Measuring traffic safety campaign results: how to know it worked
How to set a baseline, count real behavior and decide whether to keep, change or stop
Picture a volunteer group wrapping up a month of buckle up messages. They count 1,200 people at events and 5,000 web page views. Then a funder asks whether more people buckled up. No one knows, because no one counted belts beforehand.
National campaigns build that step in. The National Highway Traffic Safety Administration (NHTSA) schedules seat belt counts before and after each Click It or Ticket campaign [1].
Evaluators ask whether you did what you planned, whether behavior changed and how sure you can be that your work caused it. This guide answers those questions for small groups, with a sample logic model, a measurement plan, a belt count method, a worked example and a decision checklist.
Start with a logic model
In 2024, the Centers for Disease Control and Prevention (CDC) updated its Program Evaluation Framework, first published in 1999 [2]. In plain terms, its 6 steps are:
- Understand the setting: who has a stake and what else is going on.
- Describe the program: what you do, for whom, and what should change.
- Choose questions and a design: what you need to know and what you will compare.
- Gather credible evidence: pick measures, data sources and quality checks.
- Draw conclusions: compare results with targets and weigh other explanations.
- Act on findings: keep, adjust or rethink the work.
Step 2 usually produces a logic model, a one page map linking resources and activities to the changes you expect [2]. CDC’s guide places outputs, the counts of what you produced, in the measurement plan. But small groups often keep them in the model, as below [2].
Comparison compiled by Safety Behind the Wheel Foundation
Sample logic model: a teen seat belt campaign at one high school (hypothetical)
| Stage | What it means | Teen belt campaign example |
|---|---|---|
| Inputs | Resources you bring | Adviser time, 10 trained student observers, a small budget for printing and rewards, school and police partners |
| Activities | What you do | Student led class talks, buckle up checks at the lot exit, a hallway data wall, a family agreement night, a publicized police belt enforcement week |
| Outputs | Counts of what you delivered | Talks given, students reached, checks held, family agreements signed, page views |
| Short term outcomes | Changes in knowledge, attitudes and intentions | Students know the belt law, believe most classmates buckle and plan to buckle on every trip |
| Behavior outcomes | Changes you can see | Higher observed belt use among drivers and front passengers leaving school |
| Long term outcomes | Changes in harm | Fewer teens killed or seriously hurt in crashes, judged over 3 to 5 years alongside other programs |
| Context | Outside factors | A new state law, a local crash in the news, a schedule change |
Read it top to bottom, asking “if this, then what?” A weak link, such as class talks that never mention belts, shows up before you spend money.
Process measures versus outcome measures
Process measures show whether you did what you planned. They cover 3 things:
- Reach: who you touched, and whether they were the people you meant to reach.
- Dose: how much each person got. The AAA Foundation for Traffic Safety’s campaign toolkit separates dose delivered from dose received [3]. A poster on the wall is delivered, but only a student who reads it has received it.
- Fidelity: whether you ran activities as designed [3], such as 14 of 16 planned talks given in full.
Outcome measures show whether anything changed, climbing from knowledge and attitudes to behavior, then to crashes and injuries.
Participation counts belong on the process side. NHTSA’s Countermeasures That Work reviews youth programs such as crash reenactments and impairment goggles. Most lack strong evaluation. The studies that exist found changes in knowledge or attitudes with little or no effect on behavior [4]. A 2022 AAA Foundation review found distracted driving programs raised awareness and intentions. But how well those gains carry over to real driving remained uncertain [5].
A full gym is a good start, not a result.
Build a measurement plan before launch
List each question, the measure that answers it, how and when you will collect it, and one owner for every row.
Comparison compiled by Safety Behind the Wheel Foundation
| Level | Question | Measure | How and where | When |
|---|---|---|---|---|
| Process | Did we deliver the plan? | Activities done ÷ activities planned | Activity log | Weekly |
| Process | Whom did we reach? | Students at 1 or more activities; families signing agreements | Sign in sheets, forms | Monthly |
| Short term | Did beliefs shift? | Share who think most classmates buckle | Anonymous survey, same questions each time | Before and after |
| Behavior | Did belt use rise? | Share of drivers and front passengers belted | Lot exit counts at fixed exits, days and times, plus a comparison school | Baseline, end, 3 months later |
| Side effects | Did anything go wrong? | Complaints; shifts in which exits drivers use; who gets ticketed | Notes and partner check ins | Ongoing |
| Long term | Any change in harm? | Teen fatal and serious injury crashes in the county | State crash data, 5 year view | Yearly, as context only |
Drive With Care measurement plan template, filled in for the hypothetical belt campaign.
