Synthetic Form 1120-H - U.S. Income Tax Return for Homeowners Associations Data
Synthetic training data — no real PII, fully coherent identities
Generate synthetic Form 1120-H returns for homeowners associations — exempt function income, association expenditures, and the 60% / 90% election tests on a single dense page. Training data for HOA accounting, community-association management, and property-tech document pipelines.
62
Fields per document
1
Page
Tax
Category
What this document is
Form 1120-H is the income tax return a homeowners association files when it elects to be taxed under section 528. The whole return fits on one page: association type (condominium, residential real estate, or timeshare), exempt function income from member dues and assessments, gross income from non-exempt sources such as interest and rentals, deductions, and the tax computation. It also carries the tests that qualify the association for the election in the first place.
Why generate synthetically
Community-association management is a large, fragmented back office running on documents nobody has built training data for. There are hundreds of thousands of HOAs, each producing an annual return, and none of that data is shareable. Synthetic 1120-Hs give property-tech and association-accounting platforms a labeled corpus for the one tax form their customers all file.
What makes synthetic data useful
Each return is generated for a real entity in the identity graph, with exempt function income, non-exempt income, and expenditures scaled together rather than sampled independently — so the relationships that determine whether the association's election holds are present in the data. Association type is drawn across condominium, residential real estate, and timeshare categories, and the corresponding checkbox is set consistently with the entity's profile.
Training challenges
The 1120-H packs a complete corporate return into a single page, so currency cells make up nearly 60% of its fields and sit at very tight vertical spacing with short line labels — the classic condition for row-offset errors, where a model reads the right column but the wrong line. Roughly 44% of fields are legitimately blank: an association with no rental or royalty income leaves those lines empty, and the preparer block is unfilled on self-prepared returns. The three association-type checkboxes are mutually exclusive and sit adjacent to each other in small targets, so a checkbox classifier that fires on two of them produces a document that cannot be interpreted at all.
Generate synthetic Form 1120-H - U.S. Income Tax Return for Homeowners Associations data
Start with 500 free credits. No credit card required.
Generate NowWho uses this data
Community-association management platforms, HOA accounting and reserve-study software, property-management and proptech back offices, and the specialist CPA firms that file thousands of these returns each season. It is also a good compact benchmark for niche entity-type classification: a model that has to route a corporate filing to the right pipeline must first tell an 1120-H apart from an 1120, an 1120-POL, and an 1120-SF, all of which share a visual family.
Document complexity profile
62 fields on a single rendered page: 37 currency amounts, 15 text, and 10 checkbox targets. 62 annotation relations. The binding graph is minimal — 1 arithmetic binding, 1 function call, maximum expression depth 1 — but the page density is high: currency cells make up nearly 60% of the fields on one sheet, and about 44% of all fields are legitimately blank on a typical association return.
Key stats from our synthetic corpus
Quantitative characteristics of the Form 1120-H - U.S. Income Tax Return for Homeowners Associations documents our generator produces.
| Metric | Value | Detail |
|---|---|---|
| Population coverage | 64% | 64% of identities in our 1,000-identity corpus are attached to an entity and therefore 1120-H eligible. |
| Median exempt function income | $8,434 | Median total exempt function income is $8,434, with a p25–p75 range of $4,455 to $15,097. This is the member-dues line that distinguishes an association return from an ordinary corporate one. |
| Median total expenditures | $11,680 | Median total expenditures for the tax year is $11,680 against median gross income of $11,766 — associations that spend what they collect, which is the realistic operating pattern. |
| Fields carrying a value | 56% | 35 of the form's fields carry generated values on a typical return; the rest render blank as they do on a real filing. That mix makes this a compact corpus for training blank-versus-missing discrimination. |
| Currency share of fields | 60% | 37 of the form's fields are currency amounts on a single page. High currency density at tight line spacing is what drives row-offset errors on compact returns. |
How this document co-occurs with others
Rates at which identities in our corpus that produce a Form 1120-H - U.S. Income Tax Return for Homeowners Associations also produce other documents.
| Correlation | Rate | Detail |
|---|---|---|
| Standard corporate return | 100% | The same entity can also produce a standard Form 1120. Generating both gives the contrast pair a form-type classifier needs to route an association return away from the corporate pipeline. |
| Other niche 1120 variant (1120-POL) | 100% | Form 1120-POL is a similarly compact single-page 1120 variant for political organizations. The two look alike at a glance and are the classic confusion pair for entity-type classification. |
| Can file a 7004 extension | 100% | The same entity can produce a Form 7004 extension whose form code identifies the return being extended — direct supervision for extension-to-return matching. |
| Files quarterly payroll returns | 100% | Associations with staff file Form 941 quarterly, and every entity in the corpus does, so association tax and payroll documents can be generated as one coherent set. |
| Receives vendor invoices | 100% | Every entity in the corpus also participates in invoicing, which lets you pair an association's return with the vendor invoices behind its expenditure lines. |
All stats above are corpus-derived: they were computed on a local synthetic corpus of 1,000 generated identities produced by SymageDocs' World Simulation Engine, with 1120-H figures measured over the entity-eligible subset. No real homeowners association or member data was used. `make corpus-stats` regenerates the shipped corpus_stats.json for the site's analytics form set, which does not include this edition; reproduce these figures with `python -m symagedocs.corpus_analysis --corpus-size 1000 --form-id irs_f1120h_standard_2024`.
Frequently asked questions
- What data format do synthetic Form 1120-H documents include?
- Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes and ground-truth values for all 62 fields on the single page, including the association-type checkboxes and the exempt-function income lines. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.
- Which fields are blank on a typical return?
- About 35 of the form's fields carry values on a typical filing and the rest render blank — most commonly rentals, royalties, capital gains, the amended-return and name-change checkboxes, and the paid-preparer block. Those blanks are annotated regions with no value, so a model learns to expect them.
- Is exempt function income modeled realistically?
- Yes. Median total exempt function income in our corpus is $8,434 against median total expenditures of $11,680, with income and spending generated in proportion — the pattern of an association that collects dues and spends them on the property rather than accumulating a surplus.
- How does labeling work on a single-page form?
- Identically to the multi-page forms: every value field and every printed label is emitted with its exact bounding box and label-to-value relation. On a page this dense, those relations are what let a model resolve which of several adjacent short labels a currency cell belongs to.
- Can I use this data commercially?
- Yes. All synthetic data is generated from statistical models, contains no real association or member data, and is licensed for commercial use including ML model training and benchmarking.