Liquidity mining (LM), incentivized pools, yield farming — all terms to describe a strategy where a user provides capital to a protocol in return for its native token. Although liquidity mining is common in crypto, it by no means originated there. Electronic markets since the late 1990s have offered rebates to liquidity providers. Looking into exchange rebate programs and their structure will create more effective liquidity mining programs in DeFi.
Initially intended to enable graceful decentralization to project users and enable fair launches, token distribution programs have evolved to focus on growth marketing, increasing AUM and users through subsidies. Most liquidity mining programs have been successful in terms of temporarily increasing AUM but tend to be expensive and the growth dissipates after the program concludes. Uniswap was able to attract billions in liquidity during their incentivized pools but proceeded to lose over a billion in TVL along with market share to other DEXs when their program ended[1]. Granted Uniswap has more liquidity now than it did before the start of liquidity mining, but at the cost of 20M UNI tokens or ~60m USD. Ultimately can Uniswap, other DEXs, and DeFi projects implement a liquidity mining program that will attract users and liquidity while being sustainable indefinitely?
The Maker-Taker Model
Like Uniswap, centralized exchanges of the late 1990s faced stiff competition from each other for market share. Initially, they charged both the makers and takers of liquidity. This means no matter if you put in a limit order which adds to the total liquidity or used a market order which took away the existing liquidity, the fee would be the same. In hopes of attracting more liquidity, an exchange, Island ECN[2], implemented a maker-taker fee model in 1997. Where the taker of liquidity pays a small fee, and the maker of liquidity gets a rebate or cash back in common parlance. To clarify, a maker of liquidity creates limit orders and is rebated when their order is executed. And generally, as rebates < fees, this model is sustainable[3].
Rebates drive liquidity
Holding all variables constant, when given a choice between exchanges with and without rebates, a market maker, broker, or trader intuitively would prefer to execute their limit order on maker-taker exchange. Why give up “free” money[4]? As a result, whenever an exchange implements maker-taker pricing, it increases liquidity, tightens spreads, resulting in more volume and revenue overall. But what is liquidity exactly? When exchanges incentivize limit orders, the total amount of stock available to be taken off the market increases drastically, meaning that prices are less likely to move the market whenever there are large trades — hence less slippage and better liquidity.
This model proved so attractive that Island’s market share grew from 3% to 13% of NASDAQ’s trades in two years. Since then, the majority of exchanges and alternative trading systems (ATS) in the US have implemented a maker-taker model[5]. Though the uptake among crypto exchanges is lower, mainly due to the market structure[6].
Rebate design
Each maker-taker exchange and ATS has a different rebate and fee schedule[7]. Although arcane to the outsider, they intended to incentivize particular types of trading behavior that the exchange believes will be helpful, similar to crypto economics. Variables will include:
- Particular securities
- Who is the market participant: trader, broker, designated market maker, SLP, retail, etc
- Percentage of total volume the market participant contributes
- Proprietary or not (meaning is the market participant trading their own capital or routing client orders?)
- Time of the order
- Percentage of quotes within certain percentage of NBBO (applicable to designated market makers and SLPs only)
- Price of the security
- Routing strategy
Skim the NYSE and Cboe EDGX’s fee schedules to appreciate the complexity. (Weird DEX fee pricing schemes seem so straightforward compared to this).
Applying Maker-Taker models to DEXs
Decentralized exchanges can fall under two broad categories — central limit order books (CLOBs) and automatic market makers (AMM). The former is structured similarly to most centralized exchanges where maker-takers models can be applied directly, whereas the latter requires some creativity.
Some CLOB DEXs are currently dipping their toes into maker-taker models. Loopring charges takers 0.2%, and makers receive a rebate of 0.016%, meaning for every 100 dollars, the taker pays 20 cents, and the maker gets 1.6 cents. This rebate to fee ratio of 8% is extremely low relative to CEXs in and outside of crypto. Qualified market makers generally expect to get 80 to 97% of the fee as a rebate. In contrast, 0x v3 has a variable taker fee based on the transaction gas cost, and the fee is rebated every ten days in a convoluted way back to the maker[8]. Both these exchanges should return more value to makers to incentivize more resting limit orders to reduce spreads to attract takers[9]. Because takers pay the makers, maker-taker models do not cost the exchange anything beyond lost revenue.
