Why Trend Dates Must Match the Data
A trend factor is meaningful only when its start and end dates correspond to the loss levels being compared. If experience losses represent an average level over a year, the relevant starting date is usually the experience period’s midpoint. If the target is an average level over a prospective year, the relevant ending date is usually that year’s midpoint. Using the experience period’s end or the prospective period’s start instead can shorten the projection interval and understate the trend adjustment.
The reverse error is also possible. Projecting an average experience level all the way to the prospective period’s end uses too long an interval and overstates the trend adjustment. The dates are not bookkeeping details: they determine the exponent in the trend factor.
This tutorial assumes a constant annual trend rate over the selected interval and compounding by year, including for fractional-year intervals. The midpoint approximation treats the data for each period as representing the level at that period’s midpoint. It is a practical approximation, not a claim that every claim occurred on the midpoint date. If the underlying data are already defined at particular valuation dates, or if a different timing convention is specified, use those dates rather than imposing midpoint dates.
Set the Dates Before Calculating the Factor
A reliable correction begins by describing what each amount represents. Identify the experience period and the prospective period, determine whether each amount is an average over a period or a value at a date, and then choose the corresponding dates. Only after those dates are established should the trend interval be calculated.
State whether the experience amount represents an average over a period or a level at a specific valuation date. For an average, locate the experience midpoint under the stated convention.
Determine whether the prospective amount is an average over a period or a level at a specific date. For a prospective average, locate that period’s midpoint.
Subtract the starting date from the target date using a consistent date convention. The result may be fractional.
For a projection forward in time, multiply by \((1+r)^n\), where \(n\) is the elapsed time in years. Check that the factor exceeds 1 when \(r>0\) and \(n>0\).
Why the Midpoint-to-Midpoint Factor Holds
Suppose the average experience loss level is represented at time \(t_e\), and the desired prospective average is represented at time \(t_p\). Under the constant annual trend model, each year multiplies the level by \(1+r\). After \(n\) years, repeated compounding gives \(L(t_e)(1+r)^n\). For a fractional year, the same model extends the exponent continuously, so the projected level is \(L(t_e)(1+r)^{t_p-t_e}\). Setting \(n=t_p-t_e\) gives the factor for the dates that actually represent the two amounts.
When both levels are period averages approximated by their midpoints, substitute the experience midpoint for \(t_e\) and the prospective midpoint for \(t_p\). The result is not generally the interval from the experience period’s end to the prospective period’s start. Those boundary dates are appropriate only if the quantities being compared are defined at those boundaries.
Worked Example: Average Experience to a Calendar-Year Average
A hypothetical insurer’s average loss cost for calendar year 2024 is $420. It is to be projected to the average loss-cost level for calendar year 2026. Assume a constant annual trend rate of 8%. An analyst uses the interval from the end of 2024 to the start of 2026, or one year. Find the corrected factor and projected loss cost, and quantify the analyst’s error.
Method and date alignment. Both amounts are annual averages, so use period midpoints. The midpoint of calendar year 2024 is July 1, 2024, and the midpoint of calendar year 2026 is July 1, 2026. The elapsed time is \(2.0\) years, not one year.
Correct factor. With \(r=0.08\) and \(n=2\),
Correct projection.
Rounded to cents at the final step, the projected loss cost is $489.89.
Analyst’s calculation and error. The analyst uses one year, so the factor is \(1.08\), and the projected amount is \(420\times1.08=\$453.60\). The understatement is \(489.888-453.60=\$36.288\), or $36.29 rounded to cents. The mistake is using the gap between period boundaries even though both figures represent period averages.
Check. The correct factor, 1.1664, exceeds the analyst’s 1.08 because the midpoint-to-midpoint interval is longer. With a positive trend, the corrected projected loss cost should therefore be higher.
Experience and Prospective Periods of Different Lengths
When the experience period is longer than one year, its midpoint can be well before its end date. That difference matters even if the prospective period is a familiar calendar year. Always locate both midpoints from the full period boundaries; do not infer the interval just from the experience end and the prospective start.
Worked Example: A Two-Year Experience Period
A hypothetical book has an average loss cost of $500 over the two-year experience period January 1, 2023–December 31, 2024. It is to be projected to the average for calendar year 2026. Assume a constant annual trend rate of 6%. An analyst uses a one-year interval from the experience period’s end to the prospective period’s start. Determine the aligned interval, factor, and projected loss cost.
Method and date alignment. Treat the two-year experience average as represented at the midpoint of its full period. Using the period boundaries January 1, 2023 and January 1, 2025, that midpoint is January 1, 2024. The target calendar-year average is represented at July 1, 2026. From January 1, 2024 to July 1, 2026 is \(2.5\) years.
Correct factor. Fractional-year trend is compounded using the fractional exponent:
Correct projection.
Rounded to cents at the final step, the projected loss cost is $578.41.
