Strange Horizons's Valentin D. Ivanov has scraped Locus Magazine's "Notable Books" column going back to May 1998 and built a 10+ year dataset of genre popularity in science fiction, fantasy and horror. It's easy to get all impressionistic and say, "Oh, everything in the sf section is space opera these days," but that's as apt to be confirmation bias as fact. Here's the numbers.
How significant are these trends? Having only three measurements, we cannot provide rigorous answers, except for the major categories that are populated with sufficient numbers of books. A linear fit to the points in Figure 1 gives us the rate of increase of the number of books included in Locus Online reviews, averaged over the entire time period. For example, for SF it is 11.5 ± 0.9 books per year. In other words, the number of the reviewed SF books has increased on average by 11-12 every year between 1998 and 2007. The rate for fantasy is 26.9 ± 7.7; for horror 3.1 ± 0.7; and for other books 4.1 ± 1.0 (all in units of books per year). The uncertainty margins are the formal fitting errors. The larger uncertainty in fantasy's growth rate reflects a systematic error due to the fact that this category has inflated only recently, and the linear model is not an adequate representation of its behaviour. Fantasy's lead over SF in terms of growth rate is a margin of about 15 books per year. The probability of this happening by chance is extremely small–about 1-in-1010. Therefore, we are likely facing a statistically significant nonrandom trend here.
It appears that the invasion of the sequels is truly happening. However, this result is not as obvious as the previous one–Figure 3 suggests that the proportion of sequels included in Locus Online reviews remains nearly constant after the 2001-2002 period. In other words, the changes are well within the expected random variations, shown in the plot with error bars.
(via Making Light)