I found an interesting data source on FigShare showing how much University College London paid for the Article Processing Charges. According to the description of this data set it should include items between 1st April 2013 and 31st July 2014. However, there are some entries that are dated earlier and later than that. I guess that only the previously mentioned period is complete. I included all data anyway and then I added a linear regression line on top of the plot showing total cost by date. Judging by that, it seems like the costs keep on increasing. However, it is just a short period of time and I don’t know whether this trend held for a longer period of time.
Large journals (e.g. Nature Communications, PLOS One, and Lancet) stand out as taking more APCs than others. The vast majority of articles in this data set has been published with a CC BY licence.
Feel free to further explore the plots. They are all interactive.