Halts, LULD and Circuit Breakers

Two guardrails, two speeds

U.S. equities have two independent sets of brakes. **Market-wide circuit breakers** watch the S&P 500 against its prior close and stop everything at once: Level 1 at −7% and Level 2 at −13%, each a 15-minute halt, then Level 3 at −20%, which closes the market for the remainder of the day. Levels 1 and 2 cannot fire after **3:25 pm ET** — the rule deliberately keeps a mid-cross halt away from the closing auction, where the day’s benchmark prices are being computed. **Limit up–limit down (LULD)** works on one stock at a time. Each covered stock carries a band around a reference price — a percentage of the average price over the last five minutes — and the band is tighter for liquid names and wider for cheap or thin ones. If quotes sit outside the band and do not trade back inside within **15 seconds**, the stock enters a **limit state**: trading may continue, but no print is allowed outside the band. Fifteen seconds of that, and the name takes a **5-minute trading pause**. The asymmetry in the bands is the point rather than an accident. A Tier 1 stock above $3 gets roughly ±5%, a Tier 2 stock above $3 roughly ±10%, and a stock at or below $3 wider bands still. A $2 stock moves 5% on completely ordinary flow; a band that tight would pause it all day, and a guardrail that fires constantly is a guardrail nobody can trade around. The levels, from a prior close of 6,000 — Level 1 · −7%: 6,000 × 0.93 = 5,580 → 15-minute halt · Level 2 · −13%: 6,000 × 0.87 = 5,220 → 15-minute halt ← · Level 3 · −20%: 6,000 × 0.80 = 4,800 → market closes for the day · Cut-off: Levels 1 and 2 cannot fire after 3:25 pm ET

What happens to your order, and why the reopen is the dangerous part

A halt freezes the **matching**, not the **orders**. Exchanges stop printing executions in the halted state, and the fate of what you already sent is venue- and order-specific: resting limit orders are generally carried into the reopening process, a market order that cannot execute may be cancelled, and some venues deliberately restrict cancellation so the book is stable enough to compute a price from. The conservative assumption — the one that keeps you out of trouble — is that you cannot manage the position at all until the name reopens: no cancelling, no stop, no market order. Reopening is an **auction** (M11). Halt auctions pool everything that accumulated while the stock was frozen and print one price, chosen to clear as much volume as possible. That price is not the price you last saw. A stock halted at 40 on bad news can reopen at 28 in a single tick, because the buyers who were there at 40 were there before they had the news. Every resting order becomes a candidate to fill at the reopening price — which is why "my stop is at 38" is a statement about your intent, not about your outcome. Three habits survive the evidence. Never leave a market or stop order in a name you know is halted. If you have to react, wait for the reopening auction to print and a spread to reappear. And treat a halted stock as untradeable for your purposes, because the alternative is discovering the reopen price with your own fill. A halt is a pause, not a floor. It gives participants time to reprice; it does not promise that the repriced level is anywhere near the last trade.

Two more triggers: the short sale restriction and the news halt

Circuit breakers and LULD bands are the guardrails everyone knows. Two other mechanisms switch on without a press release, and both change what your order can do. **The short sale restriction** engages when a stock falls 10% or more from the prior close, and it stays engaged for the rest of that day and all of the next. While it is on, short sales may only be executed at a price above the best bid — the alternative uptick rule. The practical effect is narrow and worth knowing: market makers and ordinary sellers are unaffected, but a short sale that would have printed at the bid has to improve on it, so short selling gets marginally more expensive exactly when a stock is falling hardest. It was designed after the 2008 crisis and introduced in 2010, and it is the reason a fast decline in a single name can look different from a fast decline in the index. **The news halt** is a different animal entirely, because it is discretionary rather than arithmetic. A listed company is required to tell the exchange before it releases material news, and the exchange can pause trading for a few minutes to let the information disseminate evenly. There is no 7% threshold and no published formula — which means that during a news halt you cannot infer anything from the fact of the halt itself. The band mechanisms tell you a price moved a given distance. A news halt tells you only that a document is incoming. What each trigger tells you — LULD band breach: a price moved a stated distance in seconds, and mechanics caused the pause · Market-wide breaker: the index fell 7, 13 or 20 percent from the prior close · Short sale restriction: the stock is 10% below yesterday’s close; you are in a rule, not a halt ← · News halt: a document is about to be public — the cause is unknown, only the timing is ← The distinction that matters for a learner: three of these four pause trading, and only two of them carry information about price. The short sale restriction is not a halt at all, and the news halt tells you nothing you can trade on except that the information is coming evenly to everyone.

