How abstract views and downloads are counted

Authors and RePEc archive maintainers receive monthly emails with various statistics, and among the most anticipated statistics are our abstract views and download counts. It is important to understand how those statistics are collected and what they measure (and do not measure). Full statistics are available on the LogEc website managed by Sune Karlsson from Örebro University (Sweden).

Participating RePEc services (EconPapers, IDEAS, NEP and Socionet) keep a log of all activity on their sites. This allows us to count page views for the abstract pages of each items in the database (excluding NEP, as abstracts are listed in emails). Logs also record outclicks as users leave the RePEc services to the sites containing the full texts they seeks to download. This allows us to count “downloads”. Quotation marks are required as it is impossible to record whether the download was successful, for example in the case of gated publisher sites. Note also that this means that downloads that have not transited through a RePEc cannot be counted, as we do not have access to local logs.

LogEc gathers the logs from the participating services and aggregates the statistics. This involves much more than bean counting, though. Indeed, one first needs to exclude robot activity, as only human activity is of interest. Some robots declare themselves as such, but other hide their identity. One has thus to infer from various patterns what IP addresses are likely robots. This is an important step, as robots represent typically 75% of raw abstract views. Robots include spiders from many search engines as well as other initiatives on the Internet.

One needs also to weed out multiple views or downloads by the same user. This brings us to detecting attempts at increasing counts by authors. Obviously, we cannot reveal here how this is done, but let it be known that we have detected fraud even by authors using multiple Internet service providers. The methods used lead to some undercounting, though. Multiple users behind the same cache server may be counted only once, as it may for example happen to employees of the US Federal Reserve Banks that use RePEc.

And we are still not done pruning. LogEc then checks for additional patterns that need to be vetted by a human eye. Unusual activity is then checked and often reconciled with traffic from popular blogs, magazines and newspapers. But on other occasions, traffic surges cannot be explained in licit ways and need to be cleaned out.

After all these manipulations, statistics are published and disseminated. And despite substantial pruning, RePEc services still get over 2,000,000 abstract views and 600,000 downloads every month. See LogEc for details.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: