Why More Initial Capital Reduces Ruin

Why Compare Initial Capital Path by Path?

Initial capital is a buffer against claims. The precise statement is that, when two otherwise identical surplus processes experience the same premium income and the same claims, the process starting with more capital cannot be ruined on a path where the process starting with less capital survives. This is stronger than observing that a larger starting reserve is associated with a smaller ruin probability: it identifies the event-by-event reason for the comparison.

The comparison concerns initial capital alone. The premium rate, claim arrivals, and claim amounts must be held fixed across the two processes. The proof does not require claims to be small, a particular claim-size distribution, or a condition on expected premium income. Those assumptions may matter for other questions about ruin, but not for this pathwise ordering.

Definition: In the classical surplus model, let $u\geq 0$ be initial capital, $c$ the constant premium rate, $N(t)$ the number of claims by time $t$, and $X_1,X_2,\ldots$ the claim amounts. The aggregate claims by time $t$ are $S(t)=\sum_{j=1}^{N(t)}X_j$, and the surplus is $$ U_u(t)=u+ct-S(t). $$ Using the strict-negative convention, the ruin time is $T_u=\inf\{t\geq 0:U_u(t)<0\}$, and the ultimate ruin probability is $\psi(u)=P(T_u<\infty)$. Thus, ruin occurs if the surplus is negative at some finite time; reaching exactly zero is not, by itself, ruin.

For the classical compound Poisson model, one commonly assumes $N(t)$ is a Poisson process with rate $\lambda>0$, the claim amounts are independent and identically distributed nonnegative random variables, and the claims are independent of $N(t)$. The monotonicity result below needs less: it only needs both capital levels to use the same realized claim arrivals and amounts, as well as the same premium income path. In the classical model, premiums accrue between claims and claims cause downward jumps, so ruin can be checked at claim times.

The Coupling Argument

A coupling is a way to place two random processes on the same probability space so that their shared sources of randomness can be compared directly. Here, use the same claim count $N(t)$, the same claim amounts $X_j$, and the same premium rate $c$ for both initial capital levels. No new claim model is introduced for the higher-capital process.

Theorem: For $0\leq u_1\leq u_2$, under the same premium and claim experience, $\psi(u_2)\leq\psi(u_1)$. In words, ultimate ruin probability is nonincreasing in initial capital.

The key identity is that the difference between the two surplus paths never changes. For every $t\geq 0$,

U_{u_2}(t)-U_{u_1}(t) =(u_2+ct-S(t))-(u_1+ct-S(t)) =u_2-u_1\geq 0.

Therefore, at any time the higher-capital path is at least as high as the lower-capital path. If the higher-capital path is negative at some finite time, the lower-capital path is negative at that same time. This gives the event inclusion

\{T_{u_2}<\infty\}\subseteq \{T_{u_1}<\infty\}.

Taking probabilities of these nested events yields the theorem. This reasoning also applies when ruin can occur only at claim times: at every claim time the same claim is subtracted from each path, and the difference in surplus remains $u_2-u_1$. The argument is complete without calculating either ruin probability.

Share the experience.
Use identical claim arrival times, claim amounts, and premium income for both initial capitals.

Compare the paths.
At every time, the higher-capital surplus equals the lower-capital surplus plus the fixed amount $u_2-u_1$.

Compare the ruin events.
If the higher path is ever negative, the lower path must be negative then too. Hence the higher-capital ruin event is a subset of the lower-capital ruin event.

Take probabilities.
Event inclusion gives $\psi(u_2)\leq\psi(u_1)$.

Worked Comparisons

The theorem establishes an ordering, not the numerical size of a ruin probability. The following examples calculate probabilities in deliberately specified models and check that the pathwise comparison is respected.

Worked Example: One Claim and Two Capital Levels

Consider a hypothetical one-claim model. The only claim occurs at time $1$; premium income accrues at rate $2$ dollars per unit time, and there are no other claims. The claim is $6$ dollars with probability $0.4$ and $10$ dollars with probability $0.6$. Find the ultimate ruin probabilities for $u=3$ and $u=7$, using strict-negative ruin.

