Why does the Economics Job Market Rumors (EJMR) website matter?
EJMR is a website for anonymously discussing the economics profession and sharing information about the academic job market. Some people call it 4chan for economists. You can think of it as a gossip site for people in the field. Among economists, it’s a very well-known, highly trafficked platform. As of last year, as best we can tell, there were millions of unique visits to the site every month.
Unfortunately, because of the way that the site is run, it has a huge amount of problematic content—sexist, racist, and bigoted comments as well as attacks on individual people. Our research found that more than 10% of posts are “toxic.”
This matters because the people going to EJMR to post and read are also writing letters of recommendation, doing referee reports on peer review articles, teaching, and in many other ways having a significant impact on economics in academia.
Because it’s all anonymous it’s so hard to put a finger on what’s going on. It’s easy to believe, “The really problematic posts are coming from people who are not even economists. Or if they’re economists, they’re at other institutions, not my institution, or not elite institutions.”
To get past suppositions, we thought it would be valuable to quantify things. And we found that the toxic posts are coming from everywhere, including each of the top-ranked universities in the United States.
EJMR is anonymous by design. Why?
I can’t speak specifically as to why the people who run the site do it exactly this way. But it’s easy to imagine the argument that if you want people to share information or complaints without fear of consequences, anonymity helps. Anonymity might allow for a type of discussion that we wouldn’t have otherwise.
But the same anonymity that might facilitate candid discussion on EJMR also facilitates abuse, bullying, and other behavior that is deeply damaging to the field and in particular to the persons attacked on the site.
It’s anonymous in the sense not only that you don’t know who an individual is in the real world, but you don’t necessarily know it’s the same person between one thread and another.
That’s right. It’s possible to register for a fixed username, but few people do that. The way it generally worked (it’s changed now) is that when you posted you were assigned a hexadecimal username—something like “ab4f”—which was used consistently within that thread, specifically so that people can keep track who is saying what in a single thread.
But if you post from a different computer or post in a different thread, you were nearly certain to be assigned a different hexadecimal username. The result is that there would be no way to link you across those two threads.
Q: How do you know that users are posting from top-ranked universities?
We just guessed how the site assigned usernames and it turned out that the username for each post was something called a “hash,” basically a shuffled-up version of the IP addresses from which the post was made. We did this using only publicly available data. And it was possible only because the username generator was incredibly basic. In fact, I’m told our paper was spread widely among cryptography researcher whose universal reaction was shock that EJMR would use a such a bizarre scheme to choose usernames. The reaction from computer scientists was, “This is literally what I tell people not to do. Even 20 years ago this was too simple.”
So, though I wish I could say our study was proof of our astounding genius, it’s not. We were only able to do this because EJMR’s anonymous administrator had a pretty severe misunderstanding of how hashes work. They also wrote about their scheme a number of times over the years such that this was more a question of when, rather than if somebody would back-out the origin of posts.
Now, figuring out IP addresses was conceptually simple, but difficult as a practical matter. We had to write some code that was a bit challenging and use beefy computers to finish the study. We go into details in the paper for anyone interested in specifics. Using some simple statistics , we were able determine the IP address from which 66% of the 7 million posts on EJMR were made. This huge number of posts comes from a relatively small number of just about 50k IP addresses.
What does attributing a post to a given IP address tell you?
IP addresses don’t identify an individual. Many people can share one IP address. And each person will have multiple IP addresses with the various devices and networks they use.
The police or the FBI, working with an internet service provider, may be able to link a specific IP address to a specific computer, but nothing like that is part of our study.
Universities own blocks of IP addresses, so we can tell when users are on a university network. Comcast and other internet providers own blocks of IP addresses, so we can get the rough idea of the city from which each post was made.
Together, this let us demonstrate how ubiquitous posting to EJMR is. I found it shocking people that affiliated with universities were posting to the site from work or from school, sometimes writing pretty horrid things.
What do you take away from what you found?
Here’s a mental model that I think is reasonable. Every institution has a few people who are posting on EJMR. Some of what they post is toxic, some of it not. A lot of people read the posts.
For me, knowing that the behavior on this platform is a pretty ubiquitous phenomenon reflects something problematic about the culture of economics as a field.
EJMR has been around for nearly 15 years. It has a significant amount of toxic content. Posts on the site have had serious negative consequences for folks that I’ve talked to. I find it troubling that we as a profession are OK with the site being around.
What has the reaction to the paper been?
After Florian and I presented the paper, the response was the most positive either of us had ever received for a research paper.
The share of women who work in economics is still relatively small, yet a disproportionate share of the positive response to this study has come from women. I think that speaks to something about the terrible impact of EJMR on women in particular.
We’ve also received some questions about what data we will release. To be very clear, we’re not releasing anything personally identifiable, and we are following all the appropriate guidelines around privacy and research ethics, of course.
And again, the reception overall has been very positive, which I hope reflects a broad view in the profession wanting to improve on this.
It’s not up to you, but what do you think should happen to EJMR to address what you discovered?
I wish I had something insightful to say. I have only the obvious. Obviously, the toxic stuff is bad and reflects poorly on our field. Maybe less obviously, when a faculty member or a graduate student contributes useful information to EJMR, when they visit it, when they ask a colleague if a rumor there is true—they give legitimacy to the site. In so doing, they create a massive negative externality. Without that legitimacy, the site would wither into irrelevance.
So, I think we have here before us a collective action problem. I hope that our case study on EJMR is a wakeup call. Maybe this is the moment we can put this aspect of our profession behind us. I believe that if the platform were removed or reformed, it would be a net positive. I know some people think other forums would pop up to replace it. For me, there’s a lot of good evidence that deplatforming toxic sites does work. What pops up usually struggles to replicate the overall impact.
As it is, the site is run by an anonymous person who makes a good amount of money off of it. Whomever runs it has limited personal incentive to take the site down.
Whatever happens with the site itself, the core issue is the behavior which is not new. Our documenting toxicity on EJMR is not new either. What I hope we’ve done is demonstrated that the toxic behavior isn’t coming from bogeymen somewhere else. It’s real people at our institutions. Universities shouldn’t turn a blind eye to it.