Baseline, follow up and a comparison site
A baseline is your starting number, measured before launch. A follow up repeats it afterward. The AAA Foundation toolkit advises collecting both under similar conditions, leaving enough time for the campaign to work but not so much that other events intrude [3]. NHTSA’s peer to peer program guide adds that counts should match on time of day, day of week, length and method [6].
NHTSA’s 2027 Click It or Ticket calendar shows the rhythm. Baseline belt observations run April 19 to May 2. Follow up observations run June 7 to 17 [1]. Public awareness surveys bracket the campaign the same way [1].
A before and after change alone can’t tell you why belt use rose, since a new law or a widely shared crash might explain it. CDC describes comparing groups that weren’t randomly assigned as a practical option when randomizing isn’t possible [2]. For a small group, that means a comparison site: a similar school, corridor or town without the campaign, counted the same way. One enforcement study NHTSA cites counted belts on a highway corridor and at a control site [7].
Plan a later follow up too, because belt use often fell about 6 percentage points after short enforcement waves, though it stayed above where it started [7].
How to run a parking lot belt count
States run yearly seat belt surveys under federal rules in 23 CFR Part 1340 [8]. NHTSA’s national survey uses similar methods [9]. Here is how those rules translate to a school or workplace lot.
Comparison compiled by Safety Behind the Wheel Foundation
| Federal practice | Your parking lot version |
|---|---|
| Observe drivers and right front passengers; belted means the shoulder belt is in front of the shoulder [8] | Same rule. Belted means the shoulder belt crosses the chest, not tucked behind |
| Mark “unknown” when observers can’t tell, and keep unknowns at 10% or less of the survey [8] | Mark unknown for glare or dark tint. If unknowns top 10%, move your spot |
| No police vehicles, uniformed officers or survey signs at sites [8] | Use no police or signs, and never react to individual drivers |
| Observers trained within 12 months; surprise quality checks at 5% or more of sites [8] | Train in pairs on 20 practice cars, compare tallies and settle disagreements |
| Standard error of no more than 2.5 percentage points [8] | Aim for 250 or more observations per round, about what a 2.5 point standard error takes at 80% belt use (our calculation) |
| National survey watches stopped vehicles at stop signs or signals, 7 a.m. to 6 p.m. [9] | Stand where cars slow or stop, safely off the roadway, at the same weekday times each round |
In June 2025, NHTSA’s national survey observed 66,497 vehicles at 1,693 sites and found front seat belt use of 91.3% [9].
Three more habits keep counts honest. Never record plates, names or photos. Log weather and unusual events. Keep the same spots, days, times, rules and observers.
NHTSA notes that daytime surveys of front seat occupants likely overestimate overall belt use [7]. They skip night trips and back seats, where NHTSA observed 84.0% belt use in 2025 [9]. Student teams can find more ideas in our guide to student led campaigns.
Measuring speed and phone use
For speed, ask your city or county traffic engineer for a before and after speed study. Portable counters using road tubes, radar or other sensors can log each vehicle’s speed [10]. The Federal Highway Administration’s (FHWA’s) Traffic Monitoring Guide sets 48 hours as the minimum for many short term counts. It notes that 7 day counts avoid day of week adjustments [10].
FHWA’s traffic calming guide lists average and 85th percentile speeds in each direction as baseline data [10]. The 85th percentile is the speed that 85% of drivers stay at or under. Also ask for the share going 10 mph or more over the limit. At 26 speed humps in that guide, this share fell from 14% to an average of 1% [10].
For phone use, add a column to your belt count, marking whether each driver holds or looks down at a phone.
Surveys and web data: useful, with limits
Surveys capture what you can’t see from the curb, such as knowledge, beliefs about peers and awareness of enforcement. But people tend to report what sounds good. NHTSA notes that occupants in less severe crashes may tell police they were buckled to avoid a penalty [7].