Analyst’s calculation and error. A one-year factor is \(1.06\), giving \(500\times1.06=\$530.00\). The understatement is approximately \(578.4085-530=\$48.4085\), or $48.41 rounded to cents. The experience midpoint is January 1, 2024, not the experience period’s end; using the latter discards about a year of the required projection.
Check. The aligned interval is \(2.5\) years, so its factor must be greater than \(1.06^2=1.1236\). The calculated factor, approximately 1.156817, satisfies that check.
Do Not Project an Average to the Wrong Prospective Date
A common mismatch occurs when the experience midpoint is used correctly but the target is set to the prospective period’s end rather than its midpoint. This produces a factor that is too large for an average-to-average projection. The target date must describe the intended target quantity, not simply the latest date in the prospective period.
Worked Example: Correcting an Overlong Projection
A hypothetical insurer has an average loss cost of $800 for calendar year 2025 and wants the average loss-cost level for calendar year 2027. Assume a constant annual trend rate of 4%. An analyst starts at the 2025 midpoint but projects to December 31, 2027, obtaining a \(2.5\)-year interval. Find the correct projected average and compare it with the analyst’s result.
Method and date alignment. The 2025 average is represented at July 1, 2025. The target is an average over calendar year 2027, so it is represented at July 1, 2027. The correct interval is \(2.0\) years. December 31, 2027 is the end of the target period, not its midpoint.
Correct factor and projection.
The projected average loss cost is $865.28.
Analyst’s result. The \(2.5\)-year factor is
The resulting projection is \(800\times1.103020\approx\$882.42\), rounded to cents. This is about \(\$17.14\) higher than the correct result. The excess arises from adding half a year by projecting to the target period’s endpoint instead of its midpoint.
Check. The incorrect interval is longer and the trend is positive, so the analyst’s result should exceed the correct projection. The calculated values have that relationship.
Common Mistakes and Diagnostic Checks
The most useful diagnostic is to ask, “What date does this number represent?” A period label alone does not answer that question. A loss average for a year, a loss amount valued at year-end, and a prospective average for a year may all be described using the same calendar-year label, but they do not necessarily use the same trend dates.
- Using experience end to prospective start for averages. This often omits part of the trend period. For average-to-average projections, align midpoints.
- Using the prospective endpoint for a prospective average. This adds time that is not part of the target average level. Use the target midpoint.
- Assuming every amount is a period average. If an amount is explicitly valued at a particular date, that date—not a period midpoint—anchors the projection.
- Using a whole-year exponent automatically. A midpoint-to-midpoint interval can be fractional. Calculate the elapsed time from the selected dates before applying the annual rate.
- Reversing the direction. A positive trend projected forward uses a factor greater than 1. A factor below 1 may indicate a reversed interval or an unintended deflation.
A wrong calculation might take an experience average for January 1, 2023–December 31, 2024 and a target average for calendar year 2026, then use only the year from the experience end to the target start: \(1.06^1=1.06\). The correction is to use the experience midpoint, January 1, 2024, and the target midpoint, July 1, 2026: \(n=2.5\) years and \((1.06)^{2.5}\approx1.156817\). The correction follows from matching the dates to the quantities, not from adjusting the annual trend rate.
For an indication, a mismatched factor can distort the projected losses even when the stated annual trend assumption is reasonable. The effect then carries into any calculation that uses those projected losses. A disciplined date check is therefore an important part of applying trend methods aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
Check Your Understanding
For each problem, identify the appropriate dates before calculating the factor. Assume a constant annual effective trend rate and the midpoint approximation whenever the amounts are period averages.
- A hypothetical portfolio’s average loss cost for calendar year 2024 is to be projected to the average for calendar year 2025. At 5% annual trend, what interval and factor should be used? (Answer: midpoint-to-midpoint interval \(=1.0\) year; factor \(1.05\).)
- A two-year experience average covers January 1, 2022–December 31, 2023. The target is the average for calendar year 2025, and annual trend is 3%. Find the aligned interval and factor. (Answer: January 1, 2023 to July 1, 2025, or \(2.5\) years; factor \((1.03)^{2.5}\approx1.0767\).)
- A loss amount is explicitly valued at December 31, 2024, rather than representing an annual average. It must be projected to December 31, 2026 at 6% annual trend. What factor applies? (Answer: endpoint-to-endpoint interval \(=2\) years; factor \(1.06^2=1.1236\).)
- A hypothetical experience average is for calendar year 2025, and the target is the average for January 1–June 30, 2026. At 8% annual trend, find the approximate midpoint interval and factor. (Answer: July 1, 2025 to April 1, 2026, or \(0.75\) year; factor \((1.08)^{0.75}\approx1.0594\).)
- A hypothetical loss cost of $700 is an average over January 1, 2023–December 31, 2024. Project it to the average for calendar year 2026 at 6% annual trend. Give the interval, factor, and projected amount to cents. (Answer: midpoint-to-midpoint interval \(=2.5\) years; factor \((1.06)^{2.5}\approx1.156817\); projected amount \(\approx\$809.77\).)