Competition Law and Regulation of Algorithmically Coordinated Markets

1. Introduction

Algorithmically coordinated markets are markets in which firms use pricing algorithms, artificial intelligence, machine-learning systems, automated repricing tools, or shared digital infrastructure to determine prices, output, allocation, ranking, or other competitive parameters.

Algorithms can create substantial efficiencies: faster price adjustment, better demand forecasting, inventory management, fraud detection, and more accurate matching of buyers and sellers. However, the same technology can also facilitate coordination among competitors, including situations where firms reach an explicit agreement, use a common algorithm to implement coordination, or independently employ algorithms that learn to avoid aggressive competition.

The central competition-law problem is therefore:

When does algorithmic pricing or market monitoring constitute legitimate independent conduct, and when does it amount to prohibited coordination or facilitate an abuse of market power?

The issue is particularly significant in concentrated digital markets, online retail, hotel booking, ride-hailing, financial markets, commodities, and platform economies.

2. Meaning of Algorithmically Coordinated Markets

An algorithmically coordinated market exists where algorithms materially influence competitive conduct among market participants.

Common forms include:

A. Explicit algorithmic coordination

Competitors directly agree to use the same algorithm or pricing system.

Example:

• Competitor A and Competitor B agree to use the same pricing software.
• The software automatically sets prices according to common parameters.
• Prices subsequently move in parallel.

This is the clearest form of potential cartel conduct.

B. Hub-and-spoke algorithmic coordination

A platform, software provider, intermediary, or "hub" may coordinate competing businesses.

Structure:

Competitor A →
Competitor B → Common algorithm/platform
Competitor C →

The competitors may never communicate directly. Nevertheless, the platform can become a mechanism through which competitively sensitive information or pricing instructions are transmitted.

C. Algorithmic implementation of a traditional cartel

Human beings establish the agreement, while software implements it.

For example:

1. Competitors agree on minimum prices.
2. The agreement is programmed into pricing software.
3. The algorithm automatically adjusts prices.
4. Human employees no longer need to communicate about every price change.

The technological form does not necessarily change the underlying competition-law character of the agreement.

D. Tacit algorithmic coordination

The most difficult category involves firms independently adopting algorithms that learn from market behaviour and gradually converge on supra-competitive outcomes.

There may be:

• no express agreement;
• no communication between competitors;
• no common software provider;
• no direct exchange of messages.

The algorithms may nevertheless learn that aggressive price competition is undesirable and respond by maintaining elevated prices.

This creates the difficult question of whether tacit coordination alone is legally punishable.

3. Competition-Law Framework

Algorithmic coordination can potentially implicate several areas of competition law.

A. Anti-cartel rules

Competition authorities can investigate:

• price fixing;
• output restrictions;
• market allocation;
• bid rigging;
• exchange of commercially sensitive information;
• coordinated pricing;
• algorithmically facilitated agreements.

The key issue is generally not whether an algorithm was used, but whether the underlying conduct satisfies the legal requirements for an agreement, concerted practice, or equivalent prohibited coordination.

B. Abuse of dominance

Algorithms may also facilitate unilateral exclusionary conduct by dominant firms.

Potential practices include:

• algorithmic exclusion;
• discriminatory ranking;
• self-preferencing;
• discriminatory access;
• predatory pricing;
• loyalty-inducing pricing;
• tying and bundling;
• algorithmic refusal to deal;
• discriminatory recommendations.

Thus, algorithmic competition problems are not confined to cartel law.

4. Algorithmic Pricing and the Traditional Cartel Concept

Traditional cartel enforcement normally requires some form of coordination.

An important distinction is:

Independent parallel conduct

A increases price → B independently increases price.

Parallel pricing alone does not necessarily prove an unlawful agreement.

Coordinated conduct

A and B communicate or use a common mechanism to establish or maintain prices.

This may constitute prohibited coordination where the applicable legal elements are satisfied.

Algorithmically facilitated coordination

A and B provide pricing information or instructions to a common algorithm, which coordinates their conduct.

Here, the algorithm becomes evidence or mechanism of coordination rather than an independent legal actor.

5. Six Major Case Laws

Case 1: United States v. Topkins — Online Poster Cartel

United States, 2015

This is one of the most important early examples of algorithm-assisted cartel conduct.

Online sellers of posters used automated pricing software. According to the prosecution, the participants agreed to fix prices of certain products and programmed their software to implement the arrangement.

Competition-law significance

The case demonstrates that:

Using an algorithm to implement an unlawful price-fixing agreement does not immunize the conduct from cartel law.