Evaluations face the same trap, as researcher Anders af Wåhlberg found when he studied an online course for traffic offenders under 25. Their reported risky driving looked low before the course and rose afterward [11]. He linked the low starting scores to socially desirable answers, an effect that faded after the course [11]. In a related study, controlling for this bias more than halved how well driver questionnaires predicted self reported crashes [11].
For more honest answers, keep surveys anonymous and say so. Ask about specific recent behavior, such as “your last 5 trips.” Repeat the same wording each round. Compare self reported belt use with your observed rate, since a wide gap is a warning sign.
Website and resource use measures exposure, not behavior. The AAA Foundation toolkit treats web traffic as process data [3]. A view may come from a bot, a repeat visitor or someone who left in seconds. Track actions closer to behavior, such as agreement downloads or car seat check sign ups. Give each channel its own link or QR code. Even then, a click is not a buckled belt.
Why crash data rarely settles the question
Crashes are the outcome that matters most, yet they make a weak yardstick for a local campaign:
- Rare events: in a county averaging 6 traffic deaths a year, chance alone could produce anywhere from about 2 to 11 in a given year (our calculation). NHTSA cites an estimate that a driver education study would need 35,000 participants to reliably detect a 10% crash drop [4].
- Lag: NHTSA’s national overview of 2024 crashes came out in April 2026. Its counts can still change when the file is finalized [12].
- Missing crashes: NHTSA estimated that in 2019, 32% of injury crashes and about 60% of property damage only crashes never reached police [13].
- Regression to the mean: a place picked after an unusually bad year tends to improve on its own.
- Reporting changes: new crash forms, staffing or definitions can move numbers with no change on the road.
Our guide to understanding traffic safety data explains these traps. Use crash data to choose a target, as our guide to using local crash data shows, then judge a short campaign by observed behavior.
Small samples: how big a change is real?
Every count has a margin of error. For a share such as belt use, a 95% confidence interval is about:
margin = 1.96 × √(p × (1 − p) ÷ n)
Here p is the share belted, written as a decimal. The number observed is n. With 100 observations at 80% belt use, the margin is about ±8 points. If you compare two rounds of 100, a change must top about 11 points to stand out. So a 10 point rise could be noise (our calculations).
Comparison compiled by Safety Behind the Wheel Foundation
| Observations per round | Margin for one count at 80% belt use | Margin for the change between two counts |
|---|---|---|
| 50 | ±11 points | ±16 points |
| 100 | ±8 points | ±11 points |
| 200 | ±6 points | ±8 points |
| 400 | ±4 points | ±6 points |
| 800 | ±3 points | ±4 points |
Drive With Care calculations at 95% confidence, assuming each observation is independent.
Two cautions apply. Many of the same people drive past every day, so your true margin is wider than the table shows. Also plan for the change you hope to see. Detecting a rise from 80% to 90% takes about 200 observations per round. A rise from 80% to 85% takes about 900 (our calculations, at 80% power).
A worked example with numbers (hypothetical)
A student team runs the campaign above at School A, with School B as a comparison. Both are counted at one exit on the same 3 weekdays and time window, in late April and again in October.
Comparison compiled by Safety Behind the Wheel Foundation
| Measure | School A (campaign) | School B (comparison) |
|---|---|---|
| Baseline belt use | 315 of 420 belted = 75.0% | 287 of 380 belted = 75.5% |
| Follow up belt use | 353 of 410 belted = 86.1% | 304 of 395 belted = 77.0% |
| Change | +11.1 points (±5.3) | +1.4 points (±6.0) |
| Survey: “I always buckle up” | 93% before, 95% after | Not surveyed |
| Process | 14 of 16 talks given; 3 of 4 buckle up checks held | None |
| Web | 2,300 page views; 410 agreement downloads | None |
Hypothetical data. Margins are Drive With Care 95% calculations.
How an evaluator reads it:
- School A rose about 11 points, more than its margin, while School B barely moved.
- The gap between the two changes is about 10 points, give or take 8 (our calculation). The campaign likely helped, though the true effect could be small or large.
- Students reported far higher belt use than observers saw, so the survey can’t replace counts.
