In the case of literature, random sampling is arguably far less valid than selected sampling of, say, top authors.
The interactions here are complex. There are a finite set of published authors, they're selected by editors, works are bought by the public. So there's both nonrandom and popular selection. Works themselves typically contain many words either from the popular argot or not (e.g., Joyce).
And if you're going to sample from large sets of works, say, newspaper and magazine articles, you'll run into issues such as style guides which often dictate terms that must (or must not) be used.
It's a difficult field to research.
But I suppose we could ask the mods to change the submission title. How about this?—"One blogger's analysis of the size of the vocabularies employed by some of his favorite French- and English-language writers, neither claiming nor achieving statistically significant insights into the question of whether French- (on the one hand) or English-speaking peoples (on the other) use more words in common usage."
"Do the English Use More Words Than the French? I don't know, and what I did to test it doesn't tell me anything concrete"
First, the authors and books were not chosen randomly and the sample size is tiny, so the results are meaningless, although I do respect the effort that was put into this. The author is up-front and honest about this in the beginning, but unfortunately proceeds to draw unwarranted generalizations based on his study.
Second, even if a similar, but much larger, study were conducted on randomly selected authors and books, the results would still conflate the vocabulary size in common usage with the vocabulary size of single authors. Hence, what the article is attempting to study is not the same thing as the title of this post.