COE is back near its record. The real problem is what that does to used car supply.

Category A closed at $128,501 in the second bidding exercise of August, up 3.72% from $123,890 two weeks earlier, and within $500 of the all-time high of $129,000 set in July. Category B closed at $131,001. Category E reached $135,000.

The next exercise does not open until 7 September, with results on 9 September. That is a three week gap rather than the usual two, which means the market spends most of a month with no new price signal and a fair amount of speculation filling the space.

Almost all of that speculation is about where the premium goes next. That is the least useful question a used car buyer can ask, for two reasons. Nobody has a good record of answering it. And the price is not the thing that is actually changing underneath.

What a rising COE does not do

Start with the misconception, because it shapes a lot of bad advice.

A rising COE does not raise the paper value of a car already on the road. Paper value is the PARF rebate plus the COE rebate, and the COE rebate is calculated against the premium that was actually paid when the car was registered, not against today's figure. A car registered in 2019 on a $35,000 COE has its rebate worked out on that $35,000 no matter what happens in September.

So when someone tells you your car is worth more because COE went up, they are describing a real effect through the wrong mechanism.

What it does do

The effect is on demand, and it runs through three channels.

New car prices move with it. COE is the single largest component of the on-the-road price. When the premium rises by $4,600 in a fortnight, that lands on the invoice of every new car registered against it. Buyers who were close to the edge of affording new stop being able to.

Those buyers arrive in the used market. They do not leave the market. They move down into it, and they compete for the same cars everyone else is looking at. Used demand rises without used supply rising to meet it.

Trade-in offers rise too. Dealers replacing stock have to pay more to acquire it, because the seller's alternative options improved at the same time. That cost ends up in the retail price.

None of that is a forecast. It is the mechanism, and it operates whichever direction the premium moves.

The part nobody prices in

Here is the thing that does not reverse when COE falls.

The used car pool is not one pool. It is a supply of trade-ins arriving continuously, and only some fraction of it is worth putting in front of a buyer. That fraction has been getting smaller, and the reasons are structural rather than cyclical.

Private hire is a growing share of what comes back. The point-to-point vehicle population has gone from roughly 28,000 a decade ago to around 70,000 today, according to the Ministry of Transport. Those cars enter the fleet, work hard for five to seven years, and then come back into the used market as trade-ins. A car that has done PHV work is not necessarily a bad car, but it has covered two to three times the distance of a private car of the same age, most of it stop-start. As the PHV fleet grows as a proportion of the road, it grows as a proportion of what dealers are offered.

Mileage cannot be independently verified here. Singapore has no public odometer record. There is no equivalent of the UK's MOT mileage history that a buyer can look up. The number on the cluster is the number you are given, and the only checks available are indirect ones: service stamps, wear against the reading, the pattern across owners. In the US, where a central record does exist, Carfax reported 2.45 million vehicles with suspected rollbacks in 2025, a 14% increase year on year, which it attributes to cheap digital tampering tools. Those tools are not geographically restricted. What is restricted is the ability to catch them.

Owner turnover tells you something and it is rarely good. A car that has passed through four owners in five years has usually been passed on for a reason. We covered how to read that pattern in Mileage, owners and service records, how to read a car's history.

Stack those three filters and the arithmetic is unforgiving. A dealer that declines ex-PHV units, declines cars whose mileage does not hold up against the wear, and declines heavy-turnover histories is turning down a rising share of everything offered. The listings count across the market stays healthy. The count of units that clear those filters does not.

Why this creates urgency and price does not

A price move is reversible and unpredictable. Category B fell hard in late 2023 and recovered most of the drop inside a few weeks. Anyone who told you in October 2023 what December would look like was guessing. Anyone telling you now what happens on 9 September is doing the same.

So "buy before it goes up" is a claim nobody can support, and you should be sceptical of anyone making it, including a dealer.

A specific clean unit is different. It is one car. When it sells, it does not come back, and the pipeline behind it is not replacing that particular profile at the rate it used to. That is not a market forecast. It is inventory.

The practical version: if you have found a car that clears the checks, the risk in waiting three weeks for the September result is not that you pay more. It is that you are choosing from a shorter list, and the shortening is the trend rather than the noise.

The honest counterweight

Three things cut the other way, and they matter more than the urgency does.

A bad car bought quickly costs more than a good car bought late. A gearbox is $6,000 to $9,000. That is larger than any plausible price move between now and the end of the year. Speed is only an advantage if the checks are already done.

COE genuinely might fall. Quota is set to rise through 2026 as more of the 2016 cohort deregisters. If supply lands ahead of demand, premiums come down. It has happened, and the used market follows it down with a lag of a round or two.

Every remaining good car has other buyers looking at it. That is the same fact as the urgency, seen from the other side. It also means you have less negotiating room, which is a cost of buying into this market and worth naming rather than pretending away.

What to actually do with this

Nothing about the September result should change what car you buy. It changes only how long you can afford to deliberate on a specific unit you have already checked.

Four things before you commit to anything, in this order.

1. Ask directly whether the car was ever used for private hire or rental. It is on the record and a dealer can confirm it.
2. Check the mileage against the wear, not against the listing. Steering wheel, driver's seat bolster, pedal rubbers, brake discs. The number is easy to change and those are not.
3. Get the owner count and the dates. Transfers plus one. Cluster the dates and see whether anyone kept it.
4. Arrange an independent inspection at STA or a workshop of your choosing before you commit. Any seller who resists this has told you something.

If a car clears all four, the September COE result is largely irrelevant to whether it was a good buy. If it does not clear them, no COE result makes it one.

For the wider picture on how COE-era pricing works across car types, see PARF car or COE car, which one actually suits you.