By Marie Vibbert.
This article first appeared in Vector #294.
I was at a massive mixer for members of the Science Fiction Writers of America, a group I had just joined, wondering how I could even talk with these big, important people. The question everyone asked when you walked up to them was, “What type of science fiction do you write?” After mumbling some self-deprecating responses like “bad” or “oh you know like … the kind with robots and spaceships?” I tried to express what made my work different. “I write working-class science fiction,” I told the next gentleman. “Stories with waitresses and janitors in space, you know? I feel like there’s too many stories about rich guys without real problems.”
I picked the wrong man to try this tactic on. He laughed condescendingly and said, “The opposite is true. Everything is about some worker everyman. There aren’t enough stories about rich characters!”
My first thought was, Ooookay time to start never talking to this dude ever again, but my second thought was a worried, Is he right? I had this gut feeling that a lot of the science fiction I had read didn’t represent my social class, but was I just biased?1
The only answer was, of course, to collect some statistics! This paper is the culmination of my efforts to answer the question for myself, “Is there a class bias in main characters in science fiction, and if so, are poor or wealthy characters more predominant?”
Choosing the Books
The first question I had to answer was, “How do I take a sample set of science fiction?” I limited myself to novels, because novels or their detailed discussions were easy to find, and that way I’d be comparing apples to apples.
Reading every science fiction novel ever would not be feasible, especially with a staff of just me. I searched for recommended reading lists, but which to choose? Many were simply “The Best of 2019” or such. While it would be interesting to look at a specific period of SF, I wanted a cross-section of what an average reader might have in mind, and that meant including recent books as well as old classics. I googled “Top Science Fiction Novels” in an incognito browser tab (so as not to bias the results with my search history) and took the first 50 novels the search returned. I liked that list better: it felt eclectic, and included recent novels as well as Mary Shelly’s Frankenstein. Of course, the Google search results, while incognito, still would be skewed toward my location in the Midwest United States.
The British Science Fiction Association’s magazine, Vector, announced a call for papers on class and science fiction. I could hardly contain my excitement (and imposter syndrome) as I typed and re-typed my email asking if this statistical analysis was the sort of thing that maybe they’d want to see? And so, my next data set was BSFA award winners. These would skew British to balance my American bias. How better to kiss up to the editors? I started my spreadsheet!
BSFA award winners include fantasy novels with no science fictional elements, however, maintaining genre purity would open up a can of worms (how to draw the lines? Who gets to say what is or isn’t SF?). I would keep the results of each list separate, to see if there was any bias.
On accepting the paper proposal, editor Polina Levontin suggested adding the titles from the Orion SF Masterworks book series, a somewhat curated list, limited only by what titles Orion had the rights to. So now I had three piles of representative works: award winners, a hodgepodge recommended by Google, and a curated list for a total of 194 separate titles. It seemed as close as I was going to get to a reasonable sampling of notable science fiction novels.
Choosing the Characters
The next question, of course, was, “Which characters?” I didn’t want to look at antagonists or incidental characters. I focused specifically on protagonists, or “the main character.”
However, very soon I came across River of Gods by Ian McDonald, which was on both the BSFA list and SF Masterworks. A novel with ten point of view characters. Did I list all ten? Would this weigh the results unfairly toward Ian McDonald’s personal viewpoint?
Hyperion by Dan Simmons had the same problem. I re-read it, and read River of Gods for the first time, and I decided, arbitrarily, that I would limit myself to four main characters, maximum, per book. Some books would have one clear main character, but if there was an ensemble cast, I’d list only the top four mentioned in reviews of the book, or read it myself and decide who the four main-est characters were.
Books with high concept plots, like Vernor Vinge’s Fire Upon the Deep, rarely mentioned characters at all in their online summaries or reviews. (I learned to dread the words “Great book but not very good characters” as a precursor to having to struggle to find a copy of the book quickly so I could find out who the characters were). I had to dig deeper, reading review after review, asking friends who had read the book, and then reading the book itself as quickly as I could, skimming to see proper names.
