A high match rate can still mislead

A high match rate does not guarantee an unbiased dataset. If errors cluster across variables or groups, the direction and size of bias cannot be inferred from the rate alone.

False matches join records from different people. They can add noise and dilute associations between variables, often pushing effect estimates toward zero, though not always.

Missed matches leave records from the same person unlinked. That shrinks the sample and statistical power, and informative misses can undercount exposures or outcomes.

Linkage success may not be random. If some groups are harder to link, the linked data can omit important subgroups and distort comparisons.