The unit of analysis in this study is the country-year. Specifically, each unit is defined as the written constitution in force on December 31 of a given year for a given country. For example, constitutional amendments to Brazil's constitution that were put in force in August 2004 would be included in the case "Brazil 2004." Theoretically, this definition implies a universe of n countries times t years (see sampling information for our bounds on n and t). In practice, of course, the universe will be populated by only those country-years in which the constitution has changed. We call these units "constitutional events," defined as country-years for which there is a significant change in a country's written constitution. As a practical matter, the dataset will include only these constitutional events. For analytical purposes, of course, one could easily construct an uninterrupted time series by carrying past scores forward to interpolate years between events. As such, our framework is easily adaptable for those analysts interested in identifying the provisions that are in force in a given year.

An alternative unit of analysis, as opposed to "country-years," might be "constitutions" or "versions" of constitutions. For us, this alternative seems fraught with ambiguity. To be sure, some countries do have constitutions that are clearly distinct documents and are recorded as such (e.g. the Brazilian constitutions of 1988, 1968, 1945, and 1889). Other countries, however, make substantial modifications to a constitution without renaming the document (e.g. South Korea's six republics since independence have each been created through what are technically constitutional amendments). Still other countries may make only modest modifications but celebrate a "new" constitution. Finally, some countries tinker with constitutions almost constantly, a process that results in a proliferation of "constitutions." Given these concerns, we reason that country-years make for a more discrete and sensible unit of analysis. Through this strategy, "new" constitutions will be obvious to the extent that they exhibit form and content that is distinct from those in the document of the prior year.