Aligning the Experience Period With the Effective Date

Why Timeline Alignment Matters

A rate indication compares experience observed in one period with the business and losses expected under a prospective rate level. Those periods do not necessarily represent the same dates. Historical losses may be spread across a past interval, while the proposed rates apply to policies beginning on a range of future dates. A clear timeline helps identify which dates the historical and prospective amounts actually represent.

The central distinction is between a policy effective date and a loss date. A policy effective date marks the start of coverage. If a policy runs for a year, losses covered by it can occur throughout that year. Therefore, the midpoint of a set of policy effective dates is not automatically the average date of the losses generated by those policies.

Definition: The average loss date is the time-weighted average date on which the losses in a specified amount or group of expected losses occur. For losses assumed to be uniform over a historical interval, it is the interval's midpoint. For future policies, it must account for both the distribution of policy effective dates and the timing of losses during each policy term.

This tutorial uses a simple timeline convention: measure elapsed time from a stated origin in months or years, and treat a midpoint as the average date when losses are uniform over the interval. A rate effective date is a point in time; a prospective effective-date period is a range. If that range is not specified, the prospective average loss date cannot be determined from the rate effective date alone.

Historical and Prospective Average Dates

Suppose historical losses occur uniformly from date $a$ through date $b$. The average date is the midpoint. If time is measured in a consistent unit, such as years from an origin, this is the arithmetic mean of the interval endpoints.

\bar{t}_{H}=\frac{a+b}{2}

For prospective business, first find the average policy effective date. Then account for the time from policy inception to loss. Under the common simplifying assumption that losses are uniform through a policy term of length $T$, the average loss occurs $T/2$ after each policy's effective date. Thus, when all policies have the same term,

\bar{t}_{P}=\bar{t}_{E}+\frac{T}{2},

where $\bar{t}_{E}$ is the average policy effective date and $\bar{t}_{P}$ is the prospective average loss date. If effective dates themselves are uniform from $e_1$ to $e_2$, then $\bar{t}_{E}=(e_1+e_2)/2$.

Assumptions and limits: The prospective formula assumes identical policy terms and that the expected losses for each policy are uniform over its term. If losses are seasonally concentrated, coverage is not uniform, policy terms differ, or effective-date cohorts have different expected loss weights, use the appropriate weighted average or a more suitable timing assumption. Do not apply the simple midpoint formula without checking its assumptions.

Why the Prospective Formula Holds

Represent a policy's effective date by $E$, and the time from that effective date to an expected loss by $U$. The loss date is $E+U$. If losses are uniform during a term of length $T$, then $U$ is uniform from 0 to $T$, so its mean is $T/2$. Taking expectations gives the average loss date:

\operatorname{E}[E+U] =\operatorname{E}[E]+\operatorname{E}[U] =\bar{t}_{E}+\frac{T}{2}.

This also explains how to handle several effective-date cohorts. If cohort $j$ has weight $w_j$, effective date $e_j$, and the same term $T$, its average loss date is $e_j+T/2$. The combined average is the weighted average of those cohort dates. Since the same $T/2$ is added to every cohort, this equals the weighted average effective date plus $T/2$.

\bar{t}_{P} =\frac{\sum_j w_j(e_j+T/2)}{\sum_j w_j} =\frac{\sum_j w_je_j}{\sum_j w_j}+\frac{T}{2}.

The weights should represent the quantity whose average date is being calculated. In a simplified example, identical policies may be weighted by policy count. For historical losses divided into intervals, weights may be the loss amounts assigned to those intervals. Choosing an inappropriate weight changes the date being represented.

Worked Timeline Calculations

Worked Example: One-Year Policies With a Future Effective-Date Range

A hypothetical historical loss amount is assumed to occur uniformly from January 1, 2023, to January 1, 2024. A proposed group of 12-month policies has effective dates uniformly distributed from October 1, 2024, to October 1, 2025. Assume losses are uniform over each policy term. Find the historical and prospective average loss dates and the time between them.

Method. Find the midpoint of the historical loss interval. For the prospective business, find the midpoint of the effective-date range and then add half the policy term.

Historical average loss date. The interval runs from January 1, 2023, to January 1, 2024, a 12-month interval. Its midpoint is 6 months after January 1, 2023:

\bar{t}_{H}=\text{July 1, 2023}.

Prospective average effective date. The effective-date range also spans 12 months. Its midpoint is 6 months after October 1, 2024, or April 1, 2025.

Prospective average loss date. A 12-month term has an average loss time of 6 months after inception under the uniform-loss assumption. Add 6 months to April 1, 2025, giving October 1, 2025.

Interval between average loss dates. From July 1, 2023, to October 1, 2025, is 27 months, or $27/12=2.25$ years. The historical and prospective average loss dates are therefore July 1, 2023, and October 1, 2025, respectively, 2.25 years apart.

Check. The prospective effective-date midpoint is April 1, 2025, but losses occur on average six months after that midpoint. The resulting October 1, 2025 date is later than the effective-date midpoint, as expected.

Worked Example: Shorter Policy Terms

A hypothetical historical loss amount is assumed uniform from January 1, 2022, through January 1, 2024. Prospective six-month policies have effective dates uniformly distributed from April 1, 2025, through October 1, 2025. Losses are assumed uniform through each policy term. Determine both average loss dates and their separation.

