TxDOT CRIS Crash Records: What They Support

The Texas Department of Transportation is the custodian of crash records for the state, and it publishes them through the Crash Records Information System. The collection is large, it is queryable by the public, and its boundaries are unusually well documented, which makes it a good worked example of reading a dataset by its statute rather than by its interface.

What obliges a record to exist

The department states the reporting requirement in its own words, and it is worth reading before anything else, because every boundary of this dataset is in it:

Texas Transportation Code §550.062 requires any law enforcement officer who in the regular course of duty investigates a motor vehicle crash that results in injury to or the death of a person or damage to the property of any one person to the apparent extent of $1,000 or more, to submit a written report of that crash to TxDOT not later than the tenth day after the date of the crash.

Texas Department of Transportation, Crash reports and records

Three separate filters are packed into that sentence and each one shapes the data.

An officer has to have investigated. The obligation attaches to a police investigation, not to the crash. A collision nobody reported to police generates no record no matter how serious the underlying facts.

A threshold applies. Injury, death, or apparent property damage at or above the stated amount. Apparent is doing real work: it is a roadside judgement made quickly, not an assessment. Cases near the line are in or out depending on an estimate.

The damage threshold is a fixed dollar figure. A fixed nominal threshold becomes easier to cross as vehicle repair costs rise, so the population of qualifying crashes can grow over a long series without any change in what happened on the roads. Any multi-year comparison of property-damage crashes needs to account for that before it means anything.

The category that was removed

The most consequential fact about this collection is something it no longer holds.

Under Senate Bill 312 of the 85th Texas Legislature, and effective from 1 September 2017, the Driver's Crash Report, form CR-2, ceased to be retained by the department. As of 1 January 2019 the retention period expired for all such forms, and the department states that it no longer holds any Driver's Crash Reports in its custodial records, no longer hosts the form, and destroys any that are submitted.

The CR-2 was the driver-submitted report for crashes not investigated by an officer. Its removal means the collection is now, in effect, a record of police-investigated crashes only, and that the boundary moved partway through any series spanning that period. A researcher comparing a year before the change to a year after is comparing two differently defined populations, and nothing in the data itself will announce that.

Drivers given a CR-2 or an equivalent local form are advised by the department to keep it for their own records, since submitting it accomplishes nothing.

How far back it goes

The state retention schedule for crash reports and data is ten years plus the current year, and the department states plainly that data analysis for years outside that window is unavailable, though published annual reports may still exist for earlier years.

This is a rolling window, which has an implication worth making explicit: a query run today and the identical query run in three years will cover different periods, and the earlier years will have silently left. Any figure taken from this system should be stamped with the date it was retrieved. A citation without one cannot be reproduced.

What it supports, and what it does not

Within its boundaries the collection is strong. It covers crashes on public roadways and the state highway system, it is organised so that a crash can be examined together with the vehicles and people involved, and the public query interface makes location and time filtering straightforward without a data request.

It does not support any claim about crashes police never investigated, which is now the whole of the sub-threshold and unreported population. It does not support long historical series beyond the retention window. It does not support anything about outcomes after the roadside, because it is a record of a reported event and not of what followed. And the department notes that the reports themselves are treated as confidential, which governs what an individual record can be obtained for and by whom, separately from what the aggregate query tool will show.

Reading it against the federal picture

The federal fatality census and this collection will not agree, and the disagreement is structural rather than an error in either.

The federal census is built to capture fatal crashes nationally under one definition. The state collection is built to capture what Texas officers reported under a Texas threshold, including a large volume of non-fatal crashes the federal census never sees. For fatal crashes in Texas the two overlap and their counts may still differ, because inclusion rules, timing of record finalisation, and the treatment of deaths occurring some time after a crash are all defined separately.

The right instinct on finding a discrepancy is to go to both definitions before going to either dataset. Nine times in ten the definitions explain it, and the tenth is worth reporting to whoever publishes the file.

The companion guide covers the federal commercial vehicle sources this one sits alongside. For how the same reading discipline applies to court records rather than crash records, the full index of research notes connects the sections.