economists’ free riding really is caused by treatment, not selection! [more apologies to marwell and ames]

I do mostly think about things besides Freakonomics (my first two posts notwithstanding). But since I got this bee in my bonnet about the climate chapter I’ve been watching their blog to how they would respond to the avalanche of criticism. And I came across a gem today that shows why you should never trust a freakonomist, or at least why you should worry about one who took Daniel Hamermesh‘s “500-student principles [of microeconomics, I hope] indoctrination class” at the U of Texas.

Hamermesh tells us that he ran a modified prisoner’s dilemma experiment in his class. Students could collude and split $20 evenly, but had to play their moves in secret. There were eight players. Seven tried to cooperate. One student (“bless her heart”) screwed them and got the whole $20. The other seven students booed her. And here the gem. Hamermesh writes, “But I got the class to join me in applauding her, as she was the only one who understood the game.”

Call me naive, but it sounds to me like the seven screwees understood just fine. Mr. Hamermesh changed the rules ex post facto. I doubt he said ex ante, “you can collude, but know that you will be applauded by hundreds and your heart will be blessed to thousands if you defect.” They thought they lived in a world with the usual normative sanctions, and they were doing what they could to make sure she behaved the next time around. They knew they’d see each other again. They knew she might need to borrow class notes someday. Yet he — in a position of authority with some power to affect norms — sanctions the cooperators by calling them suckers and enlists the class, whose grades he controls, in that sanctioning? What’s he do when he sees a kid grab something from a playmate? Pat the grabber on the head and tell the grabbee to stand up and fight?

So I have to take back something I wrote in my last post. Freakeconomists do know that it is possible to change human behavior. And they do understand the power of social sanctions. They just don’t like sharing.

41 thoughts on “economists’ free riding really is caused by treatment, not selection! [more apologies to marwell and ames]”

  1. One of the great things about Prisoner’s Dilemma specifically and Game Theory in general is how it is ridiculously easy to show that optimal societal results almost never correspond to Nash or Subgame Perfect Equilibrium–even in the models we create.

    In a case where the girl keeps the entire $20, no one will ever be willing to negotiate with her again. Does anyone other than an economist believe that the Present Value of her choice is superior to that of everyone else?

    Would you still believe it if I told you the next round is played with 20s, then 50s, then 100s?


    1. I totally agree. So it’s baffling (well, not really, but let’s pretend) why people like Hamermesh don’t get that his in-class result is actually really interesting in a larger frame. It points to how collusion actually happens. But since he doesn’t want to understand, he casts her as the apple of his eye! Performativity in action!

      It’s not a general case, but I think that for this student, the present value just shot through the roof. I see a bright future. She gets into U of C on Hamermesh’s strong recommendation, and gets hired at Goldman where she is put in charge of negotiating with Treasury. It’s important to have someone in that job who can credibly sound at T1 like they’ll cooperate at T2.


  2. I’ve talked to a bunch of experimental economists and they tell me that when you play experimental games (ultimatum, dictator, trust, etc) with people, they will cooperate most of the time. They even do this when the games are not iterated.

    That is, unless you play with economists. They are the only group where it doesn’t work out.


  3. I took a rat choice choice with an experimentalist (Werner Raub) who told me a similar tale. People are hard wired to cooperate. He found that business people tend to cooperate a lot, which makes sense since business folks are constantly dealing with each other.


  4. Indeed, there’s research on this @2 & 3. That’s the thing about marwell and ames (granted, it’s about free riding and not prisoners’ dilemma games, but the logic is basically the same). See: Marwell, Gerald and Ruth Ames. “Economists Free Ride, Does Anyone Else?” Journal of Public Economics, 15 (1981): 295-310. They find:

    “Summarizing most of the results seems ridiculously easy: over and over again, in replication after replication, regardless of changes in a score of situational variables or subject characteristics, the strong version of the free rider hypothesis is contradicted by the evidence. People voluntarily contribute substantial portions of their resources – usually an average of between 40 and 60 percent – to the provision of a public good. This despite the fact that the conditions of the experiment are expressly designed to maximize the probability of individualized, self-interested behavior. Free riding does exist – subjects do not provide the optimum amount of the public good, and tend to reserve a meaningful fraction of their resources. The ‘weak’ free rider hypothesis is supported. Nevertheless, the amount of contribution to the public good is not easily understood in terms of current theory…

    Comparisons with the economics graduate students is very difficult. More than one-third of the economists either refused to answer the question regarding what is fair, or gave very complex, uncodable responses. It seems that the meaning of ‘fairness’ in this context was somewhat alien for this group. Those who did respond were much more likely to say that little or no contribution was ‘fair’. In addition, the economics graduate students were about half as likely as other subjects to indicate that they were ‘concerned with fairness’ in making their investment decision.

