Recently I did a quick experiment using the ADS logs; you may be interested in the result.
For three journals I looked at the two questions "How many articles were accessed by ADS users during the month?" By "accessed" I mean that for an article a single user obtained one or more types of information from the ADS, where the main types are abstract, citation list, and full text; there are a number of other types. I then asked the question "How many of these accesses involved access to the full text?" The ADS has links to the full text for all three journals. The three journals in the experiment were the Astrophysical Journal and two others. The two other journals represent the core literature for a subfield of astronomy. While the ApJ is a subscription based journal, for practical purposes it can represent an open access journal, as nearly every astronomer has access to it; also, after three years the ApJ is open access. The other two journals are certainly not open access. One is from a commercial publisher; it is normally part of a large, very expensive subscription package. The other is from a learned society; it is not bundled, but is rather expensive; also it has very restrictive terms as to which computers may access it. Via the Harvard Library I have access to the commercial journal from my desk; to read the society publication I must go to a terminal several floors away. I have had access to the ApJ online from my office for nearly a decade. The results of the experiment are that 87% of those who accessed ApJ articles obtained the full text, but only 39% and 40% of the readers of the other two journals did so. If one takes 87% as the fraction of users who wanted to obtain the full text, and assumes that readers of the other two journals share those desires to actually read the articles, then more than half of the potential reads of articles in the two journals are prevented by the restrictive licensing. I should also note that the ApJ permits its papers to be posted in the ArXiv without restriction, the other two journals forbid this entirely. The bottom line is that the cost of the restrictive access policies, in terms of access prevented, is larger than the access allowed. Best wishes, Michael Kurtz Harvard-Smithsonian Center for Astrophysics 60 Garden Street Cambridge, MA 02138
