As promised, the raw data set is available to anyone to use as they like. (I have deleted some identifying information that SurveyGizmo collected without my knowledge. I haven't made any use of that information myself, nor will I.) I hope you'll post and share your findings with the rest of us. Here is the data set in CSV format:
Here is the text of the original survey, in RTF format:
If you're not able to get the documents from the site, e-mail me privately on Live Journal, Facebook or Twitter, send me your e-mail address, and I'll send you the files directly.
Why this survey exists.
At the time, I didn't tell anyone what I was trying to find out, in order to avoid biasing the data. Here's what prompted it: In 2007, Stephen King published an essay in the New York Times Sunday Book Review called "What Ails the Short Story?" In that essay, King describes the declining prominence of short-fiction magazines in bookstores and the apparent reduction in short-fiction readership in general. At one point in the essay, he says:
What’s not so good is that writers write for whatever audience is left. In too many cases, that audience happens to consist of other writers and would-be writers who are reading the various literary magazines (and The New Yorker, of course, the holy grail of the young fiction writer) not to be entertained but to get an idea of what sells there.
This assertion -- that the readers of short fiction are increasingly writers themselves -- irritated me. I'm a short-fiction writer myself, and (much as I bask in the attention of other writers) I disliked the idea that I was writing for writers and no one else. I've wanted to disprove it.
That was the primary goal of this survey. I hoped to show that, at least in the case of science fiction, fantasy, and horror, the reading habits for short fiction were no more different for writers vs. non-writers than they were for novels. (A lot more data than that was collected, however, and presumably some other interesting results can be extracted from it.)
The results were rather the opposite of what I was hoping.
Methods and weaknesses.
After showing my draft survey to several people with survey-design experience (thanks to Dan Baxter, Tony Fruzzetti and Sherry Baisden) and revising accordingly, I posted it on SurveyGizmo on a 30-day free trial membership. I then announced the survey on Facebook, Twitter, LiveJournal and via e-mail, asking my friends to spread the word. I also asked the editors of various online markets to post news about the survey. Response was highest during the first few days, gradually declining after a week or two to a few per day. By the time the 30-day trial was over, 723 people had responded.
These methods point up some obvious flaws. First, the respondents are entirely self-selected; nobody answered unless she felt like it. Second, this is self-reported data: it reflects not what people are actually doing, but what they think they are doing. Third, the only people who had the opportunity of responding were those who got news of the survey through the communication lines I had available. Since it was spread by friends of mine to their friends, the demographics are markedly skewed.
An example: 221 respondents (29%) reported that they had published fiction during the previous two years. Obviously the population of published writers makes up nowhere near that fraction of the reading public, much less the general public, but many of my friends are writers, and many of their friends are writers, so the bias in the data isn't surprising. Such a large number of writers, though, did allow me to get some statistically significant data on writers as a group, which was part of the survey's purpose.
As another example, 405 respondents (56%) self-identified as female, while 283 (39%) self-identified as male, and the remaining 35 (5%) either declined to answer or identified one of several other gender categories, the largest being genderqueer, with 13 respondents (2%). If this were a random survey of the general population, we'd expect the numbers of male and female respondents to be closer to 1:1, rather than 5:4. (It could be, of course, that the readership of SFFH is more heavily female then the general population, but a simpler explanation is that I know more women than men, and that they also know more women than men.)
Similarly, 581 (80%) self-identified as white, higher than the general population, while only 22 (3%) self-identified as hispanic or latino, 18 (2.5%) as Asian, and 13 (2%) as black or African, all lower than in the general population. (36 (5%) declined to self-identify.) Again, while this could reflect different demographics in the readership of SFFH, my instinct is that it reflects a lack of diversity in my own extended acquaintance.
All this is by way of saying that the data should be taken with some skepticism.
There are also some flaws in the design of the survey itself. For example, when asking about recent publication history, I asked about short fiction published in magazines, online journals, audio markets, and self-published, but forgot to ask about collections or anthologies (many respondents wrote that response in under "other"). I'm sure there were other, similar failures.
I do think the survey data are still meaningful, but it's always good to have an error analysis.
Basic reading habits
Overall, respondents did report differences in their novel-reading and short-story-reading habits:
Respondents as a whole read novels more frequently than they read short fiction, with 523 (75%) reporting that they read novels daily or a few times a week, while only 226 (34%) reading short fiction that often.
When it comes to changes in reading habits over the last five years, though, the results are the opposite:
300 respondents (43%) said that their short-story reading had become more frequent over the last five years, while 163 (24%) said it had become less frequent. The novel-reading results were the opposite: only 178 (19%) reported increasing frequency of novel reading in the last five years, while 458 (49%) reported a decrease. This result, at least, suggests that the trend observed by King seven years ago may be reversing over time.
Reading habits of writers vs. nonwriters
There were three distinct groups: writers who had published fiction within the last two years, self-identified writers who had not published during that time, and self-identified nonwriters.
Their novel reading habits are remarkably similar:
Published writers, unpublished writers and nonwriters all read novels very frequently, and in about the same proportions.
The results are different in the case of short stories:
It is clearly the published writers who report the most frequent short-fiction reading, while it is the nonwriters who report the least. Looked at separately:
Nearly half of published writers read short fiction daily or a few times a week, whereas only a quarter of nonwriters do so. By contrast, only a fifth of published writers reported reading short fiction monthly or less, while nearly half of nonwriters reported doing so. This bears out King's hypothesis that it is writers, more often than nonwriters, who read short fiction, something that cannot be said of novels.
Novel writers vs. short-fiction writers
The results become even more striking when we separate those who have recently published novels from those who have published short fiction:
Whereas novel-reading habits are nearly the same for writers of novels, short stories, and both, it is the short fiction writers who read short fiction frequently, and the novel writers who read them infrequently. If we isolate just the novel writers and short story writers, the results are stunning:
56% of short-fiction writers read short fiction daily or a few times a week, whereas only 10% of novel writers do so. More striking, if you scroll back up, is that novel writers read short fiction even less frequently than nonwriters do.
Thus, King's suggestion that it is not only writers, but short-story writers in particular, who read short fiction with any frequency, would seem to be supported by this data.
Some other sources of the results?
This is as far as I want to go today, except that I've looked for other demographic bias that might have affected these results. I couldn't find any discernable difference in reading habits by gender identity. Older respondents were slightly more likely to read short fiction frequently than younger ones.
I hope to post other results later. However, as I say, the data's now freely available to anyone, and I enourage others to post their own analyses. Let me know what you find!