Download Statistics and Pitfalls

The NBER maintains a server providing download counts to subscribers of our online working papers at: http://papers.nber.org/subscriber_info.pl or visit http://papers.nber.org/ and select "Your subscription Information".

That page normally shows some basic information about your subscription (such as expiration date and contact info), based on the IP address of the client browser. So the page is only available to browsers within the subscriber's site.

We can "enable statistics", in which case the page will also show download counts by month since March 2002 for your site. We only record downloads of full-text papers, as searches and abstract views are free for all visitors anyway. You must write to Diane Birnbaum (subs at nber dotte org) to turn this feature on for your site.

Several libraries have asked for this, but I am convinced that using the logs from your own proxy server would be more accurate and would allow you to select the statistics you needed, in the format you needed. It would also protect you from counting mistakes made by publishers.

Counting downloads is more difficult than it looks, since MSIE will make multiple requests for the same PDF file for a single user request. A publisher may not discriminate between the first request and subsequent (redundant and ignored) requests, so you may end up with an overestimate of actual use. We are careful about this in our counts - duplicate requests are dropped. Among other publishers, JSTOR is the only one I have seen explicitly says it handles this problem.

If you have a good cache, it will be handing out pages without the publisher knowing, and that would result in an underestimate. This may be why some publishers forbid you from caching their documents. We encourage caching because it improves reliability and performance.

None of this is a problem if you do the counts from your own proxy server log. You may wish to discuss this with the system administration staff at your institution. I am always available to discuss these matters with you or them.

Alternatively, if your own web server tracks page views, we can arrange to hit your webserver for each page requested by a user from your site. Then you can follow your usage of our service by examining the logs for your own web server, which perhaps you already have arranged to suit your needs.

Please call if you wish to discuss your needs in this area.

Daniel Feenberg
feenberg@nber.org
617-588-0343


Last modified 4 June 2004 - http://www.nber.org/statistics.html  

 
Publications:
Main Publications Page
 
New This Week
Working Papers
Books              
Books in Progress
Older Books Online
Digest            
Reporter            
Bulletin on Aging & Health
2002 Japan Conference
Historical Bulletins
Free Subscriptions
Paid Subscriptions
 
Programs:
Program descriptions and members
 
Working Group Descriptions and Papers
 
Selected Projects:
Conference on Research in Income and Wealth
Conference on Econometrics and Mathematical Economics
Sloan Science and Engineering Workforce Project
Boston Census Research Data Center
 
Call for Papers
Submit to WP Series             
 
Data:
NBER Collection
Business Cycle Dates
Latest Business Cycle Memo
New Economic Releases
Selected Sources
Economic Organizations
US Government Agencies
Other Data Collections

Economic Report of the President
Economic Indicators
Congressional Budget Office
OECD Frequently Requested Statistics
 
About
What we are
Contact us
Non-data Links    
Search              
Site Map
Help              
Employment              
Fellowships
Early History
 
People:
Staff
Researchers
Board
Contact Us
Search
 
Search via Google:
 
printit emailit