    Perhaps these results make sense. Economists may be selected for their work by virtue of their preoccupation with the ‘rational’ allocation of money and goods. Or they may start behaving according to the general tenets of the theories they study. Confronted with a situation where others may not behave rationally, they nevertheless behave the way good economic theory predicts. Note as well the very similar responses of our ‘famous’ economists.”


  5. Thanks Shamus. Maybe I was a little cryptic w/ the offhand referencing the Marwell and Ames paper in the title. It is, uh, 28 years old at this point and not exactly on standard soc reading lists these days.

    @3 — I’d actually take issue w/ the people are “hard wired” to cooperate. Obviously, the hardware allows for cooperation. But I defy those experimentalists to find a pre-social person on whom to run those experiments. To stay with and sort of brutalize the wiring metaphor, we all have an operating system loaded onto us by our upbringing/society/etc too…; the software matters (I think I might have gotten vista; I crash a lot).

    And that’s kind of my point w/ Hamermesh’s response. It relates to what Latour/Callon and the performativity folks talk about. Hamermesh is training students in the notion of fairness. Only the defector “understood” the game. He’s enlisting his own social role, and literally hundreds of people, to reinforce that interpretation. He’s *making* homo economicus. But his results in fact make clear that it’s *way* more complicated than that *even in* his little game. He had a real teachable moment. And he (proudly) took it in a socially suboptimal direction.


  6. As my username suggests, I’m an economist. I’ve been reading and enjoying Scatterplot for several years, because I enjoy the content and appreciate the exposure to another discipline. This post, however, really turns me off. It’s an attack on a well-respected, well-liked scholar in another discipline; it casts aspirations about people who study economists; it reeks of disciplinary prejudice and is deliberately incendiary. You continue to make the issue personal and about Hammermesh himself in your responses to comments.

    Besides all of that, it’s not even well reasoned. Professor Hammermesh was teaching the economic concept of Nash equilibrium, not the sociological construct of “fairness,” as you claim in your comment on Oct. 23. The “understanding” to which Hammermesh referred was of the Nash equilibrium in the prisoner’s dilemma game, which as you know is a set of stylized facts meant to illustrate a concept. It was not understanding of a fairness norm. Hammermesh didn’t frame it that way, and neither does the discipline of economics. You did.

    Whether collusion or defection is socially optimal is not as simple as you claim, either. It depends on the objective function. If the person who defects is more desperate than the others for money — if, for example, she’s willing to defect because she needs money pay her electricity bill which has to be paid in full, whereas the others will collude and take a smaller payout because to them this is just a game — then which outcome is actually socially optimal?

    Further, students who participated in the demonstration were presumably aware that they were in an economics class. They understood that they were illustrating a concept, and that there was a “correct” answer according to the professor. If they understood the principle being illustrated by the professor, their choice was to pander to the professor and get his approval (and maybe $20) or to pander to their classmates (assuming they thought their classmates didn’t understand or were not sympathetic to the professor’s favored outcome). Hammermesh didn’t change the rules of the game at all. You are just choosing not to acknowledge that students understand that giving the answer professors want is part of the game.

    Frankly, I’d be interested in reading what sociologists think about game theory, and about how sociologists interpret lab experiments conducted with college students in order to test economic principles. However, I’m not interested in reading shallow, spiteful attacks that are poorly reasoned and seem designed to pigeon-hole a scholar and a discipline rather than promote understanding. Scatterplot is a great platform, but you’ve misused it and I’m certainly less interested in reading this blog — and therefore voices from your discipline — than I was before.


    1. I’m no economist, but the point of PD games, I thought, was not that there was “an answer” but instead that there are multiple durable equilibria. So applauding defection vs. cooperation in a PD game strikes me as odd. The the former is more in like with the majority of economic thinking, it is not the right answer. The latter is also a durable outcome. That is the point of the PD. Both outcomes are sensible.

      More importantly, the distinction is not always between collusion and defection. You can articulate it this way. But you need not. It could be called cooperation. And the point here is not a trivial one: economic relationships are embedded in all kinds of other ones. And those relationship often create ecnoomically sub-optimal outcomes. There is a world outside the game played in the class, just as there is a world outside the market. That world matters. To ignore it can get things terrible wrong. This is not just a conceptual point, it is an empirical one. This is the very point that Josh is making: that there are things outside the game that influence the game.

