# What is copy trading? How it works, the risks, and how to choose who to copy

Copy trading lets another trader’s moves run part of your money. How the copying works, what eToro, Binance, Bybit, OKX and Bitget pay the traders you follow, what research says about the results, and how to vet someone before you copy them.

- URL: https://www.ethyai.app/blog/what-is-copy-trading
- Author: Ethy Team
- Published: November 15, 2026; updated: November 15, 2026
- Publisher: Ethy AI, https://www.ethyai.app/

**The short answer:** Copy trading links part of your account to another trader, so each trade they open or close is repeated in your account automatically, sized to the money you set aside. eToro, Binance, Bybit, OKX and Bitget offer it, and the trader is paid by the platform or with a share of your profits. Their leverage becomes your risk.

## What is copy trading?

Copy trading is a way to let someone else’s trading decisions run part of your money. You pick a trader on a platform and set aside an amount. From then on, every trade they open, change or close is repeated in your account, automatically, scaled to the amount you set aside. You don’t hand the trader your money: the positions sit in your account, and you can stop copying when you choose. Vaults, covered below, are the exception: there your money joins a pool that the trader runs.

Each platform has its own name for the person you copy. eToro calls them Pro Investors, Binance and OKX lead traders, Bybit master traders and Bitget elite traders. This guide says leader. The leader gets paid for being copied, either by the platform or with a share of your profits, and that pay, together with the way your copy is sized, shapes your results as much as the leader’s skill does.

| Term | What it means |
| --- | --- |
| Leader | The trader whose trades are copied. |
| Allocation | The money you set aside for one leader. Your copy trades with it and nothing more. |
| Proportional copy | Your trade is the same share of your allocation as the leader’s trade is of their account. |
| Profit share | The cut of your copy’s profit that goes to the leader. |
| Copy stop | A loss limit on the whole copy that closes it, whatever the leader does next. |

Copy trading sits between two older ideas: trade signals, where you get an alert and place the trade yourself, and a managed account, where a manager trades your money under an agreement. A later section compares all of them.

## How does copy trading work?

Every platform runs the same loop. The leader trades, the platform sees the order, and it sends a matching order for each copier, sized by a rule you picked. eToro says copied trades are “typically executed in less than a second” after the leader’s. Four settings decide what your copy actually does.

### How much you set aside

Minimums are low: $200 per leader on [eToro](https://www.etoro.com/copytrader/), generally 100 USDT on [Bybit](https://www.bybit.com/en/help-center/article/Copy-Trading-Copy-Mode-and-Parameters-Settings), 50 USDT on [Bitget](https://www.bitget.com/support/articles/12560603826748) (USDT is a dollar stablecoin). Small amounts come with a catch. eToro doesn’t open a copied position worth less than $1, so if you copy with $200, every trade the leader makes with less than 0.5% of their account is skipped, and your copy starts to drift away from theirs.

### Proportional or fixed sizing

In proportional mode, also called ratio mode, your trade takes the same share of your allocation as the leader’s takes of their account. [Bitget’s help page](https://www.bitget.com/support/articles/12560603800576) gives the arithmetic: a leader with 1,000 USDT who puts 100 USDT of margin into a trade is using 10%, so a copier with 100 USDT puts in 10. [OKX’s proportional mode](https://www.okx.com/help/copy-traders-introduction-to-proportional-copy-trading) works with a multiplier you choose instead, from 0.01x to 10x the leader’s order: at 0.1x, a 10,000 USDT order becomes a 1,000 USDT copy. In fixed mode, every copied trade gets the same amount, whatever the leader did. [Binance](https://www.binance.com/en/support/faq/binance-copy-trading-rules-30f1e3a2835345e0b8ffc87f261a8256) calls the two modes Fixed Ratio and Fixed Amount; Bybit calls them Smart Copy and Advanced.

The choice matters more than it looks. Say a leader with $20,000 puts 2% into a small test trade and 20% into a trade they really believe in. Copying with $1,000 in ratio mode, you put in $20 and $200, the leader’s own weighting. With a fixed $100 a trade, you put 10% into the test, five times the leader’s weight, and 10% into the big idea, half of it. Fixed sizing keeps the trades and throws away the leader’s judgment about how much each one deserves.

