23
Ruin models:

23.1 Introduction

This chapter involves extending some aspects of the collective risk model to a multi-period setting. It will require a sound knowledge of the material in Chapter 18. We begin with the discrete-time case and consider another interpretation of Equation 18.3.

Consider an insurer who each period collects total premiums of c and experiences aggregate claims of ⟨N, X⟩ as defined at the end of Section 21.1. Then the gain of the insurer in the nth period is given by a random variable Gn where

numbered Display Equation

If we assume that claims each period are independent of those in other periods, we can interpret Equation (18.3) as representing a surplus process of the insurer, where Un is the surplus at time n resulting from an initial surplus of u at time 0. This will be a major application for the theory in this chapter, although it applies as well to the original gambling formulation.

Let T be the first time the surplus becomes negative. We call this the time of ruin. In the discrete-time case, we define this formally as

numbered Display Equation

The random variable T is different from the other random variables we have encountered since it is not necessarily real valued. For any realization for which the surplus is nonnegative at all times, the value of T will be ∞. The set of all such realizations can have positive probability, in which case ruin is not certain. We are interested in the probability that ruin will eventually occur. This will, of course, depend on the initial surplus u, so we denote this by ψ(u). That is,

numbered Display Equation

Note that ψ(u) is the probability of eventual ruin, which may seem to be of little interest since nobody is planning to gamble or to run an insurance company forever. From a practical point of view, one may want to compute the probability of ruin over some finite time horizon. That is, one may want the probability that ruin will occur before a fixed time t. We denote this by ψ(u, t). So,

numbered Display Equation

It is difficult to find general methods for calculating this quantity, and normally each case must be treated individually. The following example exhibits some of the possible techniques in a simple discrete time example.

Example 23.1 A special insurance company has a single contract. There can be at most one claim, and the probability that a claim does not occur by time t is 1/(1 + t). If a claim occurs, the amount is 100 with probability 0.6 or 200 with probability 0.4. Premiums are paid continuously at the rate of 20 per year. The company begins with an initial surplus of 60. What is the probability of eventual ruin?

Solution. This is really a finite-time question, since after time 7 ruin cannot occur, for the insurer will have collected the maximum claim amount of 200. We break this interval up into the relevant time periods.

  1. From time 0 to time 2, a claim will occur with probability 2/3, and the insurer will necessarily be ruined since the initial surplus and premiums collected will be under the minimum claim of 100.
  2. From time 2 to time 7, a claim will occur with probability , and ruin will occur only if the claim is for 200. So the probability of ruin in this interval is .

The total probability of ruin is therefore .

When G is finite valued, the quantity ψ(u, t) can be both childishly simple and fiendishly difficult to compute. To illustrate this paradoxical statement, consider an example. You flip a coin with probability of a head equal to p, and you win 1 for a head and lose 1 for a tail. What is ψ(1, 2)? This is answered immediately since the only way you can be ruined by time 2 is to get two tails in a row, and we conclude that ψ(1, 2) = (1 − p)2. In fact, whenever G is finite valued, we can always, in theory, compute ψ(u, t) by simply looking at all possible paths up to time t and seeing which ones lead to ruin. The problem is that if the range of G and the time t are large enough, the number of such paths could be enormous, rendering any computation infeasible. We therefore need other ways to get information. One such method is to compute ψ(u), which gives an upper bound since

numbered Display Equation

The previous discussion then serves as a motivation for the main theme of this chapter, which is to derive methods for calculating the infinite time ruin probability ψ(u). There are several ways to either compute this or estimate it, and we will discuss them in turn. Each is useful for certain cases.

Remark Some authors define ruin as the first time the surplus reaches zero, rather than the first time it becomes negative. Let denote the probability of ruin in this case. Then if either G is continuous or u is not an integer. In the discrete case, if u is a positive integer, then .

23.2 A functional equation approach

Suppose you sit down to gamble with 300 units of capital. You divide this initial stake into two piles, one with 200 and the other with the remaining 100. Now, to be ruined you first have to lose the 200 pile, and following that you have be ruined all over again starting with the 100 pile. So, it seems reasonable to conclude that ψ(300) = ψ(200)ψ(100), or, more generally, since there is nothing special about these particular amounts, that ψ(u + v) = ψ(u)ψ(v). Assuming that our reasoning here is correct, we would know already with no calculation at all (and ruling out certain degenerate cases) that ψ(u) is an exponential function. Unfortunately, our reasoning is not quite accurate. The problem is that the ruining bet, or claim, by definition, will leave us with a deficit. In the original problem, we have to draw on our second pile in order to pay for this, and we would not have the full 100 to continue the procedure. Suppose, we knew that our deficit at the time of ruin was some d > 0. That is, we would have a surplus of − d at ruin, and the appropriate equation would be

numbered Display Equation

The following trick allows us to convert this into the form above. Let ρ(u) = ψ(ud); then,

numbered Display Equation

This is the same functional equation we encountered in Section 2.6. Assuming that we know that ruin probabilities are positive, and assuming some minimal regularity condition, such as continuity at one point, we know that ρ(u) = zu for some z between 0 and 1. Therefore,

Once again, however, we have to question our assumption. Is it ever possible that we could know that the deficit at ruin had to be some fixed number d? The answer is ‘not very often’, but it does happen in one particular case. Take the discrete-time model, for which the values of G are all nonnegative integers except for a single negative value of −1 – as for a simple coin flip where G = 1 or − 1 – and for which the initial surplus u is a positive integer. In this case, the only way to be ruined is to reach a position where your surplus is 0 and then to lose 1 in the next period. The deficit at ruin can only be 1. Formula (23.1) would give us the ruin probabilities if we could only determine z. We will attempt to do so by using a recursive technique that is a basic tool in ruin theory. Let p denote the probability function of G. Suppose you start with an initial surplus of 0. If your gain is − 1 in the first period, you are immediately ruined. If your gain is k in the first period, your subsequent probability of ruin is ψ(k). Considering all possibilities, we have

(23.2) numbered Display Equation

From (23.23), with d = 1,

numbered Display Equation

and dividing this equation by z, we can write

where PG is the p.g.f. of G.

Example 23.2 Consider a game where you win 1 with probability p ⩾ 1/2, and lose 1 with probability 1 − p. What is ψ(u)?