      Just because you don’t agree with someone doesn’t mean they are spiteful and shallow, or that their positions are poorly reasoned. It strikes me as odd to critique someone’s reasoning with name-calling.


      1. The game Hammermesh proposed was a single-shot game, which means that you can’t actually enforce a cooperate-cooperate equilibrium. In a single-shot PD game, the only Nash equilibrium is defect-defect.

        As for cooperation versus collusion, with respect to the PD game this is simply terminology. Each player has two strategies to choose between. We can call them A and B.

        The things outside the game that influence the game do so by changing the payoffs. In other words, the payoff isn’t actually (20,0) versus some equal split. There’s the value of the social response to your choice that isn’t included in the payoffs written into the PD matrix. Fine, I don’t dispute that — though I might quibble with its relevance in an example to explain the mechanisms of a simple PD in an introductory economics class. But I do dispute that the students couldn’t figure out what was missing from the PD game that Hammermesh proposed. I think that college students understand full well that pleasing teachers is part of any game.

        Disagreeing with Josh isn’t why I think his comments are spiteful or shallow (note that I didn’t call Josh spiteful or shallow; I called his comments — which I characterized as attacks — spiteful and shallow. You were the one who reframed that as name-calling or calling Josh the person spiteful and shallow). I think his comments are shallow because they don’t consider the context (a class in basic economic principles), the goals of the instructor (illustrating a stylized economic behavior), they generalize broadly (you should worry about any student who took Hammermesh’s class), and precisely because they conflate single-shot versus repeated games. I think they are spiteful because they are personal, because they go beyond criticizing Hammermesh’s exercise to criticizing Hammermesh himself (he “proudly” took it in a socially suboptimal direction, per Oct. 23 comment), and because they take on a decidedly personal tone (“people like Hammermesh”) while speculating about motives (“since he doesn’t want to understand”).


    2. Look, yes the post was snarky. It was intended to be funny while making a point in way that someone might actually read it. The title was supposed to make clear that it was a play on a famous Marwell and Ames paper called “Economists free ride, does anyone else?” (referenced by Dubner himself here —

      I know that Hamermesh is a well respected scholar and I can see from his website that he is winning teaching awards. That is very very laudable for someone who could at this stage of a very decorated career just be coasting. With that said, I disagree that in the post I made an attack on his *character* in any general sense (I concede I should have been more careful in the comment – I’d take back “people like Hamermesh”). I meant to attack the *presentation* of his in-class experiment to the world, and assumed that someone with his prestige and at his stage of career has enough to stand on that what he wrote could be used to make a general point. I imagine that *in the class* he then turned to nash equilibrium – the experiment was supposed to get students to understand and remember. But *on the blog* he took the result to show how difficult collusion is to maintain. The difficulty of maintaining collusion is a hotly contested topic in the real world. It is also a topic with major policy implications.

      The sociological point was that cooperation is — or is not — a norm, and norms that are not enforced cease to be norms. In the class, the 7 students embedded the PD in life, that is, in a series of iterated interactions. She didn’t. She and they understood different games. You acknowledge as much in your own comment (“students understood they were in an economics class”), but say that the 7 were just wrong and that the rules were implicit. I’d say the rules were quite obviously contested. Definitions of the social situation often are (Ken Houghton — who is I believe an economist — recognizes this in his comment @1). The most interesting point in the actual experiment, to my read, is precisely that it is so very hard to produce an isolated one-shot PD in real life. Heck, even in an economics class, when one is proposed, most students embed that game in a larger frame. That’s really really interesting. But Hamermesh on the blog didn’t just ignore that. He wrote that they’d missed the point — which in the next graf he said was that collusion is hard to maintain (so they’d presumably best not try?). I understand that in a class, that is a fine lead-in to a discussion of nash equilibrium and may be fine pedagogy. But I still think that as regards the real world, it would have been good to notice how artificial nash equilibrium is. And I actually do hope that in class, he does actually tell that to his students.

      So the (intended) joke was that we now know where the Marwell and Ames result comes from (the inability to collude – or to cooperate – is learned in a particular way by economists).

      And the point was that collective action and PD resolution is in fact possible and really really important, and that the presumption that it isn’t can drive lots of bad policy.

      And yes, I realize that whether or not cooperation/collusion are socially optimal depends on the real world. Health insurance companies colluding to keep premiums high = not so good. Nation states colluding to punish greenhouse gas emitters = good.

      I am sorry that at least to you the humor did not come off. A little shallow was intended. I thought it was funnier that way. Spite was definitely not intended.


    3. Thanks for commenting, Token. I think this exchange is a good illustration of how people with different sensibilities can see the same comments as either a reasonable comment on an issue or outrageous/hurtful personal or otherwise biased attacks.


