- Articles whose title contains the words "is right", "agree" or "genius"
- Articles whose title contains the words "is wrong", "disagree" or "idiot"
- Articles whose content contains "is right", "agree" or "genius"
- Articles whose content contains "is wrong", "disagree" or "idiot"
Tuesday, 31 March 2009
Today: consumer perceptions of value in blog postings, and its relation to the evolution of judgment.
I have noticed recently a lot of blog posts which set out basically to shoot down other bloggers. I'm sure you share my lack of surprise at this phenomenon: it's been a staple of Internet debate for decades.
Could this be because economists (or people in general) are just disagreeable? Or is it because these postings serve a real or perceived consumer demand for conflict?
A minor tangent: Such a demand, if it exists, might arise from an evolutionary or learned response to the marginal value of information. A single new item of knowledge which acts towards confirmation of existing data is probably less valuable than one which contradicts it. This would be explained by an effect of diminishing marginal authority of data in each direction.
Let's say that Paul Krugman says the stimulus is too small. You may, for the sake of argument, infer that there's a 50% chance the stimulus is too small. If Brad DeLong agrees with Paul, that may slightly strengthen your evaluation to 70% (it would hardly increase it to 100%). On the other hand, if Greg Mankiw says the stimulus is too big, that might cancel out Paul's statement altogether, leaving you neutral as to the proper stimulus size. Thus, the degree to which your mind is changed by an agreement would be less than by a disagreement.
(This ignores the phenomenon of confirmation bias, but that applies more commonly to beliefs you have held for some time than to those formed recently from external information).
However, this is just a hypothesis about the reason that a preference for disagreement might arise. What I am interested in finding out is whether such a preference really exists.
Fortunately, we have a useful dataset with which to do so: the economics blog word cloud and the data behind it.
I wrote a program to analyse the articles and filter those which fall into the following categories:
This enables us to get a measure of how disagreeable economics bloggers are.
As I don't have access to traffic data for other people's blogs (EconDirectory.com has averages, but nothing for individual posts) I measured the number of comments on each article as a proxy for its popularity (excluding those blogs which don't accept comments).
So what results do we get?
Surprisingly (at least to me), economics bloggers are more agreeable than not. "Agree" articles (category 3) showed up more than twice as often as "disagree" (category 4). When measured by titles, the trend is not so clear, with a majority "agree" articles (category 1) when measured over the last two months but more "disagree" (category 2) when taking the last 7 days alone.
However, blog readers are not so magnanimous. On the content measure, the mean number of comments on an "is right" article (category 3) is 3.66, while there are an average of 6 comments on an "is wrong" article (category 4).
When the title filter is used, the difference is even greater: there are no comments at all on the category 1 ("genius") articles, and an average of 21.6 on category 2 ("idiot")!
(The trend for titles is less clear if we go further back, because there is a single Calculated Risk posting in the "agree" column with 166 comments. If we exclude this as an outlier, the same picture remains, but you might legitimately question why it should be excluded. If included, the average number of comments on category 1 articles jumps substantially, because there are not many in that category.)
In conclusion, bloggers can be reassured that they are quite civilised people; while blog readers are clearly baying for blood.
If you're reading this blog and you think this is incorrect, please set up a blog of your own and write an article headed "Leigh Caldwell is wrong". If enough of you do this, you may be proved right next time I run the analysis.