Richard Moorhead recently drew my attention to a fascinating draft research paper from Robert Anderson and Jeffrey Manns, both associate professors at US law schools, on the drafting of US merger agreements.
The authors used computer textual analysis to review around 12,000 public merger agreements filed in the US over 20 years. They were able to construct ‘family trees’ around these agreements, tracing the precedents used for each deal and determining how those precedents had changed over time.
The authors used this innovative approach to explore “whether transactional drafting is driven by a rational process that minimizes the cost of deal documentation and risk to clients or by an ad hoc process that increases billable hours and risk.”
I think we all know where this is going.
The authors say their findings “provide strong evidence of significant (structural) inefficiency in the drafting process which raises costs and risk to clients”.
The spidery visual maps they have produced illustrate how a precedent used in a transaction is then picked up and adapted for use in subsequent transactions in the context of specific firms, including Davis Polk and Sullivan & Cromwell. The maps offer a unique insight into how transactional lawyering is done in practice. If agreements are drafted from a single source, such as a firm standard document, the map will show lines radiating from that source. In fact, the research revealed very little radiation. Instead, each merger agreement follows its own unique, winding and particular path from its ancestors. According to the authors:
“We show that a high level of “editorial churning”, unnecessary and ad hoc edits that appear cosmetic rather than substantive, takes place in legal drafting. This churning appears to go far beyond the necessary deal-specific edits, with over half of the text of merger agreements routinely rewritten even though the substantive provisions of merger agreements have similar features.”
The research suggests that the precedent selected for any particular deal depends largely on whether the firm has previously acted for the client on another deal, rather than on whether the deal was similar. Lawyers tend to use familiar precedents and adapt them in preference to using unfamiliar precedents that need less adaptation.
As Moorhead explains: “The overall claim is that the process of precedent selection and development for M&A agreements is a rather improvised process based on haphazard choices made in the rush to get going on a deal, rather than from a basis of thoughtful calculation about ideal deal structures and contract drafting.”
I think we’ve all been there.
Commenting on the research on the IP Draughts blog, Mark Anderson noted how understandable this approach seemed, adding that: “[T]he extreme length and complexity of many agreements may make it difficult for a busy associate to fully understand the terms of his firm’s precedents. It may be tempting to use a document he recently worked on; he spent many late nights negotiating that document, so there may be a comfort factor that he understands at least some parts of that document better than the official precedent.”
There still seems to be a lot of space between a merger agreement with similar features and one that is substantially the same and the research does not go so far as to say that all the work lawyers do on merger agreements is ‘unnecessary and ad hoc’: there is still the possibility of what Moorhead calls the ‘artistry’ of deal-lawyering.
However, the authors do argue that their findings call for an industry-wide solution of creating standardised templates for merger agreements that could be used across firms. (Does this idea sound familiar to anyone?) They point not just to the benefits of efficiency and risk reduction but also to the value of allowing judicial precedent to crystallise the meaning of standardised text.
One key aspect that struck me was whether we would see a different picture if similar research could be carried out in the UK, where professional knowledge management is more deeply embedded in the culture of legal practice. When I first went out to the US in 2008 to help build the foundations of Practical Law’s US services, one of the key challenges was the fact that professional knowledge management as we know it in the UK was not nearly so widespread.
Another interesting point to note was that the median number of days between a document and its precedent is 423.5, meaning that the median precedent document is just over a year old. The authors say that this reflects a fairly strong preference for more recent precedents, with precedents over three years’ old being very unusual. There is a strong perception then that drafting needs to be updated, at least annually – but not necessarily continuously. It would be interesting to see if similar computer analysis could be used to test this, capturing how far substantive changes had to be made to precedent agreements to reflect changes in law or market practice.
