Measuring the Temperature and Diversity of the US Regulatory Ecosystem

Economies, like ecosystems, exhibit dynamic, complex behaviors resulting from the interaction of ‘organisms' inhabiting and altering their ‘environments.’  In economies, organisms and environments can be seen as companies and regulation, and just as changing environments can harm or help organisms, so too can regulation affect companies.  Yet, unlike studies of natural ecosystems, in economics we have not had a longitudinal, empirical measure of fundamental concepts like ‘temperature’ or ‘species diversity.’  Our research goal is to bridge that gap, and, in a recent paper, we find support for the common claim that regulatory exposure and complexity have been increasing in the recent past.

The key source for our research is the US Securities Exchange Commission and companies that meet its registration requirements.  These companies must file annual reports, the so-called ‘10-Ks,’ which provide information on performance and status, including current and foreseeable risks.  Critically, unlike other sources such as surveys, the statements in these 10-Ks are certified and attested to by both company officers and independent auditors under threat of criminal prosecution.  Furthermore, while the law encourages disclosure, the incentives for firms are not one-sided.  Through their annual reports, companies compete for capital by presenting themselves as better investment opportunities than their competitors. They therefore strike a balance between presenting a pessimistic future full of potential risks and an optimistic future absent any.  Given these observations, we believe that 10-K annual reports are likely to provide comprehensive and realistic descriptions of firm and environment factors.

Relying on this, we operationalize a measurement of regulatory exposure through references to regulations and regulators in 10-Ks.  References generally take two forms: either a citation to an Act, as passed by Congress, or as a reference to a state or Federal agency.  For example, Trans Energy, Inc.’s 2015 10-K filing references both the Migratory Bird Treaty Act of 1918 and Endangered Species Act of 1973 below:

The Endangered Species Act (‘ESA’) was established to protect endangered and threatened species. Pursuant to that act, [...]. Similar protections are offered to migratory birds under the Migratory Bird Treaty Act. [...]

Similarly, FedEx, in its 2014 10-K filing, references the FAA and EPA:

Environmental. Pursuant to the Federal Aviation Act, the FAA, with the assistance of the U.S. Environmental Protection Agency, is authorized to establish standards governing aircraft noise. [...]

By applying natural language processing and machine learning techniques, we are able to identify and normalize over 4.5 million references to over 400 Acts and nearly 150 Federal Agencies in over 34,000 firms.

We then define aggregate ‘regulatory exposure’ and ‘regulatory diversity.’  We calculate unnormalized exposure (‘energy’) as total references per year, and normalized exposure (‘temperature’) as the annual average rate of per-firm references .  Both measures demonstrate a substantial increase in regulatory exposure; ‘temperature’ exhibits a dramatic increase from 7.9 references per filing in 1995 to 30.8 references per filing in 2015.  While certain regulatory watershed moments like Sarbanes-Oxley or Dodd-Frank contribute to the trend, the increase is robust both before and after these well-known Acts.

Next, as measures of diversity or complexity, we examine two figures: the average number of unique Acts per filing and the average distance between regulatory bit vectors.  The former measure demonstrates that the diversity of per-firm regulatory exposure is increasing; in 1995, the firms were exposed to an average of 2.9 unique Acts per year, rising to 7.5 unique Acts per firm per year in 2015.

The regulatory bit vector diversity measure, borrowed from genomics, conceptualizes firms as vectors in a regulatory space, analogous to genomes; each firm’s vector is defined through 0s and 1s that encode the presence or absence of a certain Act or Agency.  We then measure the average distance between any two firms in this regulatory space, and find that they are becoming further apart.  In 1995, two firms chosen at random had, on average, 3.7 Acts not in common between them; by 2015, this number had risen to 8.9 Acts not in common.

In summary, these initial results represent the first large-scale, longitudinal characterization of the activity and complexity of the regulatory ecosystem.  We have identified increasing regulatory exposure per firm, providing evidence in support of the claim that the regulatory burden is increasing.  We also find that firms are becoming more diverse and complex from a regulatory perspective, as measured per firm and overall.  These conclusions are based on more than 20 years, 34,000 companies, 160,000 10-K reports, and 4.5 million references contained in uniquely comprehensive and unbiased 10-K reports.  We intend to release and maintain the source and data behind this study, and welcome collaboration with other scholars to advance the empirical study of regulation and law.

Michael Bommarito is Head of Research at the Law Lab at Chicago-Kent College of Law (Illinois Institute of Technology) and Daniel Martin Katz is Associate Professor of Law at the Chicago-Kent College of Law (Illinois Institute of Technology). Together, they are co-founders of LexPredict, LLC, a legal technology company.