Harvard Business School put out some interesting statistics back on June 1. They raise some interesting questions, but the statistics have some issues, so they should be taken with a grain of salt.
The first is their comparison with Wikipedia editors. They state that:
…the pattern of contributions on Twitter is more concentrated among the few top users than is the case on Wikipedia, even though Wikipedia is clearly not a communications tool. This implies that Twitter’s resembles more of a one-way, one-to-many publishing service more than a two-way, peer-to-peer communication network.
However the comparison is based only on Wikipedia editors, not those that use Wikipedia. Just as their study indicates many on Twitter merely follow, many Wikipedia users only read articles and do not contribute as editors. A fair comparison would exclude some fraction of Twitter users.
The gender biases are also interesting. However, they limit the gender study to “strongly gendered names”. Furthermore, as a coworker of mine pointed out, women may be more unlikely to expose their real name online. So their gender sampling method may be biased. Would it invalidate their conclusions? Perhaps not, that would require a bias of at least the order of the gender bias they measure (10% or so). But they don’t indicate what fraction of users could be assigned a gender, nor did they investigate what fraction of the population can be gender-typed using their list. The latter would be easy to do with census data.
Studies like this are never perfect, but I generally expect those that conduct studies to call out possible biases in their sampling, and report more fully on the data and their methodology. I would discard their conclusions about Twitter as a one-way publishing platform as unsupported by their data, and would take their gender bias numbers as a rough estimate with potentially significant biases.
RSS 