As inflation refuses to submit to ever-increasing interest rates, the government and the Bank of England are – finally – beginning to move their focus from wages to company profits. The Chancellor, Jeremy Hunt, has warned banks they may face tighter regulation if they do not pass on interest rate rises to savers, while the Bank’s governor, Andrew Bailey, admitted on 6 July that some retailers “have possibly been charging too much” for petrol, after the Competition and Markets Authority found that retailers had added an extra £900m to the mark-up on fuel. Grant Shapps, the Energy Secretary, has said that from now on petrol stations will be made to share their pricing data with the public. “We’ll shine a light on rip-off retailers to drive down prices,” he said.
What Shapps seems not to have grasped is that this data is already widely shared. Every petrol station in the UK connects to a data source that records the prices of every other petrol station on a daily basis. The prices that emerge from this system aren’t decided by cigar-chewing executives in pinstriped suits, but by algorithms.
The same is true of almost everything else we buy – not just online, but in supermarkets and high-street shops; when we pay our rent and our bills. Pricing has become a game played by machines, which analyse millions of data points and can adjust prices many times a day, maximising the possible returns with a constant stream of feedback. These systems are becoming the focus of a growing body of research which suggests they could upend our ideas about competition, make inflation hit consumers faster, and keep prices higher for longer.
[See also: Broke Britannia]
Petrol stations are a good place to start. Since the mid-2010s software developers have offered algorithmic pricing that can vary prices based on wholesale prices, demand and other factors. Software such as Pricecast, developed by A2i Systems, automatically lowers fuel prices to attract more customers in times of lower demand, or increases them at busy times, leading to increased margins.
Every time you buy something online, a similar system is involved; the “dynamic pricing” used by big online sellers can cause prices to change constantly, and different customers to be offered different prices. The effects of this are only beginning to be studied. In 2021 Zach Brown, an economist at Michigan University, and Alexander MacKay, an economist at Harvard, collected data from five large online retailers and tracked what happened, hour by hour, as prices changed.
Imagine two people are selling apples in a standard market, running on the accepted rules of supply and demand. Economists assume that they will try to undercut one another, paying to poach each other’s customers with lower prices until neither wants to go any cheaper. Equilibrium is achieved, apple prices have efficiently controlled themselves, and the benevolent ghost of Adam Smith high-fives the invisible hand of the market.
But wait: what if one of the apple sellers, Seller A, takes downs her “four apples for £1” sign and replaces it with a mirror that simply reflects the other seller’s sign? Now, any time Seller B reduces their prices, Seller A does too, at the speed of light. Seller B can’t undercut Seller A any more. There isn’t a morning or afternoon or even a few minutes when customers in the market will see that one stall is selling apples more cheaply than the other. There’s also, as a result, no incentive to cut prices any more.
The simplest possible pricing algorithm is effectively a mirror for other sellers’ prices. Brown says even something this simple will remove the incentive to reduce prices, and that “both prices will end up being higher” as a result. Brown says that while the real algorithms used in e-commerce are much more complex processes, more akin to game-playing strategies, “a similar logic applies”. Such algorithms are used not only by online stores and petrol stations, but supermarkets and high-street shops. Even smaller shops will take their cue from the big chains that are the price leaders; a supermarket’s algorithm might influence the cost of goods in a corner shop in another town.
Brown and MacKay’s research shows not only that the pricing of online goods is to a great extent decided by algorithms checking other sellers’ prices, but that, as Brown told me, “this doesn’t fit with our standard economic models”.
MacKay’s other work has shown that consumers already understand this intuitively: we are becoming less price sensitive, because there is less point in looking for deals in a world with no price wars. This new world requires a new kind of economics, in which the invisible hand of competition is replaced by the invisible tentacles of technology providers, and the old principles of market equilibrium are replaced by a new law: whoever has the most powerful algorithm can decide the price.
In 2011 an Amazon seller called “profnath” listed a second-hand, out-of-print academic textbook (The Making of a Fly, by the geneticist Peter Lawrence) for $9.98. And there it would have stayed, had another seller not listed the same book using automatic pricing, which caused the original listing (also set with automatic pricing) to rise in turn. Each seller just wanted to make a slightly higher margin than the market rate, so the two listings kept bidding each other upwards until the book cost well over 23 million dollars.
