Government administrative codes are composed of regulations that specify how businesses and organizations can legally operate. These codes cover everything from federal labor standards to local zoning ordinances.
Despite the substantially large role administrative codes play in our communities, there is little infrastructure in place to support the drafting and management of these policies. At present, most governments’ internal processes rely on manual drafting that can carry unintended consequences like liability, legal enforcement, and ambiguity for governments and the public.
While the shift from paper to digital is an industry standard in most private sector professions, governments are just beginning to adopt digital best practices. A 2015 AIIM survey indicates that 46% of surveyed businesses said that the “removal of paper has been the biggest single productivity improvement for most of their business processes,” (Invensis). However, a digital shift doesn’t address the administrative debt associated with possible errors throughout the code.
The rough business standard for average error rates in manual data entry is around 1%; Anything above this rate is treated as a cause for concern and change (Invensis) by businesses. Businesses employ a “1-10-100 rule” for data entry, which refers to the following cost structure:
- $1 cost to verify data accuracy at the point of entry
- $10 cost to clean up or correct data when it is in batch form
- $100 cost or more for each record if no action is taken
- The stakes are arguably larger when it comes to our administrative code: an incorrect legal reference, misplaced historical note, or even a missing letter may create legal risks, market overcorrections, and regulatory uncertainty for impacted stakeholders.
- To get a rough idea of the government costs, let’s consider a 1% error rate of 104.6 million words of federal code: a 1.046 million word error range could cost almost $10 million for batch fixes and $100 million for a non-batch fixing of the problem based on the principles above.
- There is an opportunity for sustainable solutions that address the data entry and management needs in government. Esper provides a reliable and data-driven approach for solving these challenges.
By applying natural language processing techniques, Esper identifies and flags errors and inconsistencies in public policy. Some of the most common data points we detect include:
- Legalese: Many governments have rules for allowed and unallowed words, phrases, and grammatical types in policy text. These rules were created to ensure that policies don’t have any overly complex or incomprehensible language. For example, Connecticut’s Manual for Drafting Regulations includes a table for Plain English suggestions (see below). Esper automatically detects this “legalese” and suggests fixes.
- Repealed references: Governments must regularly review regulations to ensure that they don’t contain references to laws or rules that have been repealed. The consequences for a repealed reference could create a liability for the governing agency and result in legal action. Esper automatically surfaces repealed references with a red flag and links to the repealed rule.
- Rule review: Many states with sunset review processes or general rule review processes struggle to determine which rules are approaching expiration deadlines. Esper’s platform sources important dates for rules like its creation date, last updated date, expiration date (when applicable), and the last time that the rule was reviewed.
- Comprehension statistics: Governments often want to assess whether rules are easy to read, how restrictive they are, and how long they are among other factors. Esper’s rulemaking tools analyze rules and drafts to provide authors with that information for better policymaking.
Consider the real world example of Arizona’s regulatory code. Arizona has a well documented and accessible regulatory code but errors still exist:
- Legalese: Arizona has 9,296 instances of prohibited phrases, according to their own laws.
- Repealed references: Arizona code contains 30,594 citations across its 18,767 regulations.
1,837 citations point to repealed legislation and an additional 2,356 citations point to rules that aren’t on the books.
307 cited rules that are designated as one of the following: “Renumbered”, “Recodified”, “[Reserved]”, “Transferred”, “Expired”, or “Emergency Expired”.
6-13.7% of citations in Arizona’s code point to repealed rules.
- Grammar: Aside from the 9,296 uses of prohibited phrases, Arizona’s regulatory code contains additional 3,882 grammar mistakes. Manual entry always carries the risk of an error rate beyond that of a scalable system.
Relatively simple and minute aspects of rules have wide-ranging consequences for how rules are implemented and enforced at scale. At present, governments don’t have access to tools to assess their error rates and where errors persist in their code. Esper’s platform promises to slash error rates, increase productivity, and free up government resources to better serve the public.