Synthetic W-2 Wage and Tax Statement Data
Synthetic training data — no real PII, fully coherent identities
Generate synthetic 2024 four-up W-2 sheets — the quarter-sheet layout that prints an employee's Copy B, Copy C and both Copy 2 forms on one page. Every copy on the sheet carries the same wage figures for the same employee, which makes this the layout for training copy-identification and near-duplicate handling.
152
Fields per document
1
Page
Tax
Category
What this document is
This is the 2024 four-up W-2: a single letter-size sheet divided into quadrants, each quadrant a complete W-2 Wage and Tax Statement for the same employee. It is the format employees actually receive in the mail — Copy B for the federal return, Copy C for the employee's records, and two Copy 2 forms for state and local filing — perforated so the recipient tears them apart. Payroll bureaus print it by the million and every document pipeline that ingests employee-supplied tax documents encounters it constantly.
Why generate synthetically
The four-up sheet is the W-2 format most likely to break an extraction pipeline and the least likely to appear in a curated training set, because curated sets are assembled from clean single-statement scans. When a borrower photographs the W-2 their employer mailed them, the pipeline receives four statements at once and has to decide whether it is looking at four documents, one document, or a duplicate-detection failure. Real four-up sheets carry a Social Security number, a name, an address and an income figure four times over on one page, which makes them exactly as unshareable as any other W-2 and considerably more informative to train on.
What makes synthetic data useful
Each sheet is one simulated employee's statement replicated across all four quadrants with identical values, exactly as a real perforated sheet works — the same Social Security number, employer identification number, wage figures and Box 12 codes appear four times with four distinct sets of coordinates. Box 3 Social Security wages cap at the 2024 wage base of $168,600 while Box 5 Medicare wages continue past it for high earners, and Box 17 state tax is zero for residents of states without an income tax. Because the copies are byte-identical in value and different in position, the corpus supervises the copy-identification task directly: a correct extractor returns one statement, not four.
Training challenges
Four complete statements on one page means four instances of every label, and a model that keys on label text alone will merge values across quadrants — the classic failure is a Box 1 wage from the top-left quadrant paired with a Box 2 withholding from the bottom-right. Because all four quadrants carry identical values, that error is invisible to any evaluation that compares extracted values against ground truth without checking which copy they came from. Each quadrant is a quarter of a letter page, so every box is a quarter of the size it is on a full-page statement and the Box 13 checkboxes are among the smallest targets in the entire form library. The quadrants are separated by cut lines rather than by whitespace, so segmentation has to work off thin printed rules.
Generate synthetic W-2 Wage and Tax Statement data
Start with 500 free credits. No credit card required.
Generate NowWho uses this data
Mortgage and consumer-lending platforms whose borrowers upload the sheet their employer mailed them, tax-preparation software ingesting client documents at scale, payroll-service document portals, and fraud teams training detectors on altered wage statements. It is also the standard stress test for any deduplication layer sitting in front of a document classifier.
Document complexity profile
152 fields on a single page: 72 currency amounts, 60 text, 12 checkbox targets, and 4 Social Security number and 4 employer identification number typed fields — one full statement of 38 fields replicated four times. 132 annotation relations. 16 conditional bindings and 20 function calls at a maximum expression depth of 2. The highest field count per page of any W-2 layout in the catalog, and by a wide margin the densest of the wage-statement designs.
Key stats from our synthetic corpus
Quantitative characteristics of the W-2 Wage and Tax Statement documents our generator produces.
