Projecting Historical Loss Costs to the Prospective Period

Why Historical Loss Costs Must Be Projected

An indication uses historical loss experience to estimate the cost of covering exposures in a future period. But losses from an earlier period reflect the cost conditions of that earlier time. If expected claim costs change over time, the historical loss cost must be projected to the time represented by the prospective period before it is used as an estimate of prospective costs.

The central task is to identify two dates: the date represented by the historical loss cost and the date represented by the prospective cost period. The time between those dates, not simply the number of calendar years named in the periods, determines the trend duration. Once those dates are clear, the projection is a compound-growth calculation.

The loss measure must be appropriate for the projection. For example, an ultimate loss cost per exposure can be projected if the loss amount and exposure denominator are on a consistent basis. Earlier material on on-leveling addresses adjustments for changes in rate level; trending is a separate adjustment for changes in expected claim costs over time. A trend factor applied to loss costs is not an adjustment to historical premium.

Dates and the Trend Duration

A period-wide loss cost represents experience across time rather than a single loss date. A common practical convention is to represent an annual experience period by its midpoint and a prospective cost period by its midpoint. The elapsed time between those representative dates is the trend duration. The exact dates used should follow the analysis assumptions; the key is to use the same date convention consistently.

For example, the midpoint of a calendar-year experience period is approximately July 1. If the prospective cost period is also a calendar year, its midpoint is approximately July 1. The trend duration is the time from the historical midpoint to the prospective midpoint. It is not automatically the time from the end of the historical period to the start of the prospective period.

Use a fractional exponent when the interval is not a whole number of years. For example, under a 4% effective annual trend assumption, a half-year factor is \(1.04^{0.5}\), not \(1+0.04/2\). The latter is a simple proportional approximation, not compound annual trending.

Why the Compound Factor Works

The formula follows from the meaning of an effective annual trend rate. Under a constant rate \(t\), one year of change multiplies a loss cost by \(1+t\). After two years, the second year’s change applies to the already changed amount, so the multiplier is \((1+t)(1+t)=(1+t)^2\). Repeating this for \(k\) whole years gives \((1+t)^k\).

For a fractional year, the standard compound-trend model extends the annual factor by exponentiation. In particular, a fraction \(r\) of a year has factor \((1+t)^r\), so \(n=k+r\) years has factor

This is a model assumption: it treats the effective annual trend as accruing smoothly according to a constant compound rate between annual anniversaries. If annual trend rates are instead specified separately, use the product of the applicable factors for each interval, with a suitable fractional-period factor where needed.

Projecting One Historical Period

Using Fractional Years Correctly

Projecting Several Historical Periods

When the historical data contain more than one experience period, each period has its own representative date and therefore its own trend duration to the prospective period. Project each period’s loss cost from its own date. Do not assign every historical period the same number of trend years merely because all are being used for one prospective indication.

Common Errors and Practical Checks

The projection is simple once its dates and assumptions are clear, but small setup errors can create material differences. Before calculating, check the following:

• Trend to the prospective cost date, not merely to the effective date. The target here is the period of expected costs being represented. A rate effective date and a prospective cost period are not interchangeable.
• Use the historical loss-cost date, not automatically the period endpoint. Under a midpoint convention, the loss cost represents the midpoint. Trending from a period end or to a period start changes the elapsed time.
• Keep time units consistent. If the annual trend is effective, measure \(n\) in years and use \((1+t)^n\). A rate stated for a different time unit must first be expressed consistently with the duration.
• Do not add annual percentages when compounding is assumed. Two years at 5% gives a factor of \(1.05^2=1.1025\), not 1.10.
• Do not trend a different quantity by accident. The formula projects the loss cost on the stated basis. It does not, by itself, alter exposure counts, correct an inconsistent coverage basis, or substitute for rate-level adjustments.

A useful reasonableness check is that a positive trend over a positive elapsed period should increase the projected loss cost; a negative trend should decrease it. A zero-year interval has factor 1, so the projected cost equals the historical cost. Also check that older periods generally have longer durations to the same target date than newer periods.