The Classical Surplus Process

Introduction

An insurer’s surplus changes as premiums arrive and claims are paid. A useful starting model represents these cash flows explicitly: the insurer begins with a specified amount of capital, receives premiums at a constant rate, and pays claims at random times and in random amounts. This model is called the classical surplus process.

The model is intentionally simplified. Its value is that each component has a clear actuarial interpretation, and the resulting surplus path is easy to describe. This tutorial defines the process and works through hypothetical paths; analysis of when the process becomes negative and the probability of that event belongs to the study of ruin. The material is aligned with the SOA Exam ASTAM (Advanced Short-Term Actuarial Mathematics) syllabus.

Definition: In the classical surplus process, the insurer’s surplus at time $t\geq 0$ is $$ U(t)=u+ct-S(t), \qquad S(t)=\sum_{j=1}^{N(t)}X_j. $$ Here $u\geq 0$ is initial surplus, $c\geq 0$ is the constant premium income per unit time, $N(t)$ is the number of claims by time $t$, and $X_j$ is the amount of the $j$th claim. The empty sum is zero, so $S(t)=0$ when $N(t)=0$.

The Model Components and Their Assumptions

The classical model usually assumes that claim arrivals form a homogeneous Poisson process with intensity $\lambda>0$ per unit time. In precise terms, $N(0)=0$; counts in disjoint time intervals are independent; and the number of arrivals in an interval of length $h$ has a Poisson distribution with mean $\lambda h$. Thus $N(t)$ has a Poisson distribution with mean $\lambda t$. The intensity describes the model’s average arrival rate, not a claim schedule: actual claim times remain random.

Claim amounts $X_1,X_2,\ldots$ are assumed to be independent and identically distributed positive random variables, independent of the entire arrival process $N(t)$. Their common distribution is the claim severity distribution. The model therefore separates two sources of uncertainty: how many claims arrive and how large each claim is. The assumptions do not require the claim amounts to have a particular distribution, but a specific application must state one if numerical severity probabilities or expectations are needed.

Premiums enter at the constant rate $c$, which means $ct$ of premium income has accumulated by time $t$. The model treats that income as continuous in time. It assumes no interest on capital, no investment income, no expenses, and no changes in premium rate during the period being modelled. These assumptions are modelling choices, not general descriptions of every insurer’s cash flows.

Symbol	Interpretation	Units when time is measured in years
$u$	Surplus at time zero	Dollars
$c$	Continuous premium income rate	Dollars per year
$\lambda$	Poisson claim arrival intensity	Claims per year
$N(t)$	Number of claims by time $t$	Number of claims
$X_j$	Amount of claim $j$	Dollars
$S(t)$	Total claims paid by time $t$	Dollars
$U(t)$	Surplus at time $t$	Dollars

Between claim arrivals, $S(t)$ does not change and surplus rises at rate $c$. At a claim time, surplus drops by the amount of that claim. If a claim of amount $X_j$ arrives at time $T_j$, the change across the arrival is

$$ U(T_j)-U(T_j-)= -X_j. $$

Here $U(T_j-)$ denotes the surplus immediately before the claim. The process is therefore a straight-line increase between claims with downward jumps at claim times. If $c=0$, there is no increase between claims; if $N(t)=0$ over an interval, surplus increases by exactly $c$ times the interval length.

Why the Surplus Formula Holds

The defining equation is a cash-flow balance, not a claim about the probability of any particular path. Start with $u$. By time $t$, the assumed continuous premium stream has added $ct$. The claims paid by that time are $X_1,\ldots,X_{N(t)}$, so their total is $S(t)$. Subtracting those outflows from the starting capital and premium inflows gives the surplus:

\begin{aligned} \text{surplus at time }t &=\text{initial surplus}+\text{premium income by }t-\text{claims paid by }t\\ &=u+ct-\sum_{j=1}^{N(t)}X_j\\ &=U(t). \end{aligned}

This accounting justification also shows where the formula stops applying. For example, if premiums are collected only at discrete dates, $ct$ does not represent the actual premium cash flow at every time. If investment returns, expenses, or a changing premium rate matter, those flows must be added to or substituted into the model rather than silently ignored.

Tracing a Surplus Path

For a realized path, calculate the premium accumulation to each claim time, subtract all claims paid up to that time, and then continue accumulating premiums. A claim causes an immediate jump; it does not retroactively change the premiums accumulated before its arrival.

