The Korean bus seatbelt reminder study: belt use went from 9.6% to 59%

A Korean government research institute put a buckle sensor in every passenger seat of a working interurban bus and measured what happened. Belt use rose from 9.6 per cent to 59.0 per cent. It is the only trial of its kind on a bus anywhere in the published literature — and Australia has just written a bus seatbelt rule that leaves the sensor out.

The finding, in one line

In 2023 a team from the Korea Automobile Testing and Research Institute (KATRI) reported the outcome of a pilot project to the 27th International Technical Conference on the Enhanced Safety of Vehicles. They had fitted an interurban bus in commercial service with a seat-occupancy sensor and a buckle sensor at every passenger seat, a visual warning device at each seat, and a monitoring display at the driver’s position that recorded which seats were unfastened. Then they counted belts.

Six times as many passengers wore a belt. Nothing about the route changed, no campaign ran, no fines were issued and no driver was asked to do anything differently. The only new thing in the cabin was a device that knew which seats were occupied, knew which belts were undone, and said so.

What Park et al. actually did

The study is short, and the method is unglamorous in the way good field research usually is. KATRI worked with a commercial operator, the KD Transit group, on an interurban service running between Seoul and Suwon. One bus on that route was fitted with a custom seatbelt reminder system. Wearing rates were then observed on a bus with the system and on a bus without it.

Two design choices in that table deserve to be pulled out, because they are the choices that make the result usable rather than anecdotal.

It gated the alert on occupancy

A bus is rarely full. A reminder system that reacts to an empty seat — and on a fifty-seat coach there will be plenty — fires constantly for no reason, and within a week every passenger and every driver has learned that the alert means nothing. The failure is well documented in other alarm-dense environments, where staff exposed to high false-alarm rates stop responding to true alarms as well. KATRI put a cushion sensor in alongside the buckle sensor, so the system only ever comments on a seat that actually has somebody in it. This is not a refinement; over a twenty-five year service life it is the difference between a system that still works and a system everyone has learned to ignore.

It addressed the individual seat

The warning device was at the seat, not at the front of the cabin. That converts a broadcast into a personal message. A sign at the bulkhead tells forty-eight people something in general; a lit indicator at seat 23D tells one person something specific, and it tells the people around them too. The behavioural difference between those two things is the whole subject of the car seatbelt reminder literature.

The result, placed against every other rate we know

A single number is easy to dismiss. What makes 59.0 per cent striking is where it sits relative to every other seatbelt wearing rate that has been measured in a bus or a car.

Australian bus passenger wearing rates have been reported anywhere between 7 and 45 per cent. Korea’s unequipped bus, at 9.6 per cent, sits inside the bottom of that band. Australian bus passengers are not a different species from Korean bus passengers, and the baseline says so. That is what makes the treatment arm interesting.

Why one pilot carries this much weight

Ordinarily a single-site observational comparison is weak evidence. Here it is load-bearing, for a reason that says more about the state of the field than about the study: every other measurement of a sensor-linked seatbelt reminder in the published literature was taken in a car.

Search the seatbelt reminder literature and you find more than two decades of work — observational fleet studies of Ford’s BeltMinder, Swedish and European observations of “smart” reminders, the United States field-data programme, controlled on-road experiments with part-time belt users, and fatality-risk analysis at national scale. Almost none of it is about buses. The bus-specific literature amounts to observation studies that establish how bad compliance is, and, as far as we can establish, exactly one trial that actually intervened and measured the change. That trial is Park et al.

So the honest description of the evidence base is this: the mechanism is very well established in cars, and it has been tested once on a bus, where it worked, and worked more dramatically than in any car study — because there was so much further to move. Anybody arguing that reminders on buses are unproven is arguing against the only trial there is.

In cars, this argument was settled twenty years ago

The car evidence is worth laying out properly, because it is what gives the Korean result its plausibility, and because it is the evidence Australian regulators already accepted when they wrote the car rule. Note the units column — these studies do not measure the same quantity, and the differences matter more than the headline figures.