In the end, I was able to find at least one main character for all the books, for a list of 244 separate characters across 194 novels, excepting only the handful of anthologies and short story collections on the SF Masterworks list, which I skipped, rather than have to decide whose story got to be counted and whose didn’t.
Defining Character Class
I looked each novel up on Wikipedia, where very often the Plot Summary section would begin “Joan is a physicist” or “Jane is a coal miner” or what-have-you. Clearly, jobs are important in how we define character, because the vast majority of books could provide a main character and their job through this simple method.
So, now I had a list of books, and the main characters in those books, and their jobs, if they had one. Surely jobs were tied to class, so all I had to do was list the jobs and we’d see who was working class, right?
I had 140 unique job titles across 244 characters. A pie chart with 140 wedges didn’t show me anything interesting. I realized I had to break the jobs into larger groups, and that brought me back to the initial question: What class does each character belong to?
I did a lot of searching for hard definitions of social class and found more than anything that class is complicated. A steelworker in a strong union in the 1960s could be making more than a high school teacher, yet one is seen as a member of the proletariat and the other a member of the intelligentsia. Jobs can start out highly respected, like a computer helpdesk technician in 1970, and become low-wage, low-rated work, like a computer helpdesk technician in 2010. Also, the same job pays differently, and the same wage has different purchasing powers in different countries.
Class is about more than occupation and paycheck. A low-wage worker – maybe they’re a barista at a bohemian cafe, a farm laborer at an artist’s colony, a runner on a movie set – could live rent-free in the family pied-à-terre, with a fat allowance. Part of class is in what the work is for: does it provide for the needs or wants of the worker and their dependents? How much of the profit does the worker take home, and how much is siphoned off?
Unfortunately, most novels didn’t go into the minutiae. Nonetheless, I wouldn’t rely just on characters’ jobs: sometimes I would make judgment calls based on what I could tell about the character and their family background, either from reading the novel or summaries of it.
Based on summaries I found online, I came up with these five categories2:
|1. Poor||Unable to meet basic needs. Unemployed and ‘working poor’ who work at wages below the poverty line.|
|2. Working class||Able to meet basic needs but not comfortably. May include craft workers, laborers in factories, restaurant workers, nursing home staff, workers in repair shops and garages, delivery services, military troops of low rank.|
|3. Middle class||Comfortably able to meet basic needs and some wants. May be clerical – administrative, may provide support for professionals, engage in data collection, record-keeping, paralegals, bank tellers, sales, teachers, military specialists and police officers of middle rank.|
|4. Upper middle class||Comfortably able to meet all their basic needs and many wants. May include professionals, engineers, accountants, lawyers, architects, university faculty, managers and directors, ship captains, military leaders, police chiefs, local politicians.|
|5. Wealthy||Economic elite whose wealth derives from what they own – real estate, businesses, stocks and bonds, other assets. May include heirs, top-level executives, celebrities.|
I went through my list of 140 character jobs and I sorted them into these five classes, making some judgment calls. For example, the main character of Mythago Wood by Robert Holdstock is listed in most summaries as “a military vet” – but that’s not an occupation per se. On reading further, I found that he had the wherewithal to bum around France for a year without employment before returning home, to a house with four chimneys and a name, so I put him in the upper middle class and considered that a safe bet.
Jobs could also have a different status given the society or time of the novel. For example, I had “ship captains” listed as an example of an upper-middle-class job, but Tabitha in Colin Greenland’s Take Back Plenty reads more like an owner/operator of a long-haul trucking rig in her world where space cargo ships are common, whereas an 18th century ship captain almost certainly would have been wealthy.
