White Paper Indeed: A Tragicomic Prologue

Over the course of the next 2000 words, if you stick with me, you’ll learn why this final requirement—a white paper—is ironic given my topic. It seems that what we read as Americans in our moments pursuing hardcover fiction is white paper indeed.

Introduction

For my final project in Introduction to Data Visualization and Design, I decided to analyze the last decade of New York Times bestselling hardcover fiction. I was drawn to the topic for its promise to reveal something about our national reading habits in those ten years. Was the field wide open to a range of big and small presses? Is the nation reading diverse books in this decade that begins the year it elected our first black president? Does the authorship reflect who we are as citizens, mirroring our genders, races, and countries of origin?

About the Data

Though I learned a great deal about visualizations in this process, I learned even more about data. After all, the conclusions can’t be right if the data set isn’t. The initial set looked incredibly promising: the Kaggle download was not only pulled from the Times API itself, but it provided a link to the interface as well. Further, the Times lists are public and touted by the publishing houses and authors who land on them, so I knew that corroboration, should it be needed, would be possible.

But, the set turned out to be fraught with errors. The first level of problems seemed a result of the agnostic nature of digital parsing, where St. Martin’s with a curly apostrophe was considered a separate publisher from St. Martin’s with a straight apostrophe, or where Del Rey/ Ballantine with a space after the slash was distinct from Del Rey/Ballantine without one. It was clear the data had been pulled without being cleaned. Correcting for these errors reduced the number of distinct presses from 176 to 156—already an interesting find. The authors list needed the same cleaning, as similar errors appeared, such as Annie Proulx being listed separately from E. Annie Proulx, B A Paris getting split from BA Paris, and even truncations such as Nick Horn for Nick Hornby (perhaps the parser read “by” as a preposition after “Horn”).

Cleaning the data forced me to read the entire set publisher by publisher and author by author, which brought to the surface two more layers of error. First, the publishing houses were presented as distinct presses rather than as imprints or divisions of larger houses, giving the impression that more publishers contributed to the list than actually had. I might not have noticed myself, except that, for one year right out of college, I worked for an editor who ran an imprint of a division of Random House. I saw three of my four “publishers” on the list—Shaye Areheart, Crown, and Random House—all seemingly distinct when I knew they were not. Second, when authors wrote under pen names or had co-authors, they were treated as separate authors, again seeming to widen the field. For example, a visualization of authors and their books would yield many weeks for Clive Cussler, but would separate his many more for each of the 11 bestsellers he wrote in unique partnerships, even though his name is clearly the marketing draw for all.

I researched the parent companies for each of the typo-corrected publishing houses by corroborating Wikipedia with publisher pages. I also corroborated Wikipedia pages with author and publisher pages to determine the marketing draw, any pen names, and, while I was at it, gender, race, and nationality, where available.

Collecting data on race became particularly delicate, so I looked to an outside source to help me compartmentalize: the US Census Bureau. I stuck to their race and ethnicity categories—on their own, a bizarre subject worthy of greater study—though I diverged from their labeling of Middle Easterners as white. In a post-9/11 world, that inclusion seemed to deny those from the Middle East a reality they have been living in America for the last twenty years—something readers may actively seek out and the publishing world might thereby exploit.

Where I admit some interpretive license (and I note it in my visualization) is on the question of nationality. Several authors held dual citizenship or changed citizenship over the course of their lives. I tried to be clinical in these cases sticking to the authors’ nationalities at the time they made the list, but I may have been more inconsistent than I had hoped.

Given how long all of this research took, I found myself wondering how often I’ve been persuaded by data visualizations that may have been created in a hurry. I wonder how often parsing errors or a lack of awareness about a given industry from which data spring might mislead the analysts, viewers, or even policy makers. (And that has given me great fear of and respect for the field.)

About the Visualizations

To present the data, I mimicked the layout of the New York Times’ bestseller lists themselves: large, black page titles in Times New Roman; small, gray subtitles; and a menu of genres (in this case, data topics) in sans-serif caps, separated by thin gray and black borders. I used the far, right-hand spot reserved on the lists for the date to post the data set’s years, 2008 – 2018.

The first story I’ve told is that of the publishers. The “front page” graphic relays simply the cleaned data, depicting the number of total weeks each of the publishers had, and breaking down those weeks into the authors who contributed. Following the PUBLISHERS link reveals two more graphs, the first grouping the publishers by their parent companies, and the second visualizing the number of bestselling weeks those imprints gave to their parent companies.

In each of these visualizations, I’ve used horizontal, stacked bar graphs. In his Storytelling with Data, Cole Nussbaumer Knaflic extols the form for the simple reason that it is “extremely easy to read,” playing on our natural habit of “starting at top left and making z’s with our eyes across the screen or page” (57). I also thought these horizontal graphs particularly fitting as they mimic a race—showing front runners and leads—which, in the profit-seeking industry of publishing, the bestseller list certainly is. The stacking—first of authors and then of imprints—then helps point to why the front runners have the lead.

The “front page” graphic, depicting the number of total weeks each of the publishers had, breaks down those weeks into the authors who contributed. The tool tips provide greater detail, showing how many total weeks each author had. I purposefully chose to identify the authors with individual colors to heighten what this graph seems to say: that a whole host of publishers with a range of authors are included in these lists. Sure, two publishers have a significant lead, but the graph clearly shows that Little, Brown’s success is largely due to the week-hogging James Patterson, with his total of 507 weeks on the list over the decade, while Putnam’s success comes from a combination of a few big names and many smaller ones.

