Wednesday, January 27, 2021
Assistant Professor of Business Law
David Nazarian College of Business and Economics
California State University, Northridge
Today’s post summarizes my paper’s findings regarding the factors (predictors or variables) that determine courts’ decisions in force majeure clauses. Yesterday’s post dealt with my paper’s explanation for the parties' inclusion of force majeure clauses against the backdrop of the same/similar doctrines in common law and UCC (i.e., impossibility, impracticability, and frustration of purpose).
The second inquiry of the article tries to make sense of courts’ decisions by analyzing the weight given to six types of events and three types of reasoning as explained below.
- Empirical Analysis of Courts’ Opinions
In reviewing courts’ opinions, I realized that courts generally use three factors (predictors) related to force majeure clauses: Whether the event was foreseeable (foreseeability); whether it was beyond parties’ control (controllability); and whether the language of contracts include the events before them (lexical interpretation). Although some of the inquires can be intertwined, my main question was the weight given to these factors. These factors can be outcome determinative. For instance, if foreseeability is key, the query is whether parties could foresee a global pandemic at the time of the formation of their contracts. If control is the main inquiry, the question is whether the non-performing party could avoid the consequences of Covid-19 (hence a due diligence test) by, for example, placing the restaurant tables outdoors. Last, if the language is key, the main question is whether Covid-19 falls under any of the events listed (is it an Act of God, for example?).
A corollary of this inquiry is whether the type of events (for instance, natural events as opposed to economic downturns) have any bearings on the type of reasoning the courts have employed. Accordingly, I created the model below which underpins the empirical aspect of the paper:
For the empirical approach, in summary, I used the Harvard Caselaw Access Project (they kindly featured this article). After cleaning, stemming and lemmatizing the text of the opinions related to force majeure from the year 1810 onwards, I applied natural language processing techniques including creating frequency scores based on bag-of-word model, TFIDF, and cosine similarity (in doing so I benefited from Jonathan Choi’s article, and Law as Data, edited by Michael Livermore & Daniel Rockmore, and empirical works by Jed Stiglitz, Eric Talley, Julian Nyarko, and Sarath Sanga among others). I created a dictionary of related concepts (for example, for the “governmental acts” category I included the words “regulations” “shutdown” “closure” “permit” among others.).
Last, I applied regression analysis (including bootstrapping techniques in particular Hayes’ moderation and mediation methods, which may be the first use of its kind in this context in legal scholarship).
Some of the findings reinforce what we expected and some are new. For example:
- Over time, courts have engaged increasingly with events related to (i) economic hardship (broadly defined) and (ii) governmental actions. (see Chart 2 here)
- Over time, in terms of frequency, courts have looked into intent and contractual language more. (see Chart 3 here)
- As courts engage more with the force majeure issue, they are more likely to deploy the control analysis and less so analyses related to foreseeability and contractual interpretation. Put simply, the control variable is the most important factor. (see Table 1 here)
- As to the direct effects, the results show that there are no direct effects between the types of events and the courts’ weighted score for force majeure. Put simply, no particular event triggers any particular analysis. (see Table 2 here)
- As to the indirect effects, the results show that the control variable consistently mediates the relationship between the event type and the weighted score of force majeure. Put simply, control explains the relationship between the type of event and force majeure (lexical interpretation is not significant). (see Table 3 here)
- The analysis related to control plays a more important role at the state level comparing to federal courts. Put differently, state courts tend to use the control factor more than federal courts. (See Chart 4 here)
I also hand-coded 110 decisions for their outcomes (accept or reject) (to be precise, decisions were coded as accept, reject, or not clear/not sure. I then created a dummy coded variable in which “reject” was coded as 1, “accept” was coded a 0 and “not sure” was not included.). The result is as follows:
- Acceptance of the force majeure argument is positively related to the number of times the term force majeure appeared in those cases (The same is true for rejection). Although this is subject to further research and robustness checks, it can reasonably be concluded that the frequency of the phrase “force majeure” is related to acceptance (the frequency of the phrase is a good indicator of its outcome).
In summary, the empirical aspect reveals that the type of event (e.g., natural events v. economic downturn) is not directly related to courts’ force majeure analyses. However, the control variable consistently mediates (explains) the relationship between the two. Put simply, the beyond-the-control-of-the parties standard is the most important variable. The effect of control on force majeure is moderated (more pronounced) in state courts.
Tomorrow’s post will explain the normative aspect of the article which suggests that promisee’s degree of reliance should play a bigger role in force majeure analysis.