The Anchoring Effect: Why Your First Data Point Ruins Everything

In a famous experiment, Tversky and Kahneman asked participants to estimate the percentage of African countries in the United Nations. Before answering, participants spun a rigged wheel that landed on either 10 or 65. Those who saw 10 guessed an average of 25%. Those who saw 65 guessed 45%. The initial random number, completely irrelevant, anchored their estimates.

In insurance underwriting, anchoring isn't random. It's systematic. The first risk score an underwriter sees, whether from a legacy model, a competitor's pricing, or a previous year's assessment, becomes the reference point against which all subsequent information is judged.

Imagine an underwriter presented with a wildfire risk assessment alongside a legacy score. It's likely the legacy score would anchor their perception, even if the newer model demonstrably outperforms. An address scored as 'Moderate' by the old model and 'Very High' by the new one would trigger skepticism about the new score, not the legacy model's.

This anchoring could create a compounding problem. Year after year, portfolios may accumulate small distortions as underwriters unconsciously anchor to prior assessments. A property that was incorrectly scored as 'Low Risk' five years ago is likely still being treated as lower risk than it should be, because each annual review is anchored to the previous year's assessment.

What if portfolios were re-scored from scratch, without showing underwriters the previous scores? The risk distribution would likely shift dramatically. It's plausible that 15% of policies classified as acceptable risk would be reclassified as high risk, and 12% of declined risks would be reclassified as acceptable.

This suggests anchoring isn't just a theoretical concern. It could be actively eroding portfolio quality by keeping bad risks in and good risks out. The cumulative impact on loss ratios may be 3-5 points, a massive drag that compounds over time.

De-anchoring requires deliberate process design. One promising approach is 'blind scoring': presenting the new risk assessment without the previous score visible. Once underwriters form their independent judgment, they can compare it to the prior assessment. This simple change in information sequence could reduce anchoring effects by 60-70%.

The anchor you don't see is the one that controls you. In an industry built on precise risk quantification, allowing invisible cognitive anchors to distort assessments is an unforced error, and one of the easiest to fix.