Following the PUBLISHERS link, purposefully put first in the menu, reveals a more accurate view. First, those publishers are presented as components of their parent companies, filtering the 156 presses into their mere 18 parent companies. The nature of the horizontal bar chart highlights the significantly increased lead Penguin Random House has, with an annotation that notes that it owns 50 of the imprints—the same number as its two closest competitors combined.

The final publishing graph combines the data from the first two, showing how many weeks each parent company has had over this decade as a result of its divisions. In this most accurate view, Penguin Random House, through their fifty imprints, outstrips all of its competitors combined, garnering more than half of the 10,400 possible slots.

In these two deeper dives, and in contrast to the opening chart, I chose to present all the data with a single color to mimic the uniformity of expectations and corporate practices of owned imprints. Rather than a seemingly diverse 156 independent publishers, I wanted a more monolithic feel.

The AUTHORS link from the “front page” displays four tree maps, all providing lenses through which to analyze the writers behind the bestsellers. This riff on a small multiples graph was inspired by Beatriz Wood’s interactive visualizations on presidential elections, where the bird’s-eye views of the graphs juxtaposed were as compelling as exploring them individually. The largest tree map by default portrays the proportion of the bestselling weeks each author holds as well as the proportion of those weeks given to each individual title. This multicolored view does not take into account any identity marker aside from name, making the field once more seem fairly equitable. Yes, James Patterson, David Baldacci, and John Grisham take up larger portions than others, but so do individual books, such as John Doerr’s All the Light We Cannot See and Kathryn Stockett’s The Help. In fact, the visualization looks like a multicolored quilt or a pied bookcase, with room for wide diversity.

After this exploration, my hope was that a viewer would then look at the smaller graphs to the right that break down this information by gender, race, and nationality (an order I chose based on the increasing number of variables in each). Even at a distance, these thumbnails paint  gender to be somewhat balanced and nationality to be (understandably) predominately American. But race, even from afar, is utterly dominated by a single determiner. A closer view of each of those maps (made possible simply by clicking on their titles) confirm that almost all of the hardcover bestsellers are penned by white writers.

A closer look at gender reveals that, were it not for James Patterson, the balance of NYT bestselling hardcover fiction would mimic the binary gender balance of the country. Since cleaning the data required investigating each author, I believe that all of these authors publicly identify within that binary categorization. (As Americans begin to understand a broader gender spectrum, it will be interesting to update the data in the coming decade. While one study published in 2017 in the American Journal of Public Health concludes that approximately .4% of the population was transgender as of 2016, I was unable to find similarly concrete statistics on non-binary populations beyond that. If the transgender population is similar to the nonbinary one, and if these lists mirrored the demographics of the US, we would hope to see at least 40 of the claimed bestselling weeks in a decade belonging to those outside of simple male/female identification.)

A closer look at nationality provides similarly expected results: Americans primarily read American authors. They tend, after that, to read authors who come from English speaking countries, such as England, Australia, and Canada. In this graph, it is the outliers that are most interesting, such as the sizable swath that Swedish author Stieg Larsson gets (and posthumously at that) or those weeks given to Japan’s Haruki Murakami.

The real shocker, though, is the view by race. Nearly all of the authors are white. Were it not for Celeste Ng’s Little Fires Everywhere and Colston Whitehead’s Underground Railroad, the nonwhite slice of this graph would be virtually invisible. For all of the tree maps, I included author, title, and weeks on the list in the tool tips, but for this map, I also added the publisher for a simple reason: it showed at least one upside of conglomeration—the power to publish and promote authors of color—as all but two nonwhite authors belong to giant parent companies.

Returning to the “front page,” I offer a FINAL THOUGHT, which uses a waffle grid to show the disproportion of black male authors to the national percentage of black men. This view was inspired first by Georgia Lupi’s belief that data can tell a human story. (It was also inspired by my recent obsession with waffle grids, and my professor’s helpful nudge to investigate this percentage in particular.) While the graphs by race and gender certainly suggest the reality of this final thought, they don’t relay the isolation that NYT bestselling black male authors must feel, especially in the post-Obama era. I started on a 10 x 10 grid, only to realize that, since black, male authors held less than 1% of the bestselling weeks, they would not appear within those dimensions. But, a 15 x 15 grid would reveal that .4%. I almost just presented the numbers. After all, Knaflic does say, “When you have just a number or two to share, simple text can be a great way to communicate.” (38). But, the waffle grid turns those numbers into people—simplified into icons though they may be—and that human touch here is essential.

The very last decision I made was to go through all of the visualizations with Edward Tufte in mind. He extols the virtue of Spartanism when it comes to visualizations, eager to de-clutter graphs to their minimum. What, I asked myself, could be removed from my graphs that were either already there or implied? I pared down my axes, thinned out my tool tips, and removed a few labels.

Next Steps

As interesting data always do, these visualizations raise big questions. Are these results unique to hardcover fiction? I’d love to look at paperback fiction as well as hardcover and paperback nonfiction. I’d also like to dig into audio sales. Perhaps the heavily white, mostly Penguin Random House books appeal to a certain group of readers who prefer new or hardbound novels. After all, hardbacks are more expensive and harder to carry. So, perhaps the publishers and authorship of paperback bestsellers—of books that are cheaper or easier to carry for an urban commuter like me—might better represent the country’s demographics. Or, perhaps the effects of influencers of color, such as Oprah Winfrey, might be more visible in the paperback list, as they often promote already published authors. And, of course, with the New York Times lists, there is a chicken-and-egg issue of whether sales drive the lists or the lists drive sales. So, perhaps a look into library loans, book reviews, or prize winners would be equally revealing.