Antitrust laws are designed to benefit consumers. As a result, antitrust enforcement should encourage discounting to the extent that those discounts benefit consumers. But bundled discounting by a firm with monopoly power in one market can, under certain conditions, allow the firm to extend its monopoly to a second market. At the same time, bundled discounting is pervasive even in markets without monopoly power, and it is generally recognized that bundled discounting can have procompetitive effects. As a result, economists and courts have struggled to adopt a rule that will create the incentive to offer procompetitive bundled discounts without undermining the ability of courts and plaintiffs to identify bundled discounting with the potential to harm competition.
One method, recommended by some economists and enforcement agencies (Antitrust Modernization Commission Report and Recommendations (Apr. 2007)) and adopted by some courts (Cascade Health Solutions v. PeaceHealth, 515 F.3d 883, 903 (9th Cir. 2008)), is the use of an attribution test as a safe harbor. This test, sometimes called a discount-allocation test, attributes or allocates the discounts offered on the bundling product to the cost of selling the bundled product. For example, cable TV companies offer discounts on cable TV service (the bundling product) if households also buy Internet or phone service (the bundled product). The attribution test can create an effective safe harbor for firms engaging in bundling. Because the results of the test can be interpreted as addressing the question of whether the bundled discounts are unprofitable or, alternatively, whether the bundled discounts exclude competitors that only sell the bundled product, defendants accused of anticompetitive bundled discounting that pass the test are presumed to be engaged in legal discounting because they are not excluding competitors or pricing below costs.
While some have argued that the attribution test can yield false positives, I argue that the attribution test is flawed and can be ineffective even as a safe harbor because it can generate false negatives. More specifically, firms that fail the attribution test may be offering profitable discounts without excluding competitors, which benefits consumers. The primary flaw in the attribution test is that it assumes that the quantities sold are the same at different prices. This assumption violates one of the fundamental principles of economics. I also suggest a simple modification of the test to address this flaw.
Elements of a Bundled Discount Case
The typical bundled discount case involves a defendant firm selling two products in separate markets, and a plaintiff that sells only one product. The defendant has, allegedly, monopoly power in one market. The alleged monopolist bundles the monopoly product with the other product by offering discounts to buyers that agree to buy all or most of the second product from it. Thus, the monopoly market is often referred to as the bundling market, and the other market is referred to as the bundled market.
For example, consider a local monopoly supplier of cable TV services that also sells other non-TV services such as high-speed Internet access and/or VOIP phone service. Suppose that there is a local phone company that sells the same non-TV services but does not offer TV services. If the cable company offers discounts to households that buy both cable TV and non-TV services from it, then under certain situations, the phone company can be excluded from competing for customers. If so, bundled discounting can be anticompetitive because it has the potential to allow the cable-TV monopolist to extend its monopoly power to the other markets.
Details of the Attribution Test
The attribution test comprises three steps. First, one calculates the discounts given to a buyer for agreeing to purchase the bundle. Second, one attributes that discount to the bundled product by adding the discounts to the cost of bundled product. The idea is that bundling-product discounts are a “cost” of winning the bundled-product business. Third, one subtracts the total cost, including the attributed discounts, from the revenue earned on sales of the bundled product. If the total costs exceed the revenue, then the firm has failed the attribution test for that buyer. Often, economists will conduct the test on a buyer-by-buyer basis and calculate the share of buyers that fail the test.
The following simple example will help illustrate how an attribution test is conducted. Assume the following:
• The cable company charges $20 per month for each cable TV service (e.g., basic cable, premium channels, DVR, etc.).
• The cable company charges $20 per month for each additional service (e.g., high-speed Internet, VOIP phone service, etc.).
• The cost of providing cable-TV services and non-TV service is $10 per month.
• The cable company offers a 25 percent discount on cable services if a household buys its additional services from it.
• We observe a household buying five cable-TV services plus two non-TV services from the cable company.
The bundled discount for an attribution test is 25 percent of the $100 for TV services, or $25. In this case, the cable company’s revenue from non-TV services is $40, and its costs are $20. This means that the cable company’s profits on the non-TV services are $20. However, once you attribute the bundled discounts to the cost of providing the additional services and subtract those costs, the cable company’s profits are −$5 on those services. This means the cable company fails the attribution test for that household.
Interpretations of the Attribution Test
Two interpretations are often applied to the attribution tests, and there are two (not mutually exclusive) theories about why bundled discounting can cause anticompetitive harm. Some suggest that bundled discounting can have the same impact as predatory pricing, so tests of whether bundled discounts are anticompetitive should be similar to tests of whether prices are below costs. Others suggest that bundled discounts are better viewed as a form of tying. Under this latter interpretation, tests of whether bundled discounts are anticompetitive should focus on whether the bundled discounts exclude a supplier in a bundled market from competing. See, e.g., U.S. Dep’t of Justice, Competition and Monopoly: Single-Firm Conduct Under Section 2 of the Sherman Act 95–96 (Sept. 2008).
The attribution test can be thought of as a test of whether the bundled discounts are predatory because firms that fail the test are pricing below a measure of costs to win business in the bundled market. In the example above, the discount “costs” the cable TV provider $25 to make $20 in profits selling non-TV services. Because pricing below cost is generally unprofitable, firms pricing below cost may be doing so to reduce competition.
The attribution test can also tell us whether the bundled discounts can exclude a supplier of the bundled product from competing in the market in the same way that tying can exclude a competitor that sells, for example, only one of the two tied products. Suppose that the plaintiff and the defendant have the same cost of producing the bundled product. If the defendant fails an attribution test for a particular buyer, that means that a supplier of the bundled product would need to price below its own cost to be able to compete. Because firms cannot price below cost and remain in business, the bundled discounts would exclude the competitor from competing for that buyer. If the defendant fails the attribution test for a large share of buyers, bundled discounts could foreclose competitors from the bundled-product market.