Solution. From (23.23), we have

numbered Display Equation

There are unfortunately two possible solutions to this, either z = (1 − p)/p or z = 1. We do not have a definite answer at this point but can only conclude that either

(23.4) numbered Display Equation

or

numbered Display Equation

Note that z = 1 is a solution of (23.3) for all G, so our method would seem to have accomplished little, leaving us in all cases with a possible conclusion that ruin is certain. We will show however in the following sections that we can often rule out this possibility. This will be in fact be true in the present example for p > 1/2, and we will then know that the probability of ruin is given by (23.23).

We can already deduce an interesting result in the case that p = 1/2. In that case the only root is 1, and we know definitely that ruin is certain, regardless of the initial surplus. (This is of course also true if p < 1/2.) This is one of the well-known results in ruin theory. It says that even if you are playing a perfectly fair game, if you play it long enough, you will eventually lose all your money. This may been somewhat strange since you would seem to be on equal grounds with your opponent (the casino, for example). You are not, however, since there is an implicit assumption that the opponent has unlimited resources at its disposal, while you only have the u units you started with. We will return to a variation on this problem in Example 23.4 below.

To summarize, this section has achieved only limited success in deducing ruin probabilities, but it is mainly intended as a motivation for methods to follow. One point that should be emphasized is that it illustrates the importance of considering the deficit at time of ruin (given that ruin occurs) when trying to deduce ruin probabilities. In the following sections, we will refer frequently to this random variable and denote it by D(u) (D for deficit). That is,

numbered Display Equation

In the example above D(u) was always 1, but in general it is random and can depend on the initial surplus u.

23.3 The martingale approach to ruin theory

23.3.1 Stopping times

We will motivate the concept discussed here by looking at the gambling situation. Some gamblers claim that they can overcome unfavourable odds by a clever ‘system’. This often takes the form of planning to stop at a certain point. For example, they will continue gambling until they have won $100 and then quit. That way, they claim, they are always a winner. Or, they will continue to bet on black until the wheel comes up black four times in a row, and then quit. They are using what is called a stopping time. Intuitively, a stopping time is a rule that tells you when to stop, and it must be such that you know about it when the time occurs. In other words, it depends only on the past and not the future. Stopping after four blacks in a row is a legitimate stopping time. A rule which says that whenever there are four blacks in a row, then you stop after the third one, is not a stopping time since you clearly will not be aware of that time when it occurs. (The concept is similar to that we encountered in Section 20.5 when discussing trading strategies).

Here is a more formal definition. A stopping time for a discrete-time stochastic process is a rule that assigns to each realization (xn) of the process an integer k, the stopping time, in such a way that if we assign k to a realization (xn), and (yn) is a realization such that xn = yn for n = 1, 2, …, k, then we must assign the stopping time k to (yn) as well.

To illustrate this definition, take the coin flipping example, starting with an initial surplus of 3, winning 1 for a head, and consider a rule which tells you to stop after the first head whenever you get two consecutive heads. This should not be a stopping time, and we can see that it does not satisfy the definition. A realization of the form (3, 4, 5, …) would be assigned 1, but a realization of the form (3, 4, 3, …), which agrees with the first one up to time 1, would not be assigned 1.

A stopping time will often be denoted by a letter such as S. A major example that we have already encountered is when S equals the time of ruin, which certainly satisfies the requirement for a stopping time. A particularly simple example of a stopping time is S = k, for some fixed k. That is, one stops at time k regardless of what has happened.

A fundamental fact about the fixed stopping time is that for any martingale (Xk), as defined in Section 18.3,

This is intuitively clear. In the case of a Markov chain, it is derived easily from

numbered Display Equation

and by induction we derive (23.23).

Example 23.3 Consider again the game of flipping a fair coin, winning 1 for heads, starting with an initial surplus of 3. The gambler decides to stop at time 4 or whenever two consecutive heads come up, if earlier. What is the expected surplus at the end of the game?

images

Figure 23.1 Tree for Example 23.3

Solution. This is complicated by the fact that we no longer have a Markov chain. There is, however, a useful general technique that allows us to recover the Markov property by adding states. In this example, instead of having a single state for each integer w, we insert a state wu to signify that there was a win on the previous play, following a loss on the play before; and we insert a state wd to signify a loss on the previous play. See Figure 23.1, where the shaded boxes indicate points where play stops. By counting paths, we calculate the expected surplus at stopping as

numbered Display Equation

The stopping rule has not helped to raise the expectation above the initial stake.

23.3.2 The optional stopping theorem and its consequences

This motivates the question of whether it is always true in the case of a martingale that E(XS) = E(X0)? That is, does (23.5) hold when the fixed time k is replaced by an arbitrary stopping time? The answer is no. Starting with a positive initial surplus of u, we flip a fair coin repeatedly, wagering any amount b that we choose, and receiving back 2b for a head and nothing for a tail. There are two situations we want to present. In the first, we bet 1 unit each time and continue until we lose all of our initial stake. The stopping time S is the first time the surplus reaches 0. Trivially, E(XS) = 0 ≠ u. The second case is the familiar doubling strategy. Bet 1 unit on the first toss, and double the bet each successive play. S is the first time a head comes up, and it is not hard to see that E(XS) is u + 1, so we are sure to gain 1. The problem is that neither of these strategies are feasible in practice. (Of course the first is irrational as well.) In both cases, the amount of time we need is unbounded, and in the second case the amount of capital we need is unbounded as well. We will, however, get an affirmative answer to our question if there exists a suitable bound on a combination of the stopping time and values. The following theorem gives a precise condition. Observe first that in place of XS, which is not defined if S = ∞, we want in general to consider XS|S < ∞.

Theorem 23.1 (Optional stopping theorem) Suppose that {Xn} is a martingale and S is stopping time such that

Then

numbered Display Equation

Proof. For any n,

Using a modification of Equation (A.29), the third term above can be written as

numbered Display Equation

In view of the martingale property,

numbered Display Equation

The entire third term therefore reduces to E(XS|Sn)P(Sn). We now simply take limits as n → ∞ to reach the conclusion.

inline

Corollary In the case where S is finite valued, we have E(X0) = E(XS) in either of the following cases:

  1. S is bounded. That is, for some N > 0, we have SN.
  2. The values of Xn are bounded in absolute value prior to stopping. That is, there is a constant C such that, for all n, if S > n then |Xn| ⩽ C.

