The majority of the data here is from Kevin Almeroth's systems and networking conference statistics page, so many thanks go to him. The rest comes from Jeff Erickson, the ACM, and conference proceedings and web sites. Most areas of CS are not represented, and the data for many conferences is not complete. If you send me data (conference name, list of year, submitted, accepted, and a URL documenting the authenticity of the data), I'll include it.
What's plotted for each year is the average (over all selected conferences for which there is data in that year) number of papers submitted and accepted, and the total number accepted divided by the total number submitted. Averaging partially alleviates the problem of missing data. However, there are still unfortunate effects; for example, in the all-conference graph between 1981 and 1985 the only data is from conferences that happened to have very high acceptance rates. Also, results are skewed toward larger conferences and towards areas with more conferences (systems & networking). So interpret the graphs with care.