The Long Shadow of Markets: Understanding the Secondary Market Effect and Arbitrage

Prices change in milliseconds now, and their consequences can take decades to arrive. The speed of modern markets makes it easy to forget the slower half of that sentence: value and consequence keep different clocks. A price that looks right today can prove unsustainable, and the bill for today's choices often lands after the people who made them have left the room.

Two ideas from finance describe the distance between those clocks. The secondary market effect is what happens when one value is priced in two places and the two prices disagree. Arbitrage is the trade that closes the disagreement, and the reasons it sometimes can't explain why some gaps last a microsecond and others a generation.

This page is meant to be come back to. It defines the gap and the trade that closes it, works through an arbitrage with its fees, transfer time and risks, and sets out what stops arbitrageurs from closing every gap. Then it covers which venue leads a move and which follows, the timescales a gap can last over, and how far the idea reaches beyond markets. It ends with a reference table, a case study of bitcoin's Korean premium in 2017 and 2018, where the argument fails, and a checklist.

#The Secondary Market Effect

The secondary market effect is the gap between what the same value costs in two places. "Two places" can be two exchanges, two countries, two currencies, or two instruments that track the same underlying thing: a share and its depositary receipt, an ETF and the basket it holds, a perpetual and the spot market it follows. The gap usually stays invisible until someone puts the two prices side by side.

It exists because markets never move in perfect step:

• Venues are separate auctions. Each order book clears its own buyers and sellers, so a burst of demand on one venue moves its price before anyone carries the move across.
• Liquidity is uneven. The same order moves a thin book further than a deep one.
• Information arrives at different times. A participant on one venue may see the news, or the other venue's price, later than someone on another.
• Frictions stand in the way. Fees, transfer times, capital controls, regulation, short-selling limits and trading hours all make moving value from one place to the other slow or costly.

In the most liquid markets the gap is tiny and short-lived, because automated traders find it and close it within microseconds. In slower, more structural markets such as housing and currencies, the distance between a price and the value it stands for can last for years. The mechanism is the same at every speed; what changes is how hard the gap is to trade.

#Arbitrage

Arbitrage is the trade that profits from the gap. In its cleanest form you buy where the value is cheap and sell where it's dear at the same moment, and keep the difference. You need no view about where the price goes next, because you hold no net position: the purchase and the sale cancel. Your trading also narrows the gap, since the buying lifts the cheap price and the selling presses on the dear one, so arbitrage is the force that drags venues back into line.

The clean form is rarer than the textbook suggests, because every leg costs something. Write for the price on the cheap venue, on the dear one, and for the two venues' fees as a share of the trade, and for the cost of moving one unit between them. Buying a unit on A and selling it on B makes

and the trade pays only when the gap, as a share of price, beats the round-trip cost:

So what arbitrage enforces is a band: the two prices can sit anywhere within of each other. Inside the band nobody can profit from the gap and it wanders as it likes; at the band's edge arbitrageurs trade, and their trading stops it from going further.

That band is why two venues' prices for the same thing can differ by a few hundredths of a percent all day and nobody fixes it: the difference is smaller than the cost of fixing it. It's also why cheaper trading tightens markets. Every fall in fees or transfer costs narrows the band, and the prices with it.

Arbitrage has effects beyond closing gaps. It moves volume towards the venues where it's cheapest to trade. It carries a move that starts on one venue to all the others, so a price can change on a venue where nobody has traded much at all. And because closing one gap can open another, in a related instrument or on a venue that reacts more slowly, the gaps never stop appearing.

#A Worked Arbitrage

Take bitcoin at 10,000 on venue A and 10,300 on venue B, a gap of 3%. Both venues charge a 0.10% taker fee, and withdrawing bitcoin from A costs 0.0005 BTC, about $5.

The naive version buys one bitcoin on A, withdraws it, waits for the transfer, and sells on B:

That's $274.70 on $10,010 laid out, about 2.7%, if B's price is still 10,300 when the coin arrives. It may not be. A bitcoin transfer needs confirmations, and an hour in transit is ordinary. If B's price falls 4% in that hour, because the market moves or because other arbitrageurs got there first, the trade loses money. The naive version is a bet on B's price with a fee attached.

Professional arbitrage removes the transit risk by holding inventory on both venues in advance: dollars on A, bitcoin on B. When the gap opens, it buys on A and sells on B in the same instant, and locks in

with no exposure to the price while anything is moving, before the cost of the rebalancing transfers that follow. The transfers happen afterwards, to rebalance the inventory: bitcoin from A to B, and dollars from B back to A. That converts the price risk into three other costs:

• Capital. The arbitrageur has to keep enough idle money and coins on every venue to trade the size they want, earning nothing while they wait.
• Venue risk. Inventory held on an exchange is exposed to that exchange halting withdrawals or failing.
• The slow leg. Rebalancing has to finish before the inventory runs out. If one leg takes days, the arbitrageur can only trade the gap as fast as that leg refills them.