Yves-Alexandre de Montjoye, associate professor of applied mathematics and computer science at Imperial College London, told me this was perhaps the first example of algorithms setting prices in an “adversarial” manner: using their own prices to manipulate other systems. Working with fellow scientists at the Oxford Internet Institute, de Montjoye showed that it was possible, simply by gathering data and writing a programme to respond to it, not only to understand the pricing strategies of other sellers in an online market, but to control them. “We can learn, purely from the outside, how an algorithm prices,” he said. “And we can use this knowledge to manipulate, through our prices, how a competitor is going to price.”
This suggests that it is not only economics that urgently needs to catch up to algorithmic pricing, but the law.
[See also: Will pay rises cause inflation?]
The two fundamental principles of economics are this: I want money, and you want money. Markets are driven by self-interest (I want money) and regulated by competition (you want money). I sell bread for more than it cost to make in order to profit, but if I charge too much, you’ll sell a cheaper loaf, my customers will become your customers, and I’ll have to drop my prices to get them back.
Then again, we could meet in a darkened room and agree not to lower our prices below a certain point. We’re a cartel! Now we both get higher margins and bigger profits, but not without risk: price-fixing is a criminal offence with a maximum sentence of five years.
However, it’s only a crime if we get together and agree to fix our prices deliberately. If I decide on my own to just mirror your prices, we’re not fixing prices as a cartel. In an economics department this is “conscious parallelism”; at a law school it’s “tacit collusion”. In the shop, the result – more expensive bread – is pretty much the same.
Competition regulators have wrestled with tacit collusion for a long time. Algorithmic pricing makes it a more urgent question, says Ariel Ezrachi, professor of competition law and director of the Centre for Competition Law and Policy at Oxford University. One of the key reasons for this is that, very often, the same algorithm is setting lots of prices.
Rents in the United States have risen sharply since 2020, and while there are a number of reasons for this, renters’ groups in more than 20 cities claim (in legal actions that are being combined into a class action) that some of the responsibility rested with a company called RealPage and its software, YieldStar, which is used to set the rents on tens of millions of homes. In March senators suggested YieldStar was so prevalent in US rental markets – in one area of Seattle, an investigation by the news website ProPublica found, it was used to set 70 per cent of all rents – that it had effectively become a cartel of one, engaged in “de facto price setting”.
Ezrachi calls this the “hub and spoke” problem: what looks like a market with lots of participants is in fact controlled by a single dominant algorithm. Uber is one such system: there are lots of taxi drivers but they do not compete, at all, on price. If it’s raining, or late at night, the price of an Uber goes up because Uber’s algorithm expects that customers will tolerate higher prices. If one algorithm is applied to the pricing of rents, or petrol, or supermarket goods – which are, it is worth noting, collected at a high frequency through credit and store card data – then “that mechanism of pricing doesn’t have to compete against another mechanism that might be more competitive”, says Ezrachi. “So it reinforces itself.”
Is this inflationary? The idea that prices across the economy are made “supra-competitive” by tacit collusion certainly sounds inflationary. The Competition and Markets Authority published a working paper in 2018 warning that pricing algorithms could “reduce competition and harm consumers”, and it is notable that the areas in which so-called “greedflation” is most identifiable – such as fuel – are those in which automatic pricing is most widely used.
Alexander MacKay’s work has also shown increased mark-ups on consumer products – which are significant because the mark-up is an extra cost added on top of rising input costs, compounding inflation – and while he’s cautious about attributing higher inflation to algorithms, he says it is very much the case that “companies have become more sophisticated with their pricing strategies. They’re able to change prices more quickly.” Zach Brown agrees that while they might not be a primary cause of inflation, pricing algorithms could “exacerbate” rising prices by speeding up the transmission of costs: “They could at least make consumers feel the effects of inflation faster than they would have in the past.”
Ariel Ezrachi says competition and prices are inextricably linked. “The more alignment you can create, the more you reduce the uncertainty on the market, the greater is the pressure on the prices. The dynamics of competition protect us, as consumers. Algorithms enable you to defy the dynamics of competition.”
[See also: The illusion of prosperity is over]