Proof. In (i), the second factor in (23.6) is 0 for nN, so (23.6) necessarily holds. In (ii), the second factor in (23.6) approaches 0 by the fact that S is finite, and the first factor is bounded in absolute value by C, so the product converges to 0.

inline

Example 23.4 A gambler starting with an initial fortune of a units repeatedly plays an even money game against an adversary with an initial fortune of b. Each has an equal chance of winning each game. They continue the play until one is broke. What is the probability that the gambler with a units will eventually lose his initial stake before the other does? (As an equivalent formulation, we can remove any restriction from the opponents’ initial stake and instead postulate that the gambler decides to quit upon losing a or winning b.)

Solution. In Example 23.2 above, we essentially considered the case that b was infinite, that is there was no restriction on how much the opponent might lose, and we saw that ruin was certain. Here, we deal with the more realistic case that b is finite. Even a casino has some upper bound on its available wealth. Let Uk be the fortune of our gambler at time k. In view of the fairness of the game, this is indeed a martingale. Let S be the time that the gambler either loses the initial stake of a or wins b from the opponent. This is a stopping time that certainly satisfies condition (ii) of the Corollary to Theorem 23.1 since the values of Uk range from 0 to a + b. We will show later that S must take a finite value, which allows us to use the Corollary. Let π be the probability that our gambler loses his initial stake. We can invoke the Corollary to conclude that E(US) = a. But also, considering the two possibilities for S, we have

numbered Display Equation

and by equating, we obtain

numbered Display Equation

Note that π approaches 1 as b approaches ∞, verifying our conclusion following Example 23.2.

Example 23.5 Redo Example 23.4 assuming now that the probability of a win is p ≠ 1/2. (This is a classical problem, often referred to as gambler’s ruin.)

Solution. The difficulty now is that we no longer have a martingale. However, the following ingenious trick allows us to transform the process into one. To simplify notation, let q = 1 − p. Consider the process

numbered Display Equation

For any n, Un + 1 = Un + Gn + 1, where Gn takes the value 1 with probability p and − 1 with probability q. Therefore,

numbered Display Equation

leading to

(23.8)‡ numbered Display Equation

Now, invoking the independence of Un and Gn + 1,

numbered Display Equation

showing that {Xn} is a martingale. We can now duplicate the calculations in Example 23.4, applied to Xn. We have that

numbered Display Equation

and

numbered Display Equation

Solving,

numbered Display Equation

We return to the unfinished business of showing that the stopping time S of the last two examples must assume a finite value. We have a finite-state Markov Chain with states 0, 1, 2, …, a + b, corresponding to the amount held by the person who started with a. We can see, similarly to Example 18.3, that the states 1, 2, …, a + b − 1 are transient and the states 0 and a + b are both absorbing and therefore recurrent. By Theorem 18.2 the process must reach one of these two recurrent states, implying that S cannot take a value of ∞.

Here is another striking application of this idea.

Example 23.6 A gambler plays a game in which she will either win 1000 with probability p, where p < 1, or lose 1 with probability 1 − p. Suppose, however, that whenever she accumulates more than 10 000 in winnings, a companion takes everything in excess of 10 000 to spend in the casino gift shop. What is the probability that the gambler will eventually go broke?

Solution. The probability is 1. Regardless of the initial fortune or the value of p, the gambler in this case is sure to lose everything if she plays for a sufficiently long time. Once again we have a finite Markov chain with states taking values from 0 to 10 000. All states except 0 are transient, as we showed in the previous example, and we are certain to reach the one recurrent state of 0.

Remark The key to the above example is the phrase ‘a sufficiently long time’. As a practical matter, most people would be happy to play this game for a high value of p and would expect to eventually emerge a winner.

23.3.3 The adjustment coefficient

We now wish to apply Theorem 23.1 to the discrete surplus process (18.3) as interpreted in Section 23.1. We cannot, however, expect this to be a martingale. It will only be one if E(G) = 0, that is, if c = E(N)E(X). This is unrealistic, since, as we noted in previous chapters, insurers will invariably charge an amount above this expected value to guard against the possibility that the aggregate claims will be higher than expected. That is, they take c = (1 + θ)E(N)E(X) for some θ > 0. However, we can transform this process to be a martingale by the same type of procedure as we used in Example 23.4. The following definition gives the basic tool for doing this.

Definition 23.1 An adjustment coefficient of a random variable G is a positive number R satisfying

If R is an adjustment coefficient of a discrete random variable, then for z = eR it follows that PG(z) = 1, so we have already essentially seen this idea in (23.23).

To justify the word ‘the’ in the title of this subsection, we will show that there cannot be two positive numbers satisfying (23.23). Let γ be the supremum of all points for which MG( − r) is defined. (For example, if G = cW where W ∼ Exp(β), then γ will simply be equal to β.) In many cases γ = ∞. Define a function ϕ(r) = MG( − r) − 1 = E(erG) − 1 on the interval [0, γ). Then

numbered Display Equation

so ϕ is a convex function that takes the value 0 at the point 0 and, therefore, cannot have more than one positive root (see Figure 23.2).

images

Figure 23.2 Graph of the function ϕ

An equally pertinent problem is to decide if a positive root of ϕ exists. We can show that this will almost always happen in view of the following two properties that G will satisfy in any realistic insurance context:

  1. E(G) > 0;
  2. P(G < 0) > 0.

As we mentioned above, (i) will hold due to the relative risk loading. Moreover, there must be the possibility of paying out more in claims than the premiums collected, or nobody would ever buy insurance. This will imply (ii).

Now (i) implies that ϕ′(0) = −E(G) < 0, so ϕ will start out negative. We must, therefore, have a positive root as long as limr → γMG( − r) > 1. From condition (ii), we can find a positive numbers s and δ such that P(G < −s) > δ. This means that E(erG) > δers > 1 for sufficiently large r. This guarantees a positive root whenever γ = ∞ (for example, when G is finite).

There are a few rare examples, which we will not give here, where γ is finite but limr → γMG( − r) ⩽ 1, so the adjustment coefficient will not exist. (This is aside from those cases where the m.g.f. does not exist in the first place, so we cannot even define the adjustment coefficient.)