Problems with the Attribution Test As a Test of Predatory Intent
The implicit assumption in the standard application of the attribution test is that the bundled discounts do not affect purchases of the bundling product. In other words, the only reason for offering bundled discounts is to win business in the bundled-product market (non-cable-TV services, in the hypothetical above). However, this assumption is inconsistent with basic economic principles: If you offer the bundling product at a lower price, you can sell more units of that product. So part of the reason why firms would offer bundled discounts is to sell more of the bundling product (cable-TV services, in our example).
The assumption that the quantities purchased would be the same at different prices causes two problems for using the attribution test to determine whether the bundled discounts are profitable. First, this approach incorrectly attributes discounts on the bundling product that would not have been purchased at the undiscounted prices as a cost of winning the business in the bundled product. Second, this use of the attribution test ignores the profits the firm makes on incremental units of the bundling product that it earns because of the discounts.
Consider the example above. We observe the household buying five units of cable-TV services at a price of $15 per service. Suppose that the household would have only purchased four cable-TV services at the undiscounted price of $20 per service, so the $5 discount induced the household to buy an additional service. The true cable-TV bundled discount in this case is only $20 because the discount should be calculated based on what the household would have bought at the undiscounted price (25 percent of $80 for four services) rather than what it actually bought at the discounted price. In other words, attribution of the $5 discount on the fifth service to the cost of winning the non-TV services is erroneous because that discount allowed the cable company to sell the fifth TV service. Of course, the true $20 discount still equals the $20 profit on additional services. However, the cable-TV supplier also earns an additional $5 profit on the sale of the fifth service that it would not have earned without the discount. So the true profit from the bundled discount is $5.
This example illustrates how, using the standard attribution test, one would conclude that the discount was unprofitable and, therefore, one might use the results to infer that the discounts were predatory. In fact, however, the discounts were profitable.
Problems with the Attribution Test As a Test of Foreclosure
As noted above, another interpretation of the attribution test applies to whether a firm in the bundled market can offer a price low enough to offset the discount offered in the bundling market to compete. Basic economics suggests that competition will be profitable as long as prices exceed their average variable costs (AVC). Thus, the attribution test is sometimes applied by asking whether a buyer would be willing to forgo the lost bundling-product discounts and buy from a competitor in the bundled-product market if that firm charged a price equal to its AVC.
The implicit assumption in the standard attribution test is that buyers would purchase the same number of bundled product units at a price equal to AVC as they do at the actual price. But as discussed above, this assumption is inconsistent with basic economic principle that lower prices induce buyers to purchase a greater number of units.
Assume the phone company’s AVC of providing non-TV services is $10 per service. The attribution test tells us whether, given the discounts offered by the cable TV provider, a household would be willing to buy non-TV services from the phone company for $10 per service. If not, then the bundled discounts would exclude the phone company from competing for that customer.
Using the actual purchases of the household, we could conclude the phone company was excluded. The household would have to pay $100 for cable TV services and $20 for non-TV services if it bought from the phone company for a total of $120. If it purchased the bundle from the cable TV company, it would pay $75 for cable TV services and $40 for non-TV services for a total of $115.
However, if the phone company offered additional services for $10 instead of $20, households would buy more services from the phone company. Suppose that at a price of $10 per service, the household would purchase a third service from the phone company. The total cost of buying all these services from the phone company, pricing at AVC for the non-phone services, would be $130. The cost of buying them from the cable company is $135. Thus, the household would prefer to purchase additional services from the phone company. In other words, while the standard application of the attribution test would conclude that the phone company was excluded from competing for the household’s business because of the bundled discounts, it can in fact compete because the lower prices for the non-TV services induces the household to buy more services.
A safe harbor is a screen designed to protect conduct that is clearly not anticompetitive. As a result, failing an attribution test does not imply a presumption that the firm engaged in anticompetitive bundled discounting. There are, however, two reasons why a flawed test is problematic. First, a more effective safe harbor would provide greater incentive for firms to engage in procompetitive bundled discounting. Second, judges and juries often have trouble understanding that a safe harbor is not supposed to be a demarcation between procompetitive and anticompetitive conduct. Fortunately, the analysis above suggests some simple modifications to the test that would improve the attribution test’s screen.
To use the attribution test as a screen for predatory intent, one needs to subtract the incremental profits on the bundling product from the “cost” of winning the bundled-product business. The only additional data one would need to do so would be the profit margin and the elasticity of demand for the bundling product. Data on profit margins are typically produced in litigation, and economists often use the inverse of gross-margin percentage as an estimate for the elasticity of demand. For example, if the cable company earned a 50 percent gross margin on cable services, that would imply its elasticity of demand is −2. This information would be sufficient for any economist to calculate the incremental profits on cable services from the bundled discount and conduct a modified attribution test that would be a more effective screen for predatory intent than the standard test.
To use the attribution test as a screen for exclusion by comparing the buyer’s cost of purchasing from the two sellers, one needs to include the incremental bundled-product purchases of the buyer at prices equal to the AVC. The only additional data one would need for that calculation would be the elasticity of demand for the bundled product. The elasticity of demand can be estimated using the gross margins on the bundled product. Again, this information would be sufficient for any economist to conduct a modified attribution test that would be a more effective screen to determine whether bundled discounting could exclude the plaintiff.
Keywords: litigation, antitrust litigation, bundling, discounts, exclusionary conduct, monopoly power, monopolization, attribution test, pricing below cost, foreclosure