Do not add these numbers up. One is a change in fatality risk, one is a relative increase inside a recruited sample of prior non-compliers, one is an absolute change against a regulated floor, and one is a pair of observed levels on a single route. Quoting any single figure as “the” effect of a seatbelt reminder over-reads all of them, including the 49.4 points from Korea.

Read down the design column and the pattern is consistent, and it is not about the size of the number. Every intervention that worked shares two features: a switch detects the true state of the buckle, and something the occupant would rather avoid continues until that state changes. Every intervention that only supplies information — a sign, a briefing, a recorded message, a public campaign — performs poorly and inconsistently.

The mechanism, stated plainly

A persistent reminder does not work by persuading anyone that seatbelts are important. It works because it changes the arithmetic the passenger is actually doing. Not buckling up has a cost that is deferred and probabilistic — injury in a crash that almost certainly will not happen on this trip — set against a benefit that is immediate and certain: comfort, or a nap, or not being bothered. People reliably choose the immediate certain thing. An alert at your own seat that will not stop converts the deferred cost into an immediate one, and buckling up becomes the fastest available route to being left alone. Compliance stops depending on whether the passenger agrees with you. That pattern is not specific to seatbelts: our systematic review of which behavioural mechanisms actually change passenger behaviour traces it across thirteen fields, and finds that interventions altering what happens to someone produce large effects while interventions telling someone what to do produce small ones that fade.

This is also why who hears it matters. Asked directly, part-time belt users and non-users say they are more likely to buckle in response to auditory and haptic reminders than to visual ones. That is stated intention rather than measured behaviour, so it is weak evidence — but it is the only direct comparison of alert modality we could find, and it points the same way as the gap between Korea’s visual-only 59 per cent and the near-universal compliance seen in cars with a chime that persists.

Which is exactly why ADR 4/05 makes it compulsory in cars

Australia did not sit this out. When the United Nations updated UN Regulation 16 in 2008 to require a seatbelt reminder at the driver’s position, Australia adopted it through ADR 4/05 – Seatbelts. The rule text is unambiguous:

The driver seating position of the M1 category of vehicles shall be equipped with a safety-belt reminder satisfying the requirements of this Regulation.

ADR 4/05 – Seatbelts, Appendix A (UN ECE R16), clause 8.4.1. Carried forward unchanged into ADR 4/06.

And it is not a token requirement. The rule specifies an escalating reminder with a defined floor:

• First level: a visual warning active for at least 4 seconds when the driver’s belt is unfastened and the ignition is engaged.
• Second level: a visual and audible signal, active for at least 30 seconds, once the belt is unfastened, the vehicle is in normal operation, and any of: distance driven over a threshold no higher than 500 m, speed over a threshold no higher than 25 km/h, or engine-running time over a threshold no higher than 60 seconds.
• Hard to defeat: if short-term deactivation is offered at all, it “shall be more difficult to deactivate the safety-belt reminder than buckling the safety-belt on and off”, and it must reactivate after the ignition has been off for 30 minutes.

Read that list again with the Korean study in mind. Sensing, escalation, persistence, and a deliberate refusal to let the occupant opt out cheaply. Australian regulators wrote all four of those into a car rule, and the Regulation Impact Statement gave the justification in a single sentence:

The safety benefits of installing a driver’s seat belt reminder, over the life of a vehicle, outweigh the installation costs.

Regulation Impact Statement, amendments to ADR 4/05 – Seatbelts.

Nothing in that sentence is about cars. It is a statement about seatbelt reminders, the lifetime safety benefit of fitting one, and the cost of doing so. The regulation that contains it makes the same point about its own scope, in a footnote to the very clause that mandates the reminder:

While the current requirements for a safety-belt reminder is limited to the driver’s seat of vehicles category M1, it is understood that the scope of this Regulation will be extended to vehicles of other categories and to other seating positions.

ADR 4/05 and ADR 4/06, Appendix A, footnote to clause 8.4.1.

