Transportation safety
What makes a Texas work zone risky?
A year of TxDOT crash records for Dallas and Ellis counties, filtered to the crashes that happened in construction zones.
Original project · 2025–2026PythonScikit-learnKPI ReportingGradient Boosting
The question
When and how do work-zone crashes happen, and are they more severe than other crashes?
The original project
My MS capstone analysed 93,000+ TxDOT crash records, built a KPI reporting framework and a Gradient Boosting classifier, and presented road-safety recommendations to stakeholders. This page applies the same exploratory approach to a public 2025 extract of the same TxDOT CRIS data.
At a glance
Crashes analysed53,086Dallas + Ellis, 2025
In work zones2,258
Work-zone share4.3%
With workers present721
The analysis
Work-zone crashes by month
2025, Dallas and Ellis counties
When work-zone crashes happen
Crash count by day of week and 3-hour window
12–3a3–6a6–9a9a–12p12–3p3–6p6–9p9p–12a
Mon2113544046563239
Tue1712475755503336
Wed2020685458644848
Thu1420544746674151
Fri2331515963564641
Sat5325243947435138
Sun3923212833293136
How work-zone crashes happen
Share of work-zone crashes by collision type
- Rear-end30.1%
- Sideswipe23%
- Single vehicle19%
- Angle15.7%
- Left turn across traffic6.9%
- Other5.2%
Severe-crash rate inside vs outside work zones
Fatal + suspected serious injury, as a share of crashes with known severity
- Outside work zones3.8%
- All work zones3.5%
- Work zone, no workers3.6%
- Work zone, workers present3.4%
Work-zone exposure by posted speed limit
1,626 crashes with no speed limit recorded are left out
| ≤ 30 mph | 9,632 | 238 | 2.5% | 4% |
| 35–40 mph | 17,839 | 688 | 3.9% | 2.9% |
| 45–55 mph | 8,348 | 362 | 4.3% | 4.2% |
| 60–65 mph | 6,806 | 576 | 8.5% | 2.5% |
| 70+ mph | 8,835 | 382 | 4.3% | 5.3% |
What the data shows
- 2,258 crashes (4.3%) happened in work zones; 721 of them with workers present.
- Rear-end (30.1%) and sideswipe (23%) crashes dominate — the signature of queues and lane shifts.
- Severe-crash rates were similar: 3.5% in work zones vs 3.8% outside. What sets work zones apart in this data is how crashes happen, not how severe they are.
- Weekday daytime (6am–6pm, Mon–Fri) accounts for 48.4% of work-zone crashes — about the same as for all crashes (48.6%). The busiest weekday window is 3–6p.
- Work-zone crashes were lowest in Jan and peaked in Oct (233); 60–65 mph roads have the highest work-zone share (8.5%).
Recommendations
- Add queue-warning signs and dynamic speed limits ahead of lane closures on 60–65 mph corridors.
- Avoid scheduling lane closures in the busiest weekday window (3–6p) where possible.
- Improve lane-shift delineation (cones, markings) to cut sideswipes.
- Track work-zone crashes per active project monthly as a standing KPI.
How the analysis works
- Loaded 53,086 crash records and flagged those in construction zones and with workers present.
- Built a monthly trend and a day-of-week × time-of-day heatmap of work-zone crashes.
- Grouped raw collision codes into six plain-English types.
- Compared the severe-crash rate (fatal + suspected serious injury) inside and outside work zones, excluding unknown severity.
Caveats
- Two counties for one year is a slice, not the whole state.
- Crash records show where crashes happened, not traffic volume, so rates per vehicle-mile are not possible here.
Methods
- KPI design
- Time-pattern heatmaps
- Category grouping
- Rate comparison