Distributed Exchanges and Fragmentation
Can we sustain a single distributed exchange across geographic regions?
Today’s equity markets are fragmented: In the United States, no fewer than 13 trading platforms compete to attract buyers and sellers.
Fragmentation is not unambiguously optimal. Indeed, exchanges are natural monopolies. Indeed, buyers and sellers are more likely to be matched with each other if they submit orders to the same exchange. Importantly, human traders can not participate on multiple exchanges at the same time. However, a single exchange would have monopoly power to charge high fees and commissions, distorting the incentives to trade.
In the mid-2000s, technology came to the rescue. Algorithmic trading opened the door for “cheap”, scalable, market fragmentation. Within milliseconds, computers can compare prices on multiple trading venues and choose the best one. Regulators followed technological innovation: RegNMS in the United States (2005) and Mifid in Europe (2007) directly mandated competition between platforms, setting the stage for today’s fragmented trading environment.
Can distributed exchanges (DEX) achieve the best of both worlds? The Holy Grail is a centralized marketplace where the infrastructure provider lacks monopoly power. In a DEX, computers supply trading infrastructure competitively through “mining’’ orders on a distributed ledger, e.g., a Blockchain. In contrast to a traditional exchange which can unilaterally set trading fees, such miners necessarily compete with each other. At the same time, if they use the same smart contract, a centralized market-place emerges where all buyers and all sellers participate to the same network.
In a new short technical paper with Sorin Zoican, we tackle the limits of decentralized exchanges to scale up geographically. In reality, economic activity is clustered in well-populated areas such as cities (e.g., having a market in New York and a market in Chicago, with little trading interest in between). Can we have a decentralized exchange spanning both regions?
First, we build a simple economic model to study the incentives of miners. If trading interest is unevenly distributed across regions (think a large metropolis and a smaller city), a decentralized exchange entails a value transfer from miners in the metropolis to miners in the smaller city. Otherwise, infrastructure providers in the larger cluster are better off building their own blockchain / market rather than joining forces with small-town miners. In this case, we obtain a fragmented securities market.
However, if miners in the metropolis could be convinced that they have an advantage in processing traders over their small city counterparts, a single DEX spanning both regions becomes feasible.
Such an advantage stems naturally from the way consensus is reached on peer-to-peer networks. To reach consensus, all nodes must communicate over several rounds to cross-verify information from other nodes. The consensus is consequently achieved by a qualified, rather than simple, majority of nodes. If the network consists of two asymmetric clusters, miners in the larger cluster are more likely to reach consensus first since they can reach more nearby peers. In contrast, small-city miners need to send messages across a longer distance.
We conduct a Monte Carlo experiment on an unstructured P2P computer network to calibrate this statistical advantage. We find that large-cluster miners are more likely to process the transaction first if (a) the clusters are further apart or (b) the clusters are more asymmetrical (i.e., the number of miners in the metropolis is much larger than in the small city).
Further, miners’ incentives to invest in in technology act as a centripetal force: If miners use more computing power to be locally faster, geographical distance is more important, pushing the clusters apart (from a message delay point of view). In this case, the speed arms’ race helps sustain a single DEX.
We conclude that cross-region distributed exchange blockchains are feasible as long as the regions are asymmetric both in trading services demand and in computer infrastructure supply — a natural assumption since both are driven by economic activity. Further, miners’ incentives to invest in faster CPUs to process transactions helps to sustain a unique market.
Short paper is available for download [here].