|
We measure efficiency here by the number of hours it took to code a constitutional event, both for “original” constitutional events and for “updated” ones (that is, amendments to existing constitutions). The latter should and do take significantly less time. Also, we recognize that some constitutions can be both more complicated and longer than others. However, when we adjusted for the number of words in a constitution, length did not seem to affect coding rates much at all. For example, the Indian and Brazilian constitutions -- two of the longer ones -- were coded significantly more rapidly than average. Nonetheless, the constitutions selected for the pilot are representative in terms of length, so the average rates are generally what we can expect for the future.
Coders have spent an average of 15 hours per constitution for “original” constitutions and roughly 3 hours for constitutional events that are modifications of other events. We did find that coders improved dramatically in productivity over time, with the most recent coding completed by the average coder taking a full three hours less than the first constitution the average coder completed. This leads us to believe that we will realize significant productivity gains over the course of the study. Interestingly, we did not find much of a difference in efficiency according to the coder's experience. In particular, our undergraduates were as efficient as graduate students. Of course, the undergraduates are often seasoned Political Science majors, which may explain their high level of efficiency.
|