Characters could change class over the course of a novel. At first, I listed people as whatever they were introduced as, however this didn’t always feel right. In River of Gods, Shiv is introduced in his own luxury vehicle, with a driver, a gangster having just dumped a body. He seems wealthy and I would have marked him as such, however soon he’s lost everything, we meet his impoverished family, and he takes on illicit jobs without much choice, so I marked him as working-class. (Interestingly the only one of the four main characters that I marked below middle class, and the only one who serves partially as an antagonist to the others.) In the same novel Vishram is a stand-up comedian, and listed that way in most summaries. My first instinct is to mark him poor, because who makes any money at that? But reading the novel, he turns out to be from a fabulously wealthy family and quickly becomes a top executive at his father’s energy company, so he was marked wealthy.
My marker would be whatever class the character spends the majority of the novel in, including flashbacks.
Other Statistics Tracked
While looking up their jobs and deciding on their class, I marked if the characters were male, female, or nonbinary, based on pronoun usage.3 I also marked the year the book came out, or, in the case of the BSFA list, the year it was awarded.4 I also marked if the authors were male or female. (No authors in this data set were explicitly nonbinary.)
Then, scariest of all, I tried to find out if the author had a non-writing job, and if so, what it was. Based on this I tried to class the authors, knowing it would mean a lot of guesswork, but hoping it might reveal some broad patterns. This was much harder than doing the same for fictional characters.
Most authors either did not mention their day job in their bio, or had held a variety of jobs over the course of their lives. I tried to err toward whatever job they held while writing the novel in question, or if that wasn’t clear, for the largest part of their lives, but even that wasn’t always clear.
Authors may enjoy mentioning jobs they have worked in, but they seldom reveal for how long, or whether that job was their main source of income. While I made sure to provide a job and a class for every fictional character, I allowed myself to not do so for authors, if several searches didn’t reveal enough evidence.
Finally, I included a notes column in my data, to provide details where difficult decisions were made and could, perhaps, be revised. I manually compared all duplicates and cross-checked the BSFA, SF Masterworks, and Google lists against a Master List. Then, at last, it was time to make some charts, and see what the data told.
The top twenty most common main character jobs I found were:
|No visible means of support||6|
Undifferentiated scientist was the top job even before I made every “physicist”, “geologist”, and “biochemist” also a “scientist.” (I kept research assistants separate, these were people who might be primary investigators). I felt dirty leaving the social sciences out of it (“psychologist” and “anthropologist” still make the top list) but felt that I couldn’t make “scientist” too broad a bucket. As it is, there’s no surprise a genre called “science” fiction would have a built-in bias toward scientists.
The top five jobs, other than “juvenile” are all upper-middle or wealthy class. I was not surprised to see a good turnout by “no visible means of support.” I actually expected more of that, having often complained about books where the characters drop everything to chase the plot with no concern for how they’re going to eat or pay the rent. “No Visible Means of Support” is frequently listed as wealthy, because the character does expensive things (usually travel) without concern for cost. Only one “No visible means of support” was marked middle class because, well, the character felt middle-class to me, and while he doesn’t think about money, he also doesn’t spend any. This was the protagonist of The Shrinking Man by Richard Matheson.
Almost as soon as I started recording data, I was aware of the prominence of professionals, and when I graphed the count of characters in each of the five classes, this proved correct. The largest single class represented is the upper middle class, where professionals sit, with 111 individual characters counted there. Only 13 characters were identified as poor. Double that – 26 – were working class, and over double that – 58 – middle class. Wealthy characters accounted for 36, putting them between working class and middle, in terms of representation.
I wanted to compare this against the actual class breakdown in the United States. To find hard numbers, I went with a Pew report on US income in 2016.5 The following chart shows the Pew results next to the character class breakdown in this study sample. Because Pew only used three classes, I’ve combined Poor and Working Class into “lower” and Upper Middle and Wealthy into “upper”. The difference shows that upper class characters are strongly over-represented in the study as compared to the US population, while lower class characters are strongly under-represented, Figure 1.
