Just another database format? On blockchains, governance, enterprise systems, and economics
Blockchains are typically only perceived in the context of private-token peer-to-peer payment systems, a.k.a. “cryptocurrencies”. They could have a much more significant impact under the hood of global enterprises. Here is a summary of the why and how.
Originally posted on twitter by @oliverbeige.
For the last forty years the economic profession has shown little interest in the enterprise systems that form the backbone of the global economy. Why should it change its tack now?
If one is to believe some contemporary accounts, blockchain is a fairydust-based rocket propellant that will help us all travel to the Moon in no time, and make us younger, richer, and more democratic in the process.
If you ask others, the story sounds more prosaic: we are talking about an overengineered database for keeping accounting records — a massively redundant, energy-sucking, somewhat scammy transaction log.
Given economics’ somewhat tenuous relationship with corporate accounting and its general disinterest in enterprise systems: why should economists take more than a cursory look at this apparent hype topic?
Granted, specialization in the field makes sure that someone somewhere knows everything about the economics of dental prophylaxis, but hype aside: the invention of blockchain warrants a closer look than economists usually give healthy teeth, or databases.
The nascent field of “cryptoeconomics” combines computer science (distributed systems plus cryptography) with game theory and mechanism design to calibrate the consensus process that decides which transactions go on record in which order.
This is mostly driven by computer scientists and methodologically it leans more on engineering and operations research than economics. But it is not the current participants’ fault that economists mostly remain on the sidelines at this juncture.
The problem is that much of what drove the relevant shifts in economic organization over the last twenty-plus years has happened off the radar of economists, hidden inside firms, so to properly assess the sheer magnitude a shift in data aggregation and storage might have on the economy at large, we should have a look back first.
Leonid’s one little trick
One of the dirty secrets of global capitalism is that the scientist who arguably had the biggest stake in kicking it off was also the only human to ever receive both the Economics Nobel and the Stalin Prize.
I am speaking of Leonid Kantorovich, whose invention of linear optimization was originally designed to minimize waste when cutting shapes out of plywood boards in the war-torn Soviet economy.
The western world took no notice at first, so it was later re-invented by Tjalling Koopmans from Holland and American George Dantzig. Koopmans later shared the Nobel with Kantorovich — Dantzig was overlooked.
Supported by early mainframe computing, Kantorovich’s invention fueled a short-lived attempt to solve the resource allocation problem in the entire Soviet Union using “cybernetics” — the simulation and optimization of the economy as a hermetic, circular, centrally controlled system.
In the end it didn’t work of course. The task was too daunting, the methods too novel, the mainframe computers too slow and unreliable, the ideologies too rigid and the bureaucratic corruption too deeply entrenched.
But for a brief moment in time, the Soviet economy looked like it could be competitive.
Kantorovich’s “one little trick” was to take a system of linear equations and to leave one equation — the objective function — variable. All others then become resource constraints and we can find the value which optimizes this system: maximize productivity or minimize waste, depending on framing (the two are equivalent, as John von Neumann showed).
Kantorovich’s linear optimization setup was soon expanded into more complex domains — dynamic, stochastic, combinatorial, what have you — and today it is ubiquitous.
Any time we ask ourselves whether we can improve a system and try to come up with a formal, mathematical answer, we apply some sort of constrained optimization model.
It drives finance, engineering, machine learning, and in particular it drives the estranged fraternal twins of operations research and modern economics. When Facebook tries to guess what might interest you most, they use Kantorovich’s one little trick.
Autonomous vehicles, dating websites, search engines, machine translations all optimize. Today, the idea that was supposed to modernize centralized planning in communist economies powers global commerce.
The secret history of enterprise systems
The major reason why this systematic, process-driven, and information technology-fueled way of shifting things around caught on in the West but not behind the Iron Curtain was initially more practical than political.
Enterprises are simply smaller units to model, their process flows tend to be linear rather than circular, and enterprises — unlike countries — are allowed to fail. The social costs of getting it wrong are still steep, but nowhere near as as steep as failing national economies.
The supposed flaw of capitalism, its tendency to run redundant production lines that inevitably produce overstock, became its strongest point — partly because overstock hurts companies and shortages consumers.