Other categories. Other seating positions. In 2008 the drafters wrote down that this was coming. Europe has since done it, requiring seatbelt reminders for front and rear seats under the General Safety Regulation; the United States finalised a rear-seat reminder rule in late 2024. Australia has extended it to neither — and for buses it went further than not extending it.

And ADR 68/01 does not require it in buses

ADR 68/01 – Occupant Protection in Buses was approved in December 2024, eighteen months after the Greta coach rollover, and it applies to new bus models from 1 November 2026 and to all new builds from 1 November 2027. It is a genuine improvement on what preceded it. It requires three things:

• A label on the back of each seat reading “Fasten seatbelt while seated”.
• A lit sign on each cabin bulkhead reading “Fasten seatbelt”.
• An announcement saying “Please fasten seatbelt while seated” when the last door closes.

It also closes a loophole that had allowed some coaches to avoid fitting passenger seatbelts at all, which matters. But on the question of detection, the peak industry body states the position without ambiguity:

No, seatbelt monitoring systems that detect when someone is in the seat are not required. They may, however, still be fitted as an option by the supplier or customer, though not mandated.

Bus Industry Confederation, ADR 68/01 & Seatbelts: Frequently Asked Questions, April 2025, answering whether monitoring systems are required.

The carve-out is explicit, and it is in the car rule

This is the part that is easy to miss, and it is the sharpest thing in the whole regulatory picture. ADR 4/05 and ADR 4/06 apply to omnibuses as well as cars — the applicability table lists MD1 through MD4 and ME alongside MA and MB. But both rules then say this:

Appendix A, Section 8 (Requirements concerning the installation in the vehicle), does not apply in relation to the installation of seatbelts for seats other than the driver’s seat of omnibuses complying with the Australian Design Rule 68/… – Occupant Protection in Buses.

ADR 4/06 – Seatbelts, clause 3. Materially identical wording appears in ADR 4/05.

Appendix A Section 8 is the section containing clause 8.4, Safety-belt reminder equipment. So the seatbelt rule that mandates reminders expressly switches that section off for every seat on a bus except the driver’s — and the bus rule it defers to, ADR 68/01, does not pick the requirement up. The gap is not an oversight in drafting. It is a deliberate hand-off between two rules, and nobody catches the ball.

ADR 4/05’s reminder mandate applies to the driver’s seating position of M1 vehicles. Australia has not adopted the later UN extension to passenger seats, which the EU and the United States have now both done in some form.

What the difference looks like across an actual trip

The table makes the contrast look like a list of features. It is better understood as a question of coverage in time. Put a realistic trip on an axis: a passenger boards unbelted, buckles up after about ninety seconds, then unbuckles again half an hour in to get comfortable.

The second window is the one that matters. Half an hour into the trip the ADR 68/01 equipment has nothing left to say — its announcement fired once, at the door close, and the lit sign says the same thing to everyone regardless of who is buckled. A system that senses the buckle notices, and keeps noticing until the belt is latched.

Every term of the car justification is stronger for a bus

The Regulation Impact Statement test has three moving parts: the safety benefit, the life of the vehicle, and the installation cost. Apply each one to a coach instead of a car and every term moves in favour of fitting the system, not against it.

The reasoning behind the car rule has not been rebutted for heavy vehicles. It has simply never been extended to them.

The Australian consequence, measured once

Australia has its own data point on what an unbelted coach looks like, and it did not come from a trial. At 11.17 pm on 11 June 2023 a coach carrying a driver and 35 passengers rolled at the Branxton interchange near Greta, New South Wales. Every seat had a working three-point belt; all were inspected after the crash and none had been cut during the rescue. Six passengers were wearing one.

That pattern is not peculiar to one crash. Reviewing 128 injuries across three Swedish coach rollovers, Albertsson and colleagues found seatbelts to be the single most effective countermeasure for preventing ejection and reducing injury severity. The countermeasure was present at Greta. It was fitted at every seat, and it worked when it was used. What was missing was anything that noticed it was not being used.

