Why Calendar-Year Allocation Matters
Ratemaking analyses often group experience into fixed calendar windows. A policy, however, may begin or end partway through a calendar year. Assigning all of its exposure to the year in which it began, or the year in which it ended, misstates the amount of coverage in each experience period. The correct approach is to divide covered time at the boundaries of the calendar years and allocate each portion to the year in which it falls.
The calculation is deterministic date arithmetic: it does not require an interest rate, mortality assumption, or probability distribution. It does require a clearly stated exposure basis. This tutorial uses a fixed 365-day annualization convention in its examples, including for dates in leap years. A different prescribed basis, such as an actual/actual convention, must be applied consistently instead.
Representing Dates and Counting Covered Time
Use a consistent endpoint convention before counting days. In this tutorial, a coverage interval is written as \([s,e)\): the start date \(s\) is included and the end date \(e\) is excluded. Thus, an interval from January 1 to the following January 1 covers the entire first calendar year. This convention avoids counting a boundary date twice when adjacent intervals are combined.
For each calendar year, intersect the coverage interval with that year's date interval, \([{\rm Jan\ 1}_y,{\rm Jan\ 1}_{y+1})\). Count the days in the overlap, then apply the stated exposure amount per year and annualization denominator. A leap day is a covered day when it falls within the interval; under the fixed 365-day convention, it contributes \(1/365\) of an annual exposure unit.
Here, \(E_y\) is the exposure allocated to calendar year \(y\), \(q\) is the exposure amount per year while coverage is active, and \(D_y\) is the number of covered days in that calendar year. If the exposure amount changes across intervals, calculate each interval's contribution separately and add the contributions within each year. If the exposure measure is not a simple count of identical units, \(q\) must represent the appropriate annual exposure amount in that measure.
Record whether the end date is included or excluded, the annual exposure amount, and the day-count denominator. Do not switch conventions partway through the calculation.
Identify the covered days falling in each calendar year. A coverage interval that spans several years produces one portion for each year it overlaps.
For each portion, multiply its covered days by the annual exposure amount and divide by the chosen denominator. Add separate contributions that fall in the same calendar year.
Check that the portions account for all covered days and that the annual allocations sum to the whole-interval exposure, subject only to displayed rounding.
Worked Allocations
Worked Example: One Annual Coverage Interval Across Two Calendar Years
A hypothetical equipment policy covers one exposure unit from June 1, 2024, through May 31, 2025, inclusive. Allocate its exposure to calendar years. Use the half-open interval convention and a fixed 365-day annualization basis.
Method and assumptions. Convert the dates to \([{\rm June\ 1,\ 2024},{\rm June\ 1,\ 2025})\). The exposure amount is \(q=1\) unit per year. This is deterministic date arithmetic; no interest, mortality, or probability distribution is assumed. The main trap is counting both the final covered day and the excluded end date.
2024 portion. From June 1 to January 1, 2025, the covered days are June 30, July 31, August 31, September 30, October 31, November 30, and December 31:
Therefore, 2024 exposure is \(1(214/365)=0.5863\) units.
2025 portion. From January 1 to June 1, the covered days are January 31, February 28, March 31, April 30, and May 31:
Therefore, 2025 exposure is \(1(151/365)=0.4137\) units. Rounding only the final allocations to four decimal places, the calendar-year amounts are 0.5863 and 0.4137 units. The check is \(214+151=365\) covered days and \(0.5863+0.4137=1.0000\) units. The incorrect approach would assign the full unit to 2024 because the policy began that year; the correction is to allocate by covered days in each calendar year.
Worked Example: A Long Interval Spanning a Leap Year
A hypothetical service contract provides 1.2 exposure units per year from September 15, 2023, through March 9, 2025, inclusive. Find the allocation to each calendar year using the fixed 365-day basis.
Method and assumptions. Write the interval as \([{\rm September\ 15,\ 2023},{\rm March\ 10,\ 2025})\). The annual exposure amount is constant at \(q=1.2\). The calculation is deterministic, with no interest, mortality, or probability distribution. The trap is to omit the leap day or to use a different denominator only for 2024.
2023. September 15 to October 1 is 16 days; October through December contribute \(31+30+31\) days. Thus \(D_{2023}=108\), and:
2024. The full leap year contributes 366 covered days, including February 29:
2025. January 1 to March 10 comprises 31 days in January, 28 in February, and 9 days in March, for \(D_{2025}=68\):
Rounded to four decimal places, the allocations are 0.3551, 1.2033, and 0.2236 units. The full interval has \(108+366+68=542\) covered days, so its exposure is \(1.2(542/365)=650.4/365=1.7822\) units. Before rounding, the three allocations also sum to \(650.4/365\), confirming the allocation. Using 366 as the denominator for 2024 would be a different annualization convention, not a correction to this calculation.