With improved record keeping came improved prediction, reduced redundancy, and ultimately a better flow of goods, services and finances: an improvement in the allocation of economic resources we call “innovation”.
This did not go off without a hitch of course, and management scholars like Henry Mintzberg repeatedly threw up their hands in disgust at the inability of enterprise strategic planners to shift things around properly — at times it seemed like capitalist planning was going the way of communist planning.
But with the benefit of hindsight we can look back at multiple hype-and-disappointment cycles to realize — whatever we might think of it as political animals — consumers have embraced globalization wholesale.
We expect exotic fruits year-round, and we simply don’t care that our “Made in Germany” cars or our Californian iPhones are built from components from a couple dozen source countries.
Unless disrupted by unforeseen events, be it an Icelandic volcano named Eyjafjallajökull or the sub-prime housing crisis that triggered the Great Recession, the rumblings of international goods transfers (and the financial transfers going in the opposite direction) can be perfectly ignored.
The enterprise as a knotty ball of legal entanglements
There are many ways to look at an enterprise. The standard (“textbook” or “orthodox” or “neoclassical”) economic way is to think of it as a production function: inputs of all sorts are assembled and repurposed until they turn into a product which hopefully sells for more than the cost of all inputs and efforts combined.
We could look at the sociological enterprise: a hub of collective action held together by routines and rituals, by power and influence; or the legal enterprise: a “nexus of contracts” where flows of materials, finances and human efforts are the result of agreements, promises and obligations.
The fundamental ingredient of such a contract is a transaction, typically taking on the form of “Florian to furnish Fiona with five firlots of flour for fifty-four florins by Friday.”
Transactions capture that core building block of Adam Smith’s market economy: a (hopefully) voluntary exchange of things that will (hopefully) make both sides better off. The argument is then that transactions also happen not only across markets, but also, similarly, inside firms.
Ronald Coase took this argument a step further and claimed that the complexity of a transaction decides whether it is conducted within an enterprise or across a market, since market transactions come at a price.
There are many reasons why a transaction can become more complex than the Fiona’s florins example above. Transactions can include multiple steps of delivery or payment stretched out over time— typically asynchronous and conditional.
They can contain all kinds of surprises which might become known only later, from counterfeit pharmaceuticals to exploding batteries. At any time during or after, either side can feel the urge to withdraw from the transaction or push to alter the contract — renege on it — if they feel that the cost of doing so is lower than the cost of sticking to the contract terms.
Companies internalize dispute resolution. Any transaction inside a company comes with a resolution process baked in. Right or wrong, hierarchy gets to make the final call, and oftentimes even a wrong call is preferable over neverending debate.
The other end of the spectrum is not quite as well established. If hierarchical firms are so much better at handling anything except simple on-the-spot transactions, what is the upper limit of firms? Arguably there could be one firm controlling a whole economy after all?
Coase published his argument in 1937, when business record-keeping was still done with pencil, paper, and abacus. Eight years later, Friedrich Hayek pointed out why the economy should not be run as one single all-controlling enterprise.
Nevermind the advances in operations, the economy needs not only scientific inventiveness but also opportunity recognition. The best waypointers towards lucrative opportunities are prices, and only true markets can provide true prices: the marvel of the price system.
Today firms bundle their accounting, analytics, resource tracking and planning into large-scale enterprise systems. A modern enterprise might run any number of so-called “end-to-end processes”: cross-functional, formalized if-then relationships with predefined scope for human intervention.
Processes are usually named after their endpoints, like “order-to-cash” or “procure-to-pay”. The things being shifted around within these processes are the basic components of the global economy: materials, money, information and effort.
In some processes these components run separately, like supply chain management for materials, human resources for effort, and financial consolidation for money. Elsewhere they intersect, such as in procurement, sales, or notably, accounting.
Artificial intelligence as abacus 4.0
Over the last decades, Coase and Hayek’s early mainframe-powered corporations have become faster and, surprisingly, leaner. As supply chains become more complex, they have not integrated vertically but rather disintegrated.
Even though hold-up problems haven’t gone away — quite to the contrary: the controlling entities, typically the companies that put their brand on the final product, have relied more and more on market interactions in the form of outsourcing, for multi-tier assembly.
Four or five tiers in a supply chain are not uncommon, and in a just-in-time setting suppliers sometimes have to fly in their components with helicopters to keep production lines running and avoid steep penalties.
