"Calculating Central and Initial Exposed to Risk"

Introduction

In our last module, we explored the sophisticated world of proportional hazards models, culminating in a discussion on how to choose between the robust, semi-parametric Cox model and more efficient fully parametric alternatives. That marked the end of our deep dive into modeling survival for individuals based on their specific characteristics. Now, we are making a fundamental shift in perspective. We will move from individual-level modeling to estimating mortality rates for entire populations—a core activity in actuarial work.

This transition requires a new foundational concept: exposed to risk. Before we can calculate a mortality rate, which is intuitively the number of deaths divided by the population size, we must have a rigorous method for defining and measuring the "population size." This is not as simple as counting heads on a specific day, because people enter and leave the group under observation at different times. This lesson will introduce you to the two primary methods for measuring this exposure: central exposed to risk and initial exposed to risk. Mastering these calculation methods is the first and most critical step in constructing mortality tables from raw data.

What is "Exposed to Risk"?

Imagine you want to calculate the mortality rate for 25-year-olds in an insurance portfolio during the year 2023. The basic formula for a rate is:

Rate = \frac{Number of events}{Exposure to the risk of the event}

The numerator is straightforward: you count the number of deaths of 25-year-olds in 2023. The denominator, the "exposure," is more complex. People turn 25 at different points in the year, some may buy a policy mid-year (new entrants), and others might let their policy lapse (withdrawals). A simple count of policyholders on January 1st would be inaccurate.

"Exposed to risk" is the actuarial solution to this problem. It is a precise measure of the total time that individuals in a population were exposed to the risk of an event (e.g., death) during a specified period. There are two main ways to calculate this.

Central Exposed to Risk ( $E_{x}^{c}$ ): This measures the total duration of time, usually in person-years, that the group was exposed to the risk of death while they were aged $x$ . It is the most common and conceptually direct method used today.
Initial Exposed to Risk ( $E_{x}$ ): This is an older method that essentially counts the number of individuals who were under observation at the moment they turned age $x$ . It then makes adjustments for those who die during the year.

While both methods aim to measure exposure, they do so from different perspectives. The actuarial profession has increasingly favored the central approach due to its theoretical advantages, a point we will revisit.

Central Exposed to Risk ( $E_{x}^{c}$ ): The Duration Approach

The central exposed to risk at age $x$ , denoted $E_{x}^{c}$ , is defined as the sum of the time each individual in the study was observed and at risk of death between age $x$ and age $x + 1$ .

Let's formalize this with a definition from a standard actuarial text.

[PDF] MASTER ACTUARIAL SCIENCE MASTER'S FINAL WORK

This excerpt from a Master's thesis provides a clear and concise mathematical definition of central exposed to risk and its role in calculating mortality rates.

Please read the first paragraph of section 2.1.2, starting from the definition. This establishes that E_x^c is the sum of individual observation times. Then, see how it's used to estimate the force of mortality, \mu_{x+1/2}, by reading the paragraph starting on crude mortality rates. You'll see that the crude mortality rate is simply the number of deaths divided by this central exposure.

The most direct way to understand the calculation is to work through an example. As you prefer a problem-solving approach, let's analyze a small portfolio.

Consider a one-year study from 01/Jan/2021 to 01/Jan/2022. We have five individuals with different entry and exit dates. We want to calculate the central exposed to risk for age 20 ( $E_{20}^{c}$ ) and age 21 ( $E_{21}^{c}$ ).

The setup is illustrated on the following whiteboard.

This whiteboard illustrates the calculation of central exposed to risk for five individuals over a one-year observation period. For each person, the time they spent at risk within the period is calculated and then allocated to the specific age (20 or 21) they were during that time.

Let's calculate the exposure for each person, breaking it down by age. For simplicity, we will measure exposure in months.

Life A (DOB: 01/Mar/2000): Enters the study on 01/Jan/2021 and exits on 01/July/2021.
- On 01/Jan/2021, Life A is age 20. They turn 21 on 01/Mar/2021.
- Exposure at age 20: From 01/Jan to 01/Mar = 2 months.
- Exposure at age 21: From 01/Mar to exit on 01/July = 4 months.
Life B (DOB: 01/May/2000): Enters on 01/Jan/2021 and is still in the study at the end of the year.
- On 01/Jan/2021, Life B is age 20. They turn 21 on 01/May/2021.
- Exposure at age 20: From 01/Jan to 01/May = 4 months.
- Exposure at age 21: From 01/May to 01/Jan/2022 = 8 months.
Life C (DOB: 01/July/2000): Enters on 01/Jan/2021 and exits on 01/Apr/2021.
- On 01/Jan/2021, Life C is age 20. Their 21st birthday is after their exit date.
- Exposure at age 20: From 01/Jan to exit on 01/Apr = 3 months.
- Exposure at age 21: 0 months.
Life D (DOB: 01/Sep/2000): Enters on 01/Jun/2021 and is still in at year-end.
- On entry (01/Jun/2021), Life D is age 20. They turn 21 on 01/Sep/2021.
- Exposure at age 20: From 01/Jun to 01/Sep = 3 months.
- Exposure at age 21: From 01/Sep to 01/Jan/2022 = 4 months.
Life E (DOB: 01/Nov/2000): Enters on 01/Jun/2021 and exits on 01/Dec/2021.
- On entry (01/Jun/2021), Life E is age 20. They turn 21 on 01/Nov/2021.
- Exposure at age 20: From 01/Jun to 01/Nov = 5 months.
- Exposure at age 21: From 01/Nov to exit on 01/Dec = 1 month.

