Introduction

Actuarial mathematics begins with a simple human problem:

We make promises about the future, but the future is uncertain.

An insurance company may promise to pay if a house burns, a person dies, a worker becomes disabled, or a driver causes damage. A pension plan may promise retirement income decades from now. A government, employer, insurer, or family may need to prepare for costs that are not known exactly today. Actuarial work studies these uncertain future cash flows so that decisions about premiums, reserves, benefits, capital, and risk can be made responsibly.

This book is for a beginner adult reader. You do not need to already know actuarial science. If this is your first time seeing the subject, that is completely acceptable. We will build the ideas slowly from first principles: probability, statistics, interest theory, present value, life contingencies, insurance losses, reserves, risk measures, and actuarial judgment.

What is actuarial mathematics?

An actuary is a professional who analyzes financial risk, especially when uncertain events create financial consequences. In actuarial textbooks, the field is commonly built from probability, statistics, finance, survival models, insurance mathematics, and risk modeling (Bowers et al., 1997; Dickson, Hardy, and Waters, 2020; Klugman, Panjer, and Willmot, 2019).

Let us unpack that carefully.

A risk is the possibility that an outcome will differ from what we expect, especially in a way that matters financially. For example, if an insurer sells car insurance, it does not know exactly how many claims will occur next year or how expensive each claim will be. The number of claims and the cost of each claim are uncertain. That uncertainty is risk.

A cash flow is money paid or received at a particular time. For example:

• a policyholder pays a premium of \$500 today;
• an insurer pays a claim of \$8,000 six months from now;
• a pension fund pays \$1,200 every month to a retiree;
• a life insurer pays \$100,000 when an insured person dies.

An uncertain cash flow is a cash flow whose amount, timing, or occurrence is not known in advance. Much of actuarial mathematics is about uncertain cash flows.

Here is a small example. Suppose 1,000 people each pay \$100 into a simple insurance pool. The pool collects:

\[ 1{,}000 \times 100 = 100{,}000. \]

If 8 people have claims and each claim costs \$10,000, the pool pays:

\[ 8 \times 10{,}000 = 80{,}000. \]

In this simplified situation, the pool has enough money. But if 15 people have claims, the pool pays:

\[ 15 \times 10{,}000 = 150{,}000, \]

which is more than it collected. The actuarial question is not merely “What happened last time?” but “What range of outcomes is possible, how likely are they, and what should we charge or hold in reserve so the promise is financially sound?”

That is the central spirit of this book.

Insurance as a way to share losses

Insurance is a financial arrangement in which many people or organizations contribute money, usually through premiums, so that those who suffer covered losses can receive payments. The basic idea is not that risk disappears. Instead, risk is pooled.

A risk pool is a group of similar exposures whose losses are combined. An exposure is a person, property, policy, vehicle, business, or other unit that can produce a loss. For example:

• in life insurance, each insured life is an exposure;
• in motor insurance, each insured vehicle may be an exposure;
• in property insurance, each insured building may be an exposure;
• in health insurance, each covered person may be an exposure.

Pooling works best when many exposures are combined and when the losses are not all perfectly linked. If one house burns, that does not usually mean every house burns. But if all houses are in the same flood zone, the losses may be strongly connected. This is why actuaries care not only about average cost but also about dependence, meaning the way uncertain outcomes move together.

Example: imagine two insurance portfolios.

In Portfolio A, 10,000 homes are spread across many cities. One storm is unlikely to damage all of them. In Portfolio B, 10,000 homes are all in the same coastal area. One hurricane could damage many homes at once. Even if the average expected claim cost were similar, Portfolio B may require more capital because extreme losses are more concentrated.

This distinction is a major actuarial idea: average cost matters, but risk around the average also matters.

The role of probability

Probability is the mathematical language of uncertainty. A probability is a number between 0 and 1 that describes how likely an event is within a model. A probability of 0 means the event is impossible in that model. A probability of 1 means the event is certain in that model. A probability of 0.25 means the event has a 25% chance in that model.

For example, suppose historical and underwriting information suggest that a certain type of policy has a 2% chance of producing a claim during the next year. We may write:

\[ P(\text{claim}) = 0.02. \]

This does not mean we know which particular policy will have a claim. It means the model assigns a 2% chance to the event.

A key actuarial habit is to separate two questions:

1. What does the model say?
2. How reliable is the model?

For example, saying “the probability is 2%” is not the same as saying “we are absolutely certain the true probability is 2%.” The 2% may be estimated from data, judgment, experience, or a combination of these. Later chapters will study statistical inference, credibility, model uncertainty, and professional judgment so that we do not treat uncertain estimates as if they were perfect facts.

The role of statistics

Statistics is the discipline of learning from data. In actuarial work, data may include policy records, claim counts, claim amounts, dates of loss, ages, locations, benefit amounts, survival experience, economic variables, and many other items.

Suppose an insurer observes the following claim counts over five years:

A beginner might add the numbers and divide by 5:

\[ \frac{94+103+98+110+95}{5} = 100. \]

So the average is 100 claims per year. That is useful, but actuarial thinking asks more:

• Are these years comparable?
• Did the number of policies change?
• Were there changes in policy wording?
• Was there inflation?
• Were large claims handled consistently?
• Is five years enough data?
• How variable are the outcomes?
• Could next year be very different?

