Official Secrecy of the United States Government

Description

Data on classification activity of the United States government.

Fitzpatrick (2013) notes that the dramatic jump in derivative classification activity (DerivClassActivity) that occurred in 2009 coincided with "New guidance issued to include electronic environment". Apart from the jump in 2009, the DerivClassActivity tended to increase by roughly 12 percent per year (with a standard deviation of the increase in the natural logarithm of DerivClassActivity of 0.18).

Usage

data(USclassifiedDocuments)

Format

A dataframe containing :

year

the calendar year

OCAuthority

Number of people in the government designated as Original Classification Authorities for the indicated year.

OCActivity

Original classification activity for the indicated year: These are the number of documents created with an original classification, i.e., so designated by an official Original Classification Authority.

TenYearDeclass

Percent of OCActivity covered by the 10 year declassification rules.

DerivClassActivity

Derivative classification activity for the indicated year: These are the number of documents created that claim another document as the authority for classification.

Details

The lag 1 autocorrrelation of the first difference of the logarithms of DerivClassActivity through 2008 is -0.52. However, because there are only 13 numbers (12 differences), this negative correlation is not statistically significant.

Source

Fitzpatrick, John P. (2013) Annual Report to the President for 2012, United States Information Security Oversight Office, National Archives and Record Administration, June 20, 2013

Examples

##
## 1.  plot DerivClassActivity 
##
plot(DerivClassActivity~year, USclassifiedDocuments)
#  Exponential growth?  

plot(DerivClassActivity~year, USclassifiedDocuments, 
     log='y')
# A jump in 2009 as discussed by Fitzpatrick (2013).  
# Otherwise plausibly a straight line.   

##
## 2.  First difference? 
##
plot(diff(log(DerivClassActivity))~year[-1], 
     USclassifiedDocuments)
# Jump in 2009 but otherwise on distribution 

##
## 3.  autocorrelation?  
##
sel <- with(USclassifiedDocuments, 
            (1995 < year) & (year < 2009) )
acf(diff(log(USclassifiedDocuments$
             DerivClassActivity[sel])))
# lag 1 autocorrelation = (-0.52).  
# However, with only 12 numbers, 
# this is not statistically significant.