Presenting Demand Models in Antitrust Litigation

A Q&A with Analysis Group affiliate James Levinsohn

Demand models – the statistical analyses that illustrate the complex relationships among product characteristics, product sales, prices, and consumer preferences – are essential components of an antitrust defense.

Here, Analysis Group affiliate Professor James Levinsohn, an antitrust and industrial organization expert and co-developer of the Berry–Levinsohn–Pakes (BLP) methodology for constructing demand models, discusses how best to present complex statistical models to triers of fact. 

Q: What are demand models, and why are they so important?

Professor Levinsohn: Antitrust experts, whether on the defense or the plaintiff side, often must determine which portion of a firm’s market share has resulted from alleged anticompetitive behavior. Making such a determination requires the expert to understand consumer behavior, determine what drove demand for the firm’s products, and identify whether and to what extent the conduct at issue may have influenced resulting sales.

Demand models formalize in mathematical terms the various influences that drive consumer behavior. The liability and damages portions of an antitrust matter – especially the damages portion – typically require an understanding of how a market would have behaved “but for” alleged bad acts. In this but-for world, the intensity of competition, the set of competing products, and/or the prices of products often are different from what they are in the actual world. Economists use demand models to quantify how these changes in the but-for world result in new prices and quantities.

Q: How do experts use demand models?

Professor Levinsohn: This depends a great deal on the particulars of the relevant market. Some markets are pretty simple and economists typically refer to the products in these markets as “homogeneous,” meaning everyone sells basically the same thing. A chemical used in industry or a particular type of bearing are examples. But in most markets, each firm sells its own set of products, and no two firms’ products are identical. Examples of these “differentiated products” markets include those for cars, cigarettes, beer, computers, and most consumer durables.

In an antitrust expert report, I use demand models to quantify how demand changes when a new product enters a market, an existing product exits a market, prices of competing products change, or when industry structure changes. Experts also use these models to understand how demand for a product changes when its attributes change – because of innovation, for example. Finally, experts use demand models as a foundation for merger analysis, since understanding the extent to which one product (or set of products) can substitute for other products is key to understanding the impact of a prospective merger.

The role of the expert in these matters frequently is to make the best use of the available data. In some markets, there only is product-level data (e.g., an average price and market shares), while in other markets there may exist a wealth of transaction-level data, such as point-of-sale scanner data. The expert’s challenge is to use the available data in a way that results in reliable estimates of a model of demand. Sometimes this means exploiting the variation in the data across geographic regions. Sometimes it means relying on variation within a region (or even a store) over time. And often it’s a mix of these. The guiding principle is for the expert to estimate demand reliably, and in a way that enables isolation of the effect of the challenged conduct.
 

Q: What are the challenges in presenting demand models to triers of fact?

Professor Levinsohn: Economic and statistical analyses in expert reports are useless if they cannot be explained simply to triers of fact – end of story. Being able to explain one’s analysis clearly and in terms that are intuitive to non-economists is essential. This is particularly true in complex litigation since a demand analysis typically is only one piece of the puzzle, and the triers of fact often must sort through a stunning amount of information. The challenge can be daunting, but I have found the best approach is to avoid unnecessary jargon, use examples to which anyone can relate, and highlight the key points.

I’ve also found that one technique for using demand models successfully is tailoring my approach to the question at hand. But no single approach is always best. I’ve done demand analyses that were as simple as surveying minor league baseball teams regarding their selection of bats, and as complex as modeling the American cigarette industry – with its hundreds of brands sold through outlets as varied as mom-and-pop stores, superstores, and Native American reservations. The trick is to use the right tool for the job. ■