The algorithm merely becomes the technological means of implementing the agreement.

Principle

Competition law focuses on the underlying coordination, rather than whether prices were fixed manually or technologically.

6. Case 2: In re RealPage, Inc. — Algorithmic Rental Pricing

United States

RealPage-related litigation and enforcement concerning rental-housing pricing illustrates a more complex algorithmic-coordination problem.

The allegations have concerned landlords providing competitively sensitive information to a common pricing system and using algorithmic recommendations in rental pricing.

The central competition question is whether competing landlords can effectively coordinate rental prices through a common algorithmic intermediary.

Significance

The case illustrates the potential hub-and-spoke model:

Landlord A
↓
Common pricing algorithm
↑
Landlord B

The important legal issue is whether the exchange and use of information through the algorithm constitutes unlawful coordination.

Principle

A competitor cannot necessarily avoid cartel scrutiny merely because coordination is mediated through software or an intermediary.

7. Case 3: United States v. Apple Inc. — E-Books

United States, 2013

Although this was not an artificial-intelligence pricing case, it is important for understanding technology-mediated coordination.

The litigation concerned Apple's alleged coordination with publishers regarding electronic-book pricing.

The court found that Apple had participated in a conspiracy involving publisher coordination.

Relevance to algorithmic markets

The case demonstrates an important principle:

Digital markets do not require a technologically sophisticated algorithm for traditional cartel principles to apply.

Where technology is subsequently used to implement or monitor an agreement, the underlying agreement remains legally significant.

Principle

Technology is a means of coordination, not a defence to coordination.

8. Case 4: Eturas v. Lietuvos Respublikos konkurencijos taryba

Court of Justice of the European Union, 2016

This is one of the most directly relevant cases concerning technology-mediated coordination.

Several travel agencies used the E-TURAS online booking system.

A message was transmitted through the system concerning restrictions on discounts that travel agencies could offer.

The CJEU examined whether the agencies could be held responsible for participating in a concerted practice where a common electronic system facilitated the restriction.

Importance

The case is highly relevant to algorithmic coordination because it demonstrates that:

• electronic systems can facilitate concerted practices;
• knowledge of a restrictive mechanism can be legally significant;
• participation need not necessarily involve traditional face-to-face cartel meetings.

Principle

A digital platform can constitute a mechanism through which competitors coordinate their commercial conduct.

9. Case 5: AC-Treuhand v European Commission

CJEU, 2015

AC-Treuhand involved the role of a consultancy/service provider in cartel arrangements.

The case is important because the Court recognized that an undertaking providing services that contribute to the implementation of an anticompetitive arrangement can potentially fall within EU competition rules.

Algorithmic relevance

The principle becomes particularly important in modern markets where a third-party technology provider supplies:

• pricing algorithms;
• market-monitoring systems;
• information exchanges;
• demand forecasting;
• recommendation systems.

A software provider therefore cannot automatically assume that it is outside competition law merely because it does not itself sell the final product.

Principle

Third-party facilitators can become legally relevant where their conduct intentionally contributes to an anticompetitive arrangement.

10. Case 6: Wood Pulp — Parallel Behaviour and Concerted Practices

CJEU, 1993

The Wood Pulp litigation concerned alleged coordination among producers and the evidentiary significance of parallel conduct.

The Court emphasized that parallel behaviour does not automatically establish a concerted practice.

Algorithmic importance

This principle is critical for AI-driven markets.

Suppose:

• Firm A's algorithm raises prices;
• Firm B's algorithm independently raises prices;
• Firm C's algorithm also raises prices.

The resulting uniformity does not automatically prove a cartel.

Competition authorities must distinguish:

legitimate interdependence from unlawful coordination.

Principle

Parallel market outcomes are not necessarily evidence of an unlawful agreement.

This distinction becomes even more important where autonomous algorithms generate similar outcomes without direct human coordination.

11. Case 7: T-Mobile Netherlands

CJEU, 2009

The T-Mobile Netherlands case concerned an exchange of competitively sensitive information between competitors.

The Court adopted an important approach to concerted practices and information exchange.

Relevance to algorithms

Algorithms can dramatically increase the speed and frequency of information exchange.

Competitors may potentially receive information concerning:

• current prices;
• intended prices;
• inventory;
• capacity;
• demand;
• customers;
• discounts.

An algorithm can transform occasional information exchange into continuous automated monitoring.

Principle

Information exchange can become a serious competition concern when it reduces strategic uncertainty between competitors.

