Perp Funding Rates Don't Do What They Purport
The standard funding rate is an anachronism
Synthetix’s latest perpetual swap (aka perp) market sets the funding rate based on the net long/short customer position ratio. The net customer imbalance generates a funding rate and a fee discount to incentivize an equilibrium. Synthetix's perps trade like other perp contracts, highlighting the funding rate mechanism is not required to set perp prices.
Several perps have a base funding rate of 10.95% annually (eg, Hyperliquid, Binance, Bitmex, Bybit). Previously, it just so happened that this was the funding rate that equilibrated the market, but now it is just an explicit rate determined by the protocol outside of big market moves.
These are both steps in the right direction. The perp funding rate should be used to pay for liquidity providers, who tend to be short. The perp funding rate mechanism does not, and cannot, set perp prices, which are reasonably governed by Schelling points and the self-interest of exchange market makers. The perp funding rate mechanism is a white lie that got out of hand, and while perp traders seem not to care, it's dangerous when lies become standard because it encourages more lying, which never ends well.
I have no problem with the perp market makers doing what they are doing, but the fiction these derivative markets are disciplined by arbitrage and based on Nobel prize-winning research is a lie.
Crypto perpetual futures ('perps') are a popular way to trade crypto with leverage, long or short. It started in earnest around 2016 by Bitmex when there were no stablecoins, and only Bitcoin (BTC) was used as a form of payment. How do you trade a futures when you only have one asset? In standard futures markets, someone sells a bushel of corn for delivery in three months, and cash is exchanged today while corn is exchanged at maturity.
With only one asset, bitcoin, there was nothing to deliver bitcoin into. Without a physical swap of something with Bitcoin, the price could be arbitrary or set by the good nature of the perp market makers (trust!). Bitmex contrived a story that was good enough to comfort most crypto traders who are eager for leveraged investing and just want something plausible. Remember, this was when people were putting billions into Bitconnect, a firm offering 1% daily returns for investing in their black-box volatility arbitrage strategy. Ancient Aliens fans are more skeptical than your typical crypto investor.
The perp price-funding rate equilibrium mechanism supposedly works like this. The market is bullish when the perp price is above the spot price.1 To incent traders to sell the perp price back down to the spot, the long positions are penalized while the perp is at a premium, paying a funding rate to the shorts (and vice versa). Thus, arbitrage sets perp prices in the same ruthless way that arbitrage sets prices on trad-Fi derivatives, a trustless, decentralized process similar to the blockchain's consensus mechanism.
FundingRate per day = perp premium = perp TWAP / spot TWAP – 1
In contrast, consider the arbitrage involved in a gold futures contract. Say the maturity of the futures contract is in one month, and gold is traded in USD, which has a one-month interest riskless rate of 1.0%. Gold has no dividends and negligible storage costs, so we will assume its return is zero.
If the futures price of gold is 2.0% above the spot price, an arbitrageur could buy $100 worth of gold at the spot price, sell the futures for $102 worth of gold futures, and make 2%. As the spot purchase took $100, he had to pay $1 in borrowing costs for a net profit of $1. This is arbitrage because it is riskless, as regardless of where the price moved over the month, the profit is $1.
If the futures price equals the spot price, an arbitrageur would buy the futures contract at $100, sell the gold short for $100, invest the cash, and make $1 in interest. At maturity, the arbitrageur would take delivery on his long gold position to close his short, making $1 in interest. This works because while the ratio F/S can float around, at maturity F/S=1. This is what makes it riskless.
This generates the standard futures equation
Or
Note that eq(2) looks like a perp premium. For example, with gold, r$ - rA is about the USD interest rate. The longs pay the shorts the USD interest rate, reflected in the Futures premium, just like how your average perp long pays the shorts.
Transaction costs will add friction so that the price of gold futures will have upper and lower bounds where arbitrage is infeasible. Within that range, the futures price can roam randomly. For example, if it costs 0.5% in total trading costs (fees, spread, market impact) on all the legs of the trade, in and out, the futures price is free to roam 0.5% above or below the spot price. But as the range is symmetric around a midpoint, one can be reasonably sure that if the futures price is at the top end, it will fall back down toward the middle of the range shortly, as there is no incentive for anyone to set the futures price at any particular midpoint, given the random fluctuations. Intra-day speculators anticipating arbitrage can profit by taking some risk and buying when the price is at the bottom of its range even though they do not hedge this trade, which pushes the futures price within a tighter bound than that implied by arbitrage after transaction costs.
