Algorithms and Tacit Collusion Concerns

Introduction

Tacit collusion refers to coordination between competitors that may arise without an explicit agreement to fix prices, allocate markets, restrict output, or otherwise coordinate competitive conduct. In traditional competition law, tacit coordination is difficult to establish because parallel conduct—such as competitors charging identical or similar prices—can have legitimate competitive explanations.

The increasing use of pricing algorithms, artificial intelligence, machine learning, automated monitoring systems, and common algorithmic platforms creates a new competition concern. Algorithms can rapidly observe competitors' prices, predict reactions, identify profitable coordinated outcomes, and automatically adjust prices. This may make coordination more stable even when competitors never communicate directly.

The central legal question is therefore:

When does algorithmically facilitated parallel conduct remain lawful independent behaviour, and when does it amount to an unlawful agreement, concerted practice, or abuse of market power?

1. Meaning of Algorithmic Tacit Collusion

Algorithmic tacit collusion occurs when competing firms use automated systems that facilitate mutual accommodation or coordinated market outcomes without necessarily having a conventional express agreement.

For example:

• Firm A's algorithm observes Firm B's price.
• Firm B's algorithm simultaneously observes Firm A's price.
• Both algorithms repeatedly adjust prices.
• Each algorithm learns that undercutting the competitor produces retaliation.
• Eventually both maintain prices above competitive levels.
• No employee sends an email saying "let us fix prices."

This creates a distinction between:

Traditional explicit cartel

Human communication → agreement → coordinated prices.

Algorithmic coordination

Automated observation → prediction → adaptation → stable parallel pricing.

The second situation is legally more complicated because parallel conduct itself is generally not sufficient to prove an antitrust agreement in many jurisdictions.

2. Why Algorithms Increase Tacit-Collusion Risks

A. Greater price transparency

Algorithms can monitor thousands of competitor prices continuously.

A human pricing manager might check competitors once or twice per day. An algorithm can potentially monitor prices:

• every minute;
• every second;
• across thousands of products;
• across multiple geographic markets.

This substantially reduces the time required to detect deviations.

B. Rapid retaliation

A major difficulty in sustaining tacit coordination is the possibility that one competitor will secretly reduce prices.

Algorithms can respond almost immediately.

For example:

Firm A reduces price by 5%

↓

Firm B's algorithm detects the reduction

↓

Firm B reduces price automatically

↓

Firm A detects B's response

↓

Both algorithms return to coordinated pricing

The speed of retaliation may make deviation from the coordinated outcome less attractive.

3. Algorithms as Coordination Facilitators

An algorithm does not necessarily need to "intend" to collude.

It may be programmed to maximize:

• revenue;
• profit;
• market share;
• occupancy;
• conversion;
• margin;
• inventory turnover.

If competitors use similar algorithms and face similar market conditions, their independent optimization can sometimes generate parallel outcomes.

This creates an important distinction between:

Algorithm as a tool

The algorithm merely implements an independently determined pricing strategy.

Algorithm as a coordination mechanism

The algorithm is designed or used to facilitate coordination between competitors.

The second situation raises substantially greater competition-law concerns.

4. Main Forms of Algorithmic Tacit Collusion

4.1 Monitoring Algorithms

These systems constantly monitor competitors.

They may identify:

• price changes;
• discounts;
• inventory;
• availability;
• promotional campaigns;
• delivery charges.

Monitoring itself is not necessarily unlawful.

The competition concern arises when monitoring enables firms to discipline deviations from a coordinated outcome.

4.2 Predictive Algorithms

Algorithms may predict how competitors will react to a firm's price change.

For example:

"If competitor X reduces price by 3%, reduce our price by 3%."

Over time, the system may develop highly sophisticated response strategies.

4.3 Autonomous Learning Algorithms

Machine-learning systems may identify pricing strategies without receiving explicit instructions to coordinate.

A reinforcement-learning system could discover that:

aggressive price competition → lower profits

while:

stable prices → higher long-term profits.

If multiple competitors use similar systems, stable supra-competitive pricing could potentially emerge.

This raises difficult questions concerning attribution and legal responsibility.

5. Tacit Collusion Versus Express Collusion

The distinction is fundamental.

