The act of queuing is a ubiquitous feature of modern life; from the familiar wait for the cashier at the supermarket, to a trip to the movies or our daily commute. The notion of queuing is of course much more general and can apply to more abstract situations such as the transfer of packets of data through a wireless network, the movement of goods through a supply chain, or the submission of transactions to a blockchain. In all of these cases, the flux of people or energy or information is restricted in some way by the capacity of the transmission system.
A useful example for understanding the economics of queuing is to consider a large city that decides to install public water fountains. The city government decides that since access to water is a human right, that the fountain’s water must be made freely-available. This sounds like a laudable objective and on the first morning of operation this policy seems to work well, with all citizens able to obtain water freely. After several hours of operation have passed, however, a queue of people has formed that continues to grow throughout the day.
As the queue increases in length, a progressively larger proportion of people that might wish to use the fountain instead choose not to join the queue since their perceived benefit of being provided with water is outweighed by the cost of waiting. This trade-off is captured by a concept used commonly in economics, known as utility, which involves aggregating the positive benefits and negative costs of an action to a common, typically monetary, value. In the water fountain example, individuals make the decision on whether to enter the queue on the basis of whether they consider this action to have a positive or negative utility.
The utility value of an action can vary between people and for a single person at different times and under different conditions. Water is a good example of an essential resource that can have widely different utility values depending on its availability. The willingness of citizens to queue would, for example, be much higher if the fountains were the only source of drinking water compared with typical conditions in developed countries where water can be obtained cheaply and easily. Individuals will also associate different costs to waiting in the queue because of factors such as their income and circumstances at a particular moment in time, such as whether they are on vacation or on their way to work.
If a resource has a limited capacity and excessive demand, such as in the fountain example, this implies that obtaining it involves some cost to individuals and that the provision of water does not fulfill the city government’s objective of being freely available. Worse still, the cost is absorbed by the wasteful act of queuing, which means that fountains may deliver little overall benefit or utility to the city’s residents.
One improvement that could make the system somewhat more efficient would be to introduce a flat price for using the fountain. This has the advantage that some of the cost of maintaining the resource is returned to the city, which could then use the revenue to establish a larger number of fountains to reduce congestion. However, if we assume that the number of fountains remains fixed so that the capacity is limited, users will continue to spend time inefficiently waiting in a queue. These queues will be shorter since the potential pay-off for queuing (receiving subsidized water) is lower than in the zero-price setting, but this is achieved at the cost of excluding some people who cannot or are unwilling to pay the price for water.
This system is also inefficient in the sense that some users who place a high value on the resource, but also have a strong aversion to queuing, do not make use of the fountain despite being willing to pay higher prices. A similar argument applies to users with a willingness to queue for extended periods of time but in exchange for low- or zero-price access to the resource. Both of these situations can be resolved, and the system made more efficient, by allowing financial exchanges between the queuers that wish to “jump” places in the queue and those prepared to wait in line for a longer period.
Arranging direct financial transfers between queuers is unlikely to occur in practice as the sums involved are too small to justify the effort of conducting negotiations with many different individuals. An alternative would be to arrange rounds where several people can access the resource simultaneously, and then allow users to submit public bids to be considered for the next available round. High bidders would then be able to use the fountain after a smaller number of rounds than those offering lower bids. This type of mechanism is economically efficient in the sense that it accommodates the different urgency (and willingness to pay for it) of different individuals.
Anyone familiar with the Bitcoin or Ethereum networks will note the similarity of this process to the setting of a fee payment to miners for entering a transaction into the blockchain. In a recent paper it was shown that this payment system is one of a class of Vickery-Clark-Groves (VCG) mechanisms, which means that it also possesses two other important characteristics. The first of these is that transactors are incentivized to propose a fee that reflects the true cost that they associate with waiting in the queue. The second property is that VCG mechanisms are socially optimal in the sense that the fee paid reflects the social cost that users impose on other users by creating an additional delay on their transactions being entered into a block.
This payment mechanism has the advantage that the fees that miners are prepared to accept for including a transaction in a block adjusts automatically to changes in the transaction volume. However, the lack of scalability of two of the earliest (and as of mid-2018 most valuable) cryptocurrency platforms, Bitcoin and Ethereum, leads to other issues that cannot be overcome by this fee mechanism alone. In particular, it is possible for an attacker to submit a large number of transactions with arbitrarily high fees that force other users to either accept similarly high fees or be left “locked-out” of the system.
