Housing Disrepair Index: methodology

Source

Every record comes from a determination published by the Housing Ombudsman Service, and each links to the decision it describes so any figure can be checked at source. Only published determinations are used. No personal data about residents is collected, and the Ombudsman anonymises the decisions it publishes.

The Ombudsman investigates member landlords — principally social landlords. The index therefore describes social housing, and says nothing about private landlords.

How records are produced

Stating this precisely matters, because the value of the data rests on it.

• **Assisted extraction.** Fields are drafted from the published decision by software, which refuses rather than guesses where a document does not carry the structure it needs to read.
• **Independent source-verification pass.** Every record is checked field by field against the published decision. This pass found substantial errors in the drafted data, including findings recorded as unstated that the decisions state plainly, decisions missing an entire issue, and awards reported as unreadable that the decisions set out in a list.
• **Human review and sign-off.** Corrections are applied and reviewed before anything publishes.

What that is not: nobody claims each field was transcribed from scratch by hand. Extraction was assisted, and the safeguard is the verification against source that follows it.

What the data model records, and why

An earlier version of this data held one finding per case and one compensation figure per case. Both were wrong often enough to be unusable, and the reasons are worth stating because they shape everything published here.

• A single finding per case cannot express a decision that finds no maladministration on damp and severe maladministration on complaint handling. Reporting the worst finding as though it described the disrepair would have overstated landlords repeatedly. Findings are now held per issue.
• A single compensation figure cannot distinguish a £500 total from the £400 and £100 that make it up. Adding those gives £1,000 for a decision that awarded £500. Each award now carries a scope, and totals are never summed with components.
• Money a landlord offered before the Ombudsman got involved is not money the Ombudsman ordered. Each award records who awarded it and whether it was offered, paid, ordered or recommended.
• An order binds; a recommendation does not. They are stored separately.

Where a decision is silent, or wrong

Some awards cannot be attributed to any single issue: a decision may compensate the loss of amenity of a washing machine while investigating four other things. Those are recorded as undetermined rather than assigned to a plausible issue, because assigning them would move money between issues in every figure drawn from the data.

One decision in the current set states a total that does not equal its own components — £1,844.69 broken down into sums totalling £1,844.96, a 27p difference with two digits transposed. Both figures are recorded as the decision states them, and the inconsistency is flagged. Correcting it would make the dataset tidier and untrue.

The Ombudsman publishes one date per decision. Where a separate publication date is not available from an official source, the index records the decision date only rather than manufacturing a second one.

Limitations

• The current set is small. It supports statements about the decisions in it, and does not support claims about landlords generally, league tables, or averages presented as typical outcomes.
• It is not a random sample, so the proportions in it should not be read as proportions across all Ombudsman decisions.
• It covers social landlords within the Ombudsman’s jurisdiction, not private landlords.
• A finding describes the landlord’s handling of a complaint at a point in time. It is not a statement about a landlord today.
• Decisions are anonymised by the Ombudsman, so the index holds no location data and none is inferred from a landlord’s name.

Corrections

If any record here misdescribes a decision, we will correct it and say what changed. Every record links to its source so the check can be made independently. Corrections can be reported through our contact page.