Tutorials › CAS Exam 5 - Basic Ratemaking and Estimated Claim Liabilities › Choosing the Trend Interval From Experience to Prospective Period

Home  /  Tutorials  /  CAS Exam 5 / Trend and Rate-Level Adjustments

CAS Exam 5 · Trend and Rate-Level Adjustments · Tutorial 185 of 989

Choosing the Trend Interval From Experience to Prospective Period

Learn to identify the two relevant midpoints, measure the elapsed trend duration, and apply an explicitly defined annual loss trend without confusing it with a rate change.

Intermediate 10 min read

What You'll Learn

  • Define the experience-period and prospective-period midpoints using period boundaries
  • Calculate elapsed trend duration from the two midpoints
  • Apply an annual compound loss trend over whole or fractional years
  • Distinguish the trend interval from the dates at which premiums are earned or rates change
  • Detect common endpoint and time-unit errors in trend calculations
  • Recognize when the midpoint approximation needs closer scrutiny

Why the Trend Interval Depends on Two Midpoints

A loss trend moves historical loss costs from the time they represent to the time they are intended to represent. The duration is therefore not automatically the number of years between the end of the experience period and the start of the prospective period. It depends on the representative date of each period. The standard approximation is to represent each period by its midpoint and measure the elapsed time between those two dates.

This distinction matters whenever experience and prospective periods cover a span of time. For example, losses from a calendar year are not all incurred on the last day of that year. Likewise, a prospective policy period does not begin with all policies effective on its first day. Using the two midpoints accounts for the spread of dates in both periods.

Definition: The experience-period midpoint is the date halfway through the interval represented by the loss experience. The prospective-period midpoint is the date halfway through the period for which the indicated costs are intended to apply. The trend duration is the elapsed time from the experience midpoint to the prospective midpoint, measured in years.

For date calculations, treat a period such as calendar year 2024 as the interval from January 1, 2024, up to but not including January 1, 2025. Its midpoint is July 1, 2024. This boundary convention avoids ambiguity about whether the last calendar day is counted twice.

From Midpoints to a Trend Factor

First identify the period represented by the losses and the period represented by the prospective costs. Then find each midpoint and measure the elapsed time between them. If the prospective midpoint is later, a positive loss trend moves the historical costs forward in time.

$$ t = \text{prospective-period midpoint} - \text{experience-period midpoint} $$

The subtraction in this expression means elapsed time expressed in years, not subtraction of the numerical month or day fields. For an explicitly defined annual compound trend rate \(r\), assume the loss-cost level is multiplied by \(1+r\) per full year. Under the same compound model for fractional years, the factor is:

$$ \text{Trend factor} = (1+r)^t $$

For example, a factor of \(1.06^2\) means the loss-cost level is increased by 6% in each of two years. This convention must be stated: a quoted annual trend might instead refer to a different model, such as a simple additive change. Do not apply the compound formula unless compound annual trending is the stated assumption.

Why the Midpoint Is a Reasonable Representative Date

The midpoint represents the center of a period’s timing. A short derivation shows why it is natural when the volume of experience is spread evenly over time and the trend is approximately linear over the interval. Let time \(s\) be measured from the period midpoint, and approximate the loss-cost level by \(L(s)=L(0)+bs\), where \(b\) is the rate of change per unit of time. For a period symmetric about the midpoint, from \(-h\) to \(h\), its average level is:

$$ \frac{1}{2h}\int_{-h}^{h} \bigl(L(0)+bs\bigr)\,ds = \frac{1}{2h}\left(2hL(0)+b\left[\frac{s^2}{2}\right]_{-h}^{h}\right) = L(0) $$

The linear change averages to zero around the center, so the average level equals the midpoint level under these assumptions. This is a justification for the midpoint approximation, not a claim that it is exact for every trend pattern. A compounded trend is not linear in time, and experience may not be distributed uniformly. If timing or exposure volume is materially uneven, a more detailed analysis may be needed.

Worked Example: One Calendar Year to One Prospective Calendar Year

A hypothetical insurer uses losses from calendar year 2022 to estimate loss costs for policies effective throughout calendar year 2025. Assume the annual loss-cost trend is a compound 6%. Find the midpoint-to-midpoint duration and trend factor.

Method. Identify the boundaries of each full-year period, determine both midpoints, and apply the stated annual compound trend over the elapsed years.

Experience midpoint. Calendar year 2022 runs from January 1, 2022, to January 1, 2023. Halfway through is July 1, 2022.

Prospective midpoint. Calendar year 2025 runs from January 1, 2025, to January 1, 2026. Halfway through is July 1, 2025.

Duration. July 1, 2022, to July 1, 2025, is exactly 3 years, so \(t=3\).

Factor. The compound annual factor is \(1.06\). Over 3 years:

$$ (1.06)^3 = 1.06 \times 1.06 \times 1.06 = 1.191016 $$

The trend factor is 1.191 to three decimal places, or a 19.1% increase when expressed as a percentage change from the original level. The result is consistent with three annual increases: the factor is greater than 1, and each annual increase compounds on the preceding level.

Common mistake avoided. Counting from January 1, 2023, the end boundary of experience, to January 1, 2025, the start boundary of the prospective period gives 2 years. That is not the midpoint-to-midpoint interval. The correct interval is 3 years.

