The Unintended Consequences of Blockchain Are Not Unpredictable: Respond Now Rather Than Repent Later*
In the past week the WSJ and the FT have run articles about a new bank-led initiative to move commodity trading onto a blockchain. In many ways, this makes great sense. By its nature, the process of recording and trading commodity trades and shipments is (a) collectively involve large numbers of spatially dispersed counterparties, (b) have myriad terms, and (c) can give rise to costly disputes. As a result of these factors, the process is currently very labor intensive, fraught with operational risk (e.g., inadvertent errors) and vulnerable to fraud (cf., the Qingdao metals warehouse scandal of 2014). In theory, blockchain has the ability to reduce costs, errors, and fraud. Thus, it is understandable that traders and banks are quite keen on the potential of blockchain to reduce costs and perhaps even revolutionize the trading business.
But before you get too excited, a remark by my friend Christophe Salmon at Trafigura is latent with deep implications that should lead you to take pause and consider the likely consequences of widespread adoption of blockchain:
Christophe Salmon, Trafigura’s chief financial officer, said there would need to be widespread adoption by major oil traders and refiners to make blockchain in commodity trading viable in the long term.
This seemingly commonsense and innocuous remark is actually laden with implications of unintended consequences that should be recognized and considered now, before the blockchain train gets too far down the track.
In essence, Christophe’s remark means that to be viable blockchain has to scale. If it doesn’t scale, it won’t reduce cost. But if it does scale, a blockchain for a particular application is likely to be a natural monopoly, or at most a natural duopoly. (Issues of scope economies are also potentially relevant, but I’ll defer discussion of that for now.)
Indeed, if there are no technical impediments to scaling (which in itself is an open question–note the block size debate in Bitcoin), the “widespread adoption” feature that Christophe identifies as essential means that network effects create scale economies that are likely to result in the dominance of a single platform. Traders will want to record their business on the blockchain that their counterparties use. Since many trade with many, this creates a centripetal force that will tend to draw everyone to a single blockchain.
I can hear you say: “Well, if there is a public blockchain, that happens automatically because everyone has access to it.” But the nature of public blockchain means that it faces extreme obstacles that make it wildly impractical for commercial adoption on the scale being considered not just in commodity markets, but in virtually every aspect of the financial markets. Commercial blockchains will be centrally governed, limited access, private systems rather than a radically decentralized, open access, commons.
The “forking problem” alone is a difficulty. As demonstrated by Bitcoin in 2013 and Ethereum in 2016, public blockchains based on open source are vulnerable to “forking,” whereby uncoordinated changes in the software (inevitable in an open source system that lacks central governance and coordination) result in the simultaneous existence of multiple, parallel blockchains. Such forking would destroy the network economy/scale effects that make the idea of a single database attractive to commercial participants.
Prevention of forking requires central governance to coordinate changes in the code–something that offends the anarcho-libertarian spirits who view blockchain as a totally decentralized mechanism.
Other aspects of the pure version of an open, public blockchain make it inappropriate for most financial and commercial applications. For instance, public blockchain is touted because it does not require trust in the reputation of large entities such as clearing networks or exchanges. But the ability to operate without trust does not come for free.
Trust and reputation are indeed costly: as Becker and Stigler first noted decades ago, and others have formalized since, reputation is a bonding mechanism that requires the trusted entity to incur sunk costs that would be lost if it violates trust. (Alternatively, the trusted entity has to have market power–which is costly–that generates a stream of rents that is lost when trust is violated. That is, to secure trust prices have to be higher and output lower than would be necessary in a zero transactions cost world.)
But public blockchains have not been able to eliminate trust without cost. In Bitcoin, trust is replaced with “proof of work.” Well, work means cost. The blockchain mining industry consumes vast amounts of electricity and computing power in order to prove work. It is highly likely that the cost of creating trusted entities is lower than the cost of proof of work or alternative ways of eliminating the need for trust. Thus, a (natural monopoly) commercial blockchain is likely to have to be a trusted centralized institution, rather than a decentralized anarchist’s wet-dream.
Blockchain is also touted as permitting “smart contracts,” which automatically execute certain actions when certain pre-defined (and coded) contingencies are met. But “smart contracts” is not a synonym for “complete contracts,” i.e., contracts where every possible contingency is anticipated, and each party’s actions under each contingency is specified. Thus, even with smart (but incomplete) contracts, there will inevitably arise unanticipated contingencies.