The perp market applies the perp premium (perp/spot -1) every 1 or 8 hours using an average perp/spot price in that period. Thus, a 0.03% perp premium translates into an 11% annualized rate (0.03% x 365), a 0.10% premium implies a 36.5% annualized rate, etc. Transaction costs for the most liquid equities are around 0.1%. Considering the average crypto has twice the volatility of the average ultra-liquid equity, a 0.10% transaction cost to buy or sell spot and perps is conservative. A trader would need a 0.4% premium to implement arbitrage, which would require four trades.
Yet, unlike the gold arbitrage example above, an arbitrageur selling when the perp premium is 0.4% above the spot has no guarantee that this premium will persist over the 8-hour window he bought in and certainly not the subsequent funding payment window. The perp/spot premium can become 0.0% immediately after putting on the initial ‘arbitrage’ trade, which means it is not arbitrage. Without riskless arbitrage, there are no hard edges that constrain the perp premium, just trust that the perp market makers will target the spot price.
Further complicating the funding-perp premium nexus is several degrees of freedom given to the admin generating the hourly or 8-hour funding rate, such as using level-II quotes (the ‘price impact’ bid/ask) or which ticks to sample when pulling within the time window. Considering that no perp market has independent auditors looking at its trading and messaging tape with verifiable time stamps, the standard error in this process allows the perp admins to post present arbitrary perp premiums, implying annualized funding rates +/- 30% with plausible deniability.
Bitmex initially averaged an 11% annualized funding rate, presumably via the emergence of a competitive equilibrium. Now many perps explicitly make 0.01% the default 8-hour funding rate, which annualized to 10.95%. While this is fraudulently presented due to the objective perp/spot premium, it is understandable and tolerable to perp traders. Your average perp traders want to be long, so the insider perp liquidity providers will generally be short. It is easier for the perp liquidity providers to have a flat position by having a long position off their exchange counterbalanced by a short position on the exchange (if they were short off their exchange, they would have to trust one of their competitors).
The correlation between past returns and future funding rates shows evidence of insider LP gaming. When the price rises, the market makers lose money on their perps (though offset by their long positions off the perp exchange). The perp insiders correctly calculate that long perp traders will be indifferent to paying higher funding rates over the next day, which adds a little extra to the perp market maker profits. Bitmex's market makers at least soften their traders' woes by giving the longs an actual negative funding rate (they get paid) during bear markets, while at Binance, they are not as nice and rarely post negative funding rates. In either case, it's a sweet cushion for the perp admins and their LP insiders.
Correlation between Returns and Current Funding Rates
2020-present (same result if 2022-present)
Economic Nobelist Robert Shiller wrote about perpetual futures in 1993 (see here), but his mechanism differs profoundly from the perp premium mechanism. His paper focused on creating a real-estate futures contract, and he recognized that he needed a housing dividend (rent) estimate for this to work. Most of the paper outlines how to estimate the rental return on real estate using macro data (e.g., monthly labor income). However, this is not a well-defined objective, as single-family homes do not easily map into comparable units (e.g., more square feet is better, but so is location). Any regression one could generate would have a significant standard error. This market never worked because there was no way for market makers to hedge or arbitrage the market, as neither the index (home prices) nor its return (home rental returns) had well-defined proxies that could be bought and sold.
The rental return in Shiller's perp market is nothing like a futures/spot price premium. Shiller's perp payment is exogenous, like Gehr(1988)'s description of the Singapore Gold exchange, which studied perpetual gold markets circa 1988. The crypto perp premium has no independent funding rate market as in Gehr and Shiller; it uses an average perp premium to generate a funding rate. The problem is that no riskless arbitrage mechanism bounds the perp premium as in standard fixed maturity futures contracts. In practice, each perp market has insider LPs who maintain a liquid market near the spot, an example of Aumann’s correlated equilibrium, which is like a Schelling point.
A small lie can save a lot of explanation. I can empathize with Bitmex's initial struggle to comfort traders eager to trade with leverage. Now that we have stablecoins, this fiction is no longer needed, and the perp funding rate white lie is a cancer that hits at the integrity of defi development. When naïve insiders discover their market is not working as stated, they become accustomed to the deception and do not blink when confronted with other lies. This was highlighted in the Sam Bankman-Fried trial, where insiders Ellison, Singh, and Wang made little compromises that snowballed into $8B of customer money stolen. Little lies lead to more significant lies; the best time to stop is when you can.
Bullishness is a common explanation for how the perp funding mechanism works, but this is illogical. The law of iterated expectations states that any bullishness shows up in the spot price, not the futures. The futures/spot ratio is set by the relative net interest rates on the two assets, not the spot price level. Indeed, empirically, in bull markets, the futures/spot ratio tends to decline, as these are often the result of temporary supply/demand imbalances, causing the spot to spike while the futures remain flat. This is the exact opposite of what we see in perps. Recent price jumps are an opportunity for the market makers to pull money back out of traders with house money (ie, recent unrealized pnl).