R.I.P. TimesSelect


Last week the New York Times ended its policy of charging for online access to its op-eds. There has been a lot of commentary on this (which I’m too lazy to link to) about how this was a failed experiment, and how (unlike WSJ) NYT was giving away its expensive news and charging for its cheap opinion, etc. The interesting thing to me (and I’m not claiming this is original) is that absolutely nobody reads the NYT op-ed page because it has Bob Herbert, people read Bob Herbert because he’s on the NYT op-ed page. The reason for this is not because the NYT op-ed page has so much institutional credibility that we dutifully attend to it despite the lack of compelling content (like going to church even though your preacher couldn’t sermonize his way out of a paper bag). Rather, the reason is network externalities: we used to read the op-ed page because we thought other people were reading it and we didn’t want to be left out of the conversation. Although it’s hard to remember, a few years ago a good portion of the political blogosphere was a Talmud to the op-ed page’s Torah. National Review Online even had a column called “Krugman Truth Squad” devoted to fisking the columnist, an enterprise that would seem rather pointless today. After TimeSelect, this huge cultural presence of the op-ed page crashed overnight.

This is a good example on how a shock can affect a complex system. One way to think about it is to imagine that everyone else in the chattering classes got TimeSelect, would it be worth it for you to subscribe? For myself, it would probably be worth paying $50/year to converse knowledgeably with my colleagues and the entire blogosphere. However, this was not the actual situation. If a few people drop the service this decreases the network externalities of the service, and in turn it becomes less valuable to the remaining subscribers, many of whom drop it until you’re left with only the few dozen people who actually think the op-ed page is an intrinsically riveting read. This is a dynamic threshold model with positive feedback of a type first described by Schelling and later generalized by Granovetter and still later popularized by Gladwell. The original model was to explain residential segregation. Suppose that whites have a distribution of preferences for the minimum whiteness of their neighborhood with a mean of 70% and a standard deviation of 15%. The naive prediction would be that there would be a lot of neighborhoods that were about 70% white, but in fact there are MSAs like Detroit where almost all neighborhoods are either greater than 90% or less than 10% white (you can practically see 8 Mile from space). In Schelling’s model you start with a neighborhood that is 100% white until a few non-white households move in, moving the white rate to 98%. The vast majority of white households don’t care because they have preferences of between 50% and 98%, but a few hyper-bigoted households can’t stand their new neighbors and move out, only to be replaced by more non-white households. This upsets the whites with preferences between 95% and 98% and they move out which pushes the rate down to 95%, which in turn upsets the 90-95% preference whites, etc. Since you’re still in the upper tails of the distribution this all happens fairly slowly, but eventually around +1 sigma you start reaching the fat part of the curve and the neighborhood “tips,” with all the white households leaving overnight. Thus you have segregation much higher than that implied by the mean white preference (and no doubt the mean non-white preference as well). Likewise, you can imagine the NYT op-ed page going into a downward spiral where the more readers who abandon it, the less valuable it becomes to the remaining readers, who in turn also abandon it. Alternately, imagine that you subscribe to TimesSelect, but few other people do. With whom could you discuss it? Many bloggers subscribed to TimesSelect but were reluctant to link to it for fear of making the implicit demand that their readers subscribe to the unpopular service. (As Ezra Zuckerman recently observed in the comments, network externalities can not only create value but impose obligations. I for one am glad to be through with the chore of reading the NYT op-eds).
So one puzzle is whether having fallen off the paywall, will we put Humpty Dumpty back together again? I am actually inclined to think not. As shown theoretically by Banerjee and empirically by Salganik et al, positive feedback systems are fairly arbitrary in their allocation of success. If I had to guess what will eventually serve as the top dog, my money would be on the Atlantic blogs. I RSS several of these blogs and I would prefer any of them to any part of the NYT op-ed page. Part of it is that three of the six Atlantic bloggers are my age, but the real thing is that the Atlantic lets them write posts with no minimum or maximum length whereas the NY Times standard of 800 words twice a week can turn even the Wolfe-ian style of David Brooks into a font of boredom. One indication of the Atlantic‘s success is that Megan McArdle is beginning to experience the backhanded compliment of serving as a focus of hatred for some of the left blogosphere in the same way that Paul Krugman used to for the right.