Parties will have to negotiate what to do under these contingencies. Given that this will usually be a bilateral bargaining situation under asymmetric information, the bargaining will be costly and sometimes negotiations will break down. Moreover, under some contingencies the smart contracts will automatically execute actions that the parties do not expect and would like to change: here, self-execution prevents such contractual revisions, or at least makes them very difficult.
Indeed, it may be the execution of the contractual feature that first makes the parties aware that something has gone horribly wrong. Here another touted feature of pure blockchain–immutability–can become a problem. The revelation of information ex post may lead market participants to desire to change the terms of their contract. Can’t do that if the contracts are immutable.
Paper and ink contracts are inherently incomplete too, and this is why there are centralized mechanisms to address incompleteness. These include courts, but also, historically, bodies like stock or commodity exchanges, or merchants’ associations (in diamonds, for instance) have helped adjudicate disputes and to re-do deals that turn out to be inefficient ex post. The existence of institutions to facilitate the efficient adaption of parties to contractual incompleteness demonstrates that in the real world, man does not live (or transact) by contract alone.
Thus, the benefits of a mechanism for adjudicating and responding to contractual incompleteness create another reason for a centralized authority for blockchain, even–or especially–blockchains with smart contracts.
Further, the blockchain (especially with smart contracts) will be a complex interconnected system, in the technical sense of the term. There will be myriad possible interactions between individual transactions recorded on the system, and these interactions can lead to highly undesirable, and entirely unpredictable, outcomes. A centralized authority can greatly facilitate the response to such crises. (Indeed, years ago I posited this as one of the reasons for integration of exchanges and clearinghouses.)
And the connections are not only within a particular blockchain. There will be connections between blockchains, and between a blockchain and other parts of the financial system. Consider for example smart contracts that in a particular contingency dictate large cash flows (e.g., margin calls) from one group of participants to another. This will lead to a liquidity shock that will affect banks, funding markets, and liquidity supply mechanisms more broadly. Since the shock can be destabilizing and lead to actions that are individually rational but systemically destructive if uncoordinated, central coordination can improve efficiency and reduce the likelihood of a systemic crisis. That’s not possible with a radically decentralized blockchain.
I could go on, but you get the point: there are several compelling reasons for centralized governance of a commercial blockchain like that envisioned for commodity trading. Indeed, many of the features that attract blockchain devotees are bugs–and extremely nasty ones–in commercial applications, especially if adopted at large scale as is being contemplated. As one individual who works on commercializing blockchain told me: “Commercial applications of blockchain will strip out all of the features that the anarchists love about it.”
So step back for a minute. Christophe’s point about “widespread adoption” and an understanding of the network economies inherent in the financial and commercial applications of blockchain means that it is likely to be a natural monopoly in a particular application (e.g., physical oil trading) and likely across applications due to economies of scope (which plausibly exist because major market participants will transact in multiple segments, and because of the ability to use common coding across different applications, to name just two factors). Second, a totally decentralized, open access, public blockchain has numerous disadvantages in large-scale commercial applications: central governance creates value.
Therefore, commercial blockchains will be “permissioned” in the lingo of the business. That is, unlike public blockchain, entry will be limited to privileged members and their customers. Moreover, the privileged members will govern and control the centralized entity. It will be a private club, not a public commons. (And note that even the Bitcoin blockchain is not ungoverned. Everyone is equal, but the big miners–and there are now a relatively small number of big miners–are more equal than others. The Iron Law of Oligarchy applies in blockchain too.)
Now add another factor: the natural monopoly blockchain will likely not be contestible, for reasons very similar to the ones I have written about for years to demonstrate why futures and equity exchanges are typically natural monopolies that earn large rents because they are largely immune from competitive entry. Once a particular blockchain gets critical mass, there will be the lock-in problem from hell: a coordinated movement of a large set of users from the incumbent to a competitor will be necessary for the entrant to achieve the scale necessary to compete. This is difficult, if not impossible to arrange. Three Finger Brown could count the number of times that has happened in futures trading on his bad hand.
Now do you understand why banks are so keen on the blockchain? Yes, they couch it in terms of improving transactional efficiency, and it does that. But it also presents the opportunity to create monopoly financial market infrastructures that are immune from competitive entry. The past 50 years have seen an erosion of bank dominance–“disintermediation”–that has also eroded their rents. Blockchain gives the empire a chance to strike back. A coalition of banks (and note that most blockchain initiatives are driven by a bank-led cooperative, sometimes in partnership with a technology provider or providers) can form a blockchain for a particular application or applications, exploit the centripetal force arising from network effects, and gain a natural monopoly largely immune from competitive entry. Great work if you can get it. And believe me, the banks are trying. Very hard.
