Based on a sample of 5 million webpages from 15.000 websites in China with a total of 40 million external hyperlinks (i.e. pointing outside their website), they could built the following model:

Only a tiny minority of 6% of links does leave the Chinese web space. While this figure is more or less robust he tells me that the distinction between Home vs. Other provinces is still preliminary as it is based on geo-IP location. This might tell you where the server sits but not necessarily much about the “location” of the websites its been hosting. Therefore they are now trying to explore new methods including content mining. (The results have been presented in 2005 by Jonathan J.H. Zhu and Xiaoming Li at the Annual conference of Chinese Computer-Mediated Communication)

This research is exciting to me because it links back to my research topic that is also about the extent to which Internet use (in my case for creating and maintaining friendships) is locally focused. As a matter of fact, when coming to the Oxford Internet Institute my original plan was to do something along the lines of Professor Zhu and to use hyperlinks to assess the role of national identities online (you can have a look here). I eventually abandoned the idea amongst other reasons because it requires a considerable effort in terms of hardware etc. The more I am looking forward to hear more results from these ongoing efforts.

To compare some of the findings there is a useful review article by Han Wo Park and Mike Thelwall (2002) in the Journal of Computer-Mediated Communication and a more recent study by Barnett and Sung (2005): Culture and the structure of the international hyperlink network. All point to the fact that hyperlinking behaviour seems to differ for different nations. Other notable research projects in this area have been undertaken by Robert Ackland who used to be a fellow here at the OII. He has been analysing the linking behaviour of political parties as well as environmental activist groups.

The data comes mainly from the Oxford Internet Survey (OxIS) 2005 (2186 respondents, 14+) and eBulgaria 2005 (1000, 15+) as well as some other data from the ITU and Eurostat.
A lot of the findings will not come as a huge surprise to many:

• there is a correlation between GDP/capita of a country and % Internet users
• income, education and age are significantly correlated with access with the rich, younger and more educated being more likely to have access

However, there are also some more counter-intuitive results I think:

• no real gender related differences in access
• usually who has a PC also has Internet access
• cellphone use is different (usually higher) from PC and Internet use

However, the special thing is that the findings are based on correspondence analysis (and yes, he did the regressions before) which is a neat way of observing relevant connections. In Martin’s case it basically maps different dimensions of sociodemographic categories (such as being older than 50 years, having a University degree, earning less than £10,000 etc) in a two dimensional space together with the dimensions of several dependent variables in question such as Internet use and mobile phone use.
The end result is that a certain value of a category that goes often together with Internet use (such as having a University degree) is mapped closer to the variable Internet use. Conversely, a category like “low income” is mapped closer to “no Internet use” as the two often go together.

Of course, a layperson’s descriptions won’t help you much in visualizing these diagrams but for now take it as a teaser until Martin publishes his paper. For the very interested of you I’m sure he is more than happy to share more details.

Take that all together and you can get a lot of insightful data such as that one out of 200 British Internet users was visiting a blog last week or that we should expect a lot of pirates, Darth Vaders and Wonder Women this Halloween (see also chart below). If you are interested in it, Heather Hopkins is an analyst at the company and her corporate blog reports on interesting findings for the UK with the Hitwise data.

The quality of the data is clearly related to how much traffic Hitwise is analysing and from which ISPs it comes from. They are not very open about that but in one post they claim to observe the online behavior of 25 million Internet users every day and in another place they claim to measure about 500,000 websites out of 160 industries every day. I would guess that this is the total over all the countries they cover which are the US, the UK, Australia, New Zealand, Hong Kong and Singapore.

I just got an update on that (see also comment) from Heather: Across all markets we cover, our sample is over 25 million internet users. In the UK, it is 8.43 million (as you mention). In the US it is 10 million. The rest covers Australia, New Zealand, Singapore and Hong Kong.

Early this year Matthew Hindman (Political Science Department, Arizona State University) who coined the term Googlearchy was a guest speaker at the OII and he was actually using data from Hitwise for his research on shifting patterns of information exclusivity.