As an investor, what’s more important, a company’s shareholder returns or its ranking in corporate governance tables? While most individual investors would say returns, more influential players have fixated on the forms of governance. The bad news is that this has probably cost investors; the good news is that help is on the way.
The gurus of corporate governance — dealing with things like shareholder voting rights and board elections — have convinced many that there’s a difference between “good” and “bad” corporate governance. Good governance increases “democratic” shareholder rights, like one-share/one-vote, while bad governance are things that increase “despotic” managerial power, like a CEO also chairing the board. Good governance reaps better returns for shareholders than bad governance, we are told.
As with the good and bad of consuming red meat, potatoes, and cow’s milk, real-world data and professional views evolve. In corporate governance, this next stage has arrived with a new study questioning years of conventional belief on what’s good and bad in governance.
Corporate governance data-mining dates back about two decades to pioneering work by finance professors Paul Gompers, Joy Ishi and Andrew Metrick. Using data created by the Investor Responsibility Research Center (IRRC, now part of Institutional Shareholder Services or ISS), they found that investors generally fare way better investing in “democratic” than “despotic” companies. Dubbed the G-Index, researchers proliferated numerous variations using similar datasets. Big advisers to large institutional investor advisors such as ISS and MSCI commercialized recommendations based on such data.
Now, a new study by law professors Jens Frankenreiter, Cathy Hwang, Yaron Nili and Eric Talley contends that this bedrock research contains many errors. Coders misinterpreted source material on some basic features, such as whether a company had dual-class shares, a staggered board or supermajority voting. In a multiyear effort, the law professors have built an entirely new dataset they hand-coded from the governance provisions of close to 3,000 public company charters. Comparing their findings with the original IRRC data and its offspring, the law professors report “alarming” errors in the original coding. Aggregate effects are dramatic, such as erasing most of any return premium to democratic compared to despotic companies.
Defenders of the status quo stress that the new research does not undermine the entire edifice, as much data and advice have been generated using other tools, such as relating executive compensation to corporate performance. But critics welcome the spotlight the new research shines on the often-obscure data behind the received wisdom in today’s governance debates, from takeover defenses to shareholder voting methods.
For investors, the new research highlights that it’s unwise to rely blindly on assertions of what counts as good or bad governance from any source — proxy advisors, data analytics vendors, professional service firms or academics.
Thoughtful, disciplined investors must probe the quality of the underlying datasets, particularly whether governance scores are based on these erroneous indexes. (It also pays to understand the provider’s baseline for good governance. The literature traditionally references corporate performance or shareholder returns as the baselines, whereas today’s providers may stress different priorities associated with such movements as impact investing, socially responsible investing or ESG investing.)
You may wonder how such coding errors could persist for so long. One cause is the deliberate effort to limit access to data. States such as Delaware, the leading charterer of corporations, charge hefty fees to obtain corporate charters and make them available in technologically primitive formats. The researchers estimated that the total cost of building a database from the Delaware charters alone would be $500,000. Also, proprietary services such as ISS and MSCI have incentives to maintain strict data control, selectively selling access for substantial premiums to commercial clients.
Another cause is that the chief researchers in governance data analytics have been from the fields of finance, economics and data science and not law, the logical place from which to study corporate governance. Lawyers historically have lacked training or interest in empirical data crunching. The heavy work — collecting, analyzing and labeling data — was left to professionals without appreciation of legal nuances that underlay governance topics spelled out in legal texts such as charters and statutes.
In their study, the law professors aim to mitigate both causes. They redress the data access problem by making their entire corpus of hand-coded charters freely and publicly available for all to use. They solve the competence problem by having coded everything themselves and offering a user-friendly database that anyone can navigate.
Looking forward, the new research and database both promise better understanding for investors on key topics in corporate governance. There are many hypotheses to be tested. I’m eager to test one that I’ve often asserted: that governance provisions operate differently in varying contexts, so that what’s good for one company is bad for another. Investors benefit when managers do what they think is best for their particular company, not what people generally think is good for all companies.
Lawrence A. Cunningham is a professor at George Washington University, founder of the Quality Shareholders Group, and publisher, since 1997, of “The Essays of Warren Buffett: Lessons for Corporate America.” Cunningham owns shares of Berkshire Hathaway, Constellation Software and Markel. He is a director and vice-chairman of Constellation Software. For updates on Cunningham’s research about quality shareholders, sign up here.