The LCA Disclosure Field Reference: What Each Record Proves

You are looking at a competitor's visa filings and trying to read their roles, pay, and worksites off the raw records. This reference is for strategy, research, and talent-intelligence analysts who need to turn an employer's public Labor Condition Application (LCA) disclosure into a defensible read instead of a scraped median with no warnings. It defines every field in an Office of Foreign Labor Certification (OFLC) LCA disclosure record, states what each one proves, and flags where it systematically lies.

The LCA is a Department of Labor wage attestation an employer must certify before filing an H-1B, H-1B1, or E-3 petition. OFLC publishes it as case-level disclosure data every quarter. The data is genuinely useful - but only if you know what each field covers and, more importantly, what it quietly leaves out.

What an LCA disclosure file actually contains

An LCA disclosure release is three case-level Excel files per quarter, each shipping with its own record-layout PDF. The three components are LCA_Disclosure_Data (the main file), LCA_Appendix_A, and LCA_Worksites. Every record is one certified, denied, or withdrawn application - not one worker, and not one hire.

Start every analysis from the Program Record Layout. It lists each field title and a one-line description, and it is the only authoritative way to resolve a column header to its meaning. The fields below are the ones you will lean on.

Three date fields - RECEIVED_DATE, DECISION_DATE, and ORIGINAL_CERT_DATE - let you place a filing in time and distinguish a fresh attestation from an amendment. VISA_CLASS separates H-1B from H-1B1 and E-3.

One omission dominates everything downstream: the employer's FEIN (Federal Employer Identification Number), the attorney's FEIN, and the attorney's state bar number are all excluded from the LCA file as personally identifiable information. There is no clean employer key. That single gap drives the methodology choices in the rest of this reference.

How to read the four prevailing wage levels

The prevailing wage level tells you which percentile of the Occupational Employment and Wage Statistics (OES) distribution the employer pegged the role to - and nothing about the worker's seniority. The currently effective 2005 methodology fixes four tiers by percentile.

The level is a legal floor the employer chooses by describing the role, not a measured credential of the person filling it. Filing at a lower level is cheaper and perfectly legal, which is why the distribution skews low. In FY2024, 63% of certified LCAs sat at Level I or II. Read that as a floor-setting behaviour, not as proof that the workforce is junior.

Two things make the level a moving target. The percentile tiers are set by rule, and the rule has been contested. A 2021 final rule would have raised them to 35, 53, 72, and 90; it was issued and then withdrawn. A 2025 proposed rule (NPRM) proposes 34, 52, 70, and 88 - roughly doubling the Level I floor.

The practical consequence: a Level I wage read against today's rule means something different than the same label will mean if the NPRM takes effect. Always note which methodology was in force for the fiscal year of the file you are reading, and re-check the rule state before comparing filings across years.

Why the wage field understates real pay

The LCA offered-wage field is base cash wage only, and that is a legal design choice, not an accident. Under 20 CFR 655.731, benefits provided as compensation are attested on a separate basis, so they never land in the wage number.

What is excluded from the wage field:

• Cash bonuses
• Stock options and equity
• Paid vacation and holidays
• Health, life, and disability insurance
• Retirement and savings plans

The regulation requires these to be offered to the sponsored worker on the same basis as to US workers, but it keeps them out of the wage attestation. So the number you read is a floor on cash base pay, below real total compensation by whatever the firm adds in bonus and equity.

The gap is not trivial even before equity. The OEWS mean for equivalent occupations and geographies was $130,219, about $19,000 above the prevailing wage levels those roles were filed at. Layer bonuses and stock on top and the understatement widens most at exactly the firms where equity is largest. If you read the LCA wage as total comp and conclude "this competitor underpays," you have mistaken a legal floor for a paycheck.

The LCA wage is a floor on base cash, not a paycheck, and it lies loudest where equity runs deepest.

The eight-step read, from raw file to defensible number

Work the file in a fixed order so your conclusions carry the right caveats by construction. The whole procedure runs a few hours per employer once you have the files open.

One judgment call lives inside step two. Sources disagree on whether to dedupe by case or sum TOTAL_WORKER_POSITIONS. Both are defensible: dedupe by case when you want to count filing events, sum positions when you want an upper bound on sponsored roles. Pick one, state which, and keep it consistent across competitors you compare.

LCA versus the Hub, Registration, and PERM

The LCA is one of four public H-1B-related datasets, and each answers a different question. Reaching for the wrong one is the most common structural error. The LCA is the DOL wage attestation; the USCIS Employer Data Hub is adjudication outcomes; Registration data is lottery volume; and PERM is the green-card labor certification.

The USCIS Hub reflects USCIS's first decision, with appeals and revocations excluded, aggregated by completion fiscal year, two-digit NAICS, tax ID, state, city, and ZIP. It covers FY2009 through FY2026 Q3, splits initial from continuing employment, and identifies employers by the last four digits of their tax ID. It carries no wages, so you use it to validate volume, never pay.

The gap between LCA filings and Hub decisions is the whole point of the cross-check. Microsoft filed 11,402 LCA approvals in FY2022, but USCIS total approvals and denials came to only 7,283. If you had counted LCAs as hires you would have overstated by more than half.

