There are a number of claims that this happens in the academic market as well.
Sources:
1. Peter Suber's Open Access monograph, chapter 5, p. 107-110 (itself open access, of course). https://cyber.harvard.edu/hoap/Open_Access_(the_book) argues that OA books facilitate "searching and sampling," and then people want to read the whole thing in the nicer print format.
Hypothesis: Books that already have high exposure and demand will likely be negatively impacted by making a free version available. Books that would have good sales if more people knew about them will likely benefit from the increased exposure from making free versions available.
But yeah I personally have had the experience of the first link - I've purchased many physical books that have free (and legal) PDFs online because I like print books better.
It would make sense to me, at least with textbooks, that it also helps the cream rise to the top (outside of mandated ones for college classes). You can sample several, find the one you like the most, then buy a nice copy of it to keep. I've done that with several books, checking them out from a uni library or finding them by other means and then buying once I decided it was good and I was going to stick with it. I feel like others do as well, others who would never buy them otherwise.
Also, if people can sample textbooks, you can easily start to see which ones are the best (subjective, in a way) and they can rise to the top.
The Elements of Statistical Learning and Introduction to Statistical Learning are THE textbooks for an introduction to statistical methods of data science. They are free and very high quality. Most of my class didn't buy it, but many including myself did.
Sources:
1. Peter Suber's Open Access monograph, chapter 5, p. 107-110 (itself open access, of course). https://cyber.harvard.edu/hoap/Open_Access_(the_book) argues that OA books facilitate "searching and sampling," and then people want to read the whole thing in the nicer print format.
2. University of Michigan Press says it happens sometimes thanks to increased "awareness and visibility": https://www.press.umich.edu/openaccess
3. A phd thesis reviewing studies, and finding a negative relationship between OA and sales: https://openaccess.leidenuniv.nl/handle/1887/68465
4. A swiss study of matched pairs of books, one OA, one traditional; no statistically significant difference in sales http://www.snf.ch/SiteCollectionDocuments/OAPEN-CH_schlussbe...