Unfortunately, rebates don’t map well to AMMs. AMMs are a type of DEX that uses a mathematical function instead of an order book to price its assets. The price is based on the ratio of two or more different assets deployed in a pool. There are two market participants — traders who only can make market orders and liquidity providers who deploy assets into the pool. In some sense, AMMs are the epitome of maker-taker models as LPs will receive rebates by default. Uniswap LP’s currently get 0.3% of all trades, which is significantly larger than what Loopring and 0x and centralized crypto exchanges offer. Regrettably, to reduce slippage for AMMs, there must be more assets pooled together which can be capital inefficient. If trading volumes are flat or do not grow sufficiently, the fees will be less per dollar of contributed capital. This makes new LPs wary of deploying capital into the pools as their yields will vary. Due to this dynamic, AMMs have to find ways to increase both pooled assets and volume at the same time.
AMMs should consider paying directly for order flow or volume, aka payment for order flow[10] (PFOF). Rebate DEX aggregators with the transaction fee to incentivize trades. For example, Uniswap could split the 30 bps fee to the pool’s LPs and the aggregator.
Future of rebates in crypto
Protocols need to think about the long term incentivization of their users. Indefinitely minting governance tokens can’t exist[11]. Rebate programs will become cross-subsidized, taking profits from particular actions to incentivize healthy behavior. The design space is vast, and there have been exciting designs. Thorchain has a slippage based fee that charges more to traders who create lots of slippage, resulting in more fees rebated to LPs.
Additional Notes
Caveats of rebate structures
Despite the liquidity benefits of rebates, there have been debates over second-order effects that were not initially a concern. The most consumer-facing one: given multiple venues to route a customer’s order, a broker could send the order to an exchange that offers the most generous rebate, which might not be the customer’s best execution price — a principal agent problem. Although this is explicitly banned by the SEC under NBBO as brokers have a fiduciary duty to get the best price for clients, determining the “best” price at a particular point in time is difficult. The SEC tried to indirectly reduce the rebate through a new ruling (Rule 610T in 2018), but exchanges rebelled, and the courts ruled the SEC lacked authority. (Side note an estimated 2.5 billion has been distributed through to rebates)
For now this is not a concern in decentralised markets as most trades are made directly with the exchange rather than using brokers, more commonly known as DEX aggregators, such as 1inch or Slingshot. Though in the future if DEXs start reducing fees or even giving rebates to aggregators, should their users expect the benefits to be passed on? Given the lack of any self regulating bodies, it seems unlikely.
Footnotes
[1] In itself, there is nothing wrong with LM programs that are costly and finite. They can be great ways to prime the pump — this piece is more about long term LM programs
[2] ECN or Electronic Communication Network is an Alternative Trading System (ATS), they fill a similar role to exchanges.
[3] Some exchanges offer rebates greater than the taker fee. This practice is sustainable as exchanges have multiple revenue streams such as selling data (which in practice can be equal to trading revenues)
[4] In practice, the traders are concerned about the “liquidity profile” of an exchange, e.g., quantity/quality of liquidity (fill ratios, adverse selection, the impact of HFT, etc). There’s no point in routing an order to an exchange with a high rebate if the probability of getting it filled is low. Conversely, for DEX this translates to the likelihood of failure, estimated slippage, front running, etc
[5] Some exchanges have maintained the flat fee model. Several have an inverted “taker-maker” model, where the taker gets rebated for liquidity, and the maker pays for adding liquidity.
[6] A guess is this is due to the market structure of crypto exchanges where the exchange itself owns the customer experience. In contrast, in traditional markets, a user will interface through a broker to multiple exchanges. Because switching costs are high, exchanges have more substantial pricing power. Routing orders tend to be difficult as exchanges also function as custodians; moving coins from exchange A to B takes time.
[7] RegNMS (17 CFR 242.610(c)(1)) limits “fees for access to quotations” or the cost to remove liquidity from an exchange at 0.3 cents per share; this has an indirect limit to set the max rebate to less than 0.3 cents per share. (But this is not always the case, and this only applies to RegNMS, options do not have access fee limits). The per-share pricing can be confusing initially as fees at crypto exchanges are represented as basis points rather than say per BTC.
[8] Have to include the ZRX token somehow :/
[9] But then again, if all CLOB DEXs have low rebates, market makers have no other choices. Some centralized exchanges probably would like to collude to reduce rebates so that they can capture a bigger spread on every trade.
[10] Payment for order flow is not a rebate from an exchange. It’s a rebate that market makers will give brokers the privilege to route their orders for them. (Though in turn, the market makers could get a rebate from an exchange, which is often the case)
[11] Maybe they can exist, similar to Ethereum and Monero’s tail emissions of infinite inflation
Special thanks to Yi Sun and Dmitriy Berenzon for their conversations and feedback.
You can follow me on Twitter here.