Panel 4 Discussion
Discuss Panel 4 presentations here. Feel free to start a new topic.
Hello and welcome to the discussion thread for panel 4 “Circulation Data Beyond the Libraries,” and thank you to Jennifer, Brooks and Ed for sharing their very impressive projects. As the discussant for this panel, it's my pleasure to ask a few initial questions to start the conversation.
Brooks and Ed presented a data-gathering project focused on data collected between 1919 and 1972 about the circulation of periodicals (magazines and newspapers). These data were initially gathered by third-party audit bureaus, which advertisers could use to compare periodical circulation and demographics as they made decisions about where and how their money might be best spent. The project's data and visualizations can be accessed and interacted with at https://sites.lib.jmu.edu/circulating/ and offer insight into shifting readership patterns in response to editorial changes, magazine-content trends, and outside demographic trends.
Jennifer's presentation explores the growth of viewership of a YouTube channel called Vlog Brothers, which currently has a subscription base of more than 3 million "nerdfighters", a colloquial term for Vlogbrothers fans whose communal identity comes from their engagement with Vlogbrothers content in a variety of ways. The Vlogbrothers themselves engaged in a "census" of their own community's reading habits (things their community members read, purchased, consumed), across several years and several questionnaire formats, and Jennifer's results note that Nerdfighters are aging with their community and consuming content across a much broader variety of platforms that can be linked to shared community norms.
Jennifer: I was struck by the parallels between your methods (qualitative contextualization, with your own humanities skillset, as a response to the quantitative structures of the census data) and the feedback loop between the Vlogbrothers' incoming data and their reshaped questionnaires (a stated need for the Vlogbrothers to remove themselves from the survey in order to get better external data). If you could do a follow-up survey with the Nerdfighters directly, what questions do you have based on your current analysis that might help you further explore the expansion of reading content that the Nerdfighters are consuming, particularly as they age with the Vlogbrothers and the Nerdfighers community?
Brooks and Ed: I particularly appreciated the examples of interaction between magazine content and circulation data (e.g. the visualizations that showed Black Mask dropping in circulation after the Maltese Falcon episode, Cosmopolitan's increase in circulation after Helen Gurley Brown's editorial takeover, and Ebony's circulation focus in the US South). The presentation notes future directions in the addition of international circulation data and summary data, but I also wonder what additional data you're hoping to gather--e.g. detailed demographic data (socio-economic class, income, race/ethnicity), qualitative advertiser and audience responses to content--in order to contextualize and deepen your understanding of the causal factors that help explain some of the patterns made more clear by your data and data visualizations.
Both: I'm struck by the need in both projects to tie circulation data that's contained within these datasets to quantitative and qualitative data that's outside the datasets. What kinds of data do you wish you had, and how might other researchers engaged in similar projects use your answers to help structure their methods moving forward?
I look forward to seeing these answers and I hope audience members and participants will add additional questions of their own!
Kalani: Thanks so much for your response! I'll offer a few thoughts on your questions, but I'm sure Ed will have more to share, especially when it comes to international data.
First, working with old data (and corporate data, to boot!) is a challenge in its own right. The ABC registered only the data it found valuable - really, only data that it felt it could monetize. And a lot of this ended up being inconsistent. For instance, the demographic categories (by size of location) shift over time. Initially, we thought we might include these, but a few problems popped up. First, the shifting size categories meant we'd likely have to group apples with oranges to make anything work over the long term. Second, many magazines appear to have been exempt from these more detailed reporting categories, so their forms were simply blank here. And third, the ABC also seemed to allow a magazine to use the same demographic information over multiple reporting periods, so you might have a report from 1953 with data from 1950 - not exactly the most timely. Because this was a mess, we decided to stick with what was consistent over the entire time - the geographic and issue data.
As for other categories - esp. race and gender - the ABC did not chart these at all, which is really a loss to this history. The best we can do is compare the state-level data to historical demographic state data. It's an imperfect solution, but it's the best option that these data allow. We don't currently have plans to do this - but we're hoping other scholars might make some of these demographic comparisons using the raw data on our project's OSF site.
