A Brief Review of Constitutional Datasets

April 13, 2021

Examining and forecasting political outcomes such as interstate peace, intrastate conflict, and economic development is one of the primary goals of political science researchers and practitioners. Assessing these outcomes requires data that characterizes the institutions, regimes, and behaviors that lead to these outcomes. Therein lies one of the core debates of political science. How does one best characterize political institutions? How do you quantitatively measure levels of democracy and autocracy across the global spectrum? Datasets designed to answer these questions generally fall into 2 categories: 1) a mix of composite indices and measures of institutional behaviors and outcomes coded by teams of local experts, and 2) de jure measurements of directly observable institutional characteristics.

Datasets offered by Polity (Marshall2020?) and Freedom House (FreedomHouse2021?) generally fall into the former, while direct assessments of constitutional and electoral structure like the Institutions and Elections Project (Wig2015?) and the Comparative Constitutions Project (Elkins2009?) make up the latter. Other datasets like the Varieties of Democracy (Coppedge2017?) present an interesting blend of both. Neither are better than the other; they both have their merits depending on the use case. Relying on composite indices derived by long time experts can spare you weeks of exploratory analysis and derived variable construction, while using highly disaggregated metrics or de jure measures of institutional capability may better serve narrowly focused research questions.

The Institutions and Elections Project (IAEP) captures several aspects of historical national electoral procedures, electoral events, institutional provisions, and constitutional stability for 170 countries between 1960-2012. In contrast to Freedom House and Polity V, IAEP does not utilize abstract constructs of institutions and civil liberties; most of the included variables are clear and leave little room for interpretation. Variables include binary flags for the presence of a formal constitution, whether or not the constitution was in effect, the age of the current constitution, years since it was last amended, and several additional variables describing the country’s actual electoral process. These are direct measures of constitutional and electoral institutions that permit less ambiguous causal inference between institutional structure and political behaviors and outcomes. Version 2.0 of IAEP added additional variables that describe the age of the current constitution and the number of years the current constitution has been in effect. These variables would be helpful for assessing constitutional or regime stability, but in my experience they suffer from severe coding inconsistencies that render them unusable.

Timeline of Chad's constitutional events. Data generated by the Comparative Constitution Project's Chronology of Constitutional Events dataset.

Timeline of Chad’s constitutional events. Data generated by the Comparative Constitution Project’s Chronology of Constitutional Events dataset.

The most similar dataset to IAEP is the Comparative Constitutions Project’s Characteristics of National Constitutions (CNP). The central goal of the Comparative Constitutions Project is to provide data to legal scholars assisting in the drafting of constitutions (ComparativeConstitutionsProject2016?). In doing so, they created a rich resource for political scientists and environment-security researchers looking to examine the causal relationships between institutional structure and political outcomes. In contrast to Freedom House and Polity V, CNP does not utilize abstract constructs of institutions and civil liberties; most of the included variables are clear and leave little room for interpretation. The CNP and IAEP datasets have a lot of overlap, but differ in some ways. While the CNP is almost entirely focused on constitutional structure, the IAEP contains variables that also characterize rules dictating participation in elections, candidate nomination procedures, and provisions concerning the central bank. Additionally, the CNP contains a longer historical record than IAEP (1789 vs. 1960). Both datasets receive only intermittent updates (2-3 over 15 years of existence). Subsequently, they are more applicable for static research interest, because infrequent updates limits the ability to create production pipelines or annual forecasts using the data as it becomes more dated. In these instances, DPI may be a better dataset, because it receives updates every 2-4 years.

Unlike IAEP, the Comparative Constitutions Project also provides 2 datasets that are complimentary to the CNC: Constitute (Elkins2014?; Elkins?) and the Chronology of Constitutional Events (Elkins2009a?; Elkins2020?). The Chronology of Constitutional Events (CCE) is a narrowly focused offering containing annual country-year observations of generalized “constitutional events”. There are 6 unique designations:

  1. New Constitution
  2. Amendment
  3. Interim Constitution
  4. Suspended Constitution
  5. Reinstated Constitution
  6. Non-Event (years without the above)

The limited scope of the CCE lends itself more to timeline visualizations or a quick reference, but could be helpful when used in conjunction with additional datasets or in other applications. CCE could also be used to derive quantitative metrics of constitutional stability similar to those included with version 2.0 of the Institutions and Elections Project.

Constitute is more of an investigative web-based platform than a self-contained dataset. The Constitute website contains a searchable database of verbatim transcriptions of more than 200 current and historical constitutions. The database may be filtered by clause using dozens of pre-defined categories spanning a wide range of topics including cultural, judicial, executive, oversight, etc. Although Constitute does not have a traditional centralized “file” to access, there are multiple ways to interface with the project. The project’s ontology is available as an OWL file directly from their website. The project metadata is also available in the N-Triples (.nt) format. Each country page provides links for HTML and PDF copies of the constitution in addition to N-Triple, Turtle, and RDF/XML versions of the constitutions excerpts and articles. Lastly, Constitute also maintains an API service so programmatic requests can be made to the database.

All together, the IAEP and Comparative Constitutions Project’s CNC, CCE, and Constitute provide a fantastic and comprehensive resource for researchers and practitioners across a wide range of disciplines.