### Leverage, and whose it is

Leverage lets a position be bigger than its margin, the money that backs it: at 10x, $100 of margin holds a $1,000 position. Copy in ratio mode and you take on the leader’s leverage. Some platforms let you override it. In Bybit’s Advanced mode you can copy every trade at a fixed leverage, 5x for example, even if the leader uses 100x. That shields you from the leader’s leverage, and it also means your copy stops behaving like theirs: the same price move brings a different gain, a different loss and a different liquidation price, the price at which the exchange closes a position because its margin is almost gone.

### Open positions, copy stops and slippage

- **What’s already open.** Bitget asks whether to copy the leader’s open positions or only new ones. Join mid-trade and you enter at today’s price, not at theirs.
- **A stop on the whole copy.** eToro lets you set a stop loss on the copy itself. Bybit’s trailing stop closes every position and ends the copy when your equity falls a set percentage from its high. Bitget’s Equity Guardian stops copying at a net loss or an equity level you choose.
- **Delay and slippage.** Your order goes in after the leader’s, so in a fast market you pay a little more on the way in and get a little less on the way out. Slippage is that gap between the price you expected and the one you got. Bitget lets you set a slippage limit, and copies that would fill beyond it aren’t opened. That protects your price, but the trades it skips are the ones where the price moved fastest, so your record and the leader’s drift apart again.

## Where can you copy trade, and how do the traders get paid?

Copy trading is built into the large brokers and crypto exchanges. Here is what each platform’s own pages said on November 15, 2026. The usual costs of the trades themselves (spreads, trading fees and, on futures, funding) come on top everywhere.

| Platform | What you copy | How the leader is paid | When it is settled |
| --- | --- | --- | --- |
| eToro CopyTrader | A Pro Investor’s portfolio, “in real time and direct proportion”; up to 100 leaders at once, from $200 each | By eToro: up to 1.5% a year of the money copying them, from the Champion tier up. Copiers pay no extra charge to copy | Monthly, by the 10th |
| Binance futures copy trading | A lead trader’s futures portfolio, by fixed ratio or fixed amount | Its copy-trading page cites a 10% profit share plus 10% of copiers’ trading fees; its lead-trader FAQ allows up to 30% | Weekly, on Mondays, only on profit above what was already shared |
| Bybit Copy Trading Classic | A master trader’s USDT perpetual trades, by fixed ratio or a fixed margin per trade; from 100 USDT | 10% to 15% of followers’ net profit, by the trader’s rank | Weekly; gains and losses net within each week |
| OKX copy trading | A lead trader’s trades, by a multiplier from 0.01x to 10x or a fixed amount | Up to 30% of profits, by the lead trader’s level | Weekly, once all linked positions are closed |
| Bitget futures copy trading | An elite trader’s trades, proportional or by fixed amount or multiplier; from 50 USDT | At most 10% to 20% of profits, by the trader’s tier | Daily, once positions are closed |
| Hyperliquid vaults | A shared account the leader trades for everyone in it | 10% of profits; the leader must keep at least 5% of the vault | When you withdraw |

### Why the settlement rule matters

A profit share is a cut of what your copy earns, and when it is counted changes what you pay. Say your copy loses $100 one week and makes $100 the next, so you end where you started. [Bybit’s](https://www.bybit.com/en/help-center/article/Copy-Trading-Profit-Sharing-Explained) and [OKX’s](https://www.okx.com/help/lead-traders-trader-profit-sharing-rules) pages describe netting gains against losses within each weekly cycle; they don’t describe carrying a loss into the next week, so the second week’s $100 can still carry a share: $10 at 10%. [Binance’s formula](https://www.binance.com/en/support/faq/lead-trader-benefits-in-binance-futures-copy-trading-ea9bacf82b9e4ddfae50ebc98565241b) takes the copy’s total profit since it began, applies the share, and subtracts what was already paid, so nothing is due until the copy is above its previous best. That rule is called a high-water mark. [Bitget’s rules](https://www.bitget.com/support/articles/9083311039641) net gains and losses day by day. In a choppy month, the settlement rule is the difference between paying for a round trip to nowhere and paying nothing.

