Learning in Standard Form Contracts: Theory and Evidence

Standard form contracts are known to be ‘sticky’: Although sophisticated commercial parties should be expected to update their contracts whenever doing so increases contract value, research has shown that this does not always occur. Terms that resist change, even when updating would be value enhancing, are “sticky.”

Stickiness, however, is not a general phenomenon—while some terms, like pari passu clauses in sovereign bond agreements, are resistant to change, even in the face of unfavourable judicial rulings, others are rapidly updated. In our recent paper, ‘Learning in Standard Form Contracts: Theory and Evidence’, we provide a novel account that explains why some terms are sticky, and others are not. The paper goes on to examine this theoretical account using a large, hand collected data set of online standard form contracts and empirical support for it.

Our account starts from the premise that the value of a particular term may be uncertain ex ante. Over time, however, firms can gather information and learn the relevant costs and benefits associated with certain terms. Firms learn indirectly by gathering information about terms offered by competitors, the outcomes of litigation, technological innovation, and news reports. Firms, however, also learn directly through experience with and feedback from consumers and contracting parties. Our paper focuses on the latter type of learning - experiential learning. Crucially, a firm’s ability to learn through experiential learning depends on its prior choice of contract terms.

Take, for example, a default implied warranty. If the firm chooses to offer the warranty, it may be able to charge a higher price, but it faces the risk of having to pay consumers in case of product failure. The expectation is that sellers will offer the term when doing so brings benefits that outweigh the costs. At the time that the firm must make this decision, however, the costs and benefits may be uncertain as they depend on information that is difficult to ascertain ex ante, such as the frequency of product failure, the types and amounts of consumer loss, and the frequency of consumer claims.Whether or not the firm will be able to learn over time depends on its choice of term. In the example, firms that offer the warranty gain the ability to learn the costs of the warranty over time and hence make better-informed choices in the future. In contrast, disclaiming the warranty forecloses this type of learning.

We call terms for which learning depends on initial contractual choices asymmetric learning terms. We call terms for which learning does not depend on initial contractual choices symmetric learning terms. Differently from symmetric learning terms, asymmetric learning terms - like the warranty in the example - have a “learning mode” - that is, offering the warranty - and a “non-learning mode” - that is, disclaiming the warranty. An asymmetric term’s learning mode provides value in addition to its short-term costs and benefits: it gives the firm a real option in the form of the possibility to revise the contract in response to additional information. Therefore, firms will be more likely to choose the learning mode at the outset. In essence, the firm’s ex ante choice of contract term is affected by the information-generating capabilities of a particular term.

One effect of this is that, in case of asymmetric learning, we would expect that terms initially adopted in their learning mode will, over time, be revised more frequently than those adopted in non-learning mode. The reason, of course, is that the learning mode provides firms with an opportunity to learn the true costs and benefits of the term and hence revise its initial choices. In contrast, for symmetric terms - those for which learning does not depend on initial contract choices -we should not see any correlation between initial contract choices and future revisions.Moreover, we expect an asymmetric-learning term in learning mode to be revised more frequently than a symmetric-learning term. Although one may expect the revision frequency to be the same - after all, both provide the firm with opportunities to learn - the real option embedded in the asymmetric-learning term biases the initial contract choice toward the learning mode, causing it to be revised more frequently later on.

In sum, stickiness is the result of an inability to learn.  The lack of experiential learning impairs a firm’s ability to revise its terms efficiently. For asymmetric learning terms, stickiness occurs when the firm adopts the non-learning mode. This happens when the non-learning mode of an asymmetric term is especially beneficial in the short term, which may be the case when the non-learning mode is the default contract term.

This theoretical account is borne out by our empirical analysis of a sample of End User License Agreements used by 246 firms from 2003 to 2010. As predicted, symmetric learning terms are equally likely to be revised at later dates, regardless of the initial mode. Asymmetric learning terms, on the other hand, are several times more likely to be revised late on if initially adopted in learning mode. Finally, asymmetric terms in learning mode are more likely to be revised than symmetric learning terms.

Florencia Marotta-Wurgler is professor of law at New York University School of Law and the director of NYU Women’s Leadership Network and the NYU Law Abroad in Buenos Aires.

Giuseppe Dari-Mattiacci is professor of law and professor of economics (by curtesy) at the University of Amsterdam.