Hidden Cost Of Algorithms: Addressing Supra-Competitive Prices In Non-Collusive Settings

Flights and cab services like Uber extensively use algorithmic pricing. This allows them to quickly react to the changing dynamics of the market in real-time, based on factors like supply and demand, competitor's prices, and sometimes, even personal characteristics of buyers. Algorithms can analyse large chunks of data based on the preferred factors of the firm and implement its findings on the price of a commodity or service as fast as possible, depending upon its sophistication. Sellers who leverage algorithms consistently outperform their competitors in digital markets. Large firms like Amazon have highly advanced algorithms, allowing it to change its prices around 2.5 million times a day. Third-party pricing algorithms also exist to help out smaller firms that may not have their own algorithms.

Given such a widespread use and acceptance of algorithmic pricing, it's no surprise that researchers approached it with caution. A great amount of literature in the field of algorithmic pricing is attributed to explicit and implicit algorithmic collusion. Algorithmic collusion occurs when pricing algorithms interact in ways that lead to anti-competitive outcomes, like supra-competitive prices, i.e., prices above the competitive level. Algorithms make it easier for firms to pre-mediate such behaviour, however, similar results can be achieved without such pre-mediation as well, often dubbed as implicit or tacit collusion. Such antitrust risks of algorithmic pricing have been recognised by various competition authorities, and processes are underway to tackle them.

However, there is another peril of algorithmic pricing that has similar devastating effects while being even harder to tackle through antitrust enforcement. Despite its significant impact, the issue may not have received the global attention it deserves.

How Supra Competitive Prices Are Achieved In Non-Collusive Settings?

The one thing that is often forgotten while discussing algorithmic pricing in competition law is the varying levels of sophistication between different firms' pricing algorithms. This leads to two different scenarios called 'asymmetric frequency' and 'asymmetric commitment,' as described by Brown and Mackay. These scenarios are responsible for supracompetitive prices in non-collusive settings.

Asymmetric frequency

Asymmetric frequency occurs when firms have different abilities to update their prices. A firm with an advanced pricing algorithm can adjust its prices frequently, sometimes multiple times per day, while a competitor with a less sophisticated system may only update prices once a week. This difference creates a strategic advantage for the faster firm. It can quickly lower prices in response to market conditions and competitors' prices, while the slower firm is unable to react as swiftly. Knowing that its competitor can undercut its price at any moment, the slower firm is discouraged from aggressive price competition. Instead, it sets a price above the competitive level to avoid triggering a price war it cannot win.

Meanwhile, the faster firm, recognizing that its rival has limited flexibility, sets its price just below the slower firm's price but still above the competitive level. Ultimately, this arrangement leads to higher prices for consumers because both firms price above the competitive level despite still engaging in competition.

Asymmetric commitment

Asymmetric commitment happens when some firms have more advanced pricing algorithms that can automatically adjust prices based on their competitors' changes. These algorithms follow pre-set rules and update prices multiple times before the company needs to change them. This means firms with better technology can commit to a pricing strategy in advance.

This gives them an advantage over firms with less advanced technology. The more sophisticated firm can quickly react to price changes, while the slower firm cannot. Knowing this, the slower firm avoids aggressive price competition and keeps its prices higher to prevent constant undercutting. Like asymmetric frequency, asymmetric commitment leads to higher prices for consumers because firms adjust their strategies in a way that reduces competition.

It's important to note that both scenarios do not involve collusion. Instead, the firms act in their self-interest, and still, supracompetitive prices are achieved.

Can Antitrust Laws Be Enforced Against Such Scenarios?

Asymmetric frequency and asymmetric commitment are the results of independent business decisions of firms. It leads to supercompetitive prices, but without any collusion. Due to this, enforcing antitrust laws against such behaviour becomes a difficult task.

