HMDA manufactured-home loan records: a careful reading of approval data

This article asks a bounded question: what can a researcher infer from Home Mortgage Disclosure Act records that identify manufactured-home lending? The population is reportable mortgage applications in the HMDA data universe, not every loan secured by a manufactured home. The period is the latest published material reviewed on August 13, 2026. The method is a field-definition and coverage review using CFPB and Federal Financial Institutions Examination Council material.

HMDA begins with coverage

HMDA data are collected from covered institutions and reported under defined rules. That makes them valuable for studying reported application volume, action taken, loan purpose, property characteristics, and geography. It also makes coverage the first analytical question. A record set is not automatically a census of lenders, products, borrowers, or manufactured-home transactions in a market.

The distinction is especially important where chattel lending, seller financing, personal-property transactions, or small institutions may fall outside the selected HMDA universe. A table can show what covered reporters disclosed while leaving other financing channels unobserved. “No reported applications” therefore does not mean “no financing occurred.”

The manufactured-home flag needs context

A manufactured-home indicator helps separate relevant records from the broader mortgage file, but the field should be read with the data dictionary and reporting year. Definitions, edits, and available fields can change. The researcher should preserve the exact data vintage and identify whether the query uses property type, manufactured-home characteristics, land ownership, or a combination.

Combining fields can also create an analytical trap. A land-ownership category may be missing, unknown, or not comparable across institutions. A manufactured-home record does not by itself identify a park, a site lease, a home-only loan, or the condition of the collateral. Those questions need different evidence.

Action taken is not loan performance

HMDA can describe an application action such as originated, denied, withdrawn, or incomplete, subject to the reporting definitions. That is not the same as repayment, delinquency, foreclosure, appreciation, or resident welfare. An originated application can later perform differently from another originated application, and HMDA does not turn the action field into a credit-risk score.

Denial reasons require the same care. A reported reason is a disclosure field within the application process; it is not a complete causal explanation of an applicant's experience. Comparisons should control for product, geography, loan amount, applicant and property variables where appropriate, and should avoid treating a group-level association as an individual decision rule.

Geography and denominator choices

Public HMDA analysis often compares counties, metropolitan areas, or census tracts. The denominator must match the question. Applications per occupied manufactured home, applications per household, and share of all mortgage applications answer different questions. A rate created from an incompatible denominator can look precise while measuring something else.

Small cells create disclosure and reliability concerns. Suppressed or rounded values should remain visible in the research note. If a researcher combines years, the note should explain whether a change reflects market activity, reporting coverage, rule changes, or the selected denominator. Trend language is strongest when these possibilities are tested rather than silently ignored.

What a park operator can responsibly learn

HMDA can provide market context: which covered lenders reported activity in a geography, how applications were distributed by reported purpose, and how action categories differed within the defined universe. It can help frame questions for a lender conversation or compare public market signals with internal sales records.

It cannot reveal a resident's identity, establish a specific applicant's reason for an outcome beyond the disclosed field, or prove that a lender's public pattern applies to a particular park. Internal records should be compared only in aggregate and with appropriate privacy controls. A public file is a research input, not permission to re-identify households.

Bounded conclusion

HMDA is strongest as a transparent view of covered mortgage reporting. It supports carefully defined counts and rates when the researcher names the reporting year, fields, geography, denominator, and coverage limits. It does not capture every manufactured-home financing channel and does not measure post-origination performance. Treating those boundaries as part of the result produces a more useful lending brief than a simple approval ranking.

Record quality and disclosure controls

Public HMDA files are prepared for disclosure and analysis under a privacy framework. Some fields may be modified, generalized, or suppressed. That protects applicants but limits attempts to match an individual public record to an internal file. A responsible analysis keeps the unit at an aggregate level, reports cell sizes carefully, and does not use combinations of geography and characteristics to identify a person.

The reporting institution also deserves attention. A lender name in the file identifies a reporting entity in that vintage; it does not prove that the entity offers every product to every applicant or that a future application will be handled under the same policy. Institutional mergers, acquisitions, market exits, and reporting changes can interrupt a time series. Those events should be treated as possible explanations before a market conclusion is made.

What a strong comparison looks like

A useful comparison sets one question, such as whether reported manufactured-home application volume changed in a county. It defines the years, includes the action categories, states the denominator, and shows the number of covered reporters. It then tests whether the result remains similar when unknown values, small cells, or alternate geographic groupings are handled differently. This sensitivity check is more informative than adding decimal places to an unstable estimate.

Avoiding causal language

Public lending data can reveal a pattern without revealing its cause. If denial rates differ between areas, possible explanations include applicant mix, loan amount, collateral, lender composition, geography, documentation, and other factors. The public file may contain some controls, but a descriptive table should not be rewritten as proof that one factor caused an outcome.

The same rule applies to comparisons with park records. A change in internal sales or inquiry volume can coincide with a change in HMDA reporting while arising from a separate market event. Aligning periods and definitions is necessary before comparing them, and even a careful alignment may leave the relationship descriptive. Research becomes stronger when it says exactly what was compared and what was not.

A borrower outcome is not a market statistic

The public file describes reported applications at a defined point in the lending process. It does not follow a household through moving, refinancing, servicing, delinquency, or sale. A researcher who wants those outcomes needs a different longitudinal source and a lawful method for protecting personal information. The gap should be stated rather than filled with assumptions about what happened after origination.

Institution-level comparisons need an additional denominator. A lender with a small market footprint may show a high share of a narrow category without being the dominant lender overall. A lender's reported count can also change when its covered assets, branches, or reporting obligations change. Counts should therefore be paired with the selected market definition and not presented as a universal ranking.

A final quality check asks whether every reported number can be traced to the same release and filter. If not, the result should be labeled preliminary. That modest label is preferable to combining unlike extracts and presenting a smooth trend that the underlying source does not support.

This compact distinction keeps a public disclosure field from being promoted into a private borrower outcome or a complete financing census.

Sources

1. CFPB HMDA data and documentation
2. FFIEC HMDA resources
3. HMDA data browser
4. CFPB HMDA reporting guide
5. Regulation C, 12 CFR Part 1003
6. CFPB HMDA filing instructions
7. FFIEC HMDA resources
8. CFPB data research
9. Federal mortgage disclosure information
10. eCFR mortgage disclosure