It is sometimes convenient to extend the definition of the adjustment coefficient to the extreme cases. If ϕ(t) ⩽ 0 for all t, we say that R = ∞; while if ϕ(t) ⩾ 0 for all t, we say that R = 0.

The adjustment coefficient can be taken as a measure of safety. The higher R is, the less risk there is for the insurer. This is indicated by Theorem 23.4 below, which shows that as R increases, our upper bound for the probability of ruin decreases. This principle is also seen in Examples 23.2 and 23.5. These examples show that for the case where G takes the value 1 with probability p and − 1 with probability 1 − p, the adjustment coefficient is log (p) − log (1 − p), which increases as p does. The following is yet another example.

Example 23.7 Find the adjustment coefficient for a normal distribution with mean μ and variance σ2.

Solution. The m.g.f. of the normal is given in (A.52). We have , so that − Rμ + σ2R2/2 = 0, and since R > 0,

We see that for two normals with the same mean, as the variance gets smaller, which should signify less variation and therefore less risk, the adjustment coefficient goes up.

Equation (23.10) is often used as an approximation to the adjustment coefficient for other distributions. This is justified by the following argument. The function log (MG(t)) has first derivative equal to MG(t)/MG(t), which at t = 0 equals E(G). Differentiating this latter expression yields that the second derivative at 0 is equal to Var(G). We can then write (23.9) in the form

numbered Display Equation

and ignoring powers of R higher than 2 gives (23.23).

The following example will be extensively used in the continuous-time models.

Example 23.8 Suppose that G = c − ⟨N, X⟩, where N ∼ Poisson (λ). Find an expression for R in terms of X and λ.

Solution. Using (A.50) and (A.51), MG( − r) = ecrMN, X(r), and then using (A.42) and (21.21), this is equal to . For this to equal 1, the exponent must equal 0, and dividing by λ, we get

Since c = (1 + θ)λE(X), where θ is the relative risk loading, we can also write

23.3.4 The main conclusions

A key fact about the adjustment coefficient is that the process is a martingale. This follows by exactly the same calculation as given in Example 23.5 where we used the fact that E(eRG) = 1. We now obtain a major conclusion by applying Theorem 23.1 to this martingale. We must first verify that condition (23.6) holds.

Theorem 23.2 For the process given in (18.18), suppose that E(G) > 0 and that Var(G) exists. Then, for any positive r,

numbered Display Equation

Proof. Let μ be the mean and let σ be standard deviation of G, so that

numbered Display Equation

Consider the following events referring to the situation at time n. Let An be the event that T > n and UnE(Un)/2, and let Bn be the event that T > n and 0 ⩽ Un < E(Un/2). Since T > n means that Un must be nonnegative, we see that the event T > n is a disjoint union of An and Bn.

inline

We also know from Chebyshev’s inequality (A.11) that

numbered Display Equation

so that P(Bn) → 0 as n → ∞. Then,

numbered Display Equation

As n goes to ∞, the second term approaches zero because the first factor is bounded by 1, and the first term is less than er(u + nμ)/2 and also approaches 0.

We can now apply Theorem 23.1 to the martingale e with the stopping time T to conclude that

numbered Display Equation

While stopping theorems are normally used to derive the expectation at time of stopping, in this instance we get an expression for the probability of ruin. Note that − UT|T < ∞ is just the deficit at ruin, which we termed D(u) before, so we have the following theorem.

Theorem 23.3 If the adjustment coefficient R exists, then

numbered Display Equation

Note that in the case that D(u) has a constant value of 1, we recover (23.1) with z = eR.

In general, this theorem does not give us the exact value of ψ(u) since we will not know the denominator. We can say, however, that since R is positive, the denominator is greater than 1, and we conclude the following.

Theorem 23.4 If the adjustment coefficient R exists, we have

numbered Display Equation

and therefore

numbered Display Equation

This theorem tells us that we can make the probability of ruin as small as we like by taking the initial surplus sufficiently high, which is certainly a reasonable conclusion. It also tells us that the probability of ruin reduces exponentially as a function of initial surplus. For example, if the upper bound to the probability of ruin given by Theorem 23.4 is less than 0.05, and we double the initial surplus, we then know that the probability of ruin is less than 0.0025.

It is also sometimes useful to have a lower bound for the ruin probability. This is possible if G is bounded below by − M for some M > 0. We then know that D(u) ⩽ M, and we can conclude from Theorem 23.3 that

numbered Display Equation

We conclude this section with some remarks on what happens in the rare case that the adjustment coefficient does not exist. We can still reach the second conclusion in Theorem 23.4. As long as E(G) > 0 and MG is defined on some interval of positive length about 0, we can find β > 0 such that MG( − β) < 1. We then have that is a supermartingale. We obtain a version of (23.7) with ⩾ replacing the first equality sign, and this is enough to derive the conclusion of Theorem 23.4 with β replacing R.

23.4 Distribution of the deficit at ruin

What can we say about the denominator in the expression for the ruin probability given in Theorem 23.3, other than that it is greater than 1? Can we ever evaluate it exactly in cases other than that mentioned in Section 23.1, where the values of G are nonnegative integers except for a single negative value of − 1? In this section, we will try to provide some insight into these questions. To do so, we introduce some additional random variables.

Let Y = −G|G < 0. For example, if G takes values {4, 3, −1, −2, −3} with respective probabilities {0.4, 0.1, 0.1, 0.25, 0.15}, then Y will take the values {1, 2, 3} with respective probabilities {0.2, 0.5, 0.3}. The significance of Y is that only the negative values of G can bring about ruin.

Let J(u) denote the value of the surplus in the period prior to ruin, assuming an initial surplus of u (the J stands for ‘just before’). The connection between all these is that D(u) will be equal to some value of YJ(u). (In what follows, we will just write D and J, suppressing u, which will be fixed.)

To illustrate, take G as given above, and suppose that u = 2. Given a realization G1 = 1, G2 = −1, G3 = −3, …, ruin will occur at time 3, J will equal 2, and D will equal 1. For a realization G1 = 1, G2 = −1, G3 = −2, G4 = −3, …, ruin will take place at time 4, J = 0, and D = 3.