Method. Calculate the surplus immediately before the claim, subtract each possible claim, and add the probabilities of the claim outcomes that produce negative surplus. Since the model has no later claims, this determines ultimate ruin.

For $u=3$, the pre-claim surplus is $3+2(1)=5$ dollars. A $6$-dollar claim leaves $5-6=-1$ dollar, and a $10$-dollar claim leaves $5-10=-5$ dollars. Both outcomes cause ruin, so

\psi(3)=0.4+0.6=1.

For $u=7$, the pre-claim surplus is $7+2(1)=9$ dollars. A $6$-dollar claim leaves $9-6=3$ dollars, so it does not cause ruin. A $10$-dollar claim leaves $9-10=-1$ dollar, so it does. Thus

\psi(7)=0.6.

The check is $0.6\leq 1$, as required when initial capital increases. Common mistake avoided: omitting the $6$-dollar claim from the ruin calculation at $u=3$. It exceeds the $5$-dollar pre-claim surplus and therefore causes strict-negative ruin; the probability at $u=3$ is $1$, not $0.6$.

Worked Example: Two Claims and Three Capital Levels

A hypothetical process has exactly two claims, at times $1$ and $2$, with premium income at rate $1$ dollar per unit time. Each claim independently equals $2$ dollars or $5$ dollars, each with probability $0.5$. Find the ultimate ruin probabilities for initial capitals $u=1,4,7$, using strict-negative ruin.

Method. Examine each claim sequence and check surplus immediately after each claim. Between claims surplus increases, so it cannot become negative between these downward jumps. Each of the four ordered claim pairs has probability $0.5(0.5)=0.25$.

For $u=1$, the surplus just before the first claim is $1+1=2$. A first claim of $5$ causes ruin immediately. If the first claim is $2$, the surplus becomes $0$, which is not strict-negative ruin. Just before the second claim in that case, the surplus is $0+1=1$; either possible second claim, $2$ or $5$, causes ruin. All four pairs therefore lead to ruin, and $\psi(1)=4(0.25)=1$.

For $u=4$, the first pre-claim surplus is $4+1=5$. A first claim of $5$ leaves $0$, not ruin; a first claim of $2$ leaves $3$. If the first claim is $2$, the pre-second-claim surplus is $3+1=4$: a second claim of $2$ leaves $2$, while a second claim of $5$ causes ruin. If the first claim is $5$, the pre-second-claim surplus is $0+1=1$, and either second claim causes ruin. Exactly three of four pairs cause ruin, so $\psi(4)=3(0.25)=0.75$.

For $u=7$, the first pre-claim surplus is $8$. A first claim of $2$ leaves $6$, and the pre-second-claim surplus is $7$; either second claim leaves a nonnegative surplus. A first claim of $5$ leaves $3$, and the pre-second-claim surplus is $4$; a second claim of $2$ leaves $2$, while a second claim of $5$ causes ruin. Only the pair $(5,5)$ causes ruin, giving $\psi(7)=0.25$.

The results satisfy $0.25\leq0.75\leq1$. Common mistake avoided: treating a surplus of exactly zero after a claim as ruin. At $u=4$, for example, a first claim of $5$ leaves zero, so the process must be checked at the second claim rather than declared ruined immediately.

Worked Example: A Random Claim Time

In another hypothetical one-claim model, the claim arrives at a time $T$ with an exponential distribution of rate $1$ per unit time. Its amount is fixed at $4$ dollars, premium income accrues at $2$ dollars per unit time, and there are no further claims. Calculate ultimate ruin probabilities for $u=1$ and $u=3$.

Method. The pre-claim surplus is $u+2T$. Since the claim amount is exactly $4$, strict-negative ruin occurs when $u+2T-4<0$, or $T<(4-u)/2$. For an exponential random variable of rate $1$, $P(T<t)=1-e^{-t}$ for $t\geq0$.

