Can computers catalyse centralisation?
Did communism fail for a lack of processing power?
Reflecting on past failures of centrally planned economies, Paul Mason’s Postcapitalism: A Guide to Our Future1 asks:
What if the problem with the Soviet Union was that it was too early? What if our computer processing power and behavioural data are developed enough now that central planning could outperform the market when it comes to the distribution of goods and services?
It is a fascinating counterfactual exploration. The question is especially relevant when considering the apparent disenchantment with capitalism building since the Great Recession, and an increasing trend toward more centralised firms (e.g. Google, Amazon) and institutions (e.g. global governance of taxation and climate change management). Is this sentiment not just reactionary politics but the natural consequence of a revolution in the processing and use of information?
F. A. Hayek observed in The Road to Serfdom2 that socialism (and hence Communism) requires central economic and social planning (and central planning leads to totalitarianism), but all the relevant information is decentralised. It is costly (if not impossible) to obtain all the relevant information to enable centralised authorities to organise economic and social interactions as efficiently as possible in imperfect, decentralised (i.e., market-based) alternatives.
The original syn
One attempt to centrally collect economic data to inform decision-making was Project Cybersyn, which ran 1971—73 in Chile. The Cybersyn control room was created by Chile’s president Salvador Allende as a place from which the country’s newly nationalised and socialised economy could be directed. Allende had thought that, with state-of-the-art 1970s communications and computers, it would be possible for government to optimise an industrial economy. Cybersyn ended following a military coup.
Journalist Evgeny Morozov hypothesised that Cybersyn paved the way for how technology would collect and use data.
Outperformed by a smartphone
The IBM 360/50 mainframe computer (and later, a Burroughs B3500) and telex machines that powered Cybersyn are decidedly underpowered by today’s standards — a $50 smartphone would almost certainly outperform them.
Without a doubt, advances in computers and information processing technologies have markedly altered the cost-benefit trade-offs of collecting and processing data. Specifically, with the power of artificial intelligence algorithms, and terabytes of data amassed from collecting users’ engagement with an army of online apps, central entities might now possibly know enough to tip the balance back toward the levels of centralisation observed not just in Communist states but also in many Western economies of the pre-liberalised 1980s heydays.
A greater appetite for centralised control
Equally, a greater appetite for centralised control has emerged post-2000, as politicians and populations appear engulfed in growing uncertainty on many different fronts: economic, environmental, and geopolitical, to name a few. As psychologists Daniel Kahneman and Amos Tversky and economists Cass R. Sunstein, John Kay, and Mervyn King have demonstrated, when faced with great uncertainty (or chaotic circumstances), humans tend to cede responsibility for making decisions to a “higher” or more centralised power (e.g., political leaders). Moreover, those charged with making decisions tend to be biased toward taking actions — any actions — over waiting to gain more information, even when waiting would have been the rational option. Being seen doing something is an important signal of being in control and exercising strong leadership to those on whose behalf the decision is being made. (Of course, these same scholars also warn about the tendency to over-assess the magnitude of low-probability, high-cost outcomes and under-assess higher-probability, lower-cost outcomes, often with deleterious consequences.)
So, are these two forces — lower-cost computerisation and uncertainty — leading inevitably to centralisation, reminiscent of that observed in the 1930s? And will it succeed this time, when it did not previously, because the cost-benefit equation is different?
Evidence from the centralisation “experiment” of COVID-19
The recent natural experiment with centralisation offered by Covid offers some insights. To be sure, when a crisis emerges (in the short term), there is usually a need for someone to assume coordination of the immediate response — to preclude the potential for chaos. However, the extent to which a centralised entity can continue to manage all economic and social activity efficiently or effectively from then on must be questioned.
All the information gathered and algorithms developed in the past are of little value when confronting a truly unexpected and unprecedented situation. When there simply is no legacy information on the situation faced, centralised authorities are exposed. Centralised systems operate on rules, where one size fits all. They are sustainable when they are low cost to enforce. But when the population is diverse and the rules are not a good fit, then the risk of a breach rises and enforcement costs increase. The more exceptions that are created, the higher the enforcement costs. Eventually, either enforcement costs overcome any benefits from centralised operation, and some form of planned decentralisation must occur, or revolution leads to the same result.
The primary weakness of centralised systems in the long run thus lies in their one-size-fits-all approach. When uncertainty and diversity prevail, “one size” is unlikely to be optimal. Variety, fundamental to decentralised systems, is costly, so it isn’t usual in centralised systems. Portfolio theory suggests that multiple, small low-cost options offer a better chance of finding the best response in an uncertain environment; the “best” can be promoted once identified.
Finding the sweet spot on the centralisation—decentralisation continuum is never easy, and remains a moving target. Allowing it to evolve seems better guidance for governance design than trying to prescribe it in advance, as the future is always uncertain. Joseph Stalin would be sorely disappointed.
This post was originally published at AEIdeas on 8 December 2022. It is based on an invited address to the Law and Economics Association of New Zealand, of which the author is a Fellow, on 6 December 2022. Additional material by Olivia Wills.
Paul Mason (2017). Postcapitalism: A Guide to Our Future. macmillan.
F. A. Hayek (1944). The Road to Serfdom, George Routledge & Sons.
Can I suggest that when looking for the sweet spot on the centralisation—decentralisation continuum a good place to start would be Arnold Plant (1937), 'Centralize or Decentralize?', in Arnold Plant (ed.), Some "Modern Business Problems: A Series of Studies" (pp. 3-33), London: Longmans, Green and Co.
A casual reading of this article suggests that communism failed primarily because of technological problems with centralisation. What about the lack of incentives to develop new products that consumers actually want? What also of the consequences of power being given to zealots? Convinced that they are right, communist regimes unsurprisingly confiscate property (ill-gotten gains?), leading to resistance from the owners of that property, leading to their mass murder by the regime. Convinced that they are right, communist regimes are also unsurprisingly intolerant of dissent, leading to its criminalisation, and therefore suppression of feedback, and therefore persistence with policies and leaders that do not work. Of course, communism has no monopoly on zealotry and authoritarian government, but the consequences of that combination are always adverse: just look at Iran.