Four halts in eight days: March 2020 in practice

The market-wide circuit breakers were redesigned after the 2010 Flash Crash and their current form — the 7%, 13% and 20% levels measured against the prior close of the S&P 500, with 15-minute pauses below the last level and a session close above it — was set in 2012. For eight years those levels were never reached. In March 2020 they were triggered **four times in eight trading days**, which turned a set of rules that had been theoretical into the most heavily documented test of them the market has had. What those days showed is worth separating into three findings. First, the halt worked as specified: trading paused for fifteen minutes at the 7% level and resumed, and the day continued. Second, the pause did not change the direction — each reopening auction cleared near or below where the market had stopped, and in several cases the decline continued immediately, because the information that had driven the selling had not changed during the pause. Third, and most importantly for an individual position, the **pause was not protection**: an order resting in the book at a price was still there when the session resumed, and a stop that was triggered on the way into the halt was executed on the way out at whatever the reopening auction cleared at. The mechanism does what it says — it interrupts — and the interruption is not a floor. The other thing March 2020 demonstrated was how the two guardrails interact with everything downstream of the equity market. Individual stocks hit their bands and paused repeatedly while the index-level breakers were also firing, options markets repriced with the underlying halted, and margin requirements rose as volatility did — which is the same collateral mechanic that the GameStop case describes from the other direction. A halt is an interruption in one market, and the claims on a position do not pause with it. • The current 7/13/20% framework dates from 2012 and was first used four times in March 2020. • The pause interrupts; the reopening auction sets the price, and it is usually near or below the halt level. • Resting orders and triggered stops are still live through the pause, and they execute into the reopen. • Options markets, margin and margin calls keep moving while the equity print is stopped.

A halt is not the only way to stop a market

The two mechanisms in this lesson — circuit breakers and limit up–limit down — are the American answer to an old question: what should happen when a price moves faster than anyone can react? The answer elsewhere is often a **daily price limit**, which is a structurally different instrument. Mainland Chinese exchanges stop a stock from trading beyond ±10% of the previous close, and ±20% for the technology-focused boards; India halts the index itself at 10%, 15% and 20%; several Asian and European venues operate similar bands. The crucial difference is that a price limit does not pause the market and then reopen it with an auction. It caps the price and lets orders accumulate behind the cap, so a limit-up stock is not frozen with an unknown price — it is trading, or failing to trade, at a known price with a visible queue of unfilled buyers. The queue is the information: a stock sealed at its limit with a large bid is a measurement of demand that the American halt mechanism never produces, because the American version hides the imbalance behind a pause and then resolves it with a single cross. That difference has real consequences for anyone holding the instrument. A U.S. halt prints a reopening price that can land anywhere; a Chinese limit-up stock pins at a price and simply cannot be bought for the rest of the session, which is why limit-up names sometimes run for several consecutive days of caps and why the order queue becomes a market in itself. Futures markets add a third flavour: the CME runs its own circuit breakers with the same three percentage steps but on its own hours and windows, and many contracts also carry daily **price limits** that expand on successive days of locked markets — a rule designed precisely because a limit that holds for two days in a row is no longer informing anyone. The result is that on a bad day the equity market, the futures market and the overseas market can all be in different states of interruption at the same moment, and a trader cannot assume that the S&P breaker firing means the E-mini is tradable. The practical lesson joins this lesson back to the ETF material in M16. When an index is locked — limit-locked constituents in one country, a halt in another — the creation and redemption mechanism that normally keeps an exchange-traded fund near its net asset value has nothing to work with, because the underlying cannot be bought or sold at any observable price. The fund keeps trading on the exchange while its basket is frozen, so its premium or discount to fair value can widen to several percent and stay there for days. That is not a malfunction; it is the arbitrage loop being interrupted at the point where the halts are, and it is the reason a position in a country ETF during a limit-down episode is not a position in the index. Both halves of the sentence matter: a halt buys time rather than prices, and the instruments built on top of the halted market inherit the interruption with a lag that the headline price does not show. Three ways to interrupt a market — U.S. circuit breaker: Pause the tape, then reopen with a single auction — the price is unknown until it prints · U.S. limit up–limit down: A single stock pauses in a banded state, then reopens by auction · Chinese or Indian price limit: Cap the price and let orders queue — the imbalance stays visible ← · Futures price limit: Locks with expanding bands on consecutive limit days by design ← The unifying test is what the interruption reveals. An auction reveals a price; a price limit reveals a queue; a news halt reveals only that a document is coming. When a market is interrupted, ask which of the three you are looking at before drawing any conclusion from the headline.

Sources

• Market-wide circuit breakers: the 7% / 13% / 20% levelsSEC — Market-wide circuit breaker rules (Rule 80B)
• Limit up–limit down: bands, limit states and the 5-minute pauseSEC / FINRA — Limit Up-Limit Down Plan
• Trading halts and the reopening auctionNasdaq — Trading halt and auction procedures