It's also worth pointing out another aspect of subscription agents. The second largest subscription agent, SWETS, went bankrupt in 2014. If you look at the graph for the top publishers you'll see some big changes in 2014. Ebsco (another subscription agent) revenue starts dropping and Taylor and Francis revenue shoots up dramatically. I don't know this for sure, but I'd bet that those trends are 100% related to the SWETS bankruptcy. Publishers started trying to more direct deals without subscription agents. For some reason the data only has SWETS numbers for 2010, which doesn't seem right, I'd expect numbers until 2014. But for the big changes in 2014 (like the T&F jump) I'd guess that it's just about shifting from using agents to dealing with publishers directly.

So all this to say that there are a lot of intricacies in this data, and without the context a lot of it is misleading. Taylor and Francis didn't increased their prices by 20x in 2014.

Another example is where the post cites differences among publishers when looking at "cost per citation" (which alone is a dubious metric). This is going to be entirely dependent on field of study. Certain fields, like bio and physics have massively higher citation rates per paper than fields like humanities and social sciences. Publishers (like SAGE) who focus on social sciences are going to have vastly fewer citations per article than publishers with an STM focus. But the value of a journal (either defined as the actual cost or in more theoretical ways) isn't simply directly tied to citation counts. I'd argue that's a good thing, but I suppose you could debate it. But I raise the issue only to point out that "revenue per citation" types of metrics are almost entirely related to the fields of research and not publishers over or under charging for the same content relative to each other.