Historical date. The historical interval is 24 months long. Its midpoint is 12 months after January 1, 2022, so the historical average loss date is January 1, 2023.

Prospective effective-date midpoint. April 1 to October 1 is a six-month interval. Its midpoint is July 1, 2025.

Prospective loss date. Each policy term is six months, so the mean time from inception to loss is $6/2=3$ months. Adding three months to July 1, 2025, gives October 1, 2025.

Separation. January 1, 2023, to October 1, 2025, is 33 months, or $33/12=2.75$ years. Thus the two average loss dates are 2.75 years apart.

Check. The effective-date midpoint occurs before the average loss date by half a six-month term. Using only the effective-date midpoint would understate the prospective loss date by three months.

Worked Example: Unequal Weights Across Cohorts

A hypothetical insurer's historical losses are grouped by calendar year. Assume losses are uniform within each year: $100,000 occurred in 2022, $200,000 in 2023, and $300,000 in 2024. For a prospective illustration, identical 12-month policies have effective dates of January 1, May 1, and September 1, 2026, with relative weights 1, 2, and 1. Assume uniform losses through each policy term. Find the weighted historical and prospective average loss dates and the gap between them.

Historical average. The midpoint of each calendar year is July 1. Measure time in months from January 1, 2022. The three midpoints are 6, 18, and 30 months from the origin. Weight them by the corresponding loss amounts:

\bar{t}_{H} =\frac{100(6)+200(18)+300(30)}{100+200+300} =\frac{13{,}200}{600} =22\text{ months}.

The factors of one thousand dollars cancel from the weighted average. Twenty-two months after January 1, 2022, is November 1, 2023. Therefore, the historical average loss date is November 1, 2023.

Prospective average effective date. Measure from January 1, 2026. The three effective dates are 0, 4, and 8 months from that origin. Apply the stated weights:

\bar{t}_{E} =\frac{1(0)+2(4)+1(8)}{1+2+1} =\frac{16}{4} =4\text{ months}.

The weighted average effective date is May 1, 2026. The 12-month policies have a six-month average time to loss, so the prospective average loss date is November 1, 2026.

Separation and check. November 1, 2023, to November 1, 2026, is exactly 36 months, or 3 years. The greater weight on the May effective-date cohort pulls the average effective date to May 1, rather than the unweighted midpoint of the three dates. Adding the term adjustment then gives the stated prospective loss date.

Common Timeline Errors

The most frequent error is to calculate the midpoint of prospective policy effective dates and report it as the prospective average loss date. That omits the delay from policy inception to loss. For example, in the first worked example, an incorrect calculation would use April 1, 2025, as the prospective average loss date. Comparing July 1, 2023, with April 1, 2025, gives 21 months, or 1.75 years. That answer is wrong under the stated assumptions: the one-year policies add an average six-month loss delay, moving the prospective date to October 1, 2025. The correct comparison is 27 months, or 2.25 years.

Do not confuse a rate effective date with a period. One date alone does not specify the distribution of future policy effective dates.
Do not add a full term. Under uniform losses during a policy term, add half the term, not the entire term.
Use matching time units. Convert months to years only after calculating the date interval; 27 months is 2.25 years, not 2 years.
Do not assume every average is a midpoint. A midpoint is justified for a uniform interval. Unequal cohort weights require a weighted average.
State the timing assumption. If losses are not expected to occur uniformly through the term, the half-term adjustment may not represent their average date.

Aligning the dates identifies the time span represented by the experience relative to the prospective losses. That interval can then be used consistently in the broader indication framework. The date calculation itself does not determine a trend rate or resolve other choices about the data; it establishes the timeline those choices must use.

Check Your Understanding

For each problem, calculate the historical average loss date, the prospective average loss date, and the time between them when requested. Assume the stated uniformity conditions hold.

Historical losses are uniform from January 1, 2024, to January 1, 2025. Twelve-month policies have effective dates uniform from January 1, 2026, to July 1, 2026, with losses uniform over each term. Find the two average loss dates and the gap. (Answer: July 1, 2024; October 1, 2026; 27 months or 2.25 years.)
Historical losses are uniform from January 1, 2023, to January 1, 2024. Six-month policies have effective dates uniform from October 1, 2025, to April 1, 2026, with losses uniform over each term. Find the two average loss dates and the gap. (Answer: July 1, 2023; April 1, 2026; 33 months or 2.75 years.)
Historical losses are uniform from January 1, 2024, to January 1, 2025. Prospective 12-month policies have effective dates uniform from January 1, 2027, to January 1, 2028, with losses uniform over each term. Find the prospective average loss date and its distance from the historical average loss date. (Answer: January 1, 2028; 42 months or 3.5 years.)
Historical losses are uniform within each year: $200,000 in 2023 and $400,000 in 2024. Prospective six-month policies have effective dates April 1, August 1, and December 1, 2026, with relative weights 1, 1, and 2. Losses are uniform through each term. Find the weighted historical average loss date, prospective average loss date, and gap. (Answer: March 1, 2024; December 1, 2026; 33 months or 2.75 years.)
Historical losses are uniform from January 1, 2025, to January 1, 2026. Twelve-month policies have effective dates uniform from July 1, 2026, to July 1, 2027, with losses uniform over each term. What is the effective-date midpoint, the prospective average loss date, and the gap from the historical average loss date? (Answer: January 1, 2027; July 1, 2027; 24 months or 2 years.)

What You'll Learn