A holder who reads a Binance monitoring announcement still faces most of the token's expected loss. Zhang, Kang and Sun find that tagged tokens lose 3.16% during the announcement minute. Over the next day, they lose another 8.61% relative to comparable untagged tokens.
A warning backed by private information
Binance began using the monitoring tag on 26 July 2023. The tag identifies tokens the exchange believes may cease to meet its listing criteria. Those criteria cover team commitment, development activity, liquidity, contract security and conduct. Binance reviews tagged tokens regularly and may delist them. To continue trading them, users must pass a risk quiz every 90 days. The exchange publishes tags in batches, with a millisecond time stamp.
Binance knows more than outside holders about how close a token is to removal. It speaks to project teams and can see its own internal thresholds. A tag should therefore cause a lasting price decline if it raises the probability of delisting beyond what holders had expected.
The authors assemble 167 tags from 26 announcements between July 2023 and October 2026, leaving 155 first-time tags. They match 142 of these to three untagged tokens from the same announcement. Each match has the same perpetual-futures status and the nearest logit probability of receiving a tag. A median announcement tags 5 tokens from a risk set of about 370.
The median release comes 12 seconds into the minute. During that minute, tagged tokens record a median of 368 trades; their matched peers record 5. That activity lets the authors separate trading before and after publication. The pre-release return is -0.073% across 79 tags, significant only at the 10% level. Within ten seconds of release, the return has reached -3.104% (t = -4.58), using the 119 tags that traded in that window.
Other exchanges show the same immediate fall. In the announcement minute, tagged tokens lose 4.40% on Bybit, 3.15% on Coinbase and 4.41% on KuCoin. Those subsamples contain 32, 32 and 51 tags, respectively. Another comparison uses the 20 untagged tokens most likely to receive a tag, ranked by a model fitted only to earlier announcements. The one-day gap against those peers is -9.80% (t = -5.82); the peers themselves return -0.99%.
How much remains after the announcement minute?
The residual loss is what matters to a desk. The abstract acknowledges that tagged tokens "keep falling through the first day". Two sentences later, it describes prices absorbing the warning "almost at once". Those claims pull in different directions within the abstract itself.
The body gives a clearer account. Roughly a third of the following day's loss occurs in the first minute, with the rest absorbed over the hours that follow. Relative to matched tokens, the returns are -3.14% in the minute and -8.61% over the next day. Both remain significant after a Holm adjustment across 28 return estimates.
The -1.91% five-minute step fails that adjustment, as does the -0.72% drift over the six hours before release.
There is no subsequent reversal. At 90 days, matched simple-return losses reach -24.63% when delisted tokens remain valued at their last price. Writing those tokens to zero takes the loss to -39.31%. A staggered comparison against tokens tagged later produces -23.99% by day 30 and a clean pre-trend (joint p = 0.70).
The tag forecasts removal better than the price reaction does
Against a 2% base rate among matched tokens, receiving a tag raises the probability of delisting within 90 days by 42.24 percentage points (t = 7.15). The authors acknowledge the mechanical element: Binance mostly removes tokens from the tagged pool. They argue that the tag still adds information beyond observable token characteristics. Market participants cannot know beforehand which tokens Binance will choose.
The size of the initial fall offers much less guidance. Among tagged tokens, the return from the announcement minute through the next five minutes has a coefficient of -88.35 on 90-day delisting (t = -1.49). The one-day response has the wrong sign, at 35.77 (t = 0.91). Although the initial-response estimate points in the expected direction, t = -1.49 leaves it indistinguishable from zero.
Delisting notices account for the later losses. After the announcement day, the portion of the 90-day return earned before any delisting notice is +2.47%, an insignificant result. Returns are -8.38% in the delisting window and -11.48% afterwards. Tokens eventually removed lose 53% to 58% after day 0. Tokens that remain listed gain 11.41% at 90 days, though that gain is insignificant.
The authors warn that this division conditions on the outcome. Their interpretation is that prices immediately reflect average removal risk, while investors learn which particular tokens will go mainly from Binance's subsequent notices. The figures support that interpretation.
Perps bring a faster fall and a smaller daily loss
Binance perpetual futures exist for 67.6% of the 142 matched tags. With announcement fixed effects, these tokens lose 2.34 points more during the release minute (t = -2.57). Over the following day, they lose 5.68 points less (t = 2.47). Across the minute and day together, perp tokens lose 3.3 points less.
Tokens without perps therefore carry the largest remaining drift. As the paper observes, borrowing these spot tokens is often expensive or impossible. The authors offer two explanations, which their data cannot distinguish. Tokens liquid enough to have perps may receive less new information from a tag. Alternatively, holders of thinner tokens may sell into books that require a discount to take the flow.
Perps are not randomly assigned. The comparison does, however, hold within announcements and controls for past returns, volatility, trading value and age. Adding the probability of a tag keeps the estimates significant at 5%. With 26 announcements, the evidence remains an association drawn from relatively few independent events.
Selling an existing holding
For a holder, selling spot after the announcement minute avoids an average loss of 9.6% over the day, or 8.6% relative to matched tokens, as well as the delisting tail. A short seller faces a different trade. The first ten seconds are beyond reach without sub-second infrastructure, and the largest remaining drift occurs precisely in tokens where shorting requires spot borrow.
The paper deducts no spreads or trading costs. A holder's sale would also enter the burst of activity generated by the tag. Even with those qualifications, selling an existing spot position within the day is the clearest use of the findings.
We did not replicate the event study.
One result would change my view of the risk-management case. The 13 matched tags in the very first announcement did not fall. The authors suggest investors may still have been learning what a tag meant. If later batches begin to resemble that first release, the warning will have become priced in beforehand. For now, a tag on a token you own is a reason to sell within the day.