At $u=1$, ruin occurs when $T<3/2=1.5$. Therefore,

\psi(1)=1-e^{-1.5}\approx 0.7769.

At $u=3$, ruin occurs when $T<1/2=0.5$, giving

\psi(3)=1-e^{-0.5}\approx 0.3935.

Probabilities are rounded to four decimal places; $0.3935\leq0.7769$ confirms the expected ordering. Equality at the threshold would leave surplus zero, but here $T$ has a continuous distribution, so the probability of arriving at exactly either threshold is zero. Common mistake avoided: comparing the claim amount only with initial capital. Premiums earned before the claim contribute $2T$ dollars to the available surplus.

What Monotonicity Does—and Does Not—Say

The result is nonincreasing, not necessarily strictly decreasing. If every possible claim path that can ruin the process with capital $u_1$ also ruins it with capital $u_2$, the two ruin probabilities can be equal. For instance, a sufficiently large claim can ruin both capital levels on the same path. The theorem says only that the higher capital cannot create additional ruin paths when all other experience is shared.

The coupling also makes clear why the result does not require a positive safety loading or any other net profit condition. Such conditions concern the long-run balance between premiums and claims. The pathwise identity compares the two reserves regardless of whether premiums are expected to exceed claims. It is therefore a basic ordering result, not a formula for the numerical value of ultimate ruin probability.

A useful exam check is to write the path difference explicitly before attempting a calculation. If the two processes use the same premiums and claims, that difference is the constant $u_2-u_1$. If a proposed calculation gives $\psi(u_2)>\psi(u_1)$ for $u_2>u_1$, check whether the claim paths were changed, whether ruin at zero was treated consistently, or whether probabilities were added incorrectly.

Common Mistakes

Using different claim paths. The proof compares capital levels under shared claim experience. Independent simulations may produce estimates that appear out of order from sampling variation; that does not contradict the pathwise theorem.
Confusing zero with negative surplus. Under the strict-negative convention, a claim that reduces surplus exactly to zero does not cause ruin at that instant. Check later claims before deciding whether the path is ruined.
Claiming strict decrease. Monotonicity permits equal probabilities at different capital levels. It establishes an inequality, not a strict one.
Making the result depend on a premium condition. The proof uses the same premium income in both paths, so the premium level cancels from their difference. No net profit condition is needed for this comparison.

Key takeaway: Couple the two surplus processes using identical premiums and claims. Their surplus difference is always the difference in initial capital, so ruin at the higher capital implies ruin at the lower capital. Consequently, ultimate ruin probability is nonincreasing in initial capital.

Check Your Understanding

For each question, use the stated strict-negative ruin convention and check the ordering whenever two capital levels are compared.

A hypothetical model has exactly one claim at time $2$, premium income at rate $1$ dollar per unit time, and claim amounts $5$ dollars with probability $0.3$ and $7$ dollars with probability $0.7$. Find ultimate ruin probabilities for $u=3$ and $u=5$. (Answer: $\psi(3)=0.7$, $\psi(5)=0$.)
A one-claim model has a claim of $8$ dollars at time $1$, premium income at rate $2$ dollars per unit time, and no other claims. Find the ruin probabilities for $u=2$ and $u=5$. (Answer: $\psi(2)=1$, $\psi(5)=1$.)
Exactly two claims occur at times $1$ and $2$, with premium income at rate $1$ dollar per unit time. Independently, each claim is $1$ dollar or $4$ dollars, each with probability $0.5$. Find ultimate ruin probability for $u=2$. (Answer: $0.75$.)
For the model in question 3, find ultimate ruin probability for $u=5$. (Answer: $0.25$.)
Two processes use the same premium and claim paths, and their initial capitals are $u_1=6$ and $u_2=9$. If the higher-capital process is negative at time $t=4$, what can be concluded about the lower-capital process at that time, and what event relation follows for ultimate ruin? (Answer: $U_{6}(4)=U_{9}(4)-3<0$; the higher-capital ultimate-ruin event is contained in the lower-capital event, so $\psi(9)\leq\psi(6)$.)

What You'll Learn