We can make the following observations about this example. J can take possible values of 0, 1, or 2. When J takes the value 2, then D will take the value 1 with certainty. When J takes the value of 0, then any of the three negative amounts will cause ruin, so D will have the same distribution as Y. When J takes a value of 1, then the ruining claim must be either 2 or 3, so D will take the value 1 with probability 5/8, and 2 with probability 3/8, as these are the conditional probabilities of values of Y given that Y > 1.

In the general case, if J = j, then we know that the ruining claim Y must take a value greater than j, and in particular for D to take a value of d, we need that Y = j + d. This gives

numbered Display Equation

If only we knew the distribution of J, this would give us the distribution of D. Unfortunately, deducing the distribution of J is just as difficult as getting that of D, so we would appear to have simply gone around in circles, with no gain of information.

This is not quite true, however. For one thing, our analysis indicates some features about the distribution of D. It is somewhat related to that of Y. It takes exactly the same values, and it will in fact have the same distribution as Y whenever J = 0. (In our simple example of Section 23.2 we knew in fact that J was always 0.) In general, however, it will involve more mass at the lower values than Y, for when J takes values higher than 0, there is less chance for D to assume higher values.

A second observation is that there is one case where we can indeed use the formula above. Suppose that Y ∼  Geom(p)+, using the notation we introduced at the beginning of Section 21.7. For this distribution, fY(d + j)/sY(j) = (1 − p)pd + j − 1/pj = fY(d). The second factor in the summation is independent of j, so it can be factored out, and we conclude that

numbered Display Equation

In other words, we have the following:

(23.13) numbered Display Equation

We will use this example later to motivate a result in the continuous-time case.

23.5 Recursion formulas

In some cases, we can use recursion to calculate exact ruin probabilities as well as the exact distribution of surplus at time of ruin.

23.5.1 Calculating ruin probabilities

Suppose we have the insurance claims model

numbered Display Equation

and that the following restrictions hold:

  1. c = 1.
  2. N takes the values 0 or 1 with probabilities 1 − q or q, respectively.
  3. X takes positive integer values 1, 2, …, K, with probabilities f(1), f(2), ..., f(K), respectively.
  4. 1 > qE(X), so that E(G) > 0.
  5. The initial surplus u is a positive integer.

Note that (i) and (iii) simply say that all claim amounts are integer multiples of the premium, since we can always take the amount of the premium as the unit of capital.

To simplify the notation, we will first illustrate with K = 3. Clearly, the following calculations will work for any value of K. Note that G takes the values 1, 0, −1, −2 with probabilities 1 − q, qf(1), qf(2), and qf(3), respectively.

Suppose we start with u units at time 0. In the first period, the four possible outcomes for G lead to four possible values of surplus at time 1. This gives us the following set of equations, one for each value of u:

numbered Display Equation

Note that for convenience we have included terms of the form ψ(i) for i < 0. These, of course, are just equal to 1. If we start with a negative amount, we are already ruined. Next, rearrange the equations above to give

numbered Display Equation

Sum the first n + 1 of these equations. The left-hand side adds up to ψ(0) − ψ(n + 1). To add the right-hand side, it is convenient to add by the diagonals (running northwest to southeast) because they involve the term ψ(k) for the same value of k. Since f(1) + f(2) + f(3) = 1, all of the diagonals will sum to zero except for the three on the upper right and the three on the lower left. The third diagonal on the upper right sums to qψ(0). The first two diagonals on the upper right sum to

numbered Display Equation

The three diagonals on the lower left involve terms in ψ(n + 1), ψ(n), and ψ(n − 1). These will all converge to 0 as n approaches ∞ by Theorem 23.4. Taking limits gives

numbered Display Equation

and solving

It is instructive to write this in terms of θ, the relative risk loading. Since

(23.15) numbered Display Equation

we can substitute for E(X) in (23.14) to obtain

numbered Display Equation

which tells us that

(23.16) numbered Display Equation

Now consider the general case with G taking the K values 1, 2, …, K. Formula (23.14) gives us a starting value, and all of the ruin probabilities can be calculated recursively by rearranging our first set of equations to get

(23.17) numbered Display Equation

which we can write more compactly, using the notation of Section A.12.3, as

numbered Display Equation

Remark The quantity ψ(0) is of importance since it gives us our starting value. One may think, however, that it is intrinsically of little interest, since we are not likely in practice to have a situation where the initial surplus is 0. It is, however, of great significance since given any starting value u, we can interpret ψ(0) as giving the probability that we will eventually reach some point where our surplus is less than u. Similarly, D(0) is the amount by which we are less than the initial surplus, if this occurs. We will exploit this idea to great advantage in Section 23.7.

23.5.2 The distribution of D(u)

The same approach lets us deduce the distribution of D(u). For k = 1, 2, …, K − 1, let ψk(u) be the probability that, starting with initial surplus u, ruin eventually occurs and the value of D(u) = k. It will be convenient again to consider negative values of the argument, and we note that

numbered Display Equation

Given values of ψk(u), we can immediately find ψ(u), as well as the distribution of D(u), since

(23.18) numbered Display Equation

and

(23.19) numbered Display Equation

We calculate ψk(u) by following exactly the procedure in the previous subsection. We get the same systems of equations except with ψ replaced by ψk. When we sum the second set of equations and take limits, we again will have everything on the right-hand side vanishing, except possibly for a finite number of diagonals on the upper right. Since these all involve negative values of the argument of ψ, these sums will also be zero except for the single diagonal involving the terms ψk( − k), and the sum of that will be f(k + 1) + f(k + 2) + ⋅⋅⋅ + f(K) = P(X > k). So, in place of (23.23), we get

(23.20) numbered Display Equation

From (23.23), (23.23), and (23.23), we immediately have a nice simple formula for the distribution of D(0):

(23.21) numbered Display Equation

This verifies our intuition of the previous section. In our present case, we have Y = X − 1, and we see precisely how D(0) is a type of ‘shifted to the left’ version of Y. For example, if Y takes values 1, 2, 3, with probabilities 0.2, 0.5, 0.3 respectively, then D(0) takes the values 1, 2, 3 with probabilities that are in the ratio, 1 : 0.8 : 0.3, so they are 10/21, 8/21, 3/21.