- Unbelted occupants fell from 105 to 57, a concrete number to share, but the team makes no crash claim.
Watch for unintended effects
Programs can cause harm or simply move a problem, so CDC suggests anticipating unintended outcomes when you describe a program [2].
Comparison compiled by Safety Behind the Wheel Foundation
| Effect | What can happen | What to check |
|---|---|---|
| Risk compensation and extra exposure | Driver education incentives that let teens move through graduated licensing faster may increase crashes [4] | Whether people feel so protected that they take more risks, or start driving sooner or more often |
| Backlash | Fear appeals about distracted driving may increase it among young adults [4]. In a 2013 NHTSA survey report, 70% of drivers agreed speed cameras are used to raise revenue, and 55% agreed they are used to prevent crashes [14] | Complaints, comments and support in follow up surveys |
| Stigma | Messages that single out a group can push people away | How people in that group felt about the campaign |
| Displacement | Speeds rebound quickly past feedback signs, and some camera studies found sudden speed changes around sites [14]. Speed humps cut daily traffic on treated streets by 20% on average, and the shift depends on nearby alternative routes [10] | Speeds and volumes on nearby streets and past the site |
| Unfair enforcement | The Governors Highway Safety Association (GHSA) urges collecting race and ethnicity data on stops, and reports smaller disparities when enforcement targets risky driving [15] | Who gets stopped and ticketed, by neighborhood and group |
Document lessons and decide what comes next
CDC stresses that findings don’t turn into action by themselves. Its guide suggests short findings memos, group debriefs and a clear choice to take no action, adjust the program or gather more information [2]. NHTSA’s peer to peer guide treats a disappointing result as a chance to learn and revise [6].
Within 2 weeks of each campaign wave, hold a 45 minute after action review with your team and partners. Ask what you planned, what happened, why there was a gap, and what you will keep or change. Write down the answers with the data. Store them where next year’s team will find them.
- Continue when every box is checked.
- Scale when results held up at more than one site and you can keep fidelity as you grow.
- Change when delivery fell short, reach missed or results were mixed. Fix the weak link and test again.
- Stop when a fair test shows no change or you find harm you can’t fix. Move resources to a better supported strategy.
Efforts lasting 12 months or more are more likely to change behavior, NHTSA’s peer to peer guide notes. So one semester may be a pilot rather than a verdict [6].
Frequently asked questions
What is the difference between process and outcome evaluation?
Process evaluation checks whether you delivered the program as planned and reached the right people. Outcome evaluation checks whether knowledge, behavior or harm changed [2, 3]. A program that was never fully delivered can’t fairly be judged on outcomes.
How do you measure whether a seat belt campaign worked?
Count belt use among drivers and front passengers before and after, at the same places, days and times, with the same rules [6, 8]. Add a comparison site, and aim for at least 250 observations per round.
Can we use crash data to evaluate a local campaign?
Usually not by itself, because local crash counts are small, slow to arrive and incomplete. Chance can also swamp any real effect [12, 13].
Are surveys before and after enough?
They track knowledge and beliefs, but people tend to overstate safe habits [11]. Pair surveys with observations of real behavior.
How long should we wait before measuring results?
Measure soon after the campaign ends, then again about 3 months later, because belt use often slips after short enforcement waves [7].
The bottom line
Recommendation from Safety Behind the Wheel Foundation
Measure the behavior you want to change before you start, then measure it again the same way, alongside a comparison site. Keep process counts separate from outcomes, respect the margin of error and write down what you learn. Then decide on purpose whether to continue, change, scale or stop.
