Without a breakdown by country of authorship, the study is fundamentally flawed.
The US wasn't honoring foreign copyright at all until 1891 and there wasn't practicable recognition until the 1909 Act. Thus, there was a precipitous fall in the number of public domain books available after 1910. Before that point, every new foreign works would fall into the US public domain immediately.
With the change in the law, the notion of "classic" literature also changed. Publishing houses made up "classics" so that they could continue cashing in on relatively new international works that fell into the public domain prior to the new Act. Many of the works that we still think of as being canonical just happen to fall into that category. This marketing ploy has left an indelible mark on American readership. We still read--and sell--quite a lot of Dickens.
In other words, the paper includes within its statistics a large number of works that were effectively never subject to copyright protection, to the extent that their proprietors never received any compensation whatsoever. They had the market advantage of being cheaper than their domestic contemporaries and subsequent new works. Their production never waned; a substitute wouldn't be marketable. They inhibited the production of new work. The comparison of these to a copyrighted work is apples to oranges. The simple solution is to exclude these outliers entirely and look only at US authored works.
The basic premise for the paper is great; there ought to be more empirical study of these issues. The execution could use some work though. Ultimately, it's best not to cite to "Everything You Wanted to Know About Pre-1909 Copyright But Were Too Lazy to Look Up." It hurts one's credibility.