Worked Example: Two Claims During a Policy Year

A hypothetical insurer starts with $u=\$500{,}000$ and receives premiums continuously at $c=\$72{,}000$ per year. Claim arrivals follow a Poisson process with intensity $\lambda=3$ per year. Claim amounts are independent and identically distributed, with probability $0.4$ of $\$20{,}000$ and probability $0.6$ of $\$90{,}000$, independently of arrivals. On the path considered, claims of $\$20{,}000$ and $\$90{,}000$ occur at years $0.25$ and $0.80$. Assume no interest, mortality, or expenses. Find the surplus immediately before and after each claim and at year 1.

Method. Use $U(t)=u+ct-S(t)$, with time in years and money in dollars. The Poisson intensity specifies the arrival model; the two stated event times and amounts specify the realized path.

At year $0.25$, premium income accumulated is $\$72{,}000(0.25)=\$18{,}000$. No claim has yet been paid, so the surplus immediately before the first claim is $\$500{,}000+\$18{,}000=\$518{,}000$. Subtract the $\$20{,}000$ claim to get $\$498{,}000$ immediately after it.

At year $0.80$, total premium income accumulated is $\$72{,}000(0.80)=\$57{,}600$. The first claim, $\$20{,}000$, has already been paid. Therefore the surplus immediately before the second claim is $\$500{,}000+\$57{,}600-\$20{,}000=\$537{,}600$. Subtracting the $\$90{,}000$ claim gives $\$447{,}600$ immediately after it.

At year 1, premium income is $\$72{,}000$, and total claims are $\$20{,}000+\$90{,}000=\$110{,}000$. Thus $U(1)=\$500{,}000+\$72{,}000-\$110{,}000=\$462{,}000$.

Check. From just after the second claim to year 1, no further claims occur, so surplus increases by $\$72{,}000(1-0.80)=\$14{,}400$. Indeed, $\$447{,}600+\$14{,}400=\$462{,}000$. The common mistake this avoids is subtracting a claim before its arrival time.

Worked Example: Reading the Components From a Model Description

A hypothetical equipment insurer has initial surplus $\$80{,}000$, charges premiums continuously at $\$100{,}000$ per year, and models claim arrivals by a Poisson process with intensity $6$ per year. Claim amounts are independent exponential random variables with mean $\$15{,}000$, independent of arrivals. Assume no interest, mortality, or expenses. A realized year contains claims at years $0.10$, $0.40$, and $0.90$, with respective amounts $\$10{,}000$, $\$30{,}000$, and $\$5{,}000$. Find the surplus just after each claim and at year 1.

Method. The stated exponential severity has rate $1/15{,}000$ per dollar, equivalently mean $\$15{,}000$. The observed amounts are the realized claim values; the mean is not substituted for each actual claim.

At year $0.10$, premium income is $\$100{,}000(0.10)=\$10{,}000$. The pre-claim surplus is $\$80{,}000+\$10{,}000=\$90{,}000$, and just after the $\$10{,}000$ claim it is $\$80{,}000$.

At year $0.40$, premium income is $\$40{,}000$, and $\$10{,}000$ of claims have been paid. The pre-claim surplus is $\$80{,}000+\$40{,}000-\$10{,}000=\$110{,}000$. After the $\$30{,}000$ claim it is $\$80{,}000$.

At year $0.90$, premium income is $\$90{,}000$, and prior claims total $\$10{,}000+\$30{,}000=\$40{,}000$. The pre-claim surplus is $\$80{,}000+\$90{,}000-\$40{,}000=\$130{,}000$; after the $\$5{,}000$ claim it is $\$125{,}000$.

At year 1, total premiums are $\$100{,}000$ and total claims are $\$45{,}000$, giving $U(1)=\$80{,}000+\$100{,}000-\$45{,}000=\$135{,}000$. This agrees with the post-claim surplus plus the $\$10{,}000$ of premiums earned between years $0.90$ and 1.

Worked Example: Converting Monthly Inputs to Annual Time

A hypothetical insurer begins with $\$50{,}000$. Premium income is $\$3{,}000$ per month, equivalent to $c=\$36{,}000$ per year. Claim arrivals follow a Poisson process with intensity $\lambda=3$ per year. Claim amounts are independent and identically distributed, with probability $0.5$ of $\$12{,}000$ and probability $0.5$ of $\$20{,}000$, independently of arrivals. Claims of $\$12{,}000$ and $\$20{,}000$ occur at months 2 and 7. Assume no interest, mortality, or expenses. Find the surplus just before and after each claim and at month 12.

Method. Convert times to years when using the annual premium rate: months 2 and 7 are $2/12$ and $7/12$ years. Equivalently, calculate premium income directly at $\$3{,}000$ per month.

By month 2, premiums total $\$3{,}000(2)=\$6{,}000$. The surplus immediately before the first claim is $\$50{,}000+\$6{,}000=\$56{,}000$, and immediately after it is $\$56{,}000-\$12{,}000=\$44{,}000$.

