Forget ‘price gouging’ – this is where competition is really failing
Rachel Reeves is scapegoating supermarkets for rising oil prices while ignoring algorithms that can learn ant-competitive pricing strategies, says Paul Ormerod
The government is desperately trying to convince the public that it is doing something about the potential economic crisis which is unfolding.
The public finances are a severe constraint on its ability to throw cash at the problem. The yield on the benchmark 10-year UK government bond has already been above five per cent, higher than it was under the ill-fated premiership of Liz Truss.
And no government can escape the basic fact that, as an energy consuming nation, the rise in world energy prices has reduced our national income.
In time-honoured fashion, the government is looking for scapegoats.
The gestures are led by the Chancellor, Rachel Reeves, promising to crack down on “profiteering” by retailers. Corporate chairmen and chief executives are not impressed, to say the least, with Asda’s Allan Leighton stating outright that the accusations made by Reeves have “zero credibility”.
The Chancellor has promised to introduce new powers for regulators. If necessary, the Competition and Markets Authority (CMA) will be authorised, allegedly for a limited time period, to “clamp down on price gouging if it takes place”.
A nice piece of smoke and mirrors. But completely separate from all of this, a worrying trend has been developing over the past decade which undermines the effectiveness of the CMA.
Supplanting human decision making
Increasingly, algorithms are supplanting human decision makers in the pricing of goods and services.
And as Michael Harre at the Centre for AI Trust and Governance at the University of Sydney points out in a recent note, existing competition laws, and which the CMA relies, are insufficient to capture the risks which arise from this (for transparency, I have worked with Harre on various academic projects over the years).
An important paper was published six years ago in the American Economic Review by Emilio Calavano and colleagues, of both Bologna and Toulouse.
They studied experimentally the behaviour of AI powered algorithms in a standard economic model of an oligopolistic market. In such markets, there are only a few big players. There may be tiny ones, but a relatively small number of large ones dominate the market. The petrol forecourt market in the UK is an example.
Algorithms learned to set prices at levels above those which are deemed competitive by economic theory
Each of the agents in the model set its own prices using algorithms of a type known as reinforcement learning (RL). RL was used in the famous examples of Alpha Zero and Alpha Go, in which computer programmes easily defeated the world’s best human players of chess and Go.
The key finding of Calvano and colleagues was that the algorithms learned to set prices at levels above those which are deemed competitive by economic theory.
A key aspect of the learned behaviour was to punish any agent which tried to gain an advantage by deviating from the high price equilibrium.
There was no collusion or communication between the various agents. They learned the anti-competitive strategy independently.
So a body such as the CMA looking for evidence of collusion in documents, emails, or even Morgan McSweeney’s mobile phone, would find none.
The authors readily acknowledge that the results were obtained in a laboratory-type setting, but they are robust across a range of realistic assumptions about the set-up.
Harre notes that the world has moved on since this paper was published. Six years, after all, is a long time in the tech world. AI agents have become capable of developing covert signalling channels which cannot be detected even by capable oversight systems.
Existing competition laws face problems in detecting and prosecuting anti-competitive pricing generated by autonomous algorithms. Yet they are already in use in areas such as insurance and online retailing.
This is a genuine and serious problem of potential anti-competitive behaviour which the government needs to get to grips with rather than simply posturing.
Paul Ormerod is an Honorary Professor at the Alliance Business School at the University of Manchester. You can follow him on Instagram @profpaulormerod