Load Factors Should Follow the Tail

01The evidence

• Reliability calibration for wind load shows that the required factor depends on the tail index of the fitted extreme-value distribution, with different indices implying materially different safety levels.Advanced Modeling and Simulation, 2026
• Resistance statistics for common foundations show lognormal tails with coefficients of variation large enough that the load tail dominates the reliability index.Transportation Research Board

02The tension

Codes apply partial factors that were calibrated decades ago against a small set of load models, mostly assuming a particular tail behaviour. Where the true tail is heavier, the same factor produces a lower reliability, and the shortfall is invisible because it only appears in the extreme events the code was meant to cover.

03The theory

Reliability is an integral over the tail of the load distribution. Two loads with the same mean and variance but different tail indices produce different failure probabilities at the same factor, because the integral is dominated by the far tail.

If codes are calibrated to a nominal tail index, then structures designed to those codes carry a reliability that varies with the actual tail index of their environment. The variation is systematic, not random, so it can be corrected by making the factor a function of the measured tail.

The consequence is that the same nominal safety level requires different factors for wind, snow, wave and flood, and that the factors must be updated as the climate changes the tails. A factor frozen in a code is a factor that drifts away from its target reliability.

The test is a reanalysis of failure records against the tail index at each site. Failures should cluster where the tail is heaviest, and the excess should be predictable from the index rather than from the load magnitude.

04What this is not

05Predictions

• P1Structural failures should cluster at sites with heavy-tailed load distributions, after controlling for load magnitude.
• P2Recalibrating by tail index should change required factors by a predictable amount, largest for wind and wave.
• P3Climate-adjusted tails should require larger factors even where mean loads are unchanged.

06What would refute it

If the required factor is insensitive to tail index across realistic load models, the single-factor approach is adequate and the theory is unnecessary.

07What it would change

It would make design codes site-specific in the tail rather than in the mean, and give climate adaptation a quantitative target.