The other basic type of economic interaction, the market as the agora where supply and demand meet, has shifted from a governance model to an IT operation mostly concerned with throughput. Markets themselves have become enterprised.
The New York Stock Exchange, which started as a gentlemen’s agreement under a buttonwood tree to curb the speculative tendencies of its founding members, is now the subsidiary of Intercontinental Exchange, a global IT company running two dozen stock and commodity exchanges.
The crying of the Walrasian almost-auctioneer
The Walrasian auctioneer is a figure of momentous importance in economics, as it’s he who helps price-taking buyers and sellers grope towards a market-clearing equilibrium in a perfectly competitive world. Except he never existed.
Up until 2014, the English speaking economic world (which is pretty much everybody except the French) assumed that Leon Walras described equilibrium convergence as a result of a two-sided auction conducted by said auctioneer.
When Don Walker and Jan van Daal translated the third edition of the “Elements”, they found that the Walrasian auctioneer was in fact a “crieur”, a market crier who announces the prices for all to hear but doesn’t set them. The economic community took little notice.
Somewhat contrary to its reputation, economics has developed a voluminous body of work, both theoretical and empirical, that tries to account for and propose remedies against market imperfections.
Asymmetric information, externalities, contracting hazards, public goods, market power are well researched and fairly well understood causes for market perturbations that might lead to, in econ-speak, second-best outcomes.
There is no market imperfection known as “the auctioneer trying to line his own pockets”, seeing that the market itself is not considered a real price-taking participant, but a surplus-dispensing mechanism. The auctioneer as an omniscient, selfless, costless mechanical turk.
Except no modern market, except maybe the Thai market every summer Sunday in Berlin-Wilmersdorf (come visit!), works just like its economic abstraction. Both matchmaking and governance are costly endeavors.
Over the last twenty-ish years we have witnessed the optimization of its operations in the guise of “platform business models” aka markets as enterprises, but of course banks have offered market-making services long before the internet was invented.
Economists from Vickrey to Varian have had their fingerprints all over Silicon Valley business models, and there isn’t a market imperfection that didn’t compel a startup to build its underpants collection service around.
Today, market imperfections are more likely the starting point for new ventures than for antitrust scrutiny. I call this the “Michael Porter flip”. This is not a bad thing in itself: As long as we can find good private solutions we shouldn’t invoke government fixes.
But it is a bad thing if we don’t understand the fundamental mechanics of the very thing economists should be studying, especially if we are faced with a long-term trend that rearranges the power balance between economic actors.
Even the economists writing the amicus brief for Ohio vs AmEx admit that the “economic literature analyzing two-sided platforms is new, complex, and evolving.” And two-sided markets are actually easy, because they include two distinct sides, buyers and sellers.
In a peer-to-peer world we don’t have that luxury. Any participant could be either. The platform literature goes back to the mid-1980s and is voluminous and well established compared to the population dynamics literature capturing peer interaction.
Even Robert Aumann’s Nobel-worthy conception of a Walrasian auctioneer as a game-theoretic solution concept, the correlating device (aka the Aumann machine), assumes that this device operates costlessly, frictionlessly, and devoid of selfish interests.
Our understanding of how Aumann machines, Walrasian market criers, trusted intermediaries, P2P platform models, markets-as-enterprises evolve, operate, and affect the economy at large is still rudimentary. But we should be clear that this is the central economic problem of the 21st century.
The distinguishing feature of a workable capitalist system is not that it runs on perfect competition, but that it operates close to a “Coasean optimum”: economic activity is distributed between enterprises and markets in a right balance between productivity and limitation of market power.
In the words of Douglass North & Robert Thomas, “efficient organization entails the establishment of institutional arrangements and property rights that create an incentive to channel individual economic effort into activities that bring the private rate of return close to the social rate of return.”
To the extent where markets and enterprises achieve this balance by themselves, the need for government interference is limited.
But this presumes markets and enterprises as clearly separated governance structures. What if one gets subsumed by the other?
Even Hayek’s seminal article, widely misinterpreted to say that markets are perfect and self-regulating, merely pointed out that “scientific” resource planning and “ad-hoc” resource allocation are complementary functions in an economy.