Statistics gives tools for estimating, testing, measuring uncertainty, and deciding how much trust to place in data. Actuarial judgment is needed because real data are often incomplete, biased, changing, or affected by business decisions.

The role of interest and present value

Money paid today is not financially identical to the same amount of money paid many years from now. This is the idea behind the time value of money.

If you can invest \$1,000 today and earn interest, then \$1,000 today may become more than \$1,000 in the future. Conversely, a payment of \$1,000 due in the future may be worth less than \$1,000 today, depending on interest rates and risk.

The present value of a future payment is its value measured at an earlier date, usually today. For example, if the annual effective interest rate is 5%, then receiving \$1,050 one year from now has present value:

\[ \frac{1{,}050}{1.05} = 1{,}000. \]

This calculation is called discounting. Discounting is essential in actuarial work because insurance and pension promises often involve payments far in the future. Life insurance, annuities, pensions, reserves, and long-term liabilities cannot be understood properly without present value.

In this book, we will first learn interest theory by itself, then combine it with probability. That combination gives us the expected present value of uncertain cash flows, one of the central quantities in life insurance mathematics and many other actuarial models (Bowers et al., 1997; Dickson, Hardy, and Waters, 2020).

Average cost is not the whole story

A common beginner mistake is to think that actuarial work is only about calculating averages. Averages are important, but they are not enough.

Consider two risks:

The average loss for Risk A is clearly \$1,000.

For Risk B, the expected loss is:

\[ 0.99 \times 0 + 0.01 \times 100{,}000 = 1{,}000. \]

So both risks have the same expected loss. But they are not equally risky. Risk A is predictable. Risk B is usually harmless but sometimes severe. An insurer, pension plan, or financial institution must consider not only expected cost but also variability, extreme outcomes, capital needs, and the possibility of insolvency.

Insolvency means not having enough assets to meet obligations when they come due. One reason actuarial work exists is to reduce the chance that promises are made without adequate financial support.

Later in the book, we will study risk measures, such as value at risk and tail value at risk. These tools help describe the size of adverse outcomes, especially in the tail of a distribution. In probability and statistics, the tail of a distribution refers to the extreme end of possible outcomes, such as unusually large losses.

Models: useful, not magical

A model is a simplified representation of reality designed to answer a question. A map is a model of geography. A weather forecast is a model of atmospheric behavior. An actuarial pricing formula is a model of future claims and expenses.

Models are useful because reality is too complicated to hold all at once. But every model leaves something out. This is why actuaries must understand assumptions.

An assumption is a condition or simplification accepted for the purpose of analysis. For example:

• claim counts follow a Poisson distribution;
• mortality rates will follow a selected life table;
• interest is 4% per year;
• expenses are 10% of premium;
• claims inflation is 3% per year;
• policyholders do not surrender their policies.

Some assumptions may be reasonable. Others may be dangerous. The same assumption may be reasonable in one context and unreasonable in another.

Example: assuming a constant 4% interest rate may be acceptable for a classroom exercise. It may be too simple for valuing a complex pension plan with payments over many decades. The model is not “wrong” merely because it is simplified, but the user must know what the simplification affects.

This book will repeatedly ask:

• What is the model?
• What are its inputs?
• What assumptions does it make?
• What output does it produce?
• What could make the output misleading?
• How sensitive is the answer to changes in assumptions?

That way, mathematics becomes a tool for judgment rather than a substitute for judgment.

Life contingencies and general insurance

Actuarial mathematics often divides insurance work into broad areas. Two important ones are life contingencies and general insurance.

A life contingency is a financial payment whose occurrence or timing depends on human life events, especially survival or death. Life insurance and life annuities are classic examples.

Example: a whole life insurance policy may pay \$100,000 when the insured person dies. The insurer does not know when payment will occur, but mortality data and survival models can help estimate probabilities and present values.

A life annuity is a series of payments made while a person is alive. For example, a retirement annuity may pay \$2,000 per month for as long as the retiree survives. The cost depends on interest rates and survival probabilities. If the retiree lives longer, more payments are made.

General insurance, also called property and casualty insurance in some countries, includes risks such as motor, property, liability, travel, and many commercial insurance lines. General insurance often studies claim frequency and claim severity separately.

Claim frequency means how many claims occur.
Claim severity means how large the claims are.

Example: in motor insurance, one model may estimate the number of accidents per year, while another estimates the cost of each accident. Combining frequency and severity gives an aggregate loss model, which describes the total loss for a portfolio. Loss modeling through frequency, severity, and aggregate loss distributions is a standard structure in actuarial risk modeling (Klugman, Panjer, and Willmot, 2019).

Both life contingencies and general insurance use probability, statistics, present value, and judgment. They differ in the types of data, time horizons, laws, products, and risks most commonly involved.

Premiums, reserves, and capital

Three words will appear throughout this book: premium, reserve, and capital.

