Jonathan Levin is a co-founder of Coinometrics, a premium data analytics company for digital currencies. At the company, he leads work on measuring the activity and health of the bitcoin network.
Levin was previously a postgraduate economist at the University of Oxford where his research focused on virtual currencies, creating one of the first statistical models of bitcoin transaction fees. While at Oxford, he was the convenor of the Oxford Virtual Currencies Working Group, an interdisciplinary working group focused on the economic and social implications of virtual currencies.
Levin has also consulted to government bodies, Fortune 500 companies and first-tier investment banks on the future of digital currencies.
Transaction fees on the bitcoin network are often forgotten about. Yet in bitcoin they are more important than on other payment networks.
Fees currently serve as protection against the network being flooded with arbitrary transactions. Eventually, they will also be required to incentivise bitcoin miners to continue providing security to the network.
Contrary to other payment networks, the fees will not be set centrally, but will be determined by a market between users and miners.
The lower bound of transaction fees will be determined by the marginal cost incurred by miners for including transactions. This, in turn, is determined by the speed at which information is spread around the network.
The cost of delays
Bitcoin, the protocol, is a set of rules for the construction, distribution and agreement of financial transactions. The excitement surrounding its invention stems from its ability to achieve consensus among anonymous participants on the order in which transactions were issued on the network. This permits the construction of an inclusive financial system that is free from discrepancies between market participants.
However, the network is only guaranteed to come to consensus over a period of time and, in fact, the bitcoin network is in a constant state of flux, where nodes differ in their record of the history of all transactions. Subsets of nodes that agree on a common set of transactions are known as partitions.
An event, such as the discovery of a new block in the block chain creates a partition in the network. When a miner finds a new block, it is the only node in the network that thinks that the length of the block chain has increased. The rest of the network will find out about the block consequently, with the delay largely dependent on the number of transactions included.
Each extra transaction included in the block adds to the time that it takes for blocks to reach every node in the network. Miners that are slow to hear about new-found blocks spend time working on redundant problems, which hurts their margins. Furthermore, this delay increases the possibilities for malicious attackers to double spend funds on the network.
How long is a chain?
Bitcoin miners are mainly incentivised by the 25 BTC block reward and fees play an insignificant role.
Delays in blocks spreading throughout the network mean that some miners are working on a shorter block chain than the longest chain in existence.
The bitcoin protocol ensures that all miners follow the longest chain to ensure consensus and hence a miner working on a shorter chain is likely to be wasting their efforts.
For blocks with over 1,000 transactions, it can take several minutes to reach the last node on the network. During this time the miners that are uninformed about the block in circulation will be creating a conflicting block.
Only one of these blocks can actually be included in the true record of bitcoin. Any extra data contained within the block slows the rate at which it spreads through the network, hence increasing the chances that it gets discarded along with the miner’s block reward.
This increase in the probability that the block gets discarded is the marginal cost faced by miners of including an additional transaction.
The need for speed
In my research, I found that small miners face a higher marginal cost than the minimum transaction fee 0.0001 BTC (5 cents at the time of publication). Acting rationally, they should not include any transactions in their blocks. In order to incentivise them to include transactions, fees would have to roughly quadruple to 0.0004 BTC (20 cents).
As there is a need to include more transactions, it is important to find more efficient protocols for spreading information around the bitcoin network or risk having much higher fees on transactions.
Reducing the time that it takes for information about new blocks to spread through the network would reduce time spent on redundant blocks and hence also reduce the probability that any blocks found are discarded.
However, while some engineering solutions are currently being discussed, further research around economic incentives and security risks is needed before any of these measures can be implemented.