As for the data we we wish we had, I think more granular demographic data (race, gender, income, etc.) are probably at the top of my list - just what you mention and what I've described here. The current iteration of the ABC (the Alliance for Audited Media) actually gets down, I believe, to the level of county or municipality. That would also be interesting to have for these historical magazines, though that seems like it could grow the dataset exponentially!
@kalanicraig -- thanks for your thoughts here. In some ways, I don't feel a strong need to create a survey because it's clear that Nerdfighters are clever, independent thinkers and readers, who do with surveys what they will. If I could ask Nerdfighteria more about their reading, I think I'd want to know more about the links between their reading and their making: we have examples of what Nerdfighters make in response to the stories that they find compelling because of social media and hash tag searches, the Project for Awesome and the perks the community makes to thank people for contributing to that charity fundraiser, and so on, objects on Hank Green's bookshelf that he's described as Nerdfighter-made, and so on. There's not an optimal way of searching for the phenomenon, so to me, that would be an interesting area of inquiry that could shape and inform thinking about making as a means of reader response.
@brookshefner and ed timke:
From your presentation it appears that you have some evidence of state by state/regional circulation. Elsewhere in the workshop there is a discussion of regionalism (Panel 5). Have you seen any indications of regional patterns in the work you've done with your data to date?
Also, an observation: in your presentation you attribute changes in circulation levels to editorial changes. In another sense, these changes are responses to shifts in audience demands/expectations. Cosmo, it appears, turned to Helen Gurley Brown in response to changes in the expectations of readership. In addition to the question you pose about editorial changes, your data lends itself to raising and exploring questions about changes in taste, audience demands, etc. Just a thought.
I find myself interested in the challenges related to constructing your queries. The appendix of your book lists search terms, many with multiple words, including shorthand, for the same concept. How complicated was the process of crafting queries to scrape this data? What were the ups and downs, lessons learned, etc.?
@jim-connolly: We have explored regional patterns, for sure, and we hope that more scholars will do so as they dig into the data we collected and digitized. In an article we published last year in the Journal of Modern Periodical Studies, we looked at a very small case of regional differences: the case of Scribner's vs. Harper's around 1929/1930. Ostensibly, this was to think about the presence of writers Ernest Hemingway (a Scribner's author and regular feature in Scribner's magazine) and William Faulkner (who published in Harper's). While Faulkner is considered the "regional" writer, it turned out that Scribner's was a much more "regional" magazine at this time - concentrated heavily in the Northeast - while Harper's had a much more national reach.
Also, editorship is just one of the ways to think about changes in circulation levels: the presence of certain writers (like Hammett) or cover art (as in the case of Ebony) also can correlate to sharp shifts in circulation. Magazines changed titles and cover prices, which also impacted things (not always for the better!). As for more long-term changes, you're right - the shift in audience sensibility has a big impact, though that is much harder to pinpoint and describe. Another big shift that happens at the end of our dataset is the change in postal rates for magazines, which simply killed off big magazines like Life and Look.
At the end of the day, we hope the data will allow folks to think through the many elements that impact magazine circulation and correlate with the rise and fall of circulation. Having more accurate circulation numbers can inspire scholars to start asking better questions! And these questions will hopefully generate better answers - and maybe even prompt us to complicate our conventional wisdom (such as the Faulkner = regional, Hemingway = national paradigm).
@jim-connolly -- much earlier in the book I indicated that I contracted with coders who generated those queries using the terms I generated based on my knowledge of the community and its canon.
The lesson learned, though, is that that while the variant terms left me more comfortable with the idea that we'd looked for mentions using a multiplicity of ways, they were not all equally useful. I think that using variants with some subjects (i.e., The Fault in Our Stars, which has been known as TFiOS long before its publication) and less so with other titles would be sufficient. Constructing the variant terms allows me to say that the material I discuss represents the results of a comprehensive search, but a less exhaustive search would have yielded similar results.
@jbpierce So *that's* a really interesting shift away from text analysis and toward visual analysis of the Vlog itself for object recognition. How often do the objects themselves "show up" in transcripts of the Vlogbrothers' own conversations, and is that a potential source of further inquiry?