Now, we sum the exposures for each age across all individuals:

Total $E_{20}^{c}$ : 2 (A) + 4 (B) + 3 (C) + 3 (D) + 5 (E) = 17 months
Total $E_{21}^{c}$ : 4 (A) + 8 (B) + 0 (C) + 4 (D) + 1 (E) = 17 months

To express these in person-years, we would divide by 12. So, $E_{20}^{c} = 17/12$ years.

Initial Exposed to Risk ( $E_{x}$ ): The Headcount Approach

Initial exposed to risk, $E_{x}$ , takes a different approach. Instead of summing up durations, it counts the number of people who are observed at the exact start of their year of age $x$ . A key complication is how to treat deaths.

The classic actuarial texts propose two main ways to handle the exposure contributed by those who die during the year.

[PDF] Comparison of approaches for calculating initial exposure

This technical note from the UK's Continuous Mortality Investigation (CMI) committee explains why they moved away from initial exposure, but in doing so, it provides an excellent definition of the two competing approaches for its calculation.

First, read the paragraph that defines how the observation period is determined for survivors: last day for non-deaths. This is straightforward. Next, focus on the core issue: how to treat deaths. Read the two approaches outlined in the document: Approach 1 and Approach 2.

Let's summarize these two methods:

Approach 1 (Full Year for Deaths): A person who dies at age $x$ is counted as contributing a full year of exposure to $E_{x}$ . The rationale is that they were at risk for the entire year of age, even if they didn't survive it. This method attempts to provide an unbiased estimate of $q_{x}$ directly, where $q_{x} = θ_{x} / E_{x}$ ( $θ_{x}$ being the number of deaths at age $x$ ).
Approach 2 (Partial Year for Deaths): A person who dies at age $x$ contributes exposure only up to the end of the observation period (e.g., the end of the calendar year). This is more aligned with the central exposure philosophy but is applied within an initial exposure framework.

The CMI paper you just reviewed highlights that both approaches have technical limitations, especially when dealing with changing population sizes or complex study designs. This is why central exposed to risk is now the standard. It avoids these ambiguities by simply measuring the actual time observed.

To give a concrete sense of the difference, let's reconsider our example for age 20. Initial exposed to risk, $E_{20}$ , would be the count of people who turned 20 while under observation.

Life A: Turned 20 before the study began, so does not contribute.
Life B: Same as A.
Life C: Same as A.
Life D: Turned 20 before entering the study.
Life E: Same as D.

In our specific one-year census from 01/Jan/2021, none of the individuals had their 20th birthday within the observation period. If we had a longer study period, say from 2019, then lives A through E would all have been counted in $E_{20}$ during the year 2020. This illustrates that $E_{x}$ is fundamentally a count of people passing a milestone (their birthday), whereas $E_{x}^{c}$ is a summation of time spent in an age bracket.

Conclusion

In this lesson, we have laid the groundwork for estimating population mortality rates. You've learned the definitions of the two key exposure measures and, most importantly, practiced the calculation of the one that matters most in modern actuarial work.

Key Takeaways:

Exposed to Risk is the precise actuarial measure for the "population" denominator in a mortality rate calculation.
Central Exposed to Risk ( $E_{x}^{c}$ ) is a duration-based measure, summing the total time all individuals were observed at risk while aged $x$ . This is the preferred modern method.
Initial Exposed to Risk ( $E_{x}$ ) is a headcount-based measure, counting individuals who begin their year of age $x$ under observation, with special rules for handling deaths.
The calculation of $E_{x}^{c}$ requires tracking each individual's entry date, exit date, and date of birth to correctly allocate their observed time to the correct age bracket.

You are now equipped with the fundamental tool for turning raw data on lives and deaths into meaningful exposure measures.

In our next lesson, we will build directly on this foundation. We'll explore the relationship between central and initial exposed to risk more formally and look at how to handle practical complications, such as when we only have census data at the start and end of a year instead of exact entry and exit dates for every individual.

Recommended resources

Introduction

What is "Exposed to Risk"?

Central Exposed to Risk (Exc​): The Duration Approach

Initial Exposed to Risk (Ex​): The Headcount Approach

Conclusion

Central Exposed to Risk ( $E_{x}^{c}$ ): The Duration Approach

Initial Exposed to Risk ( $E_{x}$ ): The Headcount Approach