June 28, 2008: After peaking at half a million documents a year at the end of the Cold War, U.S. government bureaucrats classifying material (confidential, secret, or top secret) declined to 105,263 by 1996. Then it began climbing again, and peaked once more in 2004 at 351,150 documents. Since then it has been declining, to 233,639 documents last year. There is also increase in "derivative classification" (reconfigured classified data, as in bits of older classified documents combined into new docs). This has been rising even faster. There were 5.6 million such actions in 1996, and 22 million last year. That's another reason to declassify old documents, to reduce these "derivative" documents.

All these secret documents cost money (to do the paperwork to determine what should be a secret, and to guard all those secrets). Last year, the secrecy bill in the U.S. government was nearly $10 billion. This is an increase of 4.6 percent from the previous year. Attempts to reduce this cost rely on classifying less, and coming up with ways to declassify this stuff inexpensively (otherwise the mass of classified data will grow, and become even more expensive to look after). More data is now being classified as "material that will automatically declassify in ten years" (easy to do with a lot of technical or operational data). Since 1980, 1.37 billion pages of classified documents have been declassified. That number has been increasing lately. From a low of 28 million documents in 2005, it increased to 37.2 million documents last year. But it's not fast enough, as the body of classified data (whose actual size is classified) continues to grow, as does the expense to protect it from unauthorized eyes.