Priced at the speed of light: algorithmic pricing, personalization, and the new inflation challenge
The article by Clare Lombardelli and Rupal Patel identifies four interconnected trends reshaping how prices are set in the UK economy and what they mean for inflation:
- Dynamic pricing is spreading rapidly. Falling menu costs, i.e. the cost of changing prices driven by digital technology, mean firms can now reprice in real time in response to demand, costs, or competitor moves. This makes prices can become more flexible and potentially more volatile, complicating the interpretation of monthly CPI readings.
- Personalized (“surveillance”) pricing is emerging. Firms increasingly use granular consumer data such as browsing history, location, demographics, and purchase patterns, to charge different prices to different individuals. This shifts pricing from uniform to discriminatory, with ambiguous welfare effects: some consumers gain access to goods they previously couldn’t afford while others pay more than they would have under uniform pricing.
- Algorithmic collusion is a growing concern. Algorithms that monitor and respond to rivals’ prices at near-zero cost may sustain supra-competitive prices without any explicit coordination simply by reacting quickly to deviations.
- CPI measurement faces new challenges. As prices become more frequent, more individualized, and more volatile, traditional price indices that are based on periodic sampling of uniform prices may not capture the inflation consumers actually experience. The UK Office of National Statistics introduced scanner data into UK CPI in March 2026 as a partial response.
The authors conclude that outcomes depend heavily on the strength of competition and consumer access to information. Where competition is robust and consumers are well-informed (e.g. via AI shopping assistants), dynamic pricing can improve efficiency and reduce average prices. Where it is not, it risks locking in higher prices and worsening distributional outcomes.
This time it’s personal: the rise of dynamic, personalised pricing and what it means for inflation
By: Clare Lombardelli, Rupal Patel – Bank of England
Coercion without coordination: pricing algorithms as market power
In this paper, Zach Y. Brown and Alexander MacKay show that a single firm with a fast, persistent pricing algorithm can unilaterally push market prices above even fully collusive levels, with no coordination, communication, or even awareness required from rivals, simply because the algorithm makes any price cut instantly unprofitable. This directly underpins Lombardelli and Patel’s concern, as highlighted above, that algorithmic pricing may drive inflation higher and elude traditional antitrust oversight, since the mechanism leaves none of the traces regulators are trained to look for.
The authors build a theoretical model of a market with two competing firms: one firm, the “algorithmic firm,” delegates pricing to an algorithm while the other sets prices directly maximizing its profits period by period without any strategic forward-looking behavior. The model rules out any voluntary cooperation on the rival’s part, so to isolate the power of the algorithm alone.
The algorithm is characterized by two features. First, its speed: how quickly it can detect and respond to the rival’s price within the same pricing period. Second, its persistence: how long a given pricing rule stays in place before the firm can revise it. The algorithm encodes a target price it wants the rival to charge, so if the rival undercuts, the algorithm instantly responds by cutting its own price sharply, making the deviation unprofitable.
The authors extend the model in three directions: simulating what happens when the rival is not even aware it faces an algorithm (it only observes its own profits and adjusts accordingly); extending to markets with more than two firms; and analyzing how an online platform would choose to design the speed and persistence of pricing tools available to sellers on its marketplace.
Findings
- One algorithm is enough to raise prices. A single firm with a fast, persistent pricing algorithm can sustain prices well above the competitive level without any agreement, communication, or even awareness on the rival’s part. The rival ends up charging high prices simply because undercutting is instantly punished and therefore never profitable.
- Outcomes can be worse for consumers than outright collusion. In the most extreme case, the algorithm pushes the rival to charge prices above what both firms would set if they were explicitly colluding together, while the algorithmic firm itself charges slightly less. The algorithmic firm earns more than it would under full collusion, and consumers are worse off than under a cartel.
- Speed and persistence work together. Speed alone makes the algorithm react like a follower as it softens competition but does not dominate. Persistence alone gives the algorithm a leadership advantage. Only the combination of both creates the full coercive effect. They are complements, not substitutes.
- Platforms have incentives to enable coercion. When an online marketplace can choose what pricing tools sellers have access to, how fast they can react to rivals and how long their rules persist, a platform that cares mainly about seller revenues will tend to equip at least one seller with the most powerful algorithm. A platform that cares equally about consumers and sellers will do the opposite and restrict both speed and persistence. A vertically integrated platform that also sells its own products has the strongest incentive to grant itself the most coercive technology while limiting rivals.
Algorithmic Coercion with Faster Pricing
Authors: Zach Y. Brown, Alexander MacKay
From: University of Michigan, University of Virginia