Competition authorities therefore face the problem of determining whether similar algorithmic conduct is merely rational independent conduct or evidence of an unlawful coordination mechanism.

6. Legal Framework

A. Agreement or Concerted Practice

Competition law generally prohibits competitors from coordinating their competitive behaviour through agreements or concerted practices.

The difficult issue is that mere conscious parallelism is normally not automatically equivalent to an agreement.

There must generally be additional evidence establishing coordination or communication, depending upon the jurisdiction.

7. United States

The principal statutory provisions are:

Sherman Act §1

Prohibits contracts, combinations and conspiracies that restrain trade.

Important distinction

US antitrust law traditionally distinguishes between:

Independent parallel conduct

and

concerted action.

Therefore, two competitors independently adopting similar algorithmic pricing strategies do not automatically establish a §1 violation.

However, algorithms could become evidence of unlawful coordination where competitors:

• share commercially sensitive information;
• agree to use a common pricing system;
• communicate algorithmic parameters;
• delegate pricing decisions to a common intermediary;
• use an algorithm specifically designed to implement an agreement.

8. European Union

The principal provisions are:

Article 101 TFEU

Prohibits agreements, decisions of associations of undertakings and concerted practices that restrict competition.

Article 101 is particularly important for algorithmic coordination because EU competition law recognizes the concept of concerted practice, which can extend beyond traditional written agreements.

However, purely parallel behaviour without the required coordination remains conceptually different from a concerted practice.

9. India

The relevant statutory framework is primarily:

Competition Act, 2002

Section 3 prohibits agreements causing or likely to cause an appreciable adverse effect on competition.

Section 3(3) specifically addresses horizontal arrangements involving:

• price fixing;
• limiting production or supply;
• market allocation;
• bid rigging.

Section 3(3) is particularly relevant where algorithms are used as instruments for implementing or facilitating horizontal coordination.

Section 4 may also become relevant where an algorithm is operated by a dominant enterprise and produces exclusionary or exploitative effects.

10. Six Important Case Laws

Because there are relatively few reported judicial decisions involving fully autonomous algorithmic tacit collusion, traditional cartel and parallel-conduct cases remain important for developing the legal principles.

Case 1: United States v. Container Corporation of America

United States Supreme Court, 1969

Facts

The case involved competitors exchanging information concerning prices and pricing intentions in the corrugated-container industry.

The Supreme Court examined whether information exchanges could facilitate coordinated pricing.

Principle

The case demonstrates that competitors do not necessarily need to exchange an explicit agreement fixing a precise price for information-sharing mechanisms to create serious competitive concerns.

Relevance to algorithms

Modern algorithms can perform information collection and dissemination much faster than traditional human information exchanges.

An algorithm that systematically collects competitors' commercially sensitive pricing information may therefore create similar competitive risks.

11. Case 2: Interstate Circuit, Inc. v. United States

US Supreme Court, 1939

Facts

A large movie exhibitor sent identical demands to several film distributors concerning contractual and pricing arrangements.

The distributors acted in a coordinated manner even though there was no conventional bilateral agreement between all participants.

Principle

The Court recognized that an agreement or concerted action can sometimes be inferred from concerted responses to a common communication and surrounding circumstances.

Algorithmic relevance

The case is useful for understanding situations where competitors knowingly participate in a common coordination mechanism.

For example:

Competitors → common platform → common algorithm → coordinated pricing.

The existence of a common technological mechanism could become relevant evidence, although technology alone would not automatically establish unlawful coordination.

12. Case 3: Theatre Enterprises, Inc. v. Paramount Film Distributing Corp.

US Supreme Court, 1954

Facts

The plaintiff alleged that movie distributors had engaged in discriminatory conduct through coordinated distribution arrangements.

The evidence largely involved parallel conduct.

Principle

The Supreme Court emphasized the important distinction between independent parallel business behaviour and concerted action.

Parallel conduct by itself is insufficient to establish an unlawful agreement.

Algorithmic relevance

This is one of the most important principles for algorithmic pricing.

Suppose:

• Company A uses Algorithm A;
• Company B uses Algorithm B;
• both independently arrive at the same price.

The similarity of prices alone should not automatically establish a cartel.