By month 7, premiums total $\$3{,}000(7)=\$21{,}000$. The first claim has been paid, so just before the second claim the surplus is $\$50{,}000+\$21{,}000-\$12{,}000=\$59{,}000$. Just after the $\$20{,}000$ claim it is $\$39{,}000$.

By month 12, premiums total $\$3{,}000(12)=\$36{,}000$, and claims total $\$32{,}000$. Thus the year-end surplus is $\$50{,}000+\$36{,}000-\$32{,}000=\$54{,}000$. The result also follows by adding $\$3{,}000(12-7)=\$15{,}000$ of premiums to the month-7 post-claim surplus: $\$39{,}000+\$15{,}000=\$54{,}000$.

The common time-unit trap is to use the annual rate $\$36{,}000$ as though it were a monthly rate. That would overstate premium income by a factor of 12. Using either $\$3{,}000$ per month with month counts or $\$36{,}000$ per year with year fractions gives the same answer.

Scope, Interpretation, and Common Traps

The classical surplus process is an aggregate cash-flow model. It does not track individual policies, premium billing dates, claim reporting delays, or the distinction between a claim incurred and a claim paid unless those features are added separately. Its Poisson arrival assumption is appropriate as a modelling starting point when arrivals are treated as random, independent across disjoint intervals, and governed by a constant intensity. A varying arrival rate, dependence between claims, or claim sizes that change systematically over time requires a different or extended model.

Do not confuse claim intensity with a guaranteed count. An intensity of $\lambda$ claims per year specifies the arrival model; it does not mean that exactly $\lambda$ claims occur each year.
Do not replace realized severities by their mean. The mean describes the severity distribution; the path formula subtracts the actual claim amounts on that path.
Do not mix time units. If $c$ is in dollars per year, multiply it by elapsed years, not by a number of months.
Apply claims at their arrival times. Between arrivals, surplus rises at rate $c$; at an arrival, it drops by the claim amount.
Do not add unmodelled cash flows implicitly. Interest, expenses, discrete premium payments, or changing rates must be represented explicitly if they matter.

Key takeaway: The classical surplus process is initial capital plus premium income accumulated over time minus cumulative claims. Its path interpretation depends on keeping the arrival model, severity model, cash-flow assumptions, and time units distinct.

Check Your Understanding

For each problem, identify the relevant model components and keep the time units consistent.

A hypothetical insurer has initial surplus $\$120{,}000$, premium income of $\$30{,}000$ per year, and one $\$18{,}000$ claim at year $0.40$. Assuming no other claims, interest, mortality, or expenses, find the surplus just before and after the claim and at year 1. (Answer: $\$132{,}000$ before, $\$114{,}000$ after, and $\$132{,}000$ at year 1.)
Claim arrivals follow a homogeneous Poisson process with intensity 5 per year. What is the probability of no arrivals during a 0.20-year interval? (Answer: $e^{-1}=0.3679$.)
An insurer starts with $\$200{,}000$, receives premiums at $\$48{,}000$ per year, and pays claims of $\$9{,}000$ at year $0.25$ and $\$40{,}000$ at year $0.75$. Find the surplus just after the second claim and at year 1, assuming no other cash flows. (Answer: $\$187{,}000$ just after the second claim and $\$199{,}000$ at year 1.)
A model specifies $N(t)$ as a Poisson process with intensity 4 per year and claim amounts as independent, identically distributed positive random variables independent of $N(t)$. State what each assumption says about the claim process and amounts. (Answer: Counts in disjoint intervals are independent, with a Poisson count of mean $4h$ in an interval of $h$ years; claim amounts are mutually independent, share one severity distribution, and are independent of arrivals.)
A hypothetical insurer has $\$70{,}000$ initial surplus and earns $\$6{,}000$ per month continuously. Claims of $\$11{,}000$ and $\$25{,}000$ occur at months 3 and 6. Find the surplus immediately after the second claim and at month 8, assuming no other cash flows. (Answer: $\$70{,}000$ immediately after the second claim and $\$82{,}000$ at month 8.)

Symbol	Interpretation	Units when time is measured in years
\(u\)	Surplus at time zero	Dollars
\(c\)	Continuous premium income rate	Dollars per year
\(\lambda\)	Poisson claim arrival intensity	Claims per year
\(N(t)\)	Number of claims by time \(t\)	Number of claims
\(X_j\)	Amount of claim \(j\)	Dollars
\(S(t)\)	Total claims paid by time \(t\)	Dollars
\(U(t)\)	Surplus at time \(t\)	Dollars

What You'll Learn