The next thing I wondered was, does this change over time? I had recently looked at a list of labor unions in science fiction on “Hugo Book Club Blog” (hugoclub.blogspot.com/2018/12/organized-labour-in-science-fiction.html) – for each example of a labor union in science fiction, they listed the date and whether the depiction was positive, negative, or mixed. Scanning the list, it seemed the first half of the list was heavily negative and the second half heavily positive. To check if I was really seeing that, I compared all the entries from before 1990 to the entries from post-1990 and I found that overwhelmingly the pre-1990 depictions of unions were negative, with only 4 positive depictions and 14 negative, and post-1990 depictions were positive – with 34 positive and only 7 negative. Both groups had around the same percentage of “mixed” depictions.6
That list included movies and TV show episodes, which creates a different genre picture than just novels, and there were more entries post-1990 than pre-1990, but still, if I could find such a shift in opinion in one list of science fiction works, why wouldn’t there be change between newer and older books as regards to character social class? I mapped the five social classes over time of publication (or award for BSFA titles.)
This … was not a very useful line chart. It looked like spaghetti. All the classes went up and down chaotically, though upper-middle class had a strong peak in the early 70s. I put in linear trend lines to clear things up, but all five classes trend slightly up over time – that implies the number of characters is going up over time, not anything about the relative frequency of class.
Could there be a bias toward what time period these books came from? I looked at the count of book titles (omitting duplicates) by year, and sure enough, the early 70s do have more titles, though in general the years after the 70s are better represented than those before. The BSFA list starts in 1969.
Separating out the BSFA list and the Google List, however, it’s only the SF Masterworks list that has such a heavy bias toward the early 70s (which raises the hypothesis that this is driven by the height of Orion publishing’s title-purchasing power.) The BSFA list by definition has one title per year, and the Google List was surprisingly evenly spread over the timeline. Perhaps if I had listed as many Google results as the SF Masterworks list, the date trend would even out?
Well, there wasn’t anything I could do about my sample set at this point, but I could make the data more visual. I converted the chart into percent of characters that year in each class instead of count. When that still looked like a mess of color, I concatenated into three classes and that became easier to read.
The “upper” class visibly dominates throughout time; however, the middle class makes strong inroads after 1995 and though the trend breaks after 2015, it could indicate a general increase in class diversity more recently. Still, it doesn’t look like an immediately comprehensible trend over time.
So I thought I’d break up the results by decade and see if there were any shifts visible that way. And there were a few.
The 1950s offer more working-class characters than the decades immediately before or after them. The 1960s are particularly abysmal for class representation, Figure 2. The 1970s and 80s are a little better. The 1990s are the most middle-class decade, and the 2000s and 2010s keep those middle-class gains, more or less, and add lower-class gains, though they don’t recapture the level of working-class representation in the 1950s.
It’s not hard to speculate about historical events impacting the preoccupations of science fiction writers each decade. Post-war progressive policies might drive the 1950s figure and then fall to consumerism in the 1960s. (The late 1930s, with the Great Depression, are a spike of poor and middle-class characters in the wealth-dominated pre-1950 world, though that is lost in Figure 2 because the early, pre-1950s data are aggregated.)
Could the sources of my data be skewing things? I hypothesized that the Google list would be more “casual” and therefore more working-class than the BSFA award-winners, which might skew toward the preoccupations of intellectuals. So, I looked at the class representation in each of my three data sources, I visualised the data as class percentages in each novel list so that the relative size of the sets wouldn’t skew the view, Figure 3.
What I found was that indeed, the BSFA had the largest percentage of both wealthy and middle-class protagonists. The Google list was the leader in working-class characters, but also the highest in upper-middle class. The Science Fiction Masters list had the highest percentage of poor characters (7%). However, all three lists followed the general trend, most characters in Upper Middle Class, least in Poor, Wealthy more common than Working Class, and less than Middle Class.