Left to develop on its own, therefore, the blockchain ecosystem will evolve to look like the exchange ecosystem of the 19th or early-20th centuries. Monopoly coalitions of intermediaries–“clubs” or “cartels”–offering transactional services, with member governance, and with the members reaping economic rents.
Right now regulators are focused on the technology, and (like many others) seem to be smitten with the potential of the technology to reduce certain costs and risks. They really need to look ahead and consider the market structure implications of that technology. Just as the natural monopoly nature of exchanges eventually led to intense disputes over the distribution of the benefits that they created, which in turn led to regulation (after bitter political battles), the fundamental economics of blockchain are likely to result in similar conflicts.
The law and regulation of blockchain is likely to be complicated and controversial precisely because natural monopoly regulation is inherently complicated and controversial. The yin and yang of financial infrastructure in particular is that the technology likely makes monopoly efficient, but also creates the potential for the exercise of market power (and, I might add, the exercise of political power to support and sustain market power, and to influence the distribution of rents that result from that market power). Better to think about those things now when things are still developing, than when the monopolies are developed, operating, and entrenched–and can influence the political and regulatory process, as monopolies are wont to do.
The digital economy is driven by network effects: think Google, Facebook, Amazon, and even Twitter. In addition to creating new efficiencies, these dominant platforms create serious challenges for competition, as scholars like Ariel Ezrachi and Maurice Stucke have shown:
Peter Thiel, the successful venture capitalist, famously noted that ‘Competition Is for Losers.’ That useful phrase captures the essence of many technology markets. Markets in which the winner of the competitive process is able to cement its position and protect it. Using data-driven network effects, it can undermine new entry attempts. Using deep pockets and the nowcasting radar, the dominant firm can purchase disruptive innovators.
Our new economy enables the winners to capture much more of the welfare. They are able to affect downstream competition as well as upstream providers. Often, they can do so with limited resistance from governmental agencies, as power in the online economy is not always easily captured using traditional competition analysis. Digital personal assistants, as we explore, have the potential to strengthen the winner’s gatekeeper power.
Blockchain will do the exact same thing.
You’ve been warned.
*My understanding of these issues has benefited greatly from many conversations over the past year with Izabella Kaminska, who saw through the hype well before pretty much anyone. Any errors herein are of course mine.
The Anarchists Blockchain – anonymous & decentralized – doesn’t make a very good currency because of the high transaction costs, and because anonymity and liquidity are natural enemies. You might as well call the units “Tulips”.
The Big Business Blockchain is just a slightly more secure version of the Git source code version control system, and really not that useful. Most contracts involve a small number of people and they’ll work just fine with digital signatures.
Neither flavor really solves the problem of what happens outside of their internally consistent bubble.
The Bitcoin bubble burst is going to be fun to watch.
Comment by Richard Harrington — April 4, 2017 @ 9:00 pm
Hi Craig,
Could you please elaborate a bit more about how this is different than existing MI structure? E.g. there is a “monopoly” in each market for CSDs, which seems like the closest equivalent in a lot of ways, as well as central bank payments infrastructure. Your typical asset manager can’t interact directly with DTCC or the Fed, which is why they need to pay State Street or BoNY. But it’s still pretty cheap, since they compete against each other.
Comment by Rob — April 5, 2017 @ 9:08 am
@Rob-That’s kinda my point. The same centripetal forces that make exchanges and CCPs and CSDs natural monopolies (or close thereto) will operate in the blockchain environment.
DTCC is an interesting case. It was mandated by the SEC in the 1960s to create an industry utility that was constrained in its ability to exercise market power. It operates on a cost basis, and is pretty much open access. OCC is a similar example. In each case, regulation constrained the natural monopoly, and made the service pretty cheap.
My point is basically that similar regulation may be required in the blockchain space.
There are nearly four dozen cryptocurrencies with market cap of $10 million or more, and in the last 4 years Bitcoin’s share has dropped from about 95% to under 75% of total market cap. That doesn’t look like inexorable centripetal force to me. Bitcoin mining has consolidated mostly because razor-thin profit margins favor economies of scale, which doesn’t sound like extraction of monopoly rents, either. Ebay can maintain an entrenched position because (a) auction sites charge listing fees, and (b) a particular item can only be listed on one auction site at a time, so a seller pays an opportunity cost to list on a non-dominant auction site. If blockchain-based trading is implemented in a way that minimizes such opportunity costs (a big “if”, I suppose) then sellers can list their wares on multiple blockchain platforms and buyers search across multiple platforms, and indeed, client software can make such actions transparent so that the trading forum is as much an afterthought as the airline you choose on Orbitz. The costs involved in proof-of-work and burgeoning blockchain size have technical solutions; the main difficulty in the cryptocoin movement is to solve them without sacrificing cypherpunk levels of anonymity.
