8
with a large number of experts over a period of many months. In addition, the assigned
scores were presented and discussed in depth in various meetings, included one with a
group of invited participants who work in diverse settings (politics, academia,
international organizations) and come from different countries in the Americas
(Argentina, Brazil, Canada, Colombia, Ecuador, Mexico, the United States and Uruguay).
These discussions led to the identification of disagreements, which led to further
research, and yet further group discussions. In the end, through this iterative process, a
large degree of consensus emerged concerning the coding of the EDI’s four dimensions.
The Generation of a Rectangular Data Set with Normalized Scales. The third set of
choices pertained to the transformation of the scores on the component scales into a
rectangular data set, that is, a data set that includes numerical scores for all cases on all
variables and all years, with normalized scales.
There were several steps taken to provide numbers in all years. First, a number of
fairly mechanical issues were tackled. Pluses and minuses were turned into numbers by
adding or subtracting 0.33 from base score (for e.g., a 3 plus was turned into a 3.33).
Dashes (-) used to indicate that the assignation of a score was not applicable, because the
government not elected, were turned into zeros (0). In addition, the scores for two of the
components that were assigned scores only for the year in which an election was held—
suffrage and free elections—were extended to the intervening years by simply carrying
over the score for a given year to subsequent years, until a new score has been assigned
(either because an election was held after a period in which there was a non-elected
government, or a new election was held, or the election process was interrupted). The
justification for this procedure is that the way in which a government originates continues
Convention