Do the results change with character gender? I found there were no poor female main characters, at all. Nonbinary/neutral/nongendered characters were exclusively in the middle or upper-middle class, and were the only gender to be mostly middle class instead of upper-middle. Though this is a tiny sample size (only 4 characters), and so this is only fanciful conjecture.
Since there are so many more male characters (out of the total 208 characters 161 were male) the changes in the smaller numbers of female (total of 43 characters) and nonbinary characters (4) are hard to see when looking at raw count, so I did the graph by percentage (so percentage of female characters in each class out of all female characters, etc.)
Other than the nonbinary characters bucking the trend, the male and female characters followed the same class pattern – most in upper-middle, then middle, then wealthy, then working, then poor. Women are slightly more likely to be depicted as wealthy than men (19% of female characters vs. 17% of male characters), and also slightly more likely to be depicted as working-class (14% to 12%), though men are much more likely to be depicted as poor, as, again, there were no poor female characters (8% of men vs. 0% of women). It seems unlikely that the gender of the character affects class representation, the differences are probably not statistically significant.
Does the class of the characters depend on the gender of the author? Probably not, Figure 4 shows that the distributions by class are very similar for female and male authors.
However, the gender of the author has an undeniable impact on the gender of the character. Female authors appear to only marginally favour female protagonists, splitting themselves pretty evenly between male and female characters and also writing nonbinary characters, while male authors are far more likely to showcase male characters – only a small proportion of the protagonists written by male authors are women (Figure 5).
While this isn’t a paper about gender itself, since I had the data, I went ahead and charted out author gender over time and character gender over time. As expected, more female authors (and female characters since it is women that tend to write them) appear more recently, though the trendline is higher in characters than in authors. Overall, gender representation is equally bad among characters as among authors – only about 20% are female. There is a veritable mountain of masculinity in the early 1970s. There are more characters than authors, but years without a female character after 1980 are rare, while years without a female author after 1980 are a more likely occurrence. Also there is a peak of female authorship in the 1990s, and a dearth of female authors in the 2000s, which echoes my findings in my paper tracking female-seeming author credits in top SF magazines.7
Interesting that the frequency of female characters somewhat tracks with the same trend, with 2006-2008 being a rare spread without female main characters and only two years in the 1990s without a female main character. It is interesting to note that only one non-binary character appeared before 2009, in 1975. The representation of female authors is a lot better on the SF Masterworks list (about 25%) than on the Google or the BSFA lists (where only 15% of the authors are women). Out of 122 unique authors only 26 (21%) are women. What about author class? As mentioned before, I couldn’t find information on all authors, so there are a number of blanks, more for women than for men. Still, it looks like regardless of gender, authors are more likely to be in the upper-middle class, just like their characters.
That one poor author is a tentative classification: Cormac McCarthy, who lived a relatively impoverished life in Tennessee in the 1960s and 1970s, although he received a major fellowship award in the early 1980s, and started seeing commercial success in the 1990s. His main character was also classed poor – a vagabond.
At first glance, I thought there was evidence for most authors writing their own social class, or at least not far from it. However, comparing individual authors to their own characters there seems to be a regression to the mean, with some poorer authors imagining richer characters and richer authors imagining poorer. More research is needed as there are very few authors in the sample who are either extremely rich or poor! What does it say about how we relate to the class of the characters we write as authors, compared to race or gender?
When I grouped the authors into three classes and compared them against the Pew survey of American households by income, I found that authors were even more skewed wealthy than characters were, so this is likely affecting their view toward characters, Figure 6.
Conclusions and Further Questions
Using these methods, there are demonstrably more upper-class and wealthy characters than middle class and below in notable works of science fiction. While the three sources of book titles showed similar trends, the BSFA and SF Masterworks lists were perhaps more elitist than titles produced from a Google search for “Top science fiction novels.” In this sample, male authors were much more likely to write male characters than other genders. While character gender seemed significantly impacted by author gender, character social class does not appear to be that affected by author gender, though female authors were less likely to write the wealthiest characters than male. Male authors are overrepresented in the sample, as are male characters, and upper class characters are overrepresented while working classes are very under-represented.
