Income Inequality in the Attention Economy

Kevin McCurley
Google Research

The full paper.

A major focus of attention in the field of welfare economics is concerned with the distribution of income in society. Multiple authors have observed that the information economy can be characterized as an "attention economy", where the scarce resource is not information, but rather attention to absorb the information.

Attention is a fundamental currency of advertising. It has value because attention is convertible into sales. In the language of a mathematician, attention is a necessary condition for sales, but not a sufficient condition for sales. Without attention there can be no sales.

It is therefore important to understand trends in how attention is being focused on the World Wide Web. In a recent paper, I describe a study of habits of web surfers, and how they distribute attention around the web. The overall conclusion is that there is a great deal of inequality in how attention is distributed around the web, and in order for "long tail markets" to develop, attention needs to be redistributed around the web.

To give you an idea, there are at least 100 million "domains" on the Web, where here I refer to or as two different domains, but is lumped in with These are the units that are assigned by the top level domain registration authorities. If you study how people surf around on these sites, it turns out that 0.2% of the domains receive about 85% of the attention (as measured by page views), and 2% of the domains receive about 95% of the attention. Moreover, the trend seems to be growing more unequal over time.

By contrast, if you look at the attention distribution that results from people clicking on search results from Google, then the attention distribution is much more equal.

The social consequences of these observations are open to speculation. I think it should be considered by those who argue about the strength of "long tail markets", since attention is a requirement of all markets, particularly when they are trying to grow.

For more information, you can read the full paper. Among other things I address questions about mathematical models for attention distribution and how to measure the effective size of the web. There are lots of URLs out there, but if nobody reads them, should we count them?