Stata 17 Merge: Should Duplicate Keys Trigger a Stop or Continue During Many‑to‑Many Merges?

Goal

We need to validate the merge output in a test environment that lacks production credentials, ensuring that downstream analysis remains deterministic.

Uncertainty

The merge command accepts duplicate key values in both master and using datasets, producing a many‑to‑many result. Documentation only issues a generic warning and does not state whether the merge should stop or continue. This ambiguity can lead to inflated observation counts and mask data quality issues.

Specific Questions

• In Stata 17, does the merge command halt execution when duplicate keys are detected, or does it proceed with all pairwise combinations?
• Is there any automatic aggregation or collapsing of duplicate rows performed by the merge, or must the user pre‑clean the data?
• Does the behavior differ between Windows and macOS platforms, or across Stata 16 and 17 releases?