### What each one does well

On eToro the platform, not the copier, pays the leader, and only members of its program can be copied at all. Members must keep to set standards: a daily risk score of 7 or less, a weekly drawdown no deeper than 25% and no single position above half the portfolio. Binance shows on each lead trader’s card what their copiers made over the last 30 days, and it caps copied orders on thin markets. Bybit lets you fix your own leverage and cap the margin per contract. OKX’s multiplier scales a copy precisely. Bitget lets you choose, leader by leader, between a proportional copy and a fixed one. Hyperliquid’s vaults run onchain, so every position and fill is public; our guide to [Hyperliquid copy trading](https://www.ethyai.app/blog/hyperliquid-copy-trading) covers vaults and the bots that mirror Hyperliquid wallets.

### What the pay rewards

A profit share pays the leader when you win and costs them nothing when you lose, apart from the copiers who leave. A payment on assets under copy pays for popularity. [Bitget’s tiers](https://www.bitget.com/support/articles/12560603792251), which set how big a share a leader may charge, rise with the leader’s 7-day trading volume and with the number or the assets of their copiers, not with their returns. None of these designs is wrong in itself, but each one rewards something besides your result.

## Is copy trading profitable?

It can be, for some copiers, some of the time. What you can’t find is a scorecard: none of the platforms above publishes an overall result for its copiers on the pages we read. The closest is per leader. Binance’s cards show what copiers made over 30 days, and Bitget lists followers’ profits next to the leader’s own. The research that exists says three things.

- **Copying can beat going it alone.** In data from a social trading platform for currency traders, MIT researchers Wei Pan, Yaniv Altshuler and Alex Pentland [found in 2012](https://dspace.mit.edu/handle/1721.1/80764) that “generally social trades outperform individual trades.” They also found that top traders’ reputations were “not completely determined by their performance.” Popular and skilled overlap, but they aren’t the same list.
- **Money chases last year’s winners.** Florian Röder and Andreas Walter [studied the money flowing](https://doi.org/10.1111/jfir.12174) into more than 5,300 social trading portfolios (Journal of Financial Research, 2019). It followed past performance, almost only into the best performers, and it followed raw returns rather than returns adjusted for the risk taken. Portfolios shown more prominently on the platform drew more money.
- **Being copied changes the trader.** Matthias Pelster and Annette Hofmann [found](https://doi.org/10.1016/j.jbankfin.2018.07.003) that traders become more prone to the disposition effect, selling winners too early and holding losers too long, once other people follow them (Journal of Banking and Finance, 2018). Among their explanations: “a fear of losing followers when admitting a poor investment decision.”

### Survivorship bias in one calculation

Survivorship bias means judging by the ones still standing. A leaderboard does exactly that: it ranks the traders who are still there by how well they did, and the ones who blew up are gone from it. That alone makes luck look like skill. Take 1,000 traders with no skill at all, each with an even chance of a winning month. After ten months, on average about one of them has won every single month, because one half multiplied by itself ten times is about 1 in 1,024. That trader tops the board. Nothing in the record tells that streak from a skilled one, which is why the checks further down look at how a result was made, not only at its size.

## Why does the leader’s leverage matter more than their returns?

Because leverage decides whether a copy survives the moves markets make all the time. A long position at 20x loses all of its margin if the price falls 5%, one divided by 20, and the exchange liquidates it before that point. At 10x it takes a 10% fall, at 5x a 20% fall. Here is how often the first of those happened on four large crypto markets in the year to November 13, 2026:

*Chart: Days from November 14, 2025 to November 13, 2026 when each market traded 5% or more below the previous day’s close: enough to erase the whole margin of a 20x long opened at that close. (Daily candles of the BTC, ETH, SOL and HYPE perpetual contracts from Hyperliquid’s public API, taken November 14, 2026; Ethy’s count.)*

Falls of 10% or more below the previous close came on 2 days for Bitcoin and 4 for Ether. Bitcoin’s worst day was February 5, 2026, when it traded 15.0% under the prior close; Ether’s was January 31, 2026, 18.7% under. A drop that size erases the whole margin of any long at 7x or more on Bitcoin, and 6x or more on Ether, and liquidation comes earlier still.

So read a leader’s leverage before their returns. A big return made at 25x and a modest one made at 2x are not the same record: the first one survived moves that would have liquidated it on a slightly worse day. Ratio copying hands you the leader’s leverage; a fixed-leverage setting hands you a different trade.

### Drawdowns

A drawdown is the fall from a peak to the next low. It takes a bigger gain to undo than it took to lose: after a 50% loss, you need a 100% gain to get back to even. Look at a leader’s worst drawdown and ask whether you would still have been copying at the bottom of it.