My initial question for our presenters was driven by new source material, but as I was reading the questions and answers here, I wondered how much each of the projects might gain in understanding circulation by drawing on the other's methodologies?
@jbpierce , I wonder if you have a sense of the Vlogbrothers' viewership numbers across a number of axes: compared to other in-genre vlogs, on a content change-over-time basis using your existing research, correlate shifts in viewership with general vlog subscription patterns, looking at regional patterns?
@brookshefner and Ed, I wonder if you've considered a larger-scale content analysis of one or more of the periodicals itself, maybe with word2vec, so that you can get an initial overview of how the specifics of content drive readership? (@jim-connolly asks a similar question about editorial content choice, and so I wonder if that would help tie editorial changes to content and thence to circulation ups and downs).
@kalanicraig I'll respond in a roundabout way. When we started this project, we had some big ambitions. One was to include a cover image with each row of issue data - which currently contains information like cover price and editor, along with reported circulation. There's a great moment in Harold Hersey's 1937 memoir Pulpwood Editor where he says, "There is an old saying in the pulpwoods that any color will do for a cover just so it's red" (17). We thought: Were more pulp magazine covers red? Did red covers correlate to higher circulation? But we ran into a couple of problems here. One was that this ambition seemed possible when our number of titles was much lower, as it was initially. But as we expanded the number of titles, we found that it was difficult to locate covers for many, many of these magazine issues. This would, of course, leave our data woefully incomplete and prevent us from making any sort of reliable case for this. (The logistics of managing this mountain of images along with the mountain of data was also beyond daunting.) But there was another problem: intellectual property and copyright. Of the 54 years of data contained in our dataset, only 7 years (1919-1925) are currently in the public domain. Cover images remain the property of the magazine's copyright holder, as do contents - assuming those contents were not separately copyrighted by the author. So, a project like the one you describe would have some significant IP hurdles - it would also be difficult to get good access to complete text of these magazine runs (and, as folks like Robert Scholes and Clifford Wulfman note, this idea of "complete" contents should include all text like advertisements, etc.). Magazines that have the most digitized contents are also likely to be magazines that value these contents more highly. It's not impossible, but the approach might need to be more targeted, unless it's easy to get permission from copyright holders. Conde Nast, for example, is notoriously difficult with this issue - and they own the rights not just to the current Conde Nast portfolio, but also to older publishers like Street & Smith.
Ultimately, we hope that other scholars will put our data to use in precisely the fashion you describe here, using word2vec or other tools. Part of our goal with Circulating American Magazines has been to make these data available for precisely this kind of investigation. We don't want to the be the only ones using it! For example, we are hoping to have a forum post coming soon from a Faulkner scholar that tracks how a story's "narrative complexity" corresponds to higher (or lower) magazine circulation. In other words, did Faulkner stories in higher-circulation magazines have simpler and more linear narratives than those in magazines with lower-circulation (and, potentially, a more discriminating readership)?
@brookshefner This is such a great answer, because there are significant IP and copyright issues at the heart of any project tackling post-1922 (now 23) projects (at least in the US). It's even more complex thinking about transnational periodical-circulation histories. Folks getting into a new computational-analysis project sometimes have this in mind, but it's valuable to highlight the specific valences it has for each project (your answer about Conde Nast owning back catalogs is a particularly pointed example here). Thank you!
@kalanicraig The last time I tried replying, the system told me, "There's a problem with your data." So let's see how this goes. There are a handful of videos that mention community-made stuff. There are other places, though, that are better in suggesting a fuller portrait of the community's making in response to words and messages in multiple media.
@jbpierce It worked! If these examples are public and you're willing to share, I'd love to see an example.
@kalanicraig One way I'd think about making, and I believe this is the tip of the iceberg, is the perks that thank contributors to the annual charity fundraiser, the Project for Awesome. (P4A benefits entitles like Partners in Health and others.). There are numerous community members who make things, often using themes and works that connect the community as inspiration, to build the items that contributors can choose as thank you's. For a single example, my work for the last five years has been creating digital knitting patterns, which are listed on Ravelry to increase findability; the patterns emerge from videos made by the brothers, their books, and community memes: https://www.ravelry.com/designers/jennifer-burek-pierce