12. Case 8: Uber — Dynamic Pricing Litigation and Competition Issues

Uber-related competition disputes in several jurisdictions illustrate the broader issue of algorithm-mediated pricing.

Ride-hailing platforms use algorithms to calculate prices based on variables such as:

• demand;
• driver availability;
• geographic location;
• time;
• congestion;
• market conditions.

The competition question becomes particularly difficult where independent service providers follow a platform-generated price.

Algorithmic coordination concern

If competing providers independently accept a platform's algorithmic price, the platform can potentially become the central mechanism determining market prices.

The legal analysis depends on the precise institutional structure and the applicable jurisdiction.

Principle

The more extensively an intermediary controls pricing among otherwise competing participants, the more important it becomes to examine whether the arrangement restricts independent price determination.

13. Tacit Algorithmic Collusion

The most difficult theoretical problem is autonomous algorithmic collusion.

Consider:

• Firm A develops Algorithm A.
• Firm B develops Algorithm B.
• Neither firm communicates with the other.
• Both algorithms observe market prices.
• Each learns that reducing prices triggers aggressive responses.
• Both algorithms eventually maintain higher prices.

There may be:

No agreement + no communication + no common algorithm.

Yet consumers may experience:

• higher prices;
• reduced output;
• reduced competitive pressure.

14. Why Tacit Collusion Is Difficult

Traditional competition law generally distinguishes between:

Lawful conscious parallelism

Firms independently respond rationally to market conditions.

Unlawful coordination

Firms communicate, agree, or engage in a legally recognized concerted practice.

An algorithm may make the market outcome look coordinated even where there is no legally cognizable agreement.

Therefore:

Economic coordination and legally prohibited coordination are not necessarily identical.

This distinction is fundamental.

15. Algorithms as "Digital Facilitators"

Algorithms can perform the functions traditionally performed by cartel participants.

This makes detection significantly more difficult.

16. Hub-and-Spoke Algorithmic Coordination

A particularly important model is:

Supplier A
↘
Platform / Algorithm
↗
Supplier B

The platform may collect:

• prices;
• inventories;
• costs;
• demand;
• customer information;
• promotional strategies.

It then provides pricing recommendations to each competitor.

Competition concern

If competitors know that their rivals are using the same system and the system is designed to reduce competitive uncertainty, authorities may investigate whether the system facilitates coordinated behaviour.

17. Algorithmic Monitoring

Algorithms can also make cartel monitoring more effective.

Suppose a cartel agrees:

Minimum price = ₹1,000.

An automated system can immediately detect:

Competitor's price = ₹950.

The cartel could then automatically respond.

This can make:

• deviation detection easier;
• retaliation faster;
• cartel stability greater.

Thus, algorithms can potentially increase the internal enforcement capability of cartels.

18. Algorithmic Transparency as a Competition Concern

Transparency normally promotes competition.

However, excessive transparency can sometimes reduce competition.

Example

If every competitor instantly knows:

• every rival's price;
• planned discounts;
• inventory;
• future pricing;
• customer movements;

firms may find it easier to coordinate.

Therefore:

More information is not always equivalent to more competition.

Competition authorities must distinguish consumer transparency from competitor transparency.

19. Pricing Algorithms and Market Power

Algorithmic coordination becomes especially significant in concentrated markets.

Factors include:

• number of competitors;
• market concentration;
• frequency of transactions;
• price transparency;
• barriers to entry;
• product homogeneity;
• algorithmic sophistication;
• switching costs;
• network effects.

A highly concentrated market with real-time pricing data can provide favourable conditions for algorithmic coordination.

20. Role of Data

Data is the fuel of algorithmic markets.

Relevant data may include:

• historical prices;
• customer demand;
• competitor prices;
• inventory;
• transaction histories;
• geographic information;
• consumer profiles.

Where several competitors obtain access to the same commercially sensitive data through a common system, competition authorities may examine whether the arrangement facilitates coordination.

21. Artificial Intelligence and Collusion

AI systems introduce additional complexity because machine-learning systems may discover pricing strategies without explicit instructions to collude.

Possible mechanisms include:

Reinforcement learning

The algorithm receives rewards for profitability and learns strategies that maximize returns.

Predictive pricing

Algorithms predict competitors' responses.

Automated retaliation

An algorithm detects a competitor's price reduction and responds immediately.

Market stabilization

Algorithms may learn that avoiding aggressive price competition produces higher long-term profits.

The legal challenge is determining whether such conduct is merely rational independent behaviour or evidence of prohibited coordination.