A premium is the amount charged for insurance coverage or a financial guarantee. If you pay \$600 per year for car insurance, that \$600 is a premium.

A reserve is an amount set aside or recognized to help meet future obligations. For example, if an insurer has already sold a life insurance policy, it may need to hold a reserve because a future death benefit may have to be paid. Reserving is not the same as pricing. Pricing asks, “What should we charge?” Reserving asks, “Given promises already made, how much value should we recognize now for future obligations?”

Capital is financial support available to absorb adverse experience beyond what is expected. If claims are higher than expected, reserves and premiums may not be enough. Capital helps protect the insurer or plan from severe but plausible outcomes.

A simplified structure is:

\[ \text{Premiums} + \text{Investment income} + \text{Capital support} \]

must be sufficient, in a suitable risk-management sense, for:

\[ \text{Claims} + \text{Expenses} + \text{Required reserves} + \text{Profit or surplus objectives}. \]

This is not a complete accounting identity, but it expresses the practical actuarial concern: money coming in and financial resources held must be adequate for money going out and promises made.

What this book will build

The chapters are arranged so that each layer supports the next.

First, we develop the mathematical language: algebra, functions, logarithms, exponentials, summation notation, and basic calculus ideas. These tools are not included to make the book difficult. They are included because actuarial formulas need a precise language.

Then we study probability: events, conditional probability, independence, Bayes’ theorem, random variables, distributions, expectation, variance, and dependence. These ideas let us describe uncertain outcomes.

Next, we study statistics and data. We learn how estimates are made, how uncertain they are, and how actuarial datasets are structured. We also discuss exposure, censoring, truncation, outliers, and credibility of data.

After that, we study the time value of money and uncertain present values. This is where probability and finance begin to work together.

Then we move into survival models, life tables, life insurance benefits, life annuities, premiums, and reserves. These chapters form the core of traditional life actuarial mathematics.

The book then expands into multiple-life models, multiple-decrement models, general insurance frequency and severity models, aggregate loss models, credibility theory, regression, predictive modeling, risk measures, solvency, communication, and professional judgment.

Finally, the capstone chapter guides you through building a simple actuarial model from start to finish.

The goal is not only that you can compute numbers. The goal is that you can explain what the numbers mean.

How to think like an actuarial beginner

As you read, try to develop four habits.

First, always identify the uncertain quantity. Is it the number of claims? The size of a claim? The lifetime of a person? The interest rate? The total present value of future payments?

Second, ask what information is available. Do we have data? A table? Expert judgment? Market information? A regulation? A contract?

Third, distinguish the calculation from the interpretation. A formula can produce a number, but the actuarial question is whether that number is useful for the decision being made.

Fourth, communicate clearly. Actuarial work often affects people who are not specialists: policyholders, employers, regulators, executives, trustees, and the public. A technically correct answer that cannot be explained may fail in practice.

Here is a small example of actuarial communication.

A poor explanation says:

The expected present value is 742.31 using \(v^t\) and the survival distribution.

A better beginner-level explanation says:

Based on the assumed interest rate and survival probabilities, the average present value of the future benefit is about \$742. This is not the exact cost for every individual policy. It is an average under the model, and actual experience may be higher or lower.

The second explanation is longer, but it is more useful. It says what was calculated, what assumptions were involved, and what the number does not mean.

A first tiny actuarial model

Let us finish the introduction with a very small model. Suppose an insurer sells one-year coverage. The policy pays \$10,000 if a claim occurs during the year and pays nothing otherwise. Suppose the probability of a claim is 3%.

Let the random loss be \(X\). A random variable is a quantity whose value depends on an uncertain outcome. Here:

\[ X = \begin{cases} 10{,}000, & \text{if a claim occurs},\\ 0, & \text{if no claim occurs}. \end{cases} \]

The expected loss is:

\[ E[X] = 0.03 \times 10{,}000 + 0.97 \times 0 = 300. \]

So the expected claim cost is \$300.

But an insurer would usually not charge exactly \$300. It may need to consider expenses, profit or surplus objectives, uncertainty, capital costs, taxes, regulation, competition, and the possibility that the 3% assumption is wrong. If the insurer sells many such policies, pooling may make the average more stable, but large deviations can still occur.

This tiny example contains many themes of the whole book:

• uncertainty is represented with probability;
• a financial outcome is represented as a random variable;
• the expected value gives an average model cost;
• the premium decision requires more than the average;
• assumptions must be stated;
• interpretation matters.

That is actuarial mathematics from first principles: careful reasoning about uncertain financial promises.

References

Bowers, Newton L., Hans U. Gerber, James C. Hickman, Donald A. Jones, and Cecil J. Nesbitt. 1997. Actuarial Mathematics. 2nd ed. Schaumburg, IL: Society of Actuaries.

Dickson, David C. M., Mary R. Hardy, and Howard R. Waters. 2020. Actuarial Mathematics for Life Contingent Risks. 3rd ed. Cambridge: Cambridge University Press.

Klugman, Stuart A., Harry H. Panjer, and Gordon E. Willmot. 2019. Loss Models: From Data to Decisions. 5th ed. Hoboken, NJ: Wiley.