The slow leg is usually money. Coins move between crypto venues in minutes; dollars, euros and won move through banks, sometimes across borders, and sometimes only with permission. The case study below is what happens when that leg nearly stops.

#Who Closes the Gap, and What Stops Them

The textbook arbitrageur needs no capital and takes no risk. Shleifer and Vishny's The Limits of Arbitrage (1997) made the case that almost no real arbitrage is like that. It needs capital, it carries risk, and it's mostly done by a small number of specialised traders using other people's money. Those investors judge them on results, so when a gap widens before it closes and the position shows a loss, the money can be withdrawn at exactly the moment the opportunity is best. A gap can therefore get wider because the people who would close it are being forced to sell.

Long-Term Capital Management showed the argument at full scale the next year. The fund was built on convergence trades, pairs of related bonds and securities whose prices it expected to come together. In 1998 the gaps widened instead, its capital ran out before they closed, and a group of fourteen banks organised by the Federal Reserve Bank of New York put in about $3.6 billion to take the fund over and wind it down. Many of the spreads narrowed again later, after the fund could no longer wait for them.

So the things that keep a gap open are the things that make arbitrage costly or risky:

Every row on that table widens the band. The widest gaps sit behind several at once.

#Venue Leadership

Moves don't start everywhere at once. Size arrives on one venue, its price moves first, and the other venues follow as arbitrage drags their prices into line. The follower's price moves with the leader's, but its volume doesn't have to, because arbitrageurs can carry the price across with comparatively little trade.

In crypto the leader is often a derivatives venue, because leverage concentrates activity there. In March 2020 Binance's BTCUSDT perpetual traded about seven times Coinbase's bitcoin volume across 12 and 13 March, and at the worst hour it closed 8% below Coinbase's spot price. Part 5's case study works through it. That's a tendency, and sometimes spot leads, as it can when fund flows arrive in the US morning.

The practical consequence is that one venue's chart can be a partial record. If the move started elsewhere, the venue you're watching shows the price without the volume that made it. Turn the Candles Off covers how to read volume with that in mind, and Tool 01 meets the same problem when it looks for a volume climax: a venue that followed the move can show none.

#From Microseconds to Decades: The Timescale of the Gap

How long a gap lasts depends on how hard it is to trade. Where both sides can be bought and sold instantly and cheaply, it closes in microseconds. Where one side can't be sold short, or the money can't move, it can last a generation.

1. Microseconds to seconds.
   Where: the same share, currency pair or coin on several electronic venues.
   Example: when a share's price moves on one US exchange, the other exchanges' quotes are stale for a few microseconds, and the fastest firms trade against them before they update.
   What closes it: speed. The advantage goes almost entirely to whoever has the fastest connections.
2. Hours to a weekend.
   Where: ETFs against home markets that are closed, futures against spot markets in other time zones.
   Example: a Japanese equity ETF trades in New York after Tokyo has closed. US news moves the ETF away from the last closing prices of the shares it holds, and the gap can't be arbitraged until Tokyo reopens.
   What closes it: the other market opening.
3. Days to months.
   Where: cross-listed companies, markets behind capital controls, illiquid small caps.
   Example: a mining company listed in London and Johannesburg can hold a persistent gap between its two prices when currencies swing or the two sets of investors feel differently about it. Bitcoin in Korea, below, held a gap of roughly 25% to nearly 60% from late December 2017 into January 2018.
   What closes it: a change in the friction, or enough capital getting through it.
4. Years to decades.
   Where: housing, currencies, sovereign debt.
   Example: a decade of very low interest rates can lift house prices well beyond what local wages and rents support, and it may take a change in credit conditions before they come back.
   What closes it: the conditions holding the gap open changing. Nobody can short a house and wait, so no arbitrageur closes it, and when it closes it tends to close all at once.

The last tier does the most damage: the imbalances there build quietly, outlast anyone trading in seconds, and wait for the conditions behind them to change.

#Beyond Markets: The Delayed Reckoning

The same logic reaches past trading. The US dollar settles much of world trade, so a monetary decision in Washington shows up in other countries' import prices, debts and reserves, often years later and long after the decision has dropped out of the news. A housing boom fuelled by cheap credit can run for years with the gap between price and what incomes support visible from early on, and close only when the credit does.

Politics runs on the same delay between a decision and the public's experience of it. A trade agreement, a regulation or a war can show its benefits within an electoral term and its costs a decade later. In that setting the "arbitrage" is usually over perception: claiming the early benefits before the later costs arrive. As in housing, nobody can sell the dear side short, so there's no trade that closes the gap early, and the reckoning arrives on its own schedule.

