In an editorial on the topic of Sci-Hub, Marcia McNutt (Editor-in-Chief of Science Journals) lists a few reasons why Sci-Hub may be bad for scientists. One of the drawbacks she lists is (emphasis mine):
Authors do not benefit from download statistics, for example, which are increasingly being used to assess the impact of their work.
This surprised me, as I have never come across any mention of download statistics being used to measure the impact of papers / scientists. Are download statistics really being used in any meaningful way when assessing the impact of scientists?
I'm also asking because sites like ArXiv does not provide download statistics, and neither do other secondary publication servers (university paper repositories, private homepages, ...), which would hurt the impact rating the same way.