### How much money a leader can take

Every copier adds to the leader’s orders. On a busy market that is noise. On a thin one, the copies push the price against each other, and late copiers get the worst fills. The platforms know it. Binance’s copy rules limit leverage on low-liquidity markets to 5x and cap the size of copied positions with a formula that includes the depth of the order book. Bitget caps each elite trader at between 500 and 1,000 copiers, depending on tier. A leader whose returns shrank as copiers arrived is showing you that limit in action.

## What are the risks of copy trading, and what have regulators said?

The main risks:

- **The leader’s losses are yours.** A copy doesn’t filter the bad trades out. When the leader loses, so do you, in proportion.
- **Leverage you didn’t choose.** In ratio mode you inherit the leader’s leverage, and with it their distance from liquidation.
- **Drift.** Delay, skipped small trades, slippage limits and different leverage all make your result differ from the leader’s, and delay works against you on both the way in and the way out.
- **Incentives.** The leader is paid for wins or for popularity, not for protecting your downside.
- **Rules that change.** Profit-share tiers, copier caps and sizing rules are the platform’s to change, and several of the help pages we read carry recent update dates.
- **Scams.** Fake platforms and “mirror trading” schemes that show invented profits. Our guide on [how to tell a real trading tool from a scam](https://www.ethyai.app/blog/is-ai-trading-legit) walks through the warning signs and real cases.

### What regulators have said

The UK’s Financial Conduct Authority (FCA) keeps a [page on copy trading](https://www.fca.org.uk/firms/copy-trading), first published in 2015 and updated in July 2026. It says: “We classify copy trading as portfolio or investment management where no manual input is clear from the account holder.” Put plainly, when trades are copied automatically, the regulator treats the service like a manager running your money, with the duties that come with it, such as checking that the service suits you and reporting to you regularly. When you must confirm each trade before it goes in, the FCA says, the activity isn’t portfolio management.

In August 2026, India’s market regulator, SEBI, [warned investors](https://www.sebi.gov.in/media-and-notifications/press-releases/aug-2026/caution-to-investors-regarding-display-of-live-trading-strategies-on-social-media-platforms_103729.html) about people on social media offering “live trading strategies” who show off the performance of their trades as they go. Its advice was not to trust their claims and not to base investment decisions on those live sessions.

## How do you choose a trader to copy?

Don’t start from the platform’s ranking. In a study of two large crypto copy-trading platforms, Daisuke Kawai and his co-authors [found](https://doi.org/10.1145/3613904.3642715) that the way each platform displayed and ranked leaders significantly swayed whom copiers picked, even with typically $500 to $2,000 or more at stake (CHI conference, 2024). Sort the list yourself, then check each candidate:

1. **The copiers’ results, not only the leader’s.** Binance and Bitget show what copiers made. A leader in profit whose copiers are in loss is showing you delay, fees and drift in a single number.
2. **Leverage and position size.** The typical leverage and the highest, and the share of the account in the largest position. On eToro, the risk score sums this up.
3. **The worst drawdown, and the time it took to recover.** Would you have kept copying through it?
4. **How long the record is.** A full year at least, including a stretch when the market fell hard, like February 5, 2026, when Bitcoin traded 15% under the previous close.
5. **How many trades made the result.** Dozens of trades with a steady edge are evidence; one or two big wins could just be luck.
6. **Behavior since the copiers came.** Compare the record before and after: bigger positions, losers held longer, or returns that shrank as the money grew.
7. **What you will pay.** The profit share, how it settles (weekly netting or a high-water mark), and the trading fees on every copied trade.
8. **Your own exit, set before you start.** The allocation, a stop on the whole copy, and a leverage cap where the platform offers one.

> **Red flags** Leverage of 20x or more on most trades. A record younger than a year with no losing month. A total that rests on one or two trades. Copiers in loss while the leader is in profit. Anyone who asks you to move money off the platform or promises a monthly return.