Related articles
- How to build a community traffic safety program that lasts
- Understanding traffic safety data: how to read crash statistics wisely
- How to use local crash data to choose community safety priorities
- Traffic safety messages that change behavior: a practical guide
- Student led traffic safety campaigns: a peer to peer guide for schools
Sources and further reading
- 1National Highway Traffic Safety Administration (NHTSA), Traffic Safety Marketing. Click It or Ticket (2027 campaign calendar, including belt observation and awareness survey windows). Accessed October 2026. trafficsafetymarketing.gov/…/click-it-or-ticket
- 2Kidder DP, Fierro LA, Luna E, et al. CDC Program Evaluation Framework, 2024. MMWR Recommendations and Reports 73(6):1 to 37. September 26, 2024; with CDC’s framework summary page (August 20, 2024) and Program Evaluation Framework Action Guide, Steps 2 to 6 (updated August 18, 2024). cdc.gov/mmwr/volumes/73/rr/rr7306a1.htm, framework summary, Step 2, Step 3, Step 6
- 3Bayne A, Siegfried A, La Rose C, Price J, Johnson-Turbes A. Evidence-Based Behavior Change Campaigns to Improve Traffic Safety: Toolkit. AAA Foundation for Traffic Safety. March 2020. aaafoundation.org/research/…toolkit and toolkit PDF
- 4NHTSA. Countermeasures That Work (11th edition, online): Alcohol-Impaired Driving, Youth Programs; Distracted Driving, Communications and Outreach; Young Drivers, Pre-Licensure Driver Education. Accessed October 2026. youth programs, communications and outreach, driver education
- 5Arnold L, Horrey WJ. Effectiveness of Distracted Driving Countermeasures: An Expanded and Updated Review of the Scientific and Gray Literatures (research brief). AAA Foundation for Traffic Safety. March 2022. aaafoundation.org/research/… and brief PDF
- 6Fischer P (for the Governors Highway Safety Association). Peer-to-Peer Teen Traffic Safety Program Guide. NHTSA, DOT HS 812 631. March 2019. ghsa.org/…/peer2peerbrochure.pdf
- 7NHTSA. Countermeasures That Work (11th edition, online): Seat Belts and Child Restraints, Data/Surveillance; Short-Term, High-Visibility Seat Belt Law Enforcement. Accessed October 2026. data/surveillance and short-term enforcement
- 8Electronic Code of Federal Regulations. 23 CFR Part 1340: Uniform Criteria for State Observational Surveys of Seat Belt Use (76 FR 18056, April 1, 2011). Current as of October 1, 2026. ecfr.gov/current/title-23/…/part-1340
- 9NHTSA. Occupant Restraint Use in 2025: Results From the NOPUS Controlled Intersection Study. DOT HS 813 820. May 2026. crashstats.nhtsa.dot.gov/…/813820
- 10Federal Highway Administration (FHWA). Traffic Monitoring Guide (2022), Chapters 2 and 3; and Traffic Calming ePrimer, Modules 4 and 7. Accessed October 2026. traffic data collection, data methods, ePrimer Module 4, ePrimer Module 7
- 11af Wåhlberg AE. Re-education of young driving offenders: effects on self-reports of driver behavior. Journal of Safety Research 41(4):331 to 338. August 2010; and Social desirability effects in driver behavior inventories. Journal of Safety Research 41(2):99 to 106. April 2010. pubmed.ncbi.nlm.nih.gov/20846549 and pubmed.ncbi.nlm.nih.gov/20497795
- 12NHTSA. Overview of Motor Vehicle Traffic Crashes in 2024. Traffic Safety Facts Research Note, DOT HS 813 791. April 2026. crashstats.nhtsa.dot.gov/…/813791
- 13Blincoe L, Miller T, Wang J-S, et al. The Economic and Societal Impact of Motor Vehicle Crashes, 2019 (Revised). NHTSA, DOT HS 813 403. February 2023. crashstats.nhtsa.dot.gov/…/813403
- 14NHTSA. Countermeasures That Work (11th edition, online): Speeding and Speed Management, Speed Safety Camera Enforcement; Dynamic Speed Display/Feedback Signs. Accessed October 2026. speed cameras and feedback signs
- 15Sprattler K, Statz L (Kimley-Horn). Equity in Highway Safety Enforcement and Engagement Programs. Governors Highway Safety Association (GHSA). August 2021. ghsa.org/…/equity_2021.pdf
This article reflects the safety priorities of program evaluation and traffic safety epidemiology. It is general education, not legal, statistical or other professional advice for any program or person. For formal studies, work with a trained evaluator. Federal rules, campaign dates and data change, so verify details with official sources such as NHTSA, CDC and your state highway safety office. Reviewed October 2026.
Keep learning
Safe Driving Resource Center
Guides, self-checks and articles for drivers, parents of new drivers, and families recovering after a crash.
