Abstract
The platform design, the process that helps potential buyers on the internet navigate toward products they may purchase, plays a critical role in reducing search frictions and determining market outcomes. We study a key trade-off associated with two important roles of efficient platform design: guiding consumers to their most desired product while also strengthening seller incentives to lower prices. We use simple theory to illustrate this, and then combine detailed browsing data from eBay and an equilibrium model of consumer search and price competition to quantitatively assess this trade-off in the particular context of a change in eBay's marketplace design.
Introduction
In digital marketplaces, search friction is a double-edged sword. While high friction makes it harder for consumers to find what they want, it can also protect seller margins by preventing instantaneous price comparisons. Conversely, a perfectly efficient platform that immediately directs users to the cheapest item might trigger a "race to the bottom" in pricing, potentially driving out high-quality sellers. This paper investigates how e-commerce platform design—specifically the layout of search results—mediates the tension between helping consumers find the "right" product and forcing sellers to compete on price.
The eBay Case Study
The study focuses on eBay's transition from a fragmented Search Results Page to a centralized Product Page. In the original design, many different versions of a product were listed together, requiring multiple clicks to compare. The redesign aggregated listings for a single item, highlighted a "Buy Box" winner, and sorted others by price. This shift provides a natural experiment to observe how design choices affect consumer search behavior and seller pricing strategies.
Structural Modeling and Estimation
To quantify these effects, the researchers employ a structural econometric approach combining three main components:
- Sequential Search Model: This explicitly models the user's decision to click. Users observe a noisy signal of a product's value and must pay a "click cost" to reveal the true utility.
- Consideration Set Formation: Using a Wallenius' non-central hypergeometric distribution, the model samples listings into a user’s view based on price and seller reputation.
- Supply-Side Equilibrium: Sellers are modeled as participants in a Nash-in-prices game. By inverting the first-order conditions of the sellers' profit-maximization problem, the authors recover unobserved marginal costs to simulate reactions to design changes.
The model is estimated using Maximum Simulated Likelihood (MSL), identifying 14 parameters that define the search environment and consumer preferences.
Key Empirical Findings
The "Buy Box" and Search Inertia
The introduction of the Buy Box—a featured, low-priced listing from a top-rated seller—was the most influential change. It captured 64% of purchases post-redesign, up from 27% for the equivalent listing previously. Interestingly, the estimated click cost ($\kappa$) is high, equivalent to approximately $7.73, while the signal correlation ($\rho$) is nearly perfect at 0.9998. This suggests that users do not click to "learn" about the product as much as they click because the platform requires it to complete a transaction.
Furthermore, the search strategy deviation costs ($d_0$ and $d_1$ in the table above) reveal significant inertia. Before the redesign, users rarely navigated to the Product Page because the Search Results Page was the default and the cost to deviate was high. Once the default flipped, users stayed on the Product Page, confirming that platform defaults are powerful tools for steering behavior.
Price Elasticity and Trust
Demand on the platform is highly elastic, with median price elasticities ranging from -8 to -11. This high sensitivity means small shifts in visibility can cause massive swings in sales volume. The model also quantifies the "trust premium": a Top-Rated Seller (TRS) could price a game at $37.17 and compete equally with a non-TRS seller pricing at $35.00.
Conclusion
The structural model successfully predicted the shift in purchase patterns, estimating a move from 40% to 65% for targeted listings, which closely matched actual data. While the "Product Page" design significantly consolidated the market and reduced search frictions, it also highlighted the platform's power to reshape competition. Future work may explore the long-term effects of such designs on seller entry, exit, and the total variety of products available in the marketplace.
Summary updated