## How is copy trading different from a managed account or trade signals?

All of these let someone else’s judgment drive your trades. They differ in whose account it is, who presses the button and how the other person earns.

|  | Copy trading | Managed account | Pooled vault or fund | Trade signals | Copying a strategy |
| --- | --- | --- | --- | --- | --- |
| Whose account | Yours | Yours, traded by a manager | The pool’s; you own a share | Yours | Yours |
| Who places each trade | The platform, mirroring the leader | The manager | The leader, for everyone at once | You | The rules, automatically |
| How the other side earns | Profit share or a platform payment | Management or performance fees under an agreement | Profit share, such as 10% on Hyperliquid user vaults | Usually a subscription | No person to pay; the tool has its own plan |
| How you get out | Stop copying | Under the agreement’s terms | Withdraw, after any lock-up | Stop following | Stop the rules |

The lines blur. The FCA treats automatic copy trading as portfolio management, the same category as a managed account. And some platforms offer both kinds: Bybit’s [Copy Trading Pro](https://www.bybit.com/en/help-center/article/Copy-Trading-Copy-Mode-and-Parameters-Settings) has investors buy shares of a Pro Master’s strategy, which then trades the pooled money, with a lock-up period and set windows for withdrawals. That is closer to a fund than to copying.

## What does it mean to copy a strategy instead of a person?

A strategy is [a set of rules](https://www.ethyai.app/blog/algorithmic-trading-for-beginners): which market, when to enter, when to exit, how much to risk. Copying one instead of a person removes the parts of copy trading the research above worries about. There is no leader to lag behind, because the rules trade in your account directly. Nobody starts trading differently because they have an audience. And the record you check describes the same fixed rules you would run, so you can tell when they stop working.

The trade-off is that [rules don’t adapt](https://www.ethyai.app/blog/ethy-vs-trading-bots#where-bots-fall-short). A person can notice that a market has changed; a rule keeps doing what it was told until you stop it. So a strategy needs the same vetting as a trader (the worst drawdown, the number of trades, the length of the record) and an exit you decide on before you start.

### How it works in Ethy

Ethy doesn’t copy other people’s accounts, wallets or vaults. What you can copy is a Top Strategy, a Hyperliquid strategy curated by Ethy.

- **What copying does.** It deploys an automation with the strategy’s published settings, from your own wallet, at the size you choose. It doesn’t mirror anyone’s account.
- **What you see first.** The [backtest](https://www.ethyai.app/blog/how-to-backtest-a-trading-strategy) the strategy was published with, and a record “since publishing”: the same frozen settings replayed every day on real candles. That record is a simulation, not money in anyone’s account. Once you copy a strategy, you also see its three newest trades.
- **Risk tiers by worst drawdown.** Conservative up to 10%, Moderate up to 20% and Aggressive up to 40%.
- **Limits you control.** A pause after a losing streak (on by default after 10 losing closes in a row), caps on daily loss and on drawdown in dollars, and a switch that archives the automation and closes its positions. If a strategy is later reviewed and taken off the shelf, your copy keeps running until you stop it, so these limits are yours to set.
- **Or a rule of your own.** Describe it in plain English, backtest the exact version that would run, and try it in paper mode, on live prices with no money moving.

> **Ask Ethy:** [Backtest a 20-day breakout on ETH on the daily and show me its worst drawdown.](https://agents.ethyai.app/chat?landing=blog&post=what-is-copy-trading&prompt=Backtest%20a%2020-day%20breakout%20on%20ETH%20on%20the%20daily%20and%20show%20me%20its%20worst%20drawdown.)

**Copy rules you can check** See a Top Strategy’s backtest and its record since publishing, then run it at your size from your own wallet. Or describe a rule of your own and test it first. [Start free](https://agents.ethyai.app/?landing=blog&post=what-is-copy-trading)

## Summary

- Copy trading repeats another trader’s trades in your account automatically, sized to the money you set aside.
- Ratio sizing keeps the leader’s weighting and leverage; fixed sizing doesn’t. Minimum sizes, delay and slippage limits make every copy drift from its leader.
- As of November 15, 2026: eToro paid its leaders up to 1.5% a year of the money copying them; leaders could take up to 15% of copiers’ profits on Bybit, 20% on Bitget and 30% on Binance and OKX; Hyperliquid vault leaders keep 10%.
- How the profit share settles matters: weekly netting can charge you in a flat month, while Binance’s formula works like a high-water mark.
- Research finds that money chases past winners and that being copied changes how traders trade, so a leaderboard mixes luck with skill.
- Leverage decides survival: in the year to November 13, 2026, Bitcoin traded 5% or more below the prior close on 16 days, enough to erase a 20x long’s margin.
- Check copiers’ results, leverage, drawdown, record length and pay before you copy, or copy a tested strategy instead of a person.

