In July 2014, we published a paper in Transportation Research Part C (online since May 2014). This journal allows to archive a post-print (i.e., final draft post-refereeing) (but not the publisher's version/PDF), which is called by Elsevier the "accepted author manuscript". In September 2014, we published an accepted author manuscript as a PDF and last week as an eboook, both on Infoscience, the institutional archive of the EPFL.
This accepted author manuscript has been downloaded more than 1000 times on Infoscience. Since the ebook was only published last week, these 1000 downloads are due to the PDF version only. Research data were published on Infoscience in March 2015, and it is not clear if the 1000 downloads are related to the accepted author manuscript or to the research data. However, it seems more reasonable that people end up on Infoscience by looking for keywords associated to the author manuscript than to the research data.
As a comparison, the publisher version was downloaded 1000 times by April 2015. The number of downloads is of the same order of magnitude and we can say that publishing the accepted author manuscript roughly doubles the number of downloads.