We now obtain the same recursion formula as in (23.23), except with ψ replaced by ψk, that is,

numbered Display Equation

Example 23.9 q = 1/15, X has a constant value of 4. Calculate ψ1(1), ψ2(1), ψ3(1), and ψ(1). Verify that your answer to the last agrees with that given by Theorem 23.3.

Solution. We note first that f(1) = f(2) = f(3) = 0, f(4) = 1. So

numbered Display Equation

Similarly,

numbered Display Equation

so that ψ(1) = 31/196, and D(1) takes the values 1,2,3 with probabilities 15/31, 15/31, 1/31, respectively. (Note that D(1) differs from D(0), which has a uniform distribution on 1, 2, 3.)

To check this by Theorem 23.3, we first solve for R, or equivalently z = eR. Since G takes the value 1 with probability 14/15 or − 3 with probability 1/15, we have

numbered Display Equation

and we can verify that z = 1/2. So,

numbered Display Equation

From Theorem 23.3,

numbered Display Equation

as above.

23.6 The compound Poisson surplus process

23.6.1 Description of the process

We now turn to ruin calculations in the continuous case. This will be based on a compound Poisson process which is simply a process corresponding to the compound Poisson distribution, which we considered in Chapter 21. That is, instead of merely counting 1 every time an event occurs, we take an observation from some given distribution. So we have a distribution X, which we can think of as a severity distribution, and we have a Poisson process, N(t). The resulting compound Poisson process is given by

numbered Display Equation

where {Xk} are independent, each has the same distribution as X, and they are independent of N(t). In other words, we can simply write

numbered Display Equation

We can now provide a model for a surplus process. Suppose that an insurer’s aggregate claims up to time t are given by a compound Poisson process S(t), as above. In the discrete case, we postulated a premium of c per period. We now assume that the insurer collects premiums at a continuous rate of c per period, so in any period of length h total premiums of ch will be collected. We assume also that the insurer begins with an initial surplus of u. The compound Poisson surplus process is the process given by

(23.22) numbered Display Equation

where U(t) is the surplus at time t. Our relative risk loading is given as in the discrete case by

numbered Display Equation

The probability of ruin is defined similarly to the definition in the discrete case, although we need an infimum to replace the minimum. That is,

numbered Display Equation

See Figure 23.3 for a typical realization of this process. The diagonal lines have slope c, and show the increase in surplus arising from the premium payments. Downward jumps then occur whenever there is a claim.

images

Figure 23.3 A realization of the continuous-time surplus process

We can approach this case by considering an approximating discrete model. Suppose that we divide up our time into periods of length h for some small h. Then, if we only view our surplus at the end of each period, we just have a discrete surplus model with the gain in each period given by

numbered Display Equation

where N ∼ Poisson(λh). We will now go over the various results we obtained in the last two sections to see how they are modified for the continuous model. In many instances, we do not give rigorous proofs. However, all the results are motivated by those of the previous sections.

23.6.2 The probability of eventual ruin

We first would like to calculate the adjustment coefficient in the continuous model. It is natural to do this by calculating the adjustment coefficient of G for the approximating discrete model as given above, and then taking the limit as h approaches 0. This turns out to be extremely simple in the Poisson case. The term c/λ that we get in (23.11) is independent of h, so we get exactly the same answer as before. The adjustment coefficient is given by (23.11) or (23.23), as it was in the discrete model. The probability of ruin is now given by Theorems 23.3 and 23.4, which carry over unaltered to the surplus model with compound Poisson aggregate claims.

23.6.3 The value of ψ(0)

In light of the Poisson assumption, and Theorem 18.3, we can view our continuous-time model as a limiting case of the particular model discussed in Section 23.5. As h goes to 0, so will q and we deduce, directly from (23.23), that

numbered Display Equation

This is somewhat surprising, since ψ(0) does not depend on the particular distribution of X, except through the expectation E(X), which affects θ. In other words, given any two distributions for X, as long as they have the same mean, they will produce the same value of ψ(0).

23.6.4 The distribution of D(0)

D(0) is necessarily continuous here in view of the fact that premiums are collected continuously. We can deduce, analogously to (23.23), that if fD(0) is the density function of D(0), then

For later purposes, we will need the m.g.f. of this distribution. Integrating by parts,

numbered Display Equation

so that

23.6.5 The case when is exponentially distributed

It is easy to deduce from the distribution of D(0) given above that when X ∼ Exp(β), then also D(0) ∼ Exp(β). However, much more than this is true. As a continuous analogue to (23.23), we obtain the following:

numbered Display Equation

The last observation tells us that we can get an exact calculation of the ruin probability when X is exponential. We first must calculate the adjustment coefficient. From (23.23), we solve

numbered Display Equation

and it is immediate that

numbered Display Equation

Substituting in (23.12) gives

Let us verify that our formula for R makes sense intuitively, recalling that our adjustment coefficient is a measure of safety. It increases as β increases, which it should since the mean of the severity distribution is decreasing. It must be 0 in the extreme case that θ = 0, and it must increase with θ.

Note now that the denominator in Theorem 23.3 is just MD(u)(R). When X ∼ Exp(β), D(u) has the same distribution, and from (23.25) we know the denominator is 1 + θ. We substitute in the statement of Theorem 20.3 to obtain a key result.

Theorem 23.5 For the surplus model, with compound Poisson aggregate claims, where the severity distribution ,

numbered Display Equation

23.7 The maximal aggregate loss

We continue with the same continuous model as in the previous section and give an alternate approach to ruin probabilities. This will provide another proof of Theorem 23.5, as well as enabling us to deduce the ruin probability when X is a mixture of gamma distributions.

Definition 23.2 The maximal aggregate loss, denoted by , is the largest amount by which our surplus will be less than the beginning surplus. In other words,

numbered Display Equation

For example, suppose we start out with an initial surplus of 10. Then, given a realization in which the smallest value of surplus ever attained is − 7, the maximal aggregate loss will be 17. It is clear that is independent of u. In fact, we could write it as

numbered Display Equation

The significance of this random variable for ruin models is that

(23.26) numbered Display Equation

Therefore, if we know the distribution of , we can immediately write down the probability of ruin. We will show in this section that for our compound Poisson model, we can find the m.g.f. of , which makes a significant step towards this goal.