Additional evidence is necessary to distinguish independent algorithmic optimization from coordinated conduct.

13. Case 4: Bell Atlantic Corp. v. Twombly

US Supreme Court, 2007

Facts

The plaintiffs alleged that telecommunications companies had engaged in parallel conduct and had unlawfully agreed not to compete in certain territories.

Principle

The Supreme Court held that allegations of parallel conduct must be accompanied by sufficient factual circumstances suggesting an actual agreement.

The famous analytical distinction is between:

parallel conduct

and

parallel conduct plus circumstances plausibly suggesting agreement.

Algorithmic relevance

This principle is directly relevant to AI-driven pricing.

Evidence such as:

• identical algorithmic parameters;
• communications between competitors;
• common software providers;
• coordinated algorithm updates;
• deliberate sharing of pricing information;

could potentially provide the additional circumstances necessary to distinguish unlawful coordination from mere parallel pricing.

14. Case 5: Apple Inc. v. Pepper

US Supreme Court, 2019

This case did not concern algorithmic collusion directly, but it is relevant to the broader legal treatment of digital-platform markets.

Relevance

The case illustrates the importance of identifying the competitive relationship created by digital platforms and intermediaries.

In algorithmic markets, a platform may potentially act as:

• intermediary;
• information provider;
• pricing infrastructure;
• transaction facilitator.

Where a common platform enables competing sellers to observe or coordinate pricing, competition authorities must examine the platform's precise economic role.

15. Case 6: United States v. Airline Tariff Publishing Co.

US Department of Justice enforcement proceeding

Facts

Airlines used a computerized fare information system through which fare changes were publicly disseminated.

The system created opportunities for competitors to observe proposed fares and coordinate their responses.

Principle

The case is particularly significant for understanding how computerized information systems can facilitate coordinated pricing.

The concern was not simply the existence of a computer system. It was the way the system could be used to communicate pricing intentions and facilitate coordination.

Algorithmic relevance

This provides an important bridge between traditional cartel law and modern algorithms:

information technology → increased transparency → faster retaliation → greater coordination capability.

16. Case 7: Eturas UAB v. Lietuvos Respublikos Konkurencijos Taryba

Court of Justice of the European Union, Case C-74/14, 2016

This is one of the most important cases for digital-platform-assisted coordination.

Facts

A common electronic booking system was used by travel agencies.

A message was distributed through the system concerning restrictions on discounts.

Legal issue

The CJEU considered when participants using a common electronic system could be regarded as participating in a concerted practice.

Principle

Knowledge of a common electronic communication concerning competitively sensitive conduct can become important evidence of participation in a concerted practice.

Algorithmic relevance

The case demonstrates that competition law can apply where a digital system serves as the communication or coordination infrastructure.

It is particularly valuable for understanding:

• platform-mediated coordination;
• algorithmic pricing;
• common software;
• automated restrictions;
• digital communications among competitors.

17. Case 8: AC-Treuhand AG v European Commission

CJEU, 2015

Facts

AC-Treuhand was a consultancy that facilitated cartel activity among producers.

It was not itself a producer competing in the relevant market.

Principle

An undertaking can potentially incur liability for facilitating a cartel, even where it is not itself an active competitor in the affected product market.

Algorithmic relevance

This is important for third-party algorithm providers.

Consider:

Competitors → common pricing software provider → algorithm → coordinated pricing.

The legal analysis may have to consider whether the software provider merely supplies neutral technology or knowingly facilitates anticompetitive coordination.

18. Case 9: Wood Pulp / Ahlström Osakeyhtiö v Commission

CJEU, Joined Cases 89/85 etc., 1988

Facts

The Commission examined parallel pricing behaviour by producers in the wood-pulp market.

Principle

The case is an important authority concerning the evidentiary distinction between parallel conduct and concerted practice.

The existence of parallel pricing alone did not automatically establish unlawful coordination.

Algorithmic relevance

This principle is particularly important in AI markets because machine-learning systems may independently converge upon similar outcomes.

Convergence therefore needs to be assessed alongside:

• communications;
• market transparency;
• pricing mechanisms;
• information exchange;
• structural conditions;
• algorithm design.