PERM solves the identity problem the LCA creates. The green-card labor certification on Form ETA-9089 - submitted through the FLAG system since June 1, 2023 - includes EMP_FEIN and EMP_NUM_PAYROLL, the total employees on payroll. When you need clean employer identity, anchor on PERM's FEIN and map the messy LCA names to it.

Registration data, the cap-season lottery volume, is worth watching for market context even though it names no wages. The eligible registration pool dropped from 758,994 in FY2024 to 470,342 in FY2025, and fell again to 343,981 for FY2026. In FY2025, USCIS selected 114,017 unique beneficiaries, about a 25.6% selection rate. A rival's sponsored hiring plan runs through that funnel, so the pool size tells you how hard sponsorship is becoming.

Refolk closes the gap the LCA cannot: it returns the people, not the filings, so you can set LCA worksite concentration against a real headcount in plain English. Where the disclosure file sees only sponsored US roles, asking Refolk for the engineers at a named employer and metro gives you the denominator the attestation data structurally lacks.

Where the numbers lie, and how to catch them

The failure modes below are the most valuable part of this reference, because each one produces a confident-sounding wrong answer. For every one, there is a specific check.

Treating the LCA wage as total comp

False positive: "the competitor underpays." The wage field is base cash only; bonuses, equity, and benefits are attested separately under 655.731. Check: label the figure base-only and never compare it to a total-comp benchmark.

Counting filings as headcount

One LCA can cover multiple positions and many records are refiles or withdrawals, so case counts overstate hiring. Microsoft's 11,402 LCAs versus 7,283 USCIS decisions is the canonical gap. Check: reconcile to the Hub before reporting any count as hires.

Reading Level I as a junior org

The level reflects the employer's role description, not the worker's credentials, and the same title appears at any level. Check: inspect the SOC code and duties, not the level alone, and remember 63% of FY2024 cases filed at Level I or II.

Name matching without FEIN

The LCA has no FEIN, so subsidiaries and aliases either fragment one employer into several or merge distinct ones. Check: enumerate the legal-entity variants and cross-check identity against PERM's FEIN.

Mistaking certified for approved

A certified LCA is not a filed or approved petition. Check: the Hub's initial and continuing approvals tell you what was actually adjudicated.

Ignoring roving and multiple worksites

The HQ or filing address is not where the work happens, and remote or roving entries scramble geo analysis. Check: always join the Worksites file and flag "multiple" entries.

Extrapolating from a stale quarter

The newest file lags about one quarter, so the most recent hiring is simply absent. Check: note the fiscal year and quarter of the file and never trend-fit the latest partial quarter.

Generalizing from IT

The LCA over-represents computer occupations and omits non-degree and citizen roles entirely. Check: compare the sponsored share to full headcount before drawing org-wide conclusions.

That last warning is the biggest blind spot. Computer and Mathematical occupations (SOC 15-XXXX) consistently dominate LCA filings, with Healthcare Practitioners (SOC 29-XXXX) and Architecture and Engineering (SOC 17-XXXX) also high. If a rival's engineering is offshore, the LCA shows you none of it.

In Refolk's index of professional profiles, the India Software Engineer population of 581,842 is roughly 1.65x the US figure of 352,540. LCA captures only the US worksites, so the larger base - and any team a competitor runs there - is invisible to the filings entirely.

Before you call the read done

Run this checklist before you circulate a conclusion. Each item defends against one of the failure modes above.

When you hand the read to a stakeholder, carry the caveats with the number. A single sentence does it.

LCA figures for [Employer], FY20XX Q#: counts are certified attestations, not hires or approvals (reconcile to USCIS Hub). Wages are base cash only; bonuses, equity, and benefits are attested separately under 20 CFR 655.731. Coverage is US worksites and sponsored specialty occupations only, dominated by SOC 15-XXXX; it excludes citizen, green-card, non-degree, and offshore staff. File lags about one quarter.

Keeping the read current

Two things change under you: the data cadence and the rule state. The data refreshes quarterly and cumulatively on a fiscal year running October 1 through September 30, and the newest file lags roughly one quarter, so re-pull the OFLC Performance page each quarter and record which fiscal-year quarter you are reading. The USCIS Hub updates on the same quarterly rhythm.

The rule state changes less often but matters more, because it redefines what a wage level means. The 2005 percentiles (17, 34, 50, 67) are in force today, a 2021 rule that would have raised them was withdrawn, and a 2025 NPRM proposes 34, 52, 70, and 88. Before you compare filings across years, confirm which methodology applied to each year's file, because a Level I under the proposed rule is nearly double the floor it is today.

Finally, remember that the LCA answers only part of the question you are asking. It tells you the sponsored, US, specialty-occupation slice of a competitor's workforce, priced at a legal base floor. For the rest - total compensation, offshore teams, the citizen and green-card majority, and clean employer identity - you reconcile against the USCIS Hub, PERM's FEIN, and a profile index of the actual people. Asking directly for the engineers at a named employer and metro gives you the headcount denominator the attestation data was never built to show.