Does Algorithmic Pricing Carry a Risk of Price Collusion?

In late 2022, Airbnb hosts using the platform’s Smart Pricing tool were consistently lowering their prices and outperforming their peers. At the same time, regulators were investigating RealPage, a company whose rent pricing algorithm allegedly contributed to rising rents through shared competitor data. Welcome to the murky world of algorithmic pricing.

Across industries, firms are turning to algorithms to set prices faster and more efficiently. But this shift raises a critical question: Can algorithms, even unintentionally, collude?

What Is Algorithmic Pricing?

Algorithmic pricing comes in several forms, each with different implications for competition:

• Rule-Based Pricing
  This is the simplest and most common approach. On platforms like Amazon, sellers often use repricing tools that peg their prices to those of competitors. For example, one seller might set their price 1 percent below a rival. But such rules can interact in strange ways. A now-famous case involved two booksellers whose pricing rules—one to undercut, one to overcut—led to a $30 book being listed for over $24 million. These tools can also cause “death spirals” where sellers continuously undercut each other to the bottom. Yet, because there’s no coordination, this behavior is legal under current antitrust laws.
• Demand Prediction (Machine Learning)
  More advanced systems use historical data to forecast demand and recommend prices. Airbnb’s Smart Pricing is a good example. Our research shows this tool often suggests lower prices, helping hosts increase bookings and earnings. Contrary to fears of price inflation, such algorithms may increase competition and make collusion harder to sustain.
• Artificial Intelligence (AI)-Powered Adaptive Pricing (Reinforcement Learning)
  The most sophisticated systems “learn” over time by testing different prices and observing outcomes. Though not widely adopted yet, research (including ours) shows these algorithms can independently learn to maintain high prices, resembling tacit collusion. Importantly, this occurs without any direct communication between firms.

When Is It Collusion?

To understand the risks, we must distinguish between explicit and tacit collusion:

• Explicit collusion involves firms agreeing to fix prices through emails, calls, or meetings. This is clearly illegal, as in the e-book price-fixing case where emails revealed publishers and Apple coordinating prices.
• Tacit collusion, by contrast, occurs when firms independently align their pricing behavior, often sustained by mutual threats of retaliation. For instance, if one firm undercuts a competitor, the competitor might respond by permanently slashing prices, a “grim trigger” strategy. This threat keeps everyone in line.

Tacit collusion is much harder to detect and not illegal unless supported by clear evidence of coordination. The challenge is that algorithmic pricing, especially with adaptive AI, can produce similar outcomes, even when firms have no intent to collude.

The RealPage Case: Data Sharing or Coordination?

RealPage, a rent-pricing software provider, is currently under scrutiny for its role in potentially driving up rental prices. Unlike many algorithms that rely solely on public data, RealPage collects sensitive, non-public information from competing landlords (such as occupancy rates and pricing), then uses it to forecast demand using ML algorithms and recommend rent levels.

The controversy lies in how those data are used. Predicting demand isn’t new or inherently anticompetitive. As seen in Airbnb’s Smart Pricing tool, better demand forecasting can lead to more competitive pricing and better outcomes for consumers and sellers alike. But RealPage’s approach differs: it relies on competitor data that individual firms would not typically share, potentially allowing landlords to align prices indirectly.

Here, the algorithm is not coordinating prices explicitly, but it may be producing similar results—higher, more uniform prices across competitors—without overt agreement. This raises a difficult legal question: Does pooling sensitive competitor data through a third party amount to coordination?

There’s no clear answer. But RealPage illustrates how collusion risks don’t always stem from complex algorithms. The concern here isn’t algorithmic intelligence, it’s that shared, sensitive data may act as a conduit for indirect coordination, even in the absence of any intent to collude.