Periods That Do Not Share the Same Length or Start Date

Experience periods can span more than one year, and prospective periods may be defined by policy effective dates rather than calendar years. The midpoint calculation still follows the same logic: find the center of each interval actually represented. Do not substitute a period’s end date for its midpoint just because the loss data are reported as a total.

Worked Example: Two-Year Experience and a One-Year Prospective Period

A hypothetical liability program has loss experience over the 24 months from January 1, 2022, to January 1, 2024. The indicated costs are for policies effective during the 12 months from January 1, 2025, to January 1, 2026. Assume a compound annual trend of 4%. Find the duration and trend factor.

Experience midpoint. The midpoint of the 24-month interval is January 1, 2023.

Prospective midpoint. The midpoint of the prospective calendar year is July 1, 2025.

Duration. January 1, 2023, to January 1, 2025, is 2 years, and January 1 to July 1 adds one-half year. Thus \(t=2.5\) years.

Factor. Apply the specified compound trend for 2.5 years:

$$ (1.04)^{2.5} \approx 1.1030 $$

The trend factor is 1.1030 to four decimal places, equivalent to an increase of about 10.30%. The experience period is two years long, but its midpoint is January 1, 2023; using the end of experience instead would use the wrong starting date.

Common mistake avoided. A candidate might calculate only one year from January 1, 2024, to January 1, 2025. That calculation compares the end of experience with the start of the prospective period, not the two midpoints. The correct duration is 2.5 years.

Worked Example: A Prospective Policy Period Starting Midyear

A hypothetical insurer uses losses from accident year 2023, covering accidents from January 1, 2023, to January 1, 2024. The prospective policies will be effective from July 1, 2025, to July 1, 2026. Assume a compound annual loss trend of 8%. Determine the midpoint interval and factor.

Experience midpoint. The accident-year interval has midpoint July 1, 2023.

Prospective midpoint. The policy effective period is one year long, from July 1, 2025, to July 1, 2026. Its midpoint is January 1, 2026.

Duration. From July 1, 2023, to July 1, 2025, is 2 years. From July 1, 2025, to January 1, 2026, is another one-half year. Therefore \(t=2.5\) years.

Factor. Applying the stated annual compound trend:

$$ (1.08)^{2.5} \approx 1.2122 $$

The factor is 1.2122 to four decimal places. The trend is applied from the accident-year midpoint to the midpoint of the policy effective period, not to its first effective date.

Common mistake avoided. Stopping at July 1, 2025, gives a two-year interval. The prospective period extends another year, so its midpoint is January 1, 2026, and the full midpoint-to-midpoint duration is 2.5 years.

Scope, Assumptions, and Exam Traps

Midpoint selection answers a timing question: which two dates define the trend interval? It does not decide which losses should be trended, whether the trend rate is appropriate, or whether a rate-level change belongs in the calculation. Keep the dates tied to the loss-cost measure and prospective cost period specified by the analysis. In particular, a rate effective date is not automatically a loss-trend endpoint.

The approximation is most defensible when the experience is reasonably spread across its stated interval and the trend is reasonably stable. It can be less suitable when experience is heavily concentrated near one end of the period, the prospective period has a different weighting pattern, or trend changes materially during the interval. In those cases, a single midpoint may not represent the timing of the underlying data well.

Check the following before finalizing a calculation:

  • Use the midpoint of the period represented by the losses, not automatically the valuation date or period end.
  • Use the midpoint of the specified prospective period, not automatically its first effective date.
  • Measure elapsed time in years, including half-years or other fractions when relevant.
  • Confirm whether the annual trend is compound before using \((1+r)^t\).
  • Check direction: a positive trend from an earlier experience midpoint to a later prospective midpoint should produce a factor greater than 1.
Key takeaway: The standard midpoint method sets the trend duration equal to the elapsed time from the experience-period midpoint to the prospective-period midpoint. The trend factor then follows from the explicitly stated annual trend model.

Check Your Understanding

For each problem, identify the two midpoints before calculating the trend factor. Assume compound annual trending whenever a compound rate is stated.

  1. A hypothetical insurer uses calendar-year 2021 losses to estimate costs for calendar-year 2024. The annual compound loss trend is 3%. Find the midpoint duration and factor, rounded to four decimal places. (Answer: 3 years; \(1.03^3=1.0927\).)
  2. A hypothetical program’s experience covers January 1, 2022, to January 1, 2024. Its prospective period covers January 1, 2025, to January 1, 2026. The annual compound trend is 4%. Find the duration and factor, rounded to four decimal places. (Answer: 2.5 years; \(1.04^{2.5}\approx1.1030\).)
  3. A hypothetical insurer uses losses from July 1, 2022, to July 1, 2023, for policies effective from July 1, 2024, to July 1, 2025. The annual compound loss trend is 6%. Find the midpoint duration and factor, rounded to four decimal places. (Answer: 2 years; \(1.06^2=1.1236\).)
  4. A hypothetical risk program uses calendar-year 2023 losses for policies effective from October 1, 2025, to October 1, 2026. The annual compound trend is 4%. Find the midpoint duration and factor, rounded to four decimal places. (Answer: 2.75 years; \(1.04^{2.75}\approx1.1139\).)
  5. A hypothetical insurer uses accident-year 2025 losses and projects to costs for calendar year 2025. The annual compound trend is 7%. Find the midpoint duration and factor. (Answer: 0 years; \(1.07^0=1.0000\).)

Previous

Next

Back to Tutorials
Clinical Biostats