#Reference

#Case Study: The Kimchi Premium

Between December 2017 and February 2018 bitcoin cost far more in South Korea than anywhere else. Makarov and Schoar studied trade-level data from 34 exchanges in 19 countries for their paper Trading and Arbitrage in Cryptocurrency Markets (Journal of Financial Economics, 2020). They found the daily average price ratio between Korea and the US was more than 15% over those months and reached 40% on several days. The press called it the "kimchi premium". My own first real trading win was a cross-exchange arbitrage at the December 2017 peak, in the same weeks.

On one Korean exchange's daily closes the premium went higher. Converting Upbit's won price to dollars at each day's exchange rate and setting it against Coinbase gives this:

The premium ran under 5% for most of November and touched 11% at the month's end. It jumped to 25% on 7 December and was close to nothing again by 12 December. From 20 December it held between 24% and 34% into the new year, then peaked at 58.7% on 8 January, when Upbit's price was about 23,800 against Coinbase's 14,994.

A gap that size should be a gift to arbitrageurs, and Makarov and Schoar estimate that at least $2 billion of potential arbitrage profit was left on the table between December 2017 and February 2018. Their evidence on why it survived is in where the gaps were. Within a region, prices across exchanges stayed close; across regions they didn't. And the price of bitcoin in ether was almost the same in Korea as in the US, about 3% apart, while the price of bitcoin in won was more than 20% above the dollar price. The coins moved freely, and the money couldn't.

The trade's loop shows where it broke. Dollars buy bitcoin on Coinbase, the bitcoin moves to Upbit on-chain, it sells there for won, and then the won has to become dollars again and leave Korea before the loop can run a second time. Korea's capital controls sat on that last leg. Anyone could make the trade once; making it at size, again and again, meant getting won out of the country, which was slow, costly and capped. Every row of the limits table above was present, but that one did the work.

The premium came down in steps as the government moved against the exchanges. On 11 January the justice minister said a bill to ban exchange trading was being prepared, as police and tax officials raided exchanges. On 23 January the financial regulator confirmed that, from 30 January, trading would require real-name bank accounts, and new accounts had been frozen in the meantime. By 2 February the premium was gone.

What the data shows is the gap and its timing. The reading I'd put on it is that the premium was Korean demand trapped behind capital controls: money that could get into bitcoin in Korea but couldn't get out to buy it cheaper abroad. When the new money stopped arriving, the premium had nothing holding it up, and arbitrage was no longer needed to close it.

#Where It Fails

• The same value isn't always the same. Two prices can differ for a reason: different contract terms, settlement dates, custody, or a venue that has stopped withdrawals. A premium on an exchange nobody can withdraw from is a price for the risk of being stuck, and buying the cheap side there doesn't lock in anything.
• Convergence isn't guaranteed. Futures converge at expiry by design. Twin shares, cross-listings and premiums behind capital controls have no date by which they must close, and they can widen further first, as LTCM found.
• Costs are larger than the visible fees. Slippage on size, withdrawal limits, banking delays, tax and the cost of idle capital all widen the real band. A gap that looks tradable from the price screen often isn't.
• Leadership changes. Derivatives venues often lead crypto moves, and sometimes spot does. Check which venue moved first for each move; don't assume it.
• The tiers are rough. Each kind of market can produce gaps far shorter or longer than its tier suggests, especially in a crash, when arbitrage capital is scarcest.

#Checklist

1. Is it really the same value? Check the contract terms, settlement, custody and whether withdrawals are open on both venues.
2. What is the round-trip cost , all in, including slippage on the size you'd trade?
3. Which leg is slow, and what can happen to the price while value is in transit?
4. Can you hold both legs at once from inventory you already have, and how much capital does that tie up?
5. What is holding this gap open, and is that thing about to change?
6. If the gap widens before it closes, how long can you hold, and who could force you out?
7. Which venue led this move, and is the chart you're watching a follower?

Since this was published: Beyond Leverage follows the same effect inside a single exchange, where three prices arithmetically bound to agree drift within the band their fees allow. The Benchmark Must Bleed asks what happens when the reference price arbitrage works against is itself controlled. Turn the Candles Off covers reading volume when the move started on another venue.

#Conclusion: Two Clocks

Every price runs on two clocks: the fast one of trades and quotes, and the slow one of the value it stands for. The secondary market effect is the gap between them, measured across places as well as time, and arbitrage is the trade that closes it, as far as costs, capital and rules allow. Where both sides can be traded freely, the gap lives for microseconds. Where something blocks one side, it lasts as long as the block does.

Seeing the gaps is useful in two ways. It tells you which price to trust, and which venue to watch, when a move starts. And it tells you which imbalances no trade will close early, so they'll wait, sometimes for years, for the conditions behind them to change.

Series note: Part 1 of Understanding Market Mechanics. Next: Part 2 · The Invisible Balance. The practical companion is Applied Market Mechanics.

---

Understanding Market Mechanics describes how markets work, published purely as education. Nothing in it is financial advice or a recommendation to buy or sell anything, and I'm not a financial adviser. Every case study is drawn from the public record and has already resolved, and nothing in a part should be read as a statement of what I hold.