Double posted from here.
So, i’m fiddling with some citation data for a few Demography journals,* and came up with a couple of weird data points that i can’t account for, so thought i’d see what teh intertubes could tell me about it. Basically, early in the process of working with this sort of data, I like to take a look at “Citation Age” information.** This gives a sense of how old the literature is that people are drawing on in a given time period. These often roughly follow linear increases (though the rate of “aging” differs across fields). Anyway, this one presented a couple of pretty extreme outliers (i’ve done this more than just a couple of times, and haven’t seen others this different). I can’t account for them, so am looking for any potential explanations.
Unfortunately the data isn’t currently in a format that would let me actually “solve” what’s accounting for this, but i should be able to soon, so it could be fun to see whose/which theories hold up. Anyway, if looking at this information by year, 1988 and 1991 are considerable outliers (see the plot after the jump). It would appear that most of the “blip” in 1988 comes from increased citations to work roughly 50 years before, while the one in 1991 comes from citations to work roughly 70 years earlier. Given that i’m not a “full fledged” demographer (training-wise), i’m guessing others of you might be able to help me out here. What happened in 1988 that led people to suddenly read/cite work from the 1930s and from 1991 to suddenly read things from the 1920s***? All potential explanations welcome.
* This currently includes Demography, Population Studies, PDR and PRPR. I’m happy to quibble with this definition of the “general” demography journals, and am definitely open to other suggestions. However, i think that’s likely a slightly separate question. I did initially aim to also include Demographic Research, but ISI (my main data source) didn’t start indexing that until 2005, so its considerable censoring made keeping it seem wonky.
** There are lots of different ways to compute “citation age” and after discussing this more times than those who work with me on these things would like, we’ve settled on typically using the option that basically depends on single cited-citing pairs contributing one observation (not a unique cited papers list, nor a citing paper-wise average). We think the others could be interesting, but for answering questions other than those we’re typically after.
***So far, we have one theory for the 1991 one which i’m looking into.