Records of transaction fees on the Ethereum blockchain suggest that an attack of this type occurred in early July 2018 when the average transaction fee rose from $0.22 on the 29th of June to a peak of $5.53 on the 2nd of July. This was dismissed as a somewhat irrational occurrence by one eminent Ethereum developer:
However, in the three months following the spam attack, Ethereum’s market capitalization collapsed by more than one-half, from $49bn dollars to $22bn. It is possible that other factors played a part in the decline of the Ethereum price, but it is clear that a leveraged short position on the value of Ether could have generated a substantial profit for the attacker. And even if this turned out not to be the case, the attack emphasizes the need for scalability if cryptocurrencies are to succeed in the multi-trillion dollar industries that could be revolutionized by the application of blockchain technology.
Having established that scalability is essential for widespread adoption of cryptocurrencies, this raises the question of the economic impact of scalability on users and operators of nodes (also known as miners). This is another situation where the water fountain analogy is instructive in understanding a system where the price of access is set for each user by the additional waiting cost that they impose on others. The consequence of extreme scalability is that these waiting costs and therefore fees would fall to zero, to the obvious benefit of users. For miners on the other hand, this scenario appears to be less appealing as they would be forced to bear the potentially infinite cost of supporting any number of transactions without a means of raising revenue.
This suggests that the design of incentives for miners is potentially of equal importance to the technical challenges of scaling decentralized ledger technologies. Perhaps the most established scalable ledger protocol is EOS, which has a stated maximum transaction processing rate of 3,500 transactions per second (TPS) and zero-fee transactions. The approach EOS takes to providing incentives to miners is to restrict their number to 21 active nodes to ensure that each miner has frequent opportunities to receive rewards from the creation of blocks. Although users do not pay transaction fees, the EOS coin inflates at a rate of 5% per year to provide block rewards and for the maintenance and extension of the protocol. The consequence of this design is that users of the EOS system are effectively transferring this percentage of their assets to miners irrespective of whether they issue transactions or not.
The principal stated advantage of blockchains is that they are free from centralized control and the extraction of rents from a single authority. Conventional payment systems, such as Visa or PayPal, typically charge transaction fees in the range of 1–5%, and, while profiting from economies of scale, are at least partially in competition with each other. It is therefore difficult to see how EOS, which replaces these centralized payment schemes with a cartel that has direct control over a large and increasing proportion of the base currency, can deliver long-term benefits to consumers in terms of cost or participation.
These limitations raise the question of which incentives should be introduced to both compensate miners and ensure truly low operating costs for users. A second question concerns how these economic aspects should in turn influence the design of scalable decentralized ledgers. While it seems certain that these issues can be resolved in many different ways, with each solution being suited to different use cases, two general guiding principles that should be applied to protect users and thereby promote more widespread adoption of blockchain technology are:
In applying these principles to scalable ledger design there are several promising avenues to creating mining incentives that do not rely on token inflation. One possibility is for the protocol to introduce fixed minimal prices for transactions that could also depend on the transaction’s value and the computational complexity of its execution. The scalability of the ledger could itself be tuned so that there is “just enough” congestion in the system to ensure that users are obliged to pay some compensation to miners. This is a difficult task as it requires that the ledger is capable of both rapidly increasing and decreasing its throughput in response to changing demand. Apart from one notable exception, this capability is not supported by most existing designs.
Another possibility is that the services contributed by a miner such as transaction execution and verification are compensated by other sources of revenue. To return to the earlier analogy, the operation of water fountains could be offered as a franchise with a requirement to provide large quantities of free or subsidized water. This could be included alongside exclusive rights to some other profitable activity, such as providing snacks or soft drinks. This might work in the context of blockchains by encouraging corporations to operate nodes to mediate low-cost and trustless financial exchange but where node-ownership confers additional benefits such as providing services to consumers or the monetization of siloed data.
Large corporations are beginning to enter the blockchain space with one example being the MOBI consortium of car manufacturers, which includes BMW, Bosch and GM. While these organisations can benefit from the opportunities of trading with each other across a shared, unpermissioned platform they can also contribute to providing the cryptoeconomic security to enable truly scalable blockchains to become a reality.