Three controls were meant to prevent that, and all three are information-only measures of exactly the type the car evidence says do not work. There was a sign, and 17 per cent compliance. There was an operator policy requiring the driver to brief passengers, the driver did not, and nothing in the system detected the omission. And belt-wearing is legally mandatory, against an enforcement record of one passenger fine issued across New South Wales in three years. ADR 68/01 adds a fourth measure of the same type.

What the Korean pilot left on the table

We should be as hard on the study as on the rule. 59.0 per cent is a large improvement and a poor absolute result. Four in ten passengers still travelled unbelted on a bus that knew they were unbelted and was telling them so.

The likely reason is in the first table: the alert was visual only. A light at your seat is easy to not look at, and trivially easy to ignore once you have looked. Nothing in the car literature suggests a light alone should be expected to perform like a chime that persists — the interventions with the strongest measured effects all make a sound, and keep making it. The Korean design also placed its full-information display at the driver’s position, where the person who can see the problem is the one person who cannot get up and do anything about it. We have since gone through that car evidence study by study in a systematic review of per-seat buckle sensing with a persistent localised audible alert, which is the configuration the Korean design stopped one step short of.

That is a gap a study could close rather than argue about, so we have written the study. Our protocol for a four-arm factorial randomised controlled trial separates the ADR 68/01 requirements from per-seat buckle sensing and measures each on its own, which is the only design that can say which of the two is doing the work.

What an Australian operator can fit today

ADR 68/01 sets a floor, and the Korean study describes a ceiling nobody has reached yet. In practice an operator specifying a bus now is choosing between three distinct things, and they are often confused with one another. The three columns below describe distinct, complementary functions.

The ADR 68/01 Kit is a separate product from the Seatbelt Warning System: the kit delivers compliance, the warning system delivers detection. Two points follow from that table, and they are the practical takeaways from the whole Korean result.

The compliance kit and the warning system solve different problems. The ADR 68/01 Kit — two cabin speakers, an illuminated bulkhead sign, a control module and a seat-back label for every seat — satisfies the rule, and it prompts people at the moment the decision is actually made: seated, doors shut, before the vehicle moves. That is the best moment there is for an information-only measure. It is still an information-only measure. Nothing in it knows whether anyone complied.

The extra buzzer and LED at each seat are the part the Korean study could not test. KATRI used a light. The available Australian configuration adds an 80 dB buzzer in the row and a bright, aisle-visible LED at the seat, so non-compliance is both audible and locatable — and so the alert reaches the people best placed to resolve it, including the passengers seated in front, whose own seats are certified on the assumption that the person behind them is restrained. On the car evidence that should outperform a light. Nobody has yet measured by how much on a bus.

Worth noting for anyone weighing timing: ADRs are not retrospective, existing buses need no retrofit, and the NSW investigator estimated it would take at least 25 years for recently introduced technologies to reach most of the fleet through replacement alone. Meanwhile the Heavy Vehicle National Law imposes a primary duty on operators now, on vehicles that will never be covered by ADR 68/01 at all.

How to read this evidence honestly

Four limits define the scope of this evidence.

• The Korean figure is a single-site comparison. One route, one country, one operator, a small number of vehicles, and an observational rather than randomised design. Passengers on a bus known to be part of a pilot may behave differently. Treat 49.4 points as evidence that the mechanism works on a bus, not as a forecast for your fleet.
• The strongest effect-size evidence is still from cars. Extending it to bus passengers is a reasonable inference from a shared behavioural mechanism. It is an inference.
• Habituation over a 25-year service life is uncharacterised. The controlled studies ran for weeks. What a buzzer means to a commuter in year eleven is genuinely unknown, and it is the main reason occupancy gating is not optional.
• Ejection risk is multi-causal. At Greta, glazing, seat design, the road environment and driver health all contributed. Raising wearing rates addresses one factor among several, and no single measure would have prevented every death.

Strip all of that back and what is left is narrow and hard to argue with. Australia accepted, in 2008, that the lifetime safety benefit of a seatbelt reminder outweighs the cost of fitting one, and wrote it into a rule for a population already buckling up at 97 per cent. The same reasoning applied to a population buckling up at 7 to 45 per cent produces a stronger answer on every term, not a weaker one. And the one time anybody actually built the thing and measured it on a bus, six times as many passengers wore a belt.

