There is no interest or investment income.

Method: sample-path surplus ledger. Immediately before the first claim, the fund has its initial $\$7$ plus one year of premiums:

$$ U(1-)=7+3(1)=10. $$

After paying the $\$8$ claim, the surplus is $10-8=\$2$. Between $t=1$ and $t=2.5$, premiums add $3(2.5-1)=3(1.5)=\$4.50$. Therefore, just before the second claim, surplus is $2+4.50=\$6.50$, and just after it, surplus is $6.50-5=\$1.50$.

Both post-claim surpluses are nonnegative, and surplus increases between claims. It is never negative on this path, so $\tau=\infty$. Common mistake avoided: counting only one year of premiums between claims would give the wrong second post-claim surplus. The elapsed time is $2.5-1=1.5$ years, not one year.

Check: Directly at $t=2.5$, cumulative premiums are $3(2.5)=\$7.50$, and total claims are $8+5=\$13$. Thus $U(2.5)=7+7.50-13=\$1.50$, matching the ledger.

Worked Example: Reaching Zero Is Not Ruin

A hypothetical insurer starts with $\$4$, receives premiums at $\$2$ per year, and has one claim of $\$7$ at $t=1.5$. There are no further claims. Determine its ruin time under the strict-negative convention. Assume no interest.

Method: compare the claim with the surplus immediately before payment. Premiums earned by $t=1.5$ equal $2(1.5)=\$3$, so

U(1.5-)=4+3=7,\qquad U(1.5)=7-7=0.

The claim leaves exactly zero, not a negative amount. After the claim, premiums continue to accumulate and there are no later claims, so the surplus remains nonnegative. Consequently, the set $\{t\geq0:U(t)<0\}$ is empty and $\tau=\infty$. Common mistake avoided: setting $\tau=1.5$ because the surplus reaches zero would use a non-strict threshold that is not the definition here.

Check: At any $t>1.5$, $U(t)=2(t-1.5)>0$; before the claim it is also nonnegative. No time on the path has negative surplus.

Why Ultimate Ruin Probability Is a Probability

For a fixed set of model assumptions and an initial surplus $u$, different claim histories can produce different ruin times. Some paths may avoid negative surplus forever, while others may eventually cross below zero. The event $\{\tau<\infty\}$ collects exactly those paths with a finite ruin time, so its probability is $\psi(u)$. It is not the probability that surplus is negative at one selected time; it concerns whether a negative surplus occurs at any finite time.

The complementary event is that the ruin time is infinite. These two events partition the possible paths: every extended-real value of $\tau$ is either finite or $+\infty$. Hence

\{\tau<\infty\}^{c}=\{\tau=\infty\}, \qquad \psi(u)=1-P(\tau=\infty\mid U(0)=u).

This identity follows directly from the definition of $\psi(u)$ and the complement rule for probabilities. It also clarifies why a single observed path does not itself determine $\psi(u)$: it tells whether ruin occurs on that path, not the proportion of all possible paths that ruin.

Worked Example: Ultimate Ruin Probability in a One-Claim Model

Consider a hypothetical model that permits exactly one claim. Its arrival time $T$ has an exponential distribution with rate $0.2$ per year, and its claim amount $X$ is independent of $T$ and exponentially distributed with mean $\$4$. The fund starts with $\$3$ and earns premiums continuously at $\$1.50$ per year. After the one claim there are no further claims. There is no interest. Find the ultimate ruin probability.

Method: condition on the claim arrival time. Before the claim, the fund has surplus $3+1.5T$. It is ruined at the claim exactly when $X>3+1.5T$. For an exponential claim amount with mean $4$, the survival function is $P(X>x)=e^{-x/4}$ for $x\geq0$. Conditional on $T=t$, therefore,

P(\text{ruin}\mid T=t)=e^{-(3+1.5t)/4}.

The density of $T$ is $f_T(t)=0.2e^{-0.2t}$ for $t\geq0$. Averaging the conditional probability over $T$ gives

\begin{aligned} \psi(3) &=\int_0^\infty e^{-(3+1.5t)/4}(0.2e^{-0.2t})\,dt\\ &=e^{-0.75}(0.2)\int_0^\infty e^{-(0.375+0.2)t}\,dt\\ &=e^{-0.75}\frac{0.2}{0.575}\\ &=e^{-0.75}\frac{8}{23}\\ &\approx 0.1643. \end{aligned}

Thus the ultimate ruin probability in this one-claim model is $0.1643$, or $16.43\%$. The equality case $X=3+1.5T$ leaves zero and is not ruin; here it has probability zero because $X$ has a continuous distribution. Common mistake avoided: using the probability that a claim exceeds the initial $\$3$ would ignore the premium accumulated before arrival. The threshold depends on $T$.

This is a deliberately specified one-claim model, not the classical model with claims continuing indefinitely. It demonstrates the meaning and calculation of an ultimate probability by conditioning; the claim process assumptions must always match the model in the question.

Common Exam Traps and Interpretation

Using a non-strict threshold. The event defining ruin is $U(t)<0$, not $U(t)\leq0$. A path that only reaches zero has not yet been ruined under this convention.
Forgetting premiums during an interval. If successive claims occur at $t_1$ and $t_2$, the premium earned between them is $c(t_2-t_1)$. It is not generally one full year of premium.
Confusing a time with a probability. $\tau$ describes the first negative-surplus time on a realized path. $\psi(u)$ describes the probability that this time is finite across the model’s possible paths.
Ignoring the model description. A probability calculation depends on the specified arrival process, claim-size distribution, premium rate, and whether claims continue. A one-claim probability is not automatically the ultimate ruin probability for a continuing-claims model.
Calling a surviving observed path proof of zero probability. A path with $\tau=\infty$ establishes only that ruin did not occur on that realization. Other paths may still have a positive probability of ruin.

The distinction between $\tau$ and $\psi(u)$ is foundational for subsequent ruin calculations: first specify the event on a path, then calculate its probability under a stated model. This treatment is aligned with the SOA Exam ASTAM (Advanced Short-Term Actuarial Mathematics) syllabus.

Key takeaway: Ruin time is the first time surplus is strictly negative, with value $+\infty$ if that never happens. Ultimate ruin probability is $P(\tau<\infty\mid U(0)=u)$, evaluated under the stated claim and premium assumptions.

Check Your Understanding

For each question, apply the strict-negative definition and keep the stated time units and model assumptions.

A fund starts at $\$6$, earns $\$2$ per year, and pays a claim of $\$8$ at $t=1$, followed by a claim of $\$3$ at $t=2$. There are no other claims. Find the post-claim surpluses and $\tau$. (Answer: The post-claim surpluses are $\$0$ and $-\$1$; $\tau=2$.)
A fund starts at $\$10$, earns $\$1.25$ per year, and pays $\$6$ at $t=0.8$ and $\$5$ at $t=2$, with no other claims. Find the post-claim surpluses and $\tau$. (Answer: The post-claim surpluses are $\$5.00$ and $\$1.50$; $\tau=\infty$.)
A model has exactly one claim. Its arrival time is exponential with rate $0.1$ per year; the claim amount is fixed at $\$9$. Initial surplus is $\$4$, and the premium rate is $\$2$ per year. Find the ultimate ruin probability. (Answer: $1-e^{-0.25}\approx0.2212$.)
A model has exactly one claim, with arrival time $T$ exponential at rate $0.4$ per year and independent claim amount exponential with mean $\$5$. Initial surplus is $\$2$, and the premium rate is $\$1$ per year. Find the ultimate ruin probability. (Answer: $e^{-0.4}(2/3)\approx0.4469$.)

What You'll Learn