22. Competition Law and Explainability

Competition authorities increasingly need to understand:

• how an algorithm makes decisions;
• what data it uses;
• whether competitors' data are incorporated;
• whether common parameters exist;
• whether employees can override decisions;
• whether the system monitors competitors;
• whether pricing instructions are shared.

This creates a strong connection between competition law and algorithmic accountability.

23. Evidence in Algorithmic-Collusion Cases

Traditional cartel evidence includes:

• emails;
• meetings;
• telephone calls;
• contracts;
• price lists.

Algorithmic cases may require:

• source code;
• system architecture;
• API records;
• server logs;
• version histories;
• training data;
• model documentation;
• audit trails;
• communications with software vendors;
• pricing histories;
• algorithmic outputs.

Digital evidence can therefore become central to cartel investigations.

24. Liability of the Algorithm Developer

An important question is:

Can a software provider be liable for its customer's anticompetitive conduct?

Not automatically.

The legal analysis should examine:

1. What service did the provider supply?
2. Did it know how the system would be used?
3. Did it intentionally facilitate coordination?
4. Did it encourage competitors to use common pricing strategies?
5. Did it transmit competitively sensitive information?
6. Did it monitor compliance?
7. Did it modify the algorithm to facilitate coordination?

The AC-Treuhand principle is particularly relevant to this question.

25. Regulatory Approaches

Competition authorities can consider several approaches.

A. Traditional enforcement

Apply existing cartel rules where algorithms facilitate conventional agreements.

B. Information-exchange regulation

Control exchanges of commercially sensitive information through digital platforms.

C. Algorithmic auditing

Require firms in particularly sensitive circumstances to maintain:

• audit trails;
• documentation;
• testing procedures;
• model governance systems.

D. Data-access controls

Prevent unnecessary sharing of competitor-sensitive data.

E. Platform regulation

Large digital platforms may face additional obligations concerning:

• interoperability;
• self-preferencing;
• data use;
• transparency;
• access.

26. Economic Analysis

Authorities should investigate whether algorithms actually affect competitive outcomes.

Relevant economic indicators may include:

Price correlation

Do firms' prices move together unusually closely?

Price dispersion

Does algorithmic pricing reduce normal competitive variation?

Margin changes

Have margins increased following adoption of a common system?

Reaction speed

Do competitors respond to each other's price changes almost instantaneously?

Market entry

Has algorithmic coordination increased barriers to entry?

Consumer effects

Have consumers experienced:

• higher prices;
• reduced output;
• fewer choices;
• reduced quality?

Economic evidence should be combined with documentary and technical evidence rather than treated as conclusive proof by itself.

27. Key Legal Distinctions

28. Important Case-Law Principles — Consolidated

29. Major Challenges for Competition Authorities

1. Establishing an agreement

The absence of emails or meetings makes conventional cartel detection more difficult.

2. Distinguishing coordination from parallelism

Similar algorithmic outcomes can arise independently.

3. Understanding complex AI

Machine-learning models may be difficult even for their developers to explain.

4. Attribution

It can be difficult to determine whether responsibility lies with:

• the firm;
• employees;
• platform;
• algorithm developer;
• data provider.

5. Cross-border enforcement

Digital pricing systems can operate simultaneously across numerous jurisdictions.

6. Rapid technological change

Competition rules can lag behind new forms of automated coordination.

30. Compliance Measures for Businesses

Companies using pricing algorithms should maintain:

1. Competition-law review before deployment
2. Documentation of algorithm design
3. Restrictions on competitor-sensitive data
4. Independent pricing parameters
5. Regular algorithmic audits
6. Human oversight
7. Employee competition-law training
8. Logging of algorithmic decisions
9. Controls on third-party software providers
10. Procedures for identifying suspicious coordinated outcomes

A particularly important principle is:

A company should not assume that outsourcing pricing decisions to software eliminates its competition-law responsibilities.

31. Conclusion

Algorithmically coordinated markets represent one of the most significant challenges at the intersection of competition law, artificial intelligence, data governance, and digital-platform regulation.

The existing case law demonstrates several established principles:

• algorithms can implement traditional cartels;
• electronic systems can facilitate concerted practices;
• information exchange can reduce competitive uncertainty;
• third-party facilitators may attract competition-law scrutiny;
• parallel conduct alone does not necessarily establish unlawful coordination.

The most difficult unresolved issue is autonomous algorithmic coordination, where independently designed algorithms may produce stable supra-competitive outcomes without an identifiable human agreement.