## Questions people ask

### Is copy trading legal?

Copy trading is offered by regulated brokers and exchanges, and regulators supervise it as an investment service. The UK’s FCA, for example, classifies automatic copy trading as portfolio management, which brings duties such as checking that the service suits you. Each platform sets out in its terms where it offers copy trading, so read them before you sign up.

### How much money do you need to start copy trading?

On November 15, 2026, the minimums were $200 per leader on eToro, generally 100 USDT on Bybit and 50 USDT on Bitget. Small amounts copy badly, though: eToro skips copied positions worth under $1, so with $200 any trade below 0.5% of the leader’s account is left out. Use money you can afford to lose entirely, and copy no more leaders than you can follow closely.

### Do you pay fees to copy a trader?

On eToro, copying itself carries no extra charge: you pay the usual spreads and fees on the trades, and eToro pays the leader. On the crypto exchanges, the leader takes a share of your profits, depending on their level: up to 15% on Bybit, 20% on Bitget and 30% on Binance and OKX, on top of normal trading fees and, on futures, funding. Hyperliquid vault leaders keep 10% of profits.

### What is a high-water mark in copy trading?

It is a rule that a profit share is charged only on gains above the copy’s previous best. If your copy falls and then climbs back to where it was, you pay nothing on the recovery. Binance’s profit-share formula works this way. Bybit and OKX describe netting gains against losses within each weekly cycle instead; neither describes carrying a loss into the next week.

### Is copy trading good for beginners?

It can show you how an experienced trader sizes and manages positions, but it doesn’t remove risk, and it adds some of its own: delay, leverage you didn’t pick and pay that rewards the leader for wins or popularity. If you try it, start small, cap leverage where the platform allows, set a stop on the whole copy, and read the leader’s worst drawdown before their best month.

### Does Ethy offer copy trading?

Not of people. Ethy doesn’t copy other traders’ accounts, wallets or vaults. You can copy a Top Strategy, a Hyperliquid strategy curated by Ethy: copying it deploys an automation with the strategy’s published settings, from your own wallet, at your size. Each one shows the backtest it was published with and a simulated record since publishing.

## Sources

- [eToro: CopyTrader (minimums, proportional copying, execution, copy stop loss)](https://www.etoro.com/copytrader/)
- [eToro: Pro Investor Program (payments on assets under copy, tiers, risk limits)](https://www.etoro.com/copytrader/popular-investor/)
- [Binance Support: Lead trader benefits in Binance futures copy trading (profit share, settlement formula)](https://www.binance.com/en/support/faq/lead-trader-benefits-in-binance-futures-copy-trading-ea9bacf82b9e4ddfae50ebc98565241b)
- [Bybit Help Center: Copy Trading, profit sharing explained](https://www.bybit.com/en/help-center/article/Copy-Trading-Profit-Sharing-Explained)
- [OKX Help: Lead trader profit-sharing rules](https://www.okx.com/help/lead-traders-trader-profit-sharing-rules)
- [Bitget Support: Elite Trader Tiers (profit share and copier limits by tier)](https://www.bitget.com/support/articles/12560603792251)
- [Hyperliquid Docs: For vault leaders (legacy)](https://hyperliquid.gitbook.io/hyperliquid-docs/hypercore/vaults/for-vault-leaders-legacy)
- [Financial Conduct Authority: Copy trading (updated July 27, 2026)](https://www.fca.org.uk/firms/copy-trading)
- [SEBI: Caution to investors regarding display of live trading strategies on social media platforms (August 17, 2026)](https://www.sebi.gov.in/media-and-notifications/press-releases/aug-2026/caution-to-investors-regarding-display-of-live-trading-strategies-on-social-media-platforms_103729.html)
- [Röder and Walter: What Drives Investment Flows into Social Trading Portfolios?, Journal of Financial Research (2019)](https://doi.org/10.1111/jfir.12174)
- [Pelster and Hofmann: About the Fear of Reputational Loss, Social Trading and the Disposition Effect, Journal of Banking and Finance (2018)](https://doi.org/10.1016/j.jbankfin.2018.07.003)
