disclosure suppression

Suppression is the price of a near-universe count. QCEW is built from state unemployment insurance tax reports rather than a survey, so its employment and wage tabulations cover virtually every covered job - the cataloged Employment Services (NAICS 5613) slice alone returns 3,623 rows for a single quarter across counties, metros, states and the nation. Where one employer dominates a county-industry cell, publishing the cell would reveal that employer's payroll, so the cell is flagged instead of filled: disclosure_code stays blank when published and becomes N when suppressed.

Census operates under a stricter statute. Title 13 confidentiality rules determine which figures may be published at all, which is why the Bureau's roughly forty documented APIs report Economic Census establishment counts, sales and payroll down to county level while never releasing identifiable business microdata.

Buy the wrong source and the holes show up in production. A county-level market sizing for automotive-parts-equipment data or computer-electronics-retail data that ignores suppression will silently sum missing cells as zeros, understating local markets wherever a handful of employers dominate.

The Census API's terms compound this: they prohibit using the data alone or combined to identify any person, household, business or other entity, so attempts to back-fill suppressed cells are not merely unreliable but out of bounds. Suppression also breaks vendor comparability - two resellers imputing the same QCEW hole differently produce two different 'county totals'. Free access helps (82.4% of the 1,744 datasets Datadory catalogs are free), but free does not mean complete; know which gaps are legal artifacts before modeling around them.

1. Treat flagged rows as absent, not zero. QCEW sets disclosure_code to blank when published and N when suppressed; summing N rows as zeros understates every thin county-industry cell.

1. Expect gaps precisely where cells are small. The BLS terms notes state that confidentiality suppression (disclosure_code N) applies to some small cells - national and state rolls stay complete while county-by-industry detail thins.

1. Recognize statutory absence as a design feature. Title 13 shapes what Census may publish at all, so a missing county figure is a legal artifact, not a data-quality defect to be patched.

1. Match your geography list to the endpoint. The Economic Census API reaches county level while the Annual Survey of Manufactures is national-only; mismatched areas masquerade as suppression.

1. Keep re-identification out of scope. Census terms prohibit using the data alone or combined to identify any person, household, business or other entity - document that constraint in any derived product.

Entries that sit next to disclosure suppression in this glossary:

• economic-census
• qcew-employment-wages