To do this, we consider what is often termed as the record low process. Imagine we are observing the surplus process and each time that we reach a new record low point in surplus, we record the amount by which we ‘beat the record’, that is, how much we are below the previous record low. Let be the amount we record on the nth occasion the record is broken, should this occur.

As an example, suppose u = 10 and c = 1. If we have a claim of size 5 at time 3, we have a new record low surplus of 8, so . Suppose the next claim is 3 at time 7. Our surplus is 9 at that time, so we do not have a record low. Then, suppose the next claim is 5 at time 8. This gives us a new record low of 5, so we have , and so on. We may of course have many more record lows, but suppose that we do not and a surplus of 5 is as low as we get. Then the maximal aggregate loss is just 5, which is the sum of the two record low increments, that is, .

Observe that the random variable is just D(0), as is clear from our remarks in the previous section. Similarly, it follows from the stationarity of the process that each will have the same distribution. How many new record lows will we encounter? Well, after any new record low has occurred, the probability of another occurrence is just ψ(0) = 1/(1 + θ). If N is the number of records lows that occur, then N ∼ Geom(1/(1 + θ)). Now , the maximal aggregate loss, is simply the sum of the Ln. So, we have

numbered Display Equation

By the independence in the process, this shows that can be expressed as a compound distribution ⟨N, D(0)⟩, where N is geometric. We have already worked out the m.g.f. for a compound geometric distribution in Example 21.2. As we noted in that example, in order to have a chance of recognizing a compound distribution from its m.g.f. we will want to get rid of the point mass at 0 and look at . From this example,

numbered Display Equation

where p = 1/(1 + θ). Substituting from (23.24) for MD(0) and simplifying gives us

Since is a mixture of 0 and with weights θ/(1 + θ) and 1/(1 + θ), respectively, we know that

numbered Display Equation

The first term in (23.27) is trivially equal to 0 for nonnegative u, yielding the key result

(23.28) numbered Display Equation

We summarize the whole procedure of finding ruin probabilities by the maximal aggregate loss procedure:

  1. Calculate MX(r) and use (23.27) to find .
  2. Try to deduce from this the survival function of .
  3. Calculate ψ(u) directly from (23.23).

Step 2 is of course the difficult one. Will we actually be able to recognize the distribution of from its m.g.f.? Here is one particularly easy example.

Example 23.10 Use the method described above to find ψ(u) when X ∼ Exp(β).

Solution. Example 17.2 showed that , and from (23.23),

numbered Display Equation

giving us an alternate proof of Theorem 18.3.

Are there any other cases where we can recognize the distribution of from its m.g.f.? We make two key observations. First, from (23.23), if MX(r) is a rational function (a quotient of two polynomials), then so is . Second, if X is a mixture of gamma distributions that have an integer for the first parameter, its m.g.f. will be a linear combination of rational functions, therefore rational, and hence will be rational. So, the steps to handle such X are as follows. For such a distribution X, calculate and decompose it into partial fractions. Suppose

numbered Display Equation

This shows that is a mixture of (Y1, Y2, …, Yn) with weights (w1, w2, …, wn), where for each i, Yi ∼ Gamma (αi, ri). From (23.23), we deduce that

numbered Display Equation

Example 23.11 Find ψ(u), given that X has the density function

numbered Display Equation

and θ = 7/9.

Solution. X is a mixture of Exp(1) and Exp(2) distributions with equal weights, so

numbered Display Equation

As a check, this quantity must take the value 0 for r = 0. Substituting MX(r) − 1 into (23.27) gives

numbered Display Equation

The partial fraction expression for this is

numbered Display Equation

So is a mixture of Exp(1/2) and Exp(7/4) distributions with weights 14/15 and 1/15, respectively. It follows that

numbered Display Equation

What happened to our adjustment coefficient? It played a prominent role in our previous method but seems to be absent here. It is not, however. The denominator of (23.27) is precisely the quantity we set equal to 0 to find R. To find the partial fraction decomposition, we must find the roots of this equation, and the adjustment coefficient appears as the smallest such root. In Example 20.11, we found the two roots r = 1/2 and r = 7/4. We then know that R = 1/2.

How do we reconcile the other root of 7/4 with our previous statement that the adjustment coefficient was uniquely defined as the only positive root of the defining equation? The point is that MX(r) = E(erX) may only be defined over a certain region. In this example, it is defined only for 0 ⩽ r < 1, and if r > 1, then the expectation will not exist. However, the algebraic expression that gives the m.g.f. over this range does make sense for values greater than 1. It just no longer represents an expectation of a function of X. The resulting algebraic expression can have other roots, as it does in this example.

The procedure followed in the previous example can be used to write a general closed-form formula for ψ(u) when X is a mixture of two exponential distributions, as given by the following theorem. We leave the details of the proof to the reader.

Theorem 23.6 Suppose that in the compound Poisson surplus process, X is a mixture of Exp(α) and Exp(kα), for some k > 1, with respective weights w and 1 − w.

  1. Let φ = 1 − w + wk. The adjustment coefficient is rα, where r is in the interval (0,1) and is a solution to the quadratic equation
    numbered Display Equation
  2. Let
    numbered Display Equation
    Then is a mixture of Exp(rα) and Exp(sα) with respective weights
    numbered Display Equation
  3.  
    numbered Display Equation

Notes and references

The inequality of Theorem 23.4 is an early result of ruin theory known as Lundberg’s inequality. See Bowers et al. (1997, Example 13.4.3) for a case where the adjustment coefficient does not exist. An alternate method of handling the case of a mixture of gamma distributions is by developing an integrodifferenial equation for ψ(u) (see Klugman et al., 2012). There is an extensive literature on ruin theory which goes beyond the treatment here. See Grandel (1991) for some of the extensions. Several results concerning the distribution of the surplus just before ruin and just after ruin were developed in Gerber and Shiu (1998). Similar results along these lines are found in Powers (1995).