19. Case 10: Competition Authority v Booking.com / online parity investigations

European competition authorities have investigated online hotel-booking and price-parity arrangements involving platforms.

Although not a pure autonomous-algorithm case, these investigations illustrate how digital platforms can influence the competitive pricing decisions of independent suppliers.

Algorithmic relevance

Platforms can potentially use algorithms to:

• monitor supplier prices;
• compare prices across platforms;
• enforce parity;
• rank suppliers;
• adjust visibility.

Consequently, algorithmic monitoring can create competitive concerns even without a conventional cartel meeting.

20. Economic Conditions Favouring Tacit Algorithmic Collusion

Algorithms are particularly capable of facilitating coordination when the market has:

1. Few competitors

Concentration makes it easier for algorithms to monitor rivals.

2. High transparency

Competitors can rapidly observe price changes.

3. Homogeneous products

Price becomes the principal competitive variable.

4. Frequent transactions

Repeated interaction allows algorithms to learn competitor responses.

5. Stable demand

Predictable market conditions make coordinated outcomes easier to maintain.

6. Rapid retaliation

Algorithms can punish deviations almost immediately.

7. High barriers to entry

New entrants are less able to disrupt coordination.

8. Similar algorithms

Competitors using comparable pricing systems may react similarly to market signals.

21. The "Algorithmic Hub-and-Spoke" Problem

One particularly important model is:

Competitor A
↓
Common Algorithm
↓
Competitor B
↓
Common Algorithm
↓
Competitor C

The algorithm provider becomes the technological hub.

If competing businesses knowingly rely on the same mechanism to coordinate commercially sensitive conduct, authorities may examine whether the arrangement constitutes a hub-and-spoke conspiracy or concerted practice.

The key question is not simply:

"Did everyone use the same software?"

It is:

Did the common software facilitate a knowing alignment of competitive behaviour?

22. Common Algorithm Providers

A software company may provide identical pricing software to thousands of competitors.

This creates several scenarios.

Scenario A — Neutral software

Each business independently chooses:

• its data;
• its objectives;
• its parameters;
• its pricing strategy.

The software merely executes those decisions.

Competition concerns are comparatively different from a coordination arrangement.

Scenario B — Common pricing recommendation

The software provider collects competitors' confidential data and generates recommendations based on the collective information.

This raises considerably greater competition-law concerns.

Scenario C — Explicit coordination

The provider designs the system to:

• maintain agreed price levels;
• punish deviations;
• communicate competitors' future pricing intentions.

This can potentially constitute direct evidence of coordination.

23. Evidence Used by Competition Authorities

Algorithmic cases require evidence beyond traditional documents.

Authorities may examine:

Technical evidence

• source code;
• algorithm architecture;
• API documentation;
• version histories;
• training data;
• system logs;
• parameter settings.

Economic evidence

• parallel price movements;
• margins;
• market concentration;
• price dispersion;
• deviation patterns;
• response times.

Communication evidence

• emails;
• WhatsApp messages;
• internal instructions;
• contracts with software providers;
• developer communications.

Behavioural evidence

Authorities may ask:

Did the algorithm systematically punish price deviations?

Did competitors modify their algorithms simultaneously?

Were commercially sensitive data exchanged?

Did the system deliberately reduce price competition?

24. The "Black Box" Problem

Machine-learning algorithms can create an attribution problem.

A company may argue:

"We did not instruct the algorithm to coordinate."

But competition authorities may ask:

Who designed the objective function?

Who selected the training data?

Who determined the reward function?

Who approved the algorithm?

Who monitored its outputs?

Who continued using it after observing coordinated pricing?

Therefore, lack of human intention does not necessarily eliminate competition-law risk where human decisions created or maintained the conditions facilitating coordination.

25. Tacit Collusion and Artificial Intelligence

AI systems create three distinct levels of concern:

Level 1 — Human-designed coordination

Humans intentionally use AI to implement a cartel.

Traditional cartel law is relatively straightforward.

Level 2 — Human-facilitated coordination

Humans create systems that predict and respond to competitors.

Greater evidentiary complexity arises.

Level 3 — Autonomous convergence

Independent AI systems independently discover coordinated strategies.

