Ministers are deciding whether learner moped riders should be barred from delivery work, or made to progress off their training licence within a set time. We insure a large number of the delivery riders that decision would affect. So we went looking in our own claims data, and found something more awkward than we expected.
Dangerous riding and learner licences are two separate problems
Two questions are being treated as one. One is about dangerous riding, meaning illegally modified e-bikes, pavements, jumped lights, and the anger about it is real and widely shared. The other is narrower: how long someone should ride on a learner licence before taking a full test. They are different questions, the evidence for each is different, and our data speaks to the second.
That second question is the one the current consultation puts on the table, as a possible time limit on Compulsory Basic Training, the one-day course that lets a rider use a moped on L-plates. Everything below is about that and nothing else.
What our claims data shows
Across two years of UK claims records, the delivery mopeds we insure made nine claims per 100 vehicle-years. The private cars we insure made 17. Delivery mopeds crash at around half the rate of ordinary private cars on the same records over the same period.
The pattern holds when you look at harm to other people. Delivery mopeds recorded one claim per 100 vehicle-years involving injury to someone else, compared with two for private cars. On damage to someone else's property the two are close, at seven compared with eight. A moped weighs a fraction of what a car weighs, and carries far less force into a collision.
Fault runs the other way from the public conversation. In 44% of moped claims, another road user caused the crash. For our car claims the figure is 28%. Riders are more often the ones being hit than the ones doing the hitting.
One thing to be clear about with these figures. They count each vehicle once a year, however far it travels. A delivery rider is out eight or nine hours a day, 35 to 40 hours a week. A private car spends most of its life parked. So the moped numbers cover far more time on the road than the car numbers do, which means the gap between them is understated here rather than overstated.
Learner riders and full-licence riders
Split the riders by licence and the picture is narrower than the debate assumes. Riders on L-plates made 10 claims per 100 vehicle-years. Riders with a full licence made eight. Both sit well below the 17 for private car drivers.
There is a gap between the two rider groups, it runs in the direction the reformers would predict, and we are not going to hide it. But it is a two-point gap in a comparison where both groups sit far closer to each other than either does to the cars around them. We have not tested what explains it, and we are not going to claim we have.
What a time limit would change
A time limit does not convert a learner rider into a full-licence rider. It requires them either to pass a full test or to stop riding on a learner licence. Only the first produces the better-trained rider the policy is aiming at, and nothing in our data tells you how many riders would take that route rather than the other one.
What claims data cannot see
We only ever see riders who are insured. Pavement riding and jumping lights do not generate a claim, so they appear nowhere in our numbers. And the riding that worries people most often comes from people working with no insurance and no licence at all, which makes them invisible to us by definition.
What is still missing
The long-term learners have been counted. Around 77,000 riders renewed their CBT rather than progressing to a full licence between March 2023 and March 2025, which is 23% of all certificates issued. That is a measure of how many people ride on a training licence. It is not a measure of how many casualties they cause, and those are different numbers.
Ours is the closest thing we have found to the second number. It shows delivery riders, including riders on L-plates, in fewer collisions than the private cars we insure, and more often on the receiving end of them. What it does not show is that taking their licences away would change any of that.
References
- DVSA and DfT, Improving moped and motorcycle training, testing and licensing. Opened 7 January 2026, closed 11 May 2026
- DVSA CBT renewal figures, March 2023 to March 2025, as reported in coverage of the joint industry consultation response. Secondary source, replace with the primary DVSA publication before publishing
- Zego claims data, UK, two years. Methodology note to follow.