Going Full Blockchain?
An analysis of price (in-)efficiency of decentralized exchanges
Written by Oleg Lukanin and Kevin Joerg — May 3, 2021
The crypto market has grown massively and is showing few signs of stopping. The digital overhaul of the global economy is unavoidable and people have no shortage of ideas to propel the world forward. Who needs pesky fiat money when you have stablecoins? Why hang a painting in your living room when you could get a sick dunk GIF from NBA Top Shot? Why don’t we trade it all on the blockchain?
Well, decentralized exchanges, such as Uniswap, have been getting more and more popular. But are they really providing you with the fairest price? Could you be paying too much? Are you missing out on some sweet arbitrage opportunities? We set out on a quest to find out.
On one side of the battle we have centralized exchanges —we picked the top 5 by volume: Binance, Huobi Global, Coinbase Pro, Kraken, and Bitfinex. As for their rivals, we chose to look into Uniswap v2 and SushiSwap.
Of course, we only looked into cryptocurrency pairs that the two groups of exchanges had in common, so we settled on 13 — ranging from the popular Bitcoin-USD Coin to the more exotic Aragon-Ethereum pair. Inverse pair listings, such as BTC-ETH instead of ETH-BTC, were also considered through a matching transformation.
The concept of price efficiency stems from the Law of One Price, which states that two identical goods must have the same price. The reasoning is just as straightforward — you should not be able to make a risk-free profit from the differences in price. In practice, all of the exchanges provide prices that are slightly different from one another.
So, how do you define a fair market price? Even though there is no perfect answer, we decided to focus on the volume-weighted average price across all of these exchanges. Needless to say, this definition punishes decentralized exchanges which, on average, enjoy smaller trading volumes. Nevertheless, our results are robust to alternative deviations, e.g. a simple average price.
Our main variable of interest is the absolute percentage deviation from the fair price for each exchange, at one-minute frequency. We regress it on the dummy variable “Decentralized”, to assess the difference in magnitude of deviations from the fair price between decentralized and centralized exchanges. Using data covering the time period since Uniswap v2’s launch in May 2020, we have the following results:
As we can see, decentralized exchanges in our sample tend to deviate from the volume-weighted average price by about 0.56% more than their centralized counterparts.
These results speak about the average deviation across the entire sample period. But is it the case that DEXs’ price efficiency has improved over time? To answer this question, we re-run our regression on different sub-periods, resulting in the following picture:
As you can see, SushiSwap experienced significant inefficiencies at first, but has improved steadily over the past months, going beyond Uniswap and coming close to the levels seen on centralized exchanges.
Market frictions, such as transaction costs, are a limitation to the self-enforcing law of one price. Currently, the commission fees on DEXs are 0.3%. In addition, miners also want their cut. To register a transaction on the blockchain, you need to pay a so called gas fee. While useful to provide an incentive for miners to include a transaction into a new block, they have become very expensive over the past few months. This is illustrated in the figure below, where we can see the average gas fees paid on SushiSwap for different transaction amounts in our sample.
In the figure, we can see that trading on the Ethereum blockchain is a costly undertaking. For a transaction of 10'000 USD people paid miners a whopping 90 basis points (0.9%) gas fee on average. As a reference, large centralised exchanges charge a trading fee of roughly 0.1% independent of the transaction size.
The high fees involved make trading on a DEX unappealing. Arbitrageurs will wait until the price differentials are worth the hassle before intervening, resulting in the high disparity seen above. These price differentials are an issue for retail investors, as they pose a risk of paying too much, or receiving too little in a transaction. And that is coupled with the much higher fees for small value transactions.
The proof of work concept limits the number of transactions that can be included in one block, and is therefore the main determinant of the high gas fees. Ethereum 2.0 will abandon the proof of work concept, and switch to a proof of stake scheme. The upgrade promises to increase capacity to 1000s of transactions per second (currently it is limited to roughly 15–20), and also to set an end to the computationally demanding and environmentally unfriendly mining procedures.
The power of a proof of stake based blockchain can be seen in the recent success of PancakeSwap and BurgerSwap, two DEXs based on the Binance Smart Chain, where gas fees are near zero. Despite being on a relatively new blockchain, they are just behind Uniswap with regards to daily volumes traded. The increase in transaction capacity should significantly reduce the percentage impact of gas fees. Together with the upcoming new features of Uniswap v3, we expect it to be the final puzzle piece that Uniswap and friends need to reign supreme.
