Over-identification occurs when there are more instruments than endogenous variables, providing more information than necessary to identify the causal effect. Over-identification tests can be conducted to assess the suitability of the instruments used.