Exercises

  1. 23.1 Consider the discrete-time surplus process Un = u + G1 + G2 + ⋅⋅⋅ + Gn, where the Gis are independent and each distributed as a random variable G which takes the values 1, 0, − 1, −2 with probabilities 0.5, 0.2, 0.2, 0.1, respectively. Let ψ(u) be the probability of eventual ruin, starting with an initial surplus of u.
    1. If you start with an initial surplus of 1, what is the probability that you will be ruined by time 2 or before?
    2. Given ψ(6) = a, ψ(5) = b, ψ(4) = c, express ψ(7) in terms of a, b, c.
  2. 23.2 You are repeatedly flipping coins and will win 1 for a head and lose 1 for a tail. You plan to stop playing when you get four consecutive wins, or after 100 tosses at the very latest. You start with 10 units. Let US denote the amount of units you will have upon stopping.
    1. Suppose that the probability of a head is 1/2. What is E(US)?
    2. Suppose again that the probability of head is 1/2, but now you are given the additional information that you began with three wins, followed by two losses. What is E(US) now?
    3. Suppose that the probability of a head is 1/3. What is ?
  3. 23.3 An insurer’s portfolio consists of a single possible claim. You are given the following information. The claim amount is uniformly distributed over (100, 500). The probability that the claim occurs after time t is e− 0.1t for t > 0. The claim time and amount are independent. The insurer’s initial surplus is 20. Premium income is received continuously at the rate of 40 per year. Determine the probability of ruin.
  4. 23.4  
    1. A random variable G takes the value 1 with probability 6/7 and − 1 with probability 1/7. Show that the adjustment coefficient of G is log(6).
    2. A random variable G takes the value 1 with probability 1/2 and 2 with probability 1/2. Show that the adjustment coefficient is ∞.
  5. 23.5 You are repeatedly playing a game in which at each stage you win 1 with probability 15/19 or lose 2 with probability 4/19.
    1. Show that the adjustment coefficient is log(3/2).
    2. If Un denotes the amount you have at time n, show that Un is not a martingale.
    3. At time 20 you have a total of 4 units. What is ?
    4. Suppose you start with 3 units at time 0. You decide to stop play when you have a total of 20 units, or after 100 plays if that occurs earlier. If US is the amount that you will have after stopping play, what is ?
  6. 23.6 You are playing a game repeatedly in which at each turn you either win 1 with probability 12/13 or lose 2 with probability 1/13.
    1. Write an equation for the adjustment coefficient R, and verify that R = log (3).
    2. Show that for an initial stake of u, the probability of eventual ruin is between (1/3)u + 2 and (1/3)u + 1.
  7. 23.7 Show that π in Example 23.5 approaches b/(a + b) as p approaches 1/2, verifying the conclusion of Example 23.4.
  8. 23.8 Refer to Example 23.4.
    1. Show that Xn = (Una)2n is a martingale.
    2. Use your result in part (a) to show that E[S] = ab.
  9. *23.9 This extends Exercise 23.8 to the case for which the probability of winning 1 is p ≠ 1/2, as in Example 23.5.
    1. Show that g(Un) − n is a martingale, where
      numbered Display Equation
    2. Use your result in part (a) to show that
      numbered Display Equation
  10. 23.10 Redo Example 20.9, now assuming that q = 1/7 and that X has a constant value of 3.
  11. 23.11 In the model of Section 23.5, let q equal 1/3, and let X take the values 1, 2, with equal probability. Show that ψ(k) = 1/4k + 1 for all k.

    The remaining exercises deal with the continuous-time surplus process:

    numbered Display Equation
  12. 23.12 Answer the following parts separately:
    1. Suppose that the initial surplus is 10. How large should the adjustment coefficient R be, to ensure that the probability of ruin will be less than 0.10?
    2. Suppose that the adjustment coefficient R = 2/9, λ = 5, and X is exponentially distributed with mean 3. Find c.
    3. You are given that c = 20, λ = 5. All you know about X is that E(X) = 3. What is the probability that the surplus will eventually drop below its initial value?
    4. Suppose that X takes some fixed value with certainty. Show that if the initial surplus does drop below its initial value, the amount of deficit the first time this occurs will be uniformly distributed.
  13. 23.13 You are given that X ∼ Exp(3), c = 2, λ = 3. You want the probability of eventual ruin to be no more than 0.05. How much initial surplus should you begin with?
  14. 23.14 You are given that X is uniformly distributed on the interval [0,2], the initial surplus u = 10, and the adjustment coefficient R = log (3).
    1. What is the probability that the surplus will eventually drop below 10?
    2. Given that the surplus does eventually drop below 10, what is the probability that the first time this happens, the surplus will be between 8.5 and 9?
  15. 23.15 Suppose that X is exponentially distributed. Is ψ(2u) equal to, strictly less than, or strictly greater than ψ(u)2?
  16. 23.16 For any random variable A, let RA be the adjustment coefficient. That is, RA is the positive solution of ϕA(r) = 0, where ϕA(r) = MA( − r) − 1. Suppose that G and H are random variables such that both adjustment coefficients exist and RGRH. If K is a mixture of G and H, show that
    numbered Display Equation

    by using properties of the graph of the function ϕA.

  17. 23.17
    1. If the adjustment coefficient R = 2, find u so that ψ(u) ⩽ 0.05.
    2. Suppose that X has a Gamma(2,3) distribution and that the adjustment coefficient R = 1. Find θ, the relative risk loading.
    3. Suppose that X is exponential with mean 2. If ψ(0) = 0.8, what is ψ(1)?
    4. Suppose that X is uniformly distributed on [0, 4] and that u = 5. Given that the surplus eventually drops below 5, what is the probability that, the first time this happens, the surplus is between 3 and 5?
  18. 23.18 Use the m.g.f. of the maximal aggregate loss to show that .
  19. 23.19 For a surplus process with a compound Poisson claim process, you are given that the adjustment coefficient is 0.25, the claim amount has a density function f(x) = e− 2x + 2.5e− 5x for x > 0. If is the maximal aggregate loss, determine .
  20. 23.20 If X is a mixture of Exp(α) and Exp(β) with weights w and 1 − w, respectively, find a formula for ψ(u) in each of the following cases:
    1. α = 3, β = 7, ω = 1/2, θ = 2/5;
    2. α = 1, β = 2, ω = 1/3, θ = 4/11;
    3. α = 3, β = 6, ω = 1/9, θ = 4/5.
  21. 23.21 Suppose that λ = 3, c = 3 and X has a density function f(x) = 36xe− 6x. Find a formula for ψ(u).
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
18.227.72.153