Ex ante versus ex post expectation damages

Zhiyong Liu*, Ronen Avraham

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


What information should courts utilize when assessing contract damages? Should they award damages that were rationally foreseeable at the ex ante stage (ex ante expected damages)? Or should they award damages at the ex post level, incorporating new information revealed after contracting (ex post actual damages)? In practice courts have varied between the two approaches, awarding damages equal to the lower, or the higher, of the two measures of damages. This article shows that ex ante expectation damages are more efficient than ex post actual damages through a simple model of costly litigation for contract breach, where there are either costs of verifying the breach victim's ex post damages, or general litigation costs such as attorneys' fees. Courts should award foreseeable flat damages, rather than seeking ex post accuracy and awarding actual damages, because actual damages lead to distortions in breach incentives once we take parties' litigation decisions as endogenous. With costly litigation, ex post expectation damages may cause over-performance or under-performance depending on whether the American or the English rule applies and on the size of the litigation cost. We find that regardless of the direction of the distortion, actual damages induce inefficiency. Ex ante damages are more efficient because of the insensitivity of parties' litigation decisions to their ex post private information under fixed damages. Our results are robust when accounting for renegotiation.

Original languageEnglish
Pages (from-to)339-355
Number of pages17
JournalInternational Review of Law and Economics
Issue number4
StatePublished - Dec 2012
Externally publishedYes


  • Asymmetric information
  • Breach of contract
  • Expectation interest
  • Renegotiation


Dive into the research topics of 'Ex ante versus ex post expectation damages'. Together they form a unique fingerprint.

Cite this