Strategic Responses to Algorithmic Recommendations: Evidence from Hotel Pricing
Algorithmic price recommendations to hotels are strategically biased because of human adjustment costs, leading to suboptimal pricing.
Summary
The study uses detailed hotel-pricing data and a structural model to examine how a pricing algorithm that advises human managers behaves. Because changing a price is costly for the manager, the algorithm anticipates that its advice will only be partly followed and skews its recommendations accordingly, so the advice is strategically biased and leads to suboptimal prices. The authors quantify the losses this causes and estimate the gain hotels would obtain by moving to fully automated pricing, which removes the mismatch between the human operator and the algorithmic adviser.
Editor's opinion
An algorithm that advises a human is not neutral: if it knows people resist changing prices, it shades its advice, and the result is worse pricing than the algorithm would choose alone. Firms should decide deliberately whether pricing is fully automated or genuinely human-led, instead of leaving an adviser to negotiate with a manager through nudges.
Key concepts
algorithmic advice · price adjustment costs · hotel pricing · human-algorithm