This presents the most difficult legal question:

Can competition law address coordination that emerges without communication or human agreement?

Traditional agreement-based competition law may have difficulty addressing genuinely autonomous tacit coordination.

26. Consumer Harm

Algorithmic tacit coordination can produce:

• higher prices;
• reduced discounts;
• reduced price competition;
• higher platform fees;
• reduced consumer choice;
• slower innovation;
• artificial price stability.

The harm can be particularly significant in markets where consumers rely heavily on online price comparison.

27. Supplier and Labour-Market Implications

The same mechanism can operate outside consumer markets.

Algorithms can coordinate:

Supplier prices

Multiple buyers use software to determine procurement prices.

Freight rates

Logistics companies use common systems to adjust transportation prices.

Wages

Employers use algorithms to determine compensation or monitor labour-market rates.

Rents

Landlords or property managers use common pricing software.

Therefore, algorithmic coordination is increasingly relevant to both:

product-market competition

and

labour-market competition.

28. Difference Between Tacit Collusion and Explicit Algorithmic Cartel

29. Regulatory Challenges

A. Attribution

Who is responsible?

• company management;
• programmers;
• software provider;
• platform;
• algorithm itself?

An algorithm has no independent legal personality, so responsibility must ultimately be attributed to human or corporate actors under applicable law.

B. Causation

Authorities must determine whether the algorithm actually caused:

higher prices → reduced competition.

Correlation between algorithmic adoption and price increases is not necessarily sufficient.

C. Intent

Traditional cartel enforcement often relies heavily on evidence of coordination.

AI systems complicate the question because outcomes can emerge from machine learning.

D. Proof

The relevant evidence may be technically complex and difficult to interpret.

Competition authorities may therefore require:

• forensic data analysis;
• source-code examination;
• economists;
• computer scientists;
• digital forensic experts.

30. Compliance Measures for Businesses

Businesses using pricing algorithms should consider:

1. independent algorithm development;
2. clear compliance policies;
3. restrictions on competitor-sensitive data;
4. auditing of algorithmic outputs;
5. documentation of pricing objectives;
6. human oversight;
7. periodic competition-law review;
8. separation of competitor datasets;
9. monitoring of common software providers;
10. procedures for suspending problematic algorithms.

31. Important Legal Test

A useful analytical framework is:

Step 1 — Identify the market

What product, service, labour or geographic market is affected?

Step 2 — Identify the algorithm

Who created and controls it?

Step 3 — Identify the data

What information does it receive?

Step 4 — Identify competitors

Does the algorithm receive competitors' confidential information?

Step 5 — Identify communication

Was there communication between competitors or through an intermediary?

Step 6 — Examine algorithmic behaviour

Does the algorithm:

• monitor competitors?
• retaliate?
• stabilize prices?
• discourage deviations?

Step 7 — Determine coordination

Is there evidence beyond mere parallel conduct?

Step 8 — Assess competitive effects

Has the system produced:

• higher prices;
• reduced output;
• reduced innovation;
• exclusion;
• reduced consumer choice?

Step 9 — Apply jurisdiction-specific law

Relevant standards may include:

• agreement;
• concerted practice;
• cartel;
• hub-and-spoke coordination;
• exchange of competitively sensitive information;
• abuse of dominance.

32. Key Case-Law Principles at a Glance

33. Conclusion

Algorithms fundamentally change the speed, precision and stability with which competitors can observe and respond to one another. They can therefore make tacit coordination easier to sustain, particularly in concentrated, transparent and repetitive markets.

However, an important legal distinction must be preserved:

Algorithmically similar prices are not automatically evidence of an unlawful cartel.

The critical inquiry is whether the algorithm is merely independently optimizing a firm's competitive strategy or whether it forms part of a mechanism through which competitors knowingly coordinate, exchange competitively sensitive information, or facilitate a common anticompetitive strategy.

The existing jurisprudence—from Theatre Enterprises, Container Corporation, Interstate Circuit, Twombly, Airline Tariff Publishing, Wood Pulp, Eturas, and AC-Treuhand—provides the conceptual foundation. The principal unresolved challenge is the genuinely autonomous scenario in which independent algorithms discover stable supra-competitive outcomes without any identifiable human communication or agreement.