In January 2014, the Competition Commission of India (‘CCI’) initiated a rather complex investigation involving the allegation of a price cartel in the airline industry (‘Airlines Cartel’)^[1]. The investigation, which was based on a letter from the Lok Sabha Secretariat, lasted seven years and included all the major domestic airlines operating at the time i.e., Jet Airways, Indigo, Go Air, Spice Jet, and the Government-owned, Air India.

With the technological advancement and unprecedented growth of air travel over the last couple of decades, airlines have implemented third party software(s) that helps them determine, implement and dynamically change the fares offered to consumers, on a real time basis. Each such software is based on an increasingly complex set of algorithms that take into consideration factors such as demand conditions, actual booking, price of competitors, seasonality, etc., to determine the airfares.

Accordingly, one of the key points of investigation and determination for the CCI was whether: such third-party software(s) have been implemented by airlines by way of a common understanding, or such software(s) result in or facilitates price collusion in some other way. The CCI recently concluded that there was no evidence to suggest that the airlines entered into a price cartel since: (i) the ultimate decision on pricing was undertaken by a revenue management teams of the respective airlines and the software only assisted in price determination; and (ii) the heavily fluctuating market shares of the respective airline during the period under investigation, strongly indicated the absence of a cartel. This article examines the key reasons for growing concern over price collusions facilitated by algorithms, in the matured antitrust jurisdictions, and examines the line of reasoning adopted by the CCI in the Airlines Cartel.

Concerns over Algorithm Facilitated Collusion: Global Perspective

In almost all the antitrust law jurisdictions, price fixing cartels are considered anti-competitive irrespective of the mode or manner in which they are implemented or operated. In the digital age, the equivalent of a ‘smoked filled room conversations’ is the use of algorithms to implement, monitor, and operate a cartel. From a legal perspective, there seems to be a general consensus that if the decision to use the same software, or sharing of algorithms, is agreed between competing enterprises to facilitate or implement price fixing arrangement, then such conduct would invariably be viewed as an anti-competitive arrangement. Therefore, while the use of algorithms by businesses to determine prices dynamically has invariably resulted in an increased level of efficiency, it has also led to potential concerns from an antitrust perspective:

i.      Hub and Spoke Scenario: Online retailers (the spokes) using common third-party algorithm software (the hub) may find themselves facing a cartel investigation, due to unintentional similarity in prices. On the other hand, the knowledge of the use of common software by competing enterprises and determination of prices, accordingly, could lead to allegations involving tacit collusion. To give an example, in a case involving the administrator of a Lithuanian online travel booking system i.e., E-turas (‘Eturas’), the question of knowledge was the focal point for EU Court of Justice. In Eturas, the administrator of the online booking travel system sent a message to all the travel agents registered with it to cap the discount rates. The Court of Justice of the EU held that travel agents who knew of the administrators’ message capping the discounts could be presumed to have participated in a cartel unless they publicly distanced themselves from the message.

Therefore, firms who independently procure services of an algorithm provider, knowing that their competitors are using the same services and that the algorithm fixes prices at a certain level, need to exercise effective due diligence to avoid any inference of collusion.

ii.    Self – Learning Algorithms: the artificial intelligence capabilities of software e., the increasing ability of algorithms to learn through experiences and make autonomous decisions may lead to potential collusive outcomes without any intent or explicit meeting of minds between humans.

iii.   Market Transparency and Tacit Collusion: the use of pricing algorithms combined with growing market transparency of current prices may also lead to collusion like effects. However, in a market where firms take unilateral decisions based on their own assessment of the input from algorithms, including by way of access to competitors’ real time pricing would not result in any form of anti-competitive arrangement or understanding absent any form of agreement or understanding. In other words, monitoring competitors’ prices in an increasingly price transparent markets and reacting to competitors’ price changes, absent any agreement or understanding, is outside the preview of legal intervention of antitrust laws.

CCI’s Reasoning in the Airline Cartel

Interestingly, CCI did recognise the presence of similar or same software used by the airlines, but more importantly, applied the very basic principle of the evidentiary standard required to prove direct or circumstantial evidence to prove a cartel conduct. The CCI reasoned that the significant fluctuation in market shares of competing airlines over a period of 4 – 5 years indicated competitive behavior in the industry, where multiple existing players lost share to a new or recent player. Further, the CCI did not find any direct evidence to the effect that the common software was implemented by competing airlines with a view to fix prices. In the relation to the modality of determination of prices, the CCI noted that the involvement of ‘human’ element to decide the final prices indicated that the use of algorithms was only to facilitate genuine price determination in an industry that requires dynamic pricing and was not done with a view of implement price cartel.

Conclusion

CCI’s approach in the Airline Cartels case appears to be in line with its global counterparts, which stress the importance of evidence-based approach. Fluctuating market shares and the fact that the ultimate decision was taken by the revenue management team helped the CCI conclude that use of common software, does not necessarily equate to a finding of collusion. However, it is yet to be seen how the CCI would approach the issue of algorithm-based collusion in facts involving self-learning algorithms, increasing market transparency and tacit collusion.

[1] In Re: Alleged Cartelization in the Airlines Industry, Suo Motu Case No. 03 of 2015.