Working Paper No. 05-36:
The Hazards of Precedent: A Parameterization of Legal Change


Charles Keckler

Date Posted: 2005

Abstract (below) | Full text (most recent) on SSRN


The overruling of prior caselaw is one of the most dramatic events in a common-law system. Unsurprisingly, therefore, the rate of overrulings is often considered an important measure of legal change. To measure this process more precisely, and to determine the variables governing it, a method is presented for the rapid collection and analysis of all binding precedents within a jurisdiction (in this prototype study the approximately fifty thousand substantive opinions of the Illinois Supreme Court between 1819 and 2005). From this study population, the 721 cases that have "failed," becoming no longer good law, and the cases or other processes causing them to fail, can be quickly separated and analyzed in order to identify factors predicting the relative durability of a judicial opinion. Given that 413 cases have thus far been overturned within the population, a long-term trend suggests the overall chance of being overruled converges toward approximately 1%, but certain types of precedents are at significantly greater risk, and certain periods of court history expose all precedents passing through them to elevated hazard. Application of engineering and epidemiological techniques of survival analysis and logistic regression allows specification of the time-dependent risk of failure experienced by a precedent. Quantification of certain factors that lower or raise the odds of failure, such as being issued with a dissent, or containing a rule of criminal law, is also possible. In addition, any concentration of precedent "mortality," when controlled for intrinsic characteristics of the precedents at risk, is a reliable indicator of an unusually hazardous judicial environment. In this Illinois population, such an environment existed between approximately 1950 and 1964, when here appears to have been a reevaluation by the court of its relative responsibility vis-á-vis the legislature for the correction of court-generated rules perceived as suboptimal but not clearly erroneous; this is correlated with, but not driven by, a change in the norm of stare decisis. A simple mathematical model relates the apparent shift in the level of self-correction with a spike in legal instability, followed by a higher equilibrium rate of court overruling. The use of logistic regression and related statistical techniques, combined with automated generation and coding of citation databases, appears to be a promising method for diagnosis of court dynamics, understanding episodes of legal change, and estimating the relative likelihood that a current rule of law will eventually be overturned.