Worked Example: Combining Separate Intervals for Multiple Items
A hypothetical policy covers three identical machines from November 1, 2022, through January 31, 2024, inclusive. A fourth machine is covered separately from July 1 through December 31, 2023, inclusive. Allocate total machine-year exposure to each calendar year using a fixed 365-day basis.
Method and assumptions. The first interval is \([{\rm November\ 1,\ 2022},{\rm February\ 1,\ 2024})\); the second is \([{\rm July\ 1,\ 2023},{\rm January\ 1,\ 2024})\). Each active machine contributes one exposure unit per year. There is no interest, mortality, or probability distribution. The trap is to treat both intervals as if they had the same start and end dates.
First interval: three machines. The 2022 portion has 30 days in November and 31 in December, or 61 days. The full 2023 portion has 365 days. The 2024 portion has 31 days in January. The contributions are:
Second interval: one machine. July through December 2023 contains \(31+31+30+31+30+31=184\) days. Its contribution is \(184/365=0.5041\) units in 2023.
Combine by year. Add the two 2023 contributions, keeping the exact fractions until the final rounding:
The final allocations, rounded to four decimal places, are 0.5014, 3.5041, and 0.2548 machine-year units. The exact total is \((183+1095+184+93)/365=1555/365=4.2603\) units, matching the sum of the separate interval exposures. The correction to the common mistake is to allocate and then aggregate each distinct interval, rather than applying one set of dates to all four machines.
Why the Allocation Reconciles
The key result is that splitting an interval does not create or remove exposure. It only assigns its covered days to separate experience windows. Suppose a constant amount \(q\) is active on an interval that overlaps calendar years \(y_1,\ldots,y_k\), and let \(D_{y_j}\) be the covered days in year \(y_j\). Because the year portions partition the interval, each covered day belongs to exactly one portion. Therefore their day counts add to the total covered days \(D\):
Applying the same exposure amount and denominator to every portion gives:
The middle equality follows because \(q/365\) is common to every portion. The final expression is precisely the exposure calculated for the entire interval under the same convention. For multiple intervals, apply this argument to each interval and add their totals. The reconciliation depends on using a consistent endpoint rule, day count, annual amount, and denominator throughout.
Common Traps and Practical Checks
Calendar-year allocation is simple arithmetic, but small date-handling choices can materially affect a data extract. Before accepting a result, check the following:
- Putting all exposure in one year. Calendar-year allocation follows covered time, not the policy's start year or end year.
- Counting a boundary date twice. With half-open intervals, the end date is excluded. Adjacent intervals \([a,b)\) and \([b,c)\) meet without overlapping on date \(b\).
- Mixing day-count bases. A fixed 365-day basis treats each covered day as \(1/365\) of an annual unit, even in leap years. If a different basis is required, use it for every portion and state it.
- Rounding too early. Keep day counts and fractions exact through aggregation; round only the final calendar-year figures. Rounded annual amounts may not sum exactly to a separately rounded total.
- Ignoring separate intervals. If an item has more than one covered interval, allocate each interval on its own dates, then add contributions by calendar year.
A reliable final check has two parts: the covered-day portions must account for the full interval, and the unrounded calendar allocations must equal the total exposure computed from that same full interval. These checks catch many endpoint, leap-day, and aggregation errors before the exposure is used in an experience analysis.
Check Your Understanding
For each problem, use half-open intervals and a fixed 365-day annualization basis unless the problem states otherwise. Round final exposures to four decimal places.
- A hypothetical policy has one annual exposure unit on \([{\rm October\ 1,\ 2025},{\rm January\ 1,\ 2026})\). How much exposure is allocated to 2025 and 2026? (Answer: 2025: 0.2521 units; 2026: 0.0000 units.)
- A contract provides one annual unit on \([{\rm December\ 1,\ 2023},{\rm March\ 1,\ 2024})\). Allocate exposure by calendar year, counting February 29 as covered. (Answer: 2023: 0.0849 units; 2024: 0.1644 units.)
- A hypothetical fleet exposure amount is 2.5 units per year on \([{\rm May\ 1,\ 2022},{\rm August\ 1,\ 2023})\). Find the allocation for each year and the total. (Answer: 2022: 1.6781 units; 2023: 1.4521 units; total: 3.1301 units.)
- One item is covered on \([{\rm January\ 1,\ 2024},{\rm April\ 1,\ 2024})\), and a separate item is covered on \([{\rm October\ 1,\ 2024},{\rm January\ 1,\ 2025})\). Each contributes one unit per year. Find total exposure by calendar year. (Answer: 2024: 0.5014 units; 2025: 0.0000 units.)
- A hypothetical contract provides 0.75 annual units on \([{\rm November\ 15,\ 2023},{\rm February\ 15,\ 2025})\). Find the allocation by calendar year and the unrounded-basis total, allowing for February 29, 2024. (Answer: 2023: 0.0966 units; 2024: 0.7521 units; 2025: 0.0925 units; total: 0.9411 units.)