Composition example. Real-shape academic paper page using the research-publishing cluster end-to-end — page header + author byline + paper meta + structured abstract + body prose + numbered figures / tables + inline footnotes + sticky Cite menu + acknowledgments. Composes ~10 Tux* components into the canonical paper rhythm.
article · transportation research record
Enduring infrastructure cues vs novelty effects
FHWA
HRDS-30-2022-04
TxDOT
0-7042
Abstract
Background
Earlier evaluations of rural intersection treatments often suffered from short follow-up windows that conflated treatment effects with novelty effects. We extend the follow-up window to 36 months across twelve treated sites with matched controls.
Methods
Twelve treated rural intersections in central Texas were instrumented for 36 months post-construction; matched controls were observed in parallel. Stop-line non-compliance rates were sampled in 6-month windows; difference-in-differences with site-level fixed effects estimated treatment effects.
Results
Treated sites showed a 37 percent reduction in stop-line non-compliance vs controls; the reduction persisted across all 36 months, with no detectable decay over time (slope = −0.04 pp/month, p = 0.78).
Conclusion
The persistence of the effect supports the infrastructure-cue hypothesis over the novelty hypothesis. Implications: treatment-effect attribution in safety evaluation should not rely on 12-month-or-shorter follow-up windows; persistence is itself a finding.
Keywordsrural roadway safety,long-term evaluation,stop-line compliance,difference-in-differences,TxDOT
1. Introduction
Roadway-safety treatments — pavement markings, signage, geometric changes, rumble strips — are routinely evaluated in 6- to 12-month windows post-construction. This window captures the immediate behavior change but cannot distinguish infrastructure-cue effects from novelty effects1. If treatment effects observed at 12 months reflect novelty rather than durable infrastructure cues, the cost-effectiveness frameworks built on those evaluations systematically overstate long-term benefits.
We extend the follow-up window to 36 months across twelve treated rural intersections in central Texas. Our central finding — no detectable decay in treatment effect across the full 36-month window — supports the infrastructure-cue hypothesis2 and has direct implications for safety-treatment evaluation methodology.
2. Methods
Twelve treated rural intersections were selected from the TxDOT 2024 Q1 treatment program (geometric realignment + high-visibility signage). Matched controls were drawn from the same TxDOT district using propensity-score matching on traffic volume, lane configuration, and prior crash rate.
- Treated (n=12)
- Control (n=12)
Figure 1 shows the canonical comparison: treated sites stabilized near 38–42% non-compliance within the first 6-month window and held there through month 36; controls oscillated around 77–81% with no comparable trend.
2.1 Corridor sub-study (I-35 mile 174–195)
A parallel sub-study extracted segment-level results along the I-35 corridor in the same TxDOT district. Treatment zone A (mile 178–184) and treatment zone B (mile 188–192) bracket a construction zone; baseline segments precede and follow.
Northbound →
I-35 corridor sub-study, mile 174–195
3. Results
| Window | Treated | Control | Δ (pp) |
|---|---|---|---|
| 0–6 mo | 42 | 79 | −37 |
| 6–12 mo | 41 | 81 | −40 |
| 12–24 mo | 39 | 78 | −39 |
| 24–36 mo | 38 | 77 | −39 |
Across the four 6-month windows, the treatment-control gap is statistically indistinguishable from constant (test of equal slopes across windows: F(3, 23) = 0.41, p = 0.75).
4. Discussion
The persistence finding has two methodological implications3. First, the standard 6- to 12-month evaluation window is sufficient to detect the treatment effect but not to distinguish it from novelty; researchers using shorter windows should be explicit about which claim they're making. Second, the cost-effectiveness frameworks that amortize treatment benefits over 5-, 10-, or 20-year horizons may be more defensible than the methodological literature suggests — at least for this class of treatments.
Open peer-review
TRR opted into open-review for this manuscript. Anonymized reviewer threads anchored to specific sections appear below; author responses are co-located. Resolved threads collapse behind a toggle.
Acknowledgments & declarations
Funding
- Federal Highway Administration (grant
HRDS-30-2022-04) - Texas Department of Transportation (grant
0-7042)
Acknowledgments
The authors thank the TTI MovementLab field team — Jordan Kim, Mariana Vega, and Daniel Park — for instrumentation and data quality assurance across the 36-month follow-up. Statistical review by D. Park.
Conflicts of interest
The authors declare no competing financial or non-financial interests.
Ethics & data access
All study protocols were reviewed and approved by the Texas A&M IRB (#IRB-2022-0418). Aggregated corridor data are available on request; per-subject data are not released to protect operator privacy.
Notes
- The novelty hypothesis: drivers attend differently to anything new in their environment; once the environment becomes familiar, attention reverts to baseline. The infrastructure-cue hypothesis: the new physical configuration carries persistent legibility advantages independent of novelty. ↩
- Decay would be expected if the effect were novelty-driven; the 12–15 month and later windows should show a regression toward control rates. ↩
- And a third practical one: TxDOT can plan capital deployment with confidence that the treatment package's effects don't require maintenance / reinforcement to persist. This bears on benefit-cost ratios in the agency's standard evaluation framework. ↩
Methods §2.3 · sample selection
Strong design overall. One concern: the 36-month window straddles a construction-phase shift in 2024 on the I-35 corridor. A sensitivity analysis dropping those quarters would strengthen the persistence claim. @ed-hassan thoughts?
We ran the analysis with and without 2024-Q2 through Q4. Headline coefficient moved <4%. We'll add a sensitivity-analysis footnote in the revised proofs.
Discussion §4 · paragraph 2
The cost-effectiveness extrapolation in para 2 is reasonable but feels under-cited — Hauer (2015) and the FHWA H-SIP framework both bear directly. Suggest one sentence + 2 cites.