Whitepaper

Thomas Smyth · Jonathan Dach · Maya Rosen — Concourse Research · July 21, 2026 · 7 pages

Method, validation record, the standing candidate set, and an appendix on replicating the analysis for New York City agency payrolls — what data is needed and how the pipeline adapts.

Download the PDF (7 pages)

Abstract

Screening New York State certified payrolls against published prevailing-wage schedules

New York State publishes two datasets that, taken together, permit a systematic compliance screen that today is performed only by complaint: the certified-payroll wage lines filed with the Department of Labor’s Bureau of Public Work, published on data.ny.gov, and the Department’s own prevailing-wage rate schedules.

We joined 588,505 public-work wage lines (January–July 2026) against a rate library built from the effective-dated schedules, using conservative rules at every step. Of 554,606 screenable journeyworker lines, 24,947 (4.5%) report a cash wage below the lowest scheduled journeyworker rate mappable to their trade, county and week. A second screening pass built entirely from the same public sources explains away much of that signal — 26.0% of flagged lines exactly match published apprentice rates, and 5.5% exactly match a different classification’s journeyworker rate — leaving a standing set of 6,835 high-confidence findings across 468 contractors, approximately $1.75M in indicative cash-wage shortfall.

The screen independently flagged the one firm in the data with an adjudicated willful-violation record (EGL Drainage Solutions & Repairs, debarred July 2025), with every cited rate hand-verified against the schedule PDF. All findings are allegations that prioritize human review; none is a determination. An appendix describes how the same analysis is performed for New York City, where a completed pilot demonstrates the method on the Comptroller’s §220 certified-payroll database.

Contents

What the paper covers

  1. Background
  2. Data — certified-payroll wage lines; rate schedules
  3. Method — county attribution; Tier 1, the floor screen; Tier 2, removing what public data can explain
  4. Validation — an adjudicated corroboration, honestly bounded; false-positive controls and hand verification
  5. The standing candidate set
  6. Limitations
  7. Recommendations
  8. Replicating the analysis for New York City
Read inline

Full text

The paper reads best as the PDF. If you would rather stay on the page, load it inline below — some mobile browsers will still hand it to their own viewer.