The evidence behind each step of that argument is set out in full in our research papers.

Regulatory context: Review our detailed analysis on why ADR 68/01 has no seatbelt warning system requirement and what steps operators must take.

Frequently asked questions

What did the Korean seatbelt reminder study find?

The Korea Automobile Testing and Research Institute fitted an interurban bus in commercial service between Seoul and Suwon with a seat-occupancy sensor and a buckle sensor at every passenger seat, a visual warning device at each seat and a monitoring display at the driver’s position. Observed seatbelt wearing was 9.6 per cent on a bus without the system and 59.0 per cent on the bus with it, a difference of 49.4 percentage points (Park et al., 2023, ESV Paper 23-0190).

Is this the only study of seatbelt reminders on buses?

It is the only published trial of a per-seat seatbelt reminder on a bus that we have been able to locate. The wider seatbelt reminder literature — more than two decades of observational fleet studies, on-road experiments and fatality analyses — is almost entirely about passenger cars. That is why a single bus pilot carries unusual weight.

Does ADR 68/01 require a seatbelt warning system on buses?

No. ADR 68/01 requires a label on the back of each seat, a lit “Fasten seatbelt” sign on the cabin bulkhead, and one recorded announcement when the last door closes. It does not require a buckle sensor, an occupancy sensor or any per-seat detection. The Bus Industry Confederation states it plainly: seatbelt monitoring systems that detect when someone is in the seat are not required.

Why are seatbelt reminders compulsory in cars but not on buses?

ADR 4/05 and its successor ADR 4/06 require a seatbelt reminder at the driver’s seating position of a passenger car, and both rules expressly disapply the section containing that requirement to seats other than the driver’s seat of an omnibus. The reasoning used to justify the car rule — that the safety benefit over a vehicle’s life outweighs the fitting cost — has never been rebutted for buses. It was simply never extended to them.

How much better could a bus system do than 59 per cent?

Nobody has measured it. The Korean pilot used a visual indicator only, and 41 per cent of passengers still travelled unbelted. Car research consistently finds that people who do not always buckle up respond more strongly to a sound that persists than to a light, which is the basis for pairing per-seat buckle sensing with an audible alert at the unbelted seat. That combination has not yet been trialled on a bus.

Do buses already in service have to be retrofitted?

No. Australian Design Rules are not retrospective, so ADR 68/01 applies to new-model buses from 1 November 2026 and all new builds from 1 November 2027. Existing buses are not required to be retrofitted, and the NSW investigator estimated it would take at least 25 years for recently introduced technologies to reach most of the fleet through replacement alone.

Related research

• What actually changes passenger behaviour — fifteen behavioural mechanisms across thirteen fields, tested against the conditions inside a bus cabin.
• Per-seat buckle sensing with a persistent localised audible alert — the passenger car evidence reviewed study by study, and what transfers to a bus.
• Information or consequence? — a four-arm factorial trial protocol separating the ADR 68/01 requirements from per-seat buckle sensing.

Related reading

• ADR 68/01 compliance guide — what the rule requires, who is exempt, and the 2026 and 2027 dates.
• Bus seatbelt monitoring systems: buyer’s guide — how to compare detection systems on specification rather than claim.
• Retrofitting an existing fleet — why ADRs being non-retrospective is an operator problem, not a reprieve.
• Occupancy detection without cameras — why passive switches beat imaging for this job.
• Australian bus seatbelt regulation — how the ADRs, state law and the HVNL fit together.
• Seatbelt fines and penalties by state — why enforcement does almost no work on a bus.

Hardware operational guide: Discover how driver consoles and sensors coordinate in our guide to on-vehicle safety monitoring systems.