This balance has shifted significantly over the years due to the increased power of enterprise systems and the rise of the internet and cloud computing.
It is slated to shift even more as transactional and analytical functions are bound to merge, bringing together production and accounting data. “Scientific planning” closes in on “ad-hoc”.
The effect of such a shift, the subsumption of a market into an enterprise and the replacement of “good governance” with “profit-maximizing throughput” is subtle. After all, a market-as-enterprise can only operate as a long as it keeps the participants happy. To a degree.
But “optimize market participant welfare subject to operating cost constraints” and “optimize returns to operation subject to market participation constraints” are fundamentally different objectives.
The Satoshi white paper, written at the low point of the Great Recession, not only pinpointed the “middleman problem” of modern capitalism, but also proposed a solution: to give markets their own accounting system, separate from the enterprise systems that run the middleman economy.
Self-driving governance mechanisms
To assess the relevance of blockchain to economics, it might be useful to sidestep the politically charged topics for the moment, to postpone judgment on the feasibility of any currently proposed solutions, and to focus entirely on the problem it tries to tackle.
We can, if we want, divide the “machine age” into five phases that align with the automation of particular enterprise functions: production, administration, supply, demand. One function that still largely resists automation is governance.
This might be somewhat counterintuitive since accounting, a major component of enterprise governance, was one of the earliest adopters of mainframe computing.
But the role of automation in accounting is largely restricted to administration, the inward-looking part. The outward-looking part of the enterprise, the right side of the balance sheet where the enterprise has to demonstrate good custodianship of the assets it holds and employs on behalf of external claimants (creditors & shareholders), requires both human judgment and discretion.
The reason why we are seeing automation happen in this order is largely due to the difficulty of the underlying tasks. Internal is easier to automate than external, goods are easier than services, quantities are easier than quality. It’s easier to maximize corporate profits than human wants.
But each phase not only transforms the ground rules of the function it automates, it also triggers a fundamental shift in the adjacent functions, and in the objective of the enterprise itself. In turn it also puts pressure on these adjacent functions to automate.
The last two major phases of enterprise automation, business process automation (ca. 1970–2000) and business model automation (since 2000) were both enabled by technological shocks (enterprise computing & databases, internet & distributed/cloud computing) but the formal groundwork had been laid before — at times even decades before.
Business processes go all the way back to Kantorovich and Dantzig, business models go all the way back to Vickrey and Shapley.
One could claim that governance automation goes all the way back to Coase, who framed enterprises as legal rather than economic entities, as contractual arrangements to bundle all collective activities which, if conducted on a spot market, would lead to friction and misallocation.
This is something of a counterpoint to Hayek’s “marvel of the price system” (itself expressed as a counterpoint to the cybernetic proposition that all economic activity could optimally be located within a single entity).
Under certain conditions, price acts as a check on planning. It is the most succinct signal that a business might be in for a negative (or positive) surprise.
The price system is a marvel, but it has its scope. And what exactly the scope of the enterprise vs the market is has been the focus of the field of industrial organization (and its fraternal twin, corporate strategy).
The current shift from dominance of processes to models, from scale economies on the supply side (production and operations) to the demand side (interaction and network effects) is largely the result of the shift from hardware to software, from physical goods to digital services.
It is also the result of the increasing availability of preference data (“likes”) in addition to choice data (“checkout swipes”) which turn out to be a powerful early prediction tool for those who can integrate these data into their value chain.
With business processes having reached maturity and business models at the peak of their economic power, we now not only understand their benefits. We are also more and more confronted with their drawbacks, which feed into the impetus for creating credible governance mechanisms.
It is still early enough for all kinds of predictions about the ultimate fate of blockchains to be in play: world-changing technology or biggest scam in history. It is conceivable that shared ledgers emerge that don’t rely on a hashchain of transaction blocks.
Many of the early mental models that were coined without considering the exigencies of economic efficiency will fall by the wayside. The firm is still the strongest means to orchestrate the production of economic value. Maybe a fraction of the world population wants to be its own bank, but the large majority neither has the experience nor the inclination.
But what is emerging is the need for governance forms that on the one side fit in with the need for automated operations, and on the other side for the clear and credible commitment by the provider to safeguard the user against opportunism, “greenwashing”, and expropriation.
Whoever is able to create these forms will win the next stage of economic development.