It's not mentioned in the article, but this could also be a sampling issue. Samples from previous decades might've been more limited or self-selected. The internet may mean that a broader, more representative sample of the population is being tested, and you're just seeing regression to the mean.