Constitutional Amendment Versus Constitutional Replacement: An Empirical Comparison

Abstract

The distinction between constitutional amendments and constitutional replacements figures prominently in the study and practice of constitutional change, but it is also deeply problematic. Scholars have struggled to distinguish between mere “amendments” and true “replacements” in a manner that is both conceptually satisfying and amenable to empirical study. A common approach is simply to look to whether particular constitutional changes are formally labeled as “amendments” or “replacements.” It is widely known that the application of these labels can be arbitrary or idiosyncratic, with the result that the labels do not necessarily correspond to the actual importance or magnitude of the changes in question. Little is known, however, about how reliably–or unreliably–these formal labels reflect the actual degree of change. Nor do the methods traditionally used by empirical scholars offer convenient alternatives for gauging the magnitude of constitutional change.

Natural language processing techniques drawn from computer science offer powerful new tools for measuring textual change and investigating various aspects of constitutions, such as the empirical difference between amendments and replacements. Latent semantic analysis of 1,000+ constitutional changes in the world over a fifteen-year period shows that the formal “amendment” and “replacement” labels paint a systematically misleading picture of the magnitude of constitutional change and cannot be used to reliably distinguish between major and minor changes. On average, so-called “amendments” involve changes of a lesser magnitude than so-called “replacements,” measured in terms of the proportion of the text that changes. However, these labels are prone to both overstating and understating the degree of constitutional change. In some cases, amendments can be massive and amount in substance to stealth replacements. More frequently, replacements can be inconsequential and amount in substance to overblown amendments. In reality, most constitutional changes—regardless of how they are labeled—are relatively minor in textual magnitude.

Automated analysis of the constitutional texts also reveals an empirical relationship between the magnitude of constitutional change and the passage of time. The more time that elapses between constitutional changes, the bigger that the changes tend to be. At one end of the spectrum, some constitutions evolve in a gradual way through incremental and frequent tinkering. At the opposite end, other constitutions are characterized by punctuated equilibrium, or long periods of stasis interrupted by spurts of drastic change. These findings support the view that the need or demand for constitutional change tends to accumulate over time and leads to periodic major changes if it is not instead satisfied by smaller, more frequent adjustments.

Keywords: constitutions, constitutional amendment, constitutional replacement, empirical, topic model, natural language processing, latent semantic analysis, automated content analysis, constitutional change, constitution-making, constitutional evolution, punctuated equilibrium

JEL Classification: K00, K19