Written by GR

September 24, 2007 at 3:46 pm

7 Responses

Subscribe to comments with RSS.

  1. A few points worth noting:

    1) TimesSelect was free with a NYT subscription.
    2) TS eventually became free (maybe a year ago?) for anyone with a .edu email address. Even before this, many academics could read TS content through their libraries.
    3) It wasn’t hard to find blogs that reproduced NYT op-ed content, or friends who would email them to you.

    Your analysis is interesting, but I’m not convinced that TS really limited the reach of the NYT columnists, for the reasons noted above.


    Dave P.

    September 24, 2007 at 4:41 pm

  2. dave p,
    good point. i guess my answer is that in practice these things ameliorated but did not eliminate the cost of reading the op-ed page. for instance, getting a column through your university’s lexis-nexis subscription is a lot of hassle. even a blog reproducing the content is still costly in that it’s a less obvious place to look for the content than and it either wouldn’t occur to many people to google it or they would just feel uncomfortable with pirated columns.



    September 24, 2007 at 4:55 pm

  3. Gabriel,

    One thing I’ve noticed (in a very rudimentary and unscientific way) since the NYT canceled Times Select is that their op-ed columnists seem to be ranking higher on their website’s list of “most e-mailed” stories. They are easy to spot on these lists because the columnist’s name precedes the column headline.

    Obviously, this is less than one week of fuzzy observations, but it may be worth keeping an eye on as an indicator of how many people are actually looking at the op-eds. (They also have a counter for “most blogged articles” which I have not followed until reading this post. As of this posting, two columns are in the top 10 most e-mailed but none are in the most blogged list.)

    PS: If anyone has some other observations about this, please share.



    September 24, 2007 at 5:10 pm

  4. I suppose you could settle this empirically. If the blog citations to NYT op-eds went down after the Times Select wall was set up and never recovered their pre-wall prominence, I think you’d find support for Gabriel’s hypothesis.



    September 24, 2007 at 5:12 pm

  5. One empirical question that would be interesting to investigate is something that is left relatively understated in the Schelling Hypothesis. In the residential preference question, what is the difference between making the decision to move out and the decision to move into a different neighborhood. In other words, there might be a stronger inertia to keep a white resident in a neighborhood that was becoming increasingly non-white so that resident actually ends up residing in a neighborhood in which they would never consider moving into if they were selecting a new place of residence.

    The same question is interesting here. As Brayden notes, the citations to Times Op-Eds decreased after the Times Select wall went up. But, how much persistence (or, I like to think of it as inertia) exists for bloggers still citing Times Op Eds. And, do bloggers who come to blogging after the Times Select wall use it less because it was never available to them in the first place. Because blogging was expanding at the same time that the Times Select wall went up, might be part of the explanation. Eventually, people are exposed to new resources , and other firms use that exposure to their advantage (i.e. the Atlantic), and stop relying on their “old-standby” Times Op Ed.

    This also raises interesting questions relating to Ezra’s point about academics being coerced to respond to blogs. If both bloggers and non-bloggers tend to remain in their persistent state and you add grad students who have no initial preferences; they could end up preferring/liking blogs because they have no modus operandi in which they can persist, then they can drive the field to blogging by the kinds of externalities Ezra mentions.

    I am not sure any of this makes sense – I am not really an Econ Soc person, I actually study residential preferences and segregation. But the question of inertia has come up for me as a consideration of an important dynamic that is understudied in the residential segregation literature but might be more articulated in the Org Studies or Econ Soc literature.



    September 24, 2007 at 5:32 pm

  6. NYTimes also features blogs of some of its columnists. They brought the Freakonomics blog on board a while ago, and after TS died, Paul Krugman has started one (or, his already existing blog — which was also behind the TS paywall! — became publicly accessible). Nick Kristof has one too.

    When it comes to newspaper-hosted blogs, I think a comparison with Guardian’s “Comment is Free” is more appropriate than the one with the Atlantic blog.



    September 25, 2007 at 8:26 am

  7. […] of abandonment until finally the tipping point was achieved and mass abandonment ensued. (See this post for an explanation of tipping points, network externalities, and threshold models). Part of the […]


Comments are closed.

%d bloggers like this: