Skip to content

Benchmark beta and correlation are different

Slope depends on covariance relative to benchmark variance; correlation also divides by asset dispersion.

On this page

Compare the denominators

For compatible return series, fitted slope is covariance divided by benchmark variance. Pearson correlation is covariance divided by both positive standard deviations. The denominator change means the two quantities need not have the same magnitude.

Read the original pair

The original five-period fit has beta 2.1 and correlation approximately 0.986666. A slope above one is permitted; correlation must remain between −1 and 1.

Check an asset-unit transformation

Multiplying every asset return by a positive factor multiplies slope by that factor but leaves correlation unchanged. This is a scale property, not evidence that the transformed numbers are appropriate financial data.

Handle a constant series

If benchmark returns are constant, a slope cannot be identified by this estimator. If only asset returns are constant, slope is zero while correlation is undefined because asset dispersion is zero.

Further references