| Metric | Value | Detail |
|---|---|---|
| Median Box 1 wages | $45,600 | Across 641 eligible synthetic identities the median Box 1 wage is $45,600, with a p25–p75 range of $27,200 to $74,400 and a full span of $4,000 to $316,900. The distribution is right-skewed to match real wage data rather than centred on a training-convenient mean. |
| 2024 Social Security wage base | $168,600 | Box 3 Social Security wages top out at exactly $168,600, the 2024 wage base, while Box 5 Medicare wages continue to $316,900 for the same high earners. Documents where Box 3 and Box 5 legitimately disagree are the case that trips rule-based validators, and the corpus contains them by construction. |
| Statements per sheet | 4 | Every sheet carries four complete statements with identical values and four distinct coordinate sets. Copy-level annotation means you can measure whether an extractor returned one wage record or four — an error class that value-only evaluation cannot see. |
| Blank field rate | 42% | 64 of the annotated fields render blank on a typical sheet — allocated tips, dependent care, nonqualified plans, Box 12c and 12d, the statutory-employee and sick-pay checkboxes, Box 14, and the local wage block, all repeated across four copies. Blank-versus-zero discrimination is the most commonly skipped evaluation in W-2 extraction. |
| Retirement plan participation | 57% | 57.1% of sheets check the Box 13 retirement plan box, and the same 57.1% carry a Box 12a elective-deferral code — 71% code D, 17% code E, 12% code G. At a quarter of page size the Box 13 checkboxes are among the smallest detection targets in the catalog. |
How this document co-occurs with others
Rates at which identities in our corpus that produce a W-2 Wage and Tax Statement also produce other documents.
| Correlation | Rate | Detail |
|---|---|---|
| Following-year sheet, same layout | 100% | The same employee receives the 2025 four-up sheet with the higher Social Security wage base. Two-year lookback is standard in mortgage underwriting, so the pair is the realistic submission rather than either year alone. |
| Half-page single-copy layout | 100% | The same wage data rendered as a single half-page statement. Training on both layouts is the cleanest test of whether a model learned the W-2's semantics or memorised one sheet geometry. |
| Issuing employer's quarterly returns | 100% | Every synthetic W-2 is issued by an employer that also files Form 941. Box 1 wages across an employer's W-2 set aggregate against that employer's quarterly wage totals — the canonical payroll reconciliation task. |
| Employee's personal return | 100% | The wages on the sheet flow to the employee's Form 1040. Generating both for the same identity produces the most common multi-document bundle in consumer tax processing. |
Wage and rate figures above are corpus-derived: they were computed over the 641 employment-eligible identities inside a local synthetic corpus of 1,000 identities generated by SymageDocs' World Simulation Engine at seed 20260421. Field, type and relation counts come from the shipped 2024 four-up W-2 definition in the SymageDocs form library. No real employee, employer, or payroll data was used at any stage.
Frequently asked questions
- What data format do synthetic four-up W-2 documents include?
- Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes, field types, and ground-truth values for all 152 fields on the sheet — 38 fields per copy across four copies — plus 132 label-to-value relations. Every annotation records which copy it belongs to, so copy-level accuracy is measurable. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.
- Are the four copies really identical?
- Identical in value, distinct in position. That is the whole point of the layout: a correct extractor must recognise that it is looking at one wage statement printed four times, not four statements from four employers. Systems that emit four records per sheet are a common and expensive production bug in income-verification pipelines, and this corpus is the only practical way to measure it.
- How does the 2024 sheet differ from the 2025 one?
- The layout is the same; the Social Security wage base differs. Box 3 caps at $168,600 on the 2024 sheet and $176,100 on the 2025 sheet, so the same high-earning employee shows a different Box 3 value in each year while Box 5 continues unchanged. Mortgage underwriting looks back two years, so both revisions arrive in the same submission constantly — train on the pair.
- What does Box 1 represent on this layout?
- On the four-up and vertical-stack layouts Box 1 carries the employee's gross annual wage, so Box 1, Box 3 and Box 5 agree below the Social Security wage base. The half-page single-copy W-2 in this catalog binds Box 1 to federal wages net of elective deferrals instead, which produces a lower Box 1 median. If your evaluation depends on the Box 1 versus Box 12 deferral relationship, use the single-copy layout for it.
- Can I use this data commercially?
- Yes. Every Social Security number, name, address, employer and wage figure is synthetic, contains no real personal data, and is licensed for commercial use including model training, benchmarking, and redistribution inside your own products.