Proxy Test
A New York DFS method for detecting whether a prohibited protected-class characteristic influences an insurance decision through a correlated proxy variable.
The proxy test is a statistical method used by New York DFS under Circular Letter No. 7 to detect unlawful discrimination in insurance underwriting and pricing. It asks whether a model output or input variable is a proxy for a protected class, even if that protected class is not explicitly used in the model.
The test has three parts:
- Identify protected classes: such as race, color, creed, national origin, sex, marital status, or sexual orientation.
- Identify proxy variables: variables that are correlated with a protected class strongly enough to predict it.
- Measure impact: determine whether the use of the proxy produces a differential outcome that would be unlawful if the protected class were used directly.
A variable can fail the proxy test even if it is facially neutral. New York DFS expects insurers to document how they test for proxy effects and what remedial action they take when a proxy is found. Our guide to NY DFS Circular Letter No. 7 explains the test in underwriting practice, and our AI in underwriting guide covers the broader compliance framework.