As Mr. Harrington noted in his comment, a full-on cyperpunky blockchain is overengineered for the problems faced by commodity traders. Not many traders are going to want to pay for pregnancy test kits or Prozac or some other confidential purchase with a slice of oil futures. The main features I see are non-deniability and being able to publicly audit that inventory isn’t oversold (or that a buyer isn’t overdrawn), and for that a notary-like authentication is sufficient. No need to compute billions of SHA256 hashes every second.
Comment by M. Rad. — April 5, 2017 @ 12:12 pm
I never really understood this “centralized blockchain” thing.
The blockchain technology’s main feature point is that it’s decentralized. For this feature it pays price in complexity, high latency and insane amount of wasted computational power. If you don’t need to use the “decentralized” feature, there are hunrdeds of systems existing today (heh, even 30 years ago) that can do what blockchain does, just simpler, faster and waaaaaay cheaper.
Comment by tegla — April 5, 2017 @ 1:37 pm
Like tegla said ?
Comment by Tony — April 6, 2017 @ 2:56 am
Really insightful post. Many thanks for your thinking on this.
Comment by Bernard Golden — April 6, 2017 @ 8:37 am
[…] here is Craig Pirrong on the blockchain in financial […]
Pingback by Lacker Leaks and Model Errors — April 7, 2017 @ 3:10 am
@The Professor
Okay yeah I see where you’re coming from in that case. I agree that banks are seeing blockchain systems as just another financial market utility, albeit with more potential to cut down on BO reconciliation costs.
To me though, it doesn’t make me that concerned – inter-bank utilities don’t generally have extortionate profit margins. Is anyone really worried about the fact that e.g. CLS, or the varous CSDs, or even LCH and other CCPs are capturing way too much profit from the financial system?
Comment by Rob — April 7, 2017 @ 3:27 am
Kudos for your quotes in the WSJ article (4/7) on rumours of the sale of Gunvor
Comment by AndyEss — April 7, 2017 @ 11:49 am
[…] The Unintended Consequences of Blockchain Are Not Unpredictable: Respond Now Rather Than Repent Late… […]
Pingback by Links, 7 April 2017 | illiquid ideas — April 7, 2017 @ 12:51 pm
[…] Here is a link to the entire post, and a quote. Read it. […]
Pingback by Innovation Has Unintended Consequences | Points and Figures — April 8, 2017 @ 6:46 am
really great blogpost. I have looked at bitcoin in all kinds of ways and wound up with a positive view of it (mostly). this post brings into play lots of things I hadn’t thought about. One place I think bitcoin/blockchain could be very powerful is inside government. voting using a blockchain for example. could it be structured in a way to end voter fraud?
Comment by pointsnfigures — April 8, 2017 @ 6:50 am
@pontsnfigures, I don’t know what a blockchain can do for the casting of ballots that isn’t better with plain old paper, but there could be a scheme to use cryptographic signatures to control and track the issuance of blank ballots, so that convenient discoveries of ballots in the trunk of some car can be authenticated or denied cryptographically. (I’m thinking of a protocol where each ballot has a nonce signed by a key whose public component is announced to a blockchain when the package is opened. If a trove of ballots shows up in a suspiciously convenient circumstance, the signatures on the nonces can be checked against the announced public keys.) There may be a way to use zerocoin-style blockchains to allow voter verification, but there are ways to do this with paper.
Comment by M. Rad. — April 9, 2017 @ 8:27 pm
[…] Источник […]
Pingback by Блокчейны: лучше реагировать сейчас, чем сожалеть потом - — September 9, 2017 @ 2:09 am
[…] “The Unintended Consequences of Blockchain Are Not Unpredictable: Respond Now Rather Than Repe… was published by Streetwise Professor […]
Pingback by Market structure implications of blockchain – Tango with the Future — September 12, 2017 @ 1:08 pm
[…] Источник […]
Pingback by Блокчейны: лучше реагировать сейчас, чем сожалеть потом | Pinme.club — October 4, 2017 @ 5:00 am
[…] Streetwise Professor: The Unintended Consequences of Blockchain Are Not Unpredictable: Respond Now Rather Than Repent Late… […]
Pingback by The Weekend Read: May 21 – R3 — November 6, 2017 @ 12:57 pm