Popularity of class representation seems to change over time. Also, looking at the class of authors, as much as could be determined from online biographies, found that lower class writers are also severely under-represented. These groups of novels may represent a general sense of science fiction as a genre, because award-winning and popular titles dominate our perceptions over lesser-known titles, but this data set could still be a biased one. Much of what we consider when we say “science fiction in general” isn’t novels. What if I could include movies? TV shows? I would have loved to have done short fiction, but there is so much of it that creating a reasonably representative sample set was too daunting for the scope of this project.
While the three lists showing the same general trends is a hopeful clue that the trends are informative beyond the sample set, these three lists could just be equally biased! What about bestsellers? Hugos? Nebulas?
What could we learn by including an analysis of race?
Given how prominent scientists are in science fiction, how would omitting them change the pattern of class or gender (most scientists in SF are male) representation? Just how much of that upper middle class bias is made up of Principal Investigators?
I would have liked fewer arbitrary choices by me in the data, especially in class attribution. Perhaps a crowd-sourced site could be set up.
Or I could rest on my laurels, having worked up enough numbers to prove some guy in a cocktail party wrong. It isn’t the worker everyman who dominates science fiction, it can safely be said if any archetype does, it is a male scientist at the top of his profession.
If anyone would like to play with the data and come to their own conclusions, the full data set is available on the Open Science Framework: osf.io/ptqys/
1. It is odd to footnote one’s own class identity, but I feel it is necessary to point out that I am now comfortably middle class, a computer programmer by day, but much of my self-identity comes from growing up below the poverty line, raised by a single father who was a union laborer.
2. This is basically similar to how the sociologists W. Thompson and J. Hickey divide it up in their book Society in Focus (2005).
3. One exception being the case of the AI main character of Ancillary Justice; I felt the ubiquitous she/her pronoun usage in the book rendered the main character neutral, not female.
4. This was an oversight caused by copying the BSFA award list directly into my spreadsheet, but novels are awarded the year after they come out, so it’s close enough for this level of general inquiry.
5. From Pew Research “The American middle class is stable in size, but losing ground financially to upper-income families” by RAKESH KOCHHAR, September 6, 2018 pewresearch.org/fact-tank/2018/09/06/the-american-middle-class-is-stable-in-size-but-losing-ground-f inancially-to-upper-income-families/
6. The list was heavily weighted toward titles post-1980, since it focused on television and movies. The only 1930s entry was positive. Obviously, I suspect the virulently anti-union 1980s defined by Ronald Reagan’s policies caused all those negative portrayals in the 80s, rampant corruption in the 1970s lead to mostly mixed and negative portrayals in that decade, and I would hypothesize that the positive portrayals after 1990 shows authors yearning for something that has been destroyed.
7. “The Women We Can See in Analog” tracked female-seeming author credits across six magazines from 1926 to 2010 and appeared in Analog Science Fiction and Fact in November 2020. I found the 1990s, particularly the early 90s, to be a time of great female representation in author credits, with a sudden and alarming drop off shorting after 2000, perhaps due to increased competition for more limited page space as online magazines challenged print.
MARIE VIBBERT HAS SOLD OVER 70 SHORT STORIES TO PROFESSIONAL MARKETS, OFTEN FOCUSING ON WORKING CLASS CHARACTERS AND SITUATIONS, REFLECTING HER BACKGROUND. HER PAPER ON THE FREQUENCY OF FEMALE NAMES IN TOP SF MAGAZINES OVER THE 90-YEAR HISTORY OF ASTOUNDING/ ANALOG WAS WELL-RECEIVED AT THE CONFERENCE COVERING THAT HISTORY IN NEW YORK IN 2019, AND WOULD LATER APPEAR IN ANALOG. BY DAY SHE’S A COMPUTER PROGRAMMER IN CLEVELAND, OHIO.