For example, under Section 3 of the Competition Act, 2002 ('The Act'), liability requires proof of a concerted practice or an agreement between firms. However, firms using pricing algorithms do not communicate or coordinate directly; instead, they independently react to market signals and competitor's prices. Since Indian laws do not penalize supracompetitive prices directly or even tacit collusion, such behaviour escapes antitrust scrutiny. Similarly, Section 4 of the act only applies if a firm is dominant and engages in exclusionary or exploitative practices. However, pricing asymmetries occur even among non-dominant firms, making it difficult to establish abuse of market power. Furthermore, it will be challenging to hold that using highly sophisticated algorithms by dominant firms is itself an abuse of market power.

Even advanced antitrust regulators like the European Commission (EU) and the U.S. Department of Justice face similar difficulties. The EU's Article 101 (anti-collusion) and Article 102 (abuse of dominance) require proof of coordination, similar to Section 1 of the Sherman Act, which is absent in these scenarios.

What Can Be Done?

As explored above, the two factors that lead to supracompetitive prices from algorithmic pricing are asymmetric frequency and asymmetric commitment. Mackay and Weinstein suggest that to regulate supra-competitive prices, both factors or even one of them, need to be eliminated through governmental regulations.

Regulation of Pricing Frequency

The main cause of asymmetric frequency is the difference in the level of sophistication of algorithms used by firms, creating an imbalance in how many times firms can change their prices. Firms with less sophisticated pricing technology recognize that whatever price they set can be beaten by those that can adjust prices more often. Hence, this kills any incentive for these firms to lower prices.

However, if all firms had the same frequency to update prices, the incentive to lower prices to compete in a price competition would not be eliminated.

This can be established through regulations that permit when prices can be updated. A real-life example of it can be the 2009 Austrian law that regulated when gas stations can increase their prices; however, there were no restrictions on decreasing them. This would prevent firms from establishing a leader-follower pattern and instead encourage them to offer their best price upfront. Algorithms would still be useful for analysing market trends, demand, and other factors, but firms with slower technology would no longer be discouraged from competing on price. In a well-functioning market, this could help bring prices closer to the competitive level.

Moreover, reducing the time between price changes would force firms to use larger, more noticeable price adjustments to maintain higher prices. This would make it easier for regulators and consumer groups to detect and challenge pricing strategies that harm competition.

Regulation of Pricing Algorithms

Essentially, asymmetric frequency in pricing creates situations that lead to supracompetitive prices only because algorithms factor in the competitor's prices. If this is not factored in, firms with superior pricing algorithms won't be able to undercut smaller firms solely on the basis of their prices. Algorithms will still be able to determine prices on other significant data like market conditions, seasonal conditions, inventory, etc.

This regulation will allow repricing as often as firms want, helping firms react quickly to market conditions without any restriction.

The problem with this regulation is the difficulty of enforcement. While regulators could easily monitor when firms set prices, it would be much harder to determine how they are setting them or whether algorithms are referencing competitors' prices. Effective enforcement would likely require firms to submit their pricing algorithms for review, ensuring they do not rely on rivals' prices. This would necessitate creating a new regulatory body, increasing government size, cost, and oversight power. This solution aligns with scholars like Andrew Tutt, who have already recommended the formation of centralized authority to oversee algorithms.

Algorithms are going to be a core part of the economy. Its effects can already be observed whenever products or services are bought through digital platforms. It's imperative that its benefits and drawbacks are scrutinised carefully.

This article has discussed a major issue with algorithmic pricing, i.e., supra-competitive prices, that can result without any collusion between firms. As no collusion is involved, the two provided solutions are also regulatory in nature, as antitrust enforcement will not apply to them.

If asked which solution is better, then regulating pricing frequency takes the spot because it is much easier to enforce and implement. Regulating the algorithms directly will require a great amount of investment; however, it will secure the ability of firms to react to market conditions quickly. Further research can look into the strengths and weaknesses of both approaches deeply to reach a more constructive answer.

Views are personal.