References

1. Park J, et al. The effectiveness of seat belt reminders by vehicle types — analysing the result of the pilot project of an interurban bus with SBR. Proceedings of the 27th International Technical Conference on the Enhanced Safety of Vehicles, Paper 23-0190. Korea Automobile Testing and Research Institute; 2023. Full paper (PDF)
2. Vehicle Standard (Australian Design Rule 4/05 – Seatbelts) 2012, Authorised Version F2019C00042. Clause 3 and Appendix A (UN ECE R16/06) clauses 8.4.1 to 8.4.2.6.1. legislation.gov.au
3. Vehicle Standard (Australian Design Rule 4/06 – Seatbelts) 2018, F2019L00026. Clause 3, including the omnibus carve-out of Appendix A Section 8, and Appendix A clause 8.4. legislation.gov.au
4. Department of Infrastructure and Transport. Amendments to Vehicle Standard (Australian Design Rule 4/05 – Seatbelts) 2006: Regulation Impact Statement. Office of Impact Analysis. oia.pmc.gov.au
5. Bus Industry Confederation. ADR 68/01 & Seatbelts: Frequently Asked Questions, April 2025. bic.asn.au (PDF)
6. Vehicle Standard (Australian Design Rule 68/01 – Occupant Protection in Buses). Approved December 2024. See also Department of Infrastructure, Transport, Regional Development, Communications and the Arts, New bus design rules set to reduce fatalities. infrastructure.gov.au
7. Office of Transport Safety Investigations (NSW). Bus safety investigation report: rollover of coach 4666MO, Greta, NSW, 11 June 2023. Source for the Greta casualty and belt-use figures, Australian bus and car wearing rates, the NSW enforcement record, the seat-anchorage dual-loading finding, and the 25-year fleet turnover estimate.
8. Williams AF, Wells JK, Farmer CM. Effectiveness of Ford’s belt reminder system in increasing seat belt use. Injury Prevention. 2002;8(4):293–296.
9. Krafft M, Kullgren A, Lie A, Tingvall C. The use of seat belts in cars with smart seat belt reminders: results of an observational study. Traffic Injury Prevention. 2006;7(2):125–129.
10. Freedman M, Levi S, Zador P, Lopdell J, Bergeron E. The effectiveness of enhanced seat belt reminder systems: observational field data collection methodology and findings. DOT HS 810 844. NHTSA; December 2007.
11. Lie A, Krafft M, Kullgren A, Tingvall C. Intelligent seat belt reminders: do they change driver seat belt use in Europe? Traffic Injury Prevention. 2008;9(5):446–449.
12. Farmer CM, Wells JK. Effect of enhanced seat belt reminders on driver fatality risk. Journal of Safety Research. 2010;41(1):53–57.
13. Kidd DG, Singer J. The effects of persistent audible seat belt reminders and a speed-limiting interlock on the seat belt use of drivers who do not always use a seat belt. Journal of Safety Research. November 2019.
14. Kidd DG, McCartt AT, Oesch NJ. Attitudes toward seat belt use and in-vehicle technologies for encouraging belt use. Traffic Injury Prevention. 2014;15(1):10–17.
15. Anund A, Forward S, Sjörs Dahlman A. Seat belt usage in buses: an observation study of usage and travellers’ perspectives. Accident Analysis and Prevention. 2023;190:107138.
16. Albertsson P, Falkmer T, Kirk A, Mayrhofer E, Björnstig U. Case study: 128 injured in rollover coach crashes in Sweden — injury outcome, mechanisms and possible effects of seat belts. Safety Science. 2006;44(2):87–109.
17. Regulation (EU) 2019/2144 of the European Parliament and of the Council of 27 November 2019 on type-approval requirements for motor vehicles as regards their general safety and the protection of vehicle occupants and vulnerable road users.
18. National Highway Traffic Safety Administration. Federal Motor Vehicle Safety Standards; Occupant Crash Protection, Seat Belt Reminder Systems, Controls and Displays. Final rule, 49 CFR Part 571, Docket No. NHTSA-2024-0071. Federal Register; January 2025.
19. Cvach M. Monitor alarm fatigue: an integrative review. Biomedical Instrumentation & Technology. 2012;46(4):268–277.
20. AITSC. Seatbelt Warning System, ADR 68/01 Kit and per-seat buzzer and LED indicator: technical summary. 2026.