The Diworsification of Centralized Exchanges
Entering the crypto-sphere as a civilian, one of the first things you notice is that there are a lot of exchanges and that their names seem to be constructed from a finite set of nouns. For example, if you take the largest 100 exchanges on CoinGecko, a full 18% of them start with the word “Bit”. As in bitFlyer, Bitvavo, Bitfinex and so on. Whilst it’s true that the permutational space constructed from “bit”, “ex”, “coin” and so on is probably quite large, it does make for a sort of monotony.
But, of course, once you start to look a little closer, it’s also remarkable how different all these exchanges are – particularly when viewed through the lens of market specialisation. Some exchanges focus on the most popular currency pairs. Others service only certain countries and languages. This idea is captured in the graphic below, where we plot how evenly spread each exchange is across coins on the vertical axis and countries on the horizontal axis.
The measure we’ve used is entropy, a concept defined originally in Thermodynamics and later co-opted by Information Theory. If you’re trading three coin pairs with 33% of your volume in each, the entropy is about 1.1. If, in practice, all your volume is in one of those pairs, the entropy is zero. So the graphic shows how Kraken is trading more evenly across the various coins it covers whereas WazirX has most of its volume in XRP-USDT, according to Coingecko, even though it makes a market in other pairs.
Likewise, based on something like Similarweb, you can measure the extent to which each exchange has a geographic specialisation. Upbit and Bithumb are almost entirely South Korean. WazirZ is predominantly Indian. Conversely, Gate.io and Binance service a range of countries with no single location accounting for more than 10% of their volumes.
It’s interesting to note that these diversification measures are somewhat, but not entirely, correlated. Indeed, what correlation exists is largely down to a small cluster of “one-country, one coin” exchanges. If you remove these niche firms, the top-right cluster is basically a cloud. Some trade a lot of pairs but mainly in one country (Coinbase), some trade fewer pairs but in more places (LBank), and so on.
Are we bothered? Given Agio’s mission is to measure counterparty risk, we’re interested if more diversified exchanges are safer places to trade. And when you contrast performing and defaulted exchanges based on these entropy metrics you find statistically significant effects. Specifically, trading more coin-pairs does generate a lower volatility income stream that reduces default risk. More coin-pairs generate diversification. However, country diversification has the opposite effect. It turns out that covering fewer countries leads to better branding, greater customer loyalty, and thereby more financial stability. In other words, more countries generate diworsification.