    4. Methinks thou dost protest too much! While a bit snarky – as Scatterplot tends to be in many domains – it’s hardly the outrageous ad hominem attack you frame it as. Further, the idea that students know that “pleasing the teacher” is part of the game is quite external to the game! By analogy, if I can trick my students into administering electric shocks to one another to please me, that doesn’t demonstrate that people rationally administer shocks to one another.

      Shamus’s suggestion of cooperation as opposed to collusion is not just terminological – it’s about framing the behavior with positive or negative valence. If it’s collusion, then making it rational requires the iteration you discuss. If cooperation, then making it rational just requires a cooperative impulse or desire.


      1. My point about “pleasing the teacher” is that it is no more external than currying favor with one’s classmates. If the students are assumed to understand the latter, then why not the former?


      2. tokeneconomist: it’s not (necessarily) about “currying favor” with classmates – it’s about cooperating with them! (Thanks, shakha, for the insight.) And the claim isn’t that students don’t understand either of these, but that if we have to posit either of them as motivations then we can’t use the outcome as evidence that the behavior is rational or normal outside the classroom context. Hence my electric shocks example.


      3. Josh’s original claim was that by having the class applaud the student who did not cooperate, Hamermesh changed the rules of the game after the fact. My point is that it’s reasonable to believe that students understood in advance that giving the answer the teacher was looking for would also have implications for the game. I wasn’t proposing either of these as motivations, only claiming that if Josh assumes one, it’s not reasonable to claim that the other constitutes changing the rules ex post.

        As for what any of these imply about the interpretation of the results of the game, I don’t agree with your claim that “if we have to posit either of them as motivations then we can’t use the outcome as evidence that the behavior is rational or normal outside the classroom context.” Rather, I think that one implication is that there are often unstated additional costs or benefits to actions. These costs or benefits may be imposed by social norms. Individuals take these into account when making decisions, so when these additional costs or benefits — or the possibility they exist — are not included in our models, then the predictions of our models will not hold in the real world. Indeed, this is my area of research. I am in complete agreement with the idea that payoffs are not simply monetary and that they may have a social component. I even agree with Josh’s original claim that the payoffs in Hamermesh’s demonstration were not as stated. I don’t agree, though, that the students obviously did not understand the true payoffs in advance, and as I think I’ve made abundantly clear, I don’t think the tone of the original post or Josh’s early comments was at all justified.


      4. tokeneconomist, I agree with you here:

        “…I think that one implication is that there are often unstated additional costs or benefits to actions. These costs or benefits may be imposed by social norms…” (and by various other things, etc.)

        But, I think it provokes a question: when (if ever) would you choose to sacrifice the rationality assumption rather than the criticizing the mapping of observable rewards into utility payoffs?

        And – does a great willingness to extend the categories of utility-providing sources not seem to you to threaten the normative status of utility maximization in economics? Sam Bowles makes a similar point in his behavioral Microeconomics text, arguing briefly that utility cannot at once represent something normatively desirable and something which is, in general practice, maximized. That is: if upon finding our predictions do not hold in reality, our first inclination is to assume our payoff representations are wrong and not that our rationality assumption is mistaken, does that not make you worry that what must be meant by “utility” is no longer something which people would recognize as good?


      5. rabidaltruism, my take on that would be that there are differences between behavioral explanations and behavior that is consistent with standard rational choice when constraints and payoffs are accounted for properly. One example of the difference would be the concept of time-inconsistent preferences (hyperbolic discounting). Sustenance constraints may lead to spending patterns that would also be consistent with hyperbolic discount rates if those constraints were ignored. I think there’s room for both sorts of explanations, and laboratory and field experiments are being designed to tease out these subtle but important differences. One reason understanding which explanation fits better is that they often have different policy prescriptions (e.g. commitment savings accounts are good tools for (sophisticated) hyperbolic discounters, but do not help smooth consumption in the presence of sustenance constraints.)


      6. Are you familiar at all with the Matching Law (, tokeneconomist? It’s an asymptotic/equilibrium description of behavior drawn from behavioral psychology, and has an interesting history in relation to hyperbolic discounting (hyperbolic discounting can be derived from the Matching Law, and I believe hyperbolic discounting was first popularized, if not discovered, by researchers working on the Matching Law, though I’m not certain about that latter claim).

        I’m not familiar with sustenance constraint explanations; I’ll check those out.


  7. Though I don’t hold a PhD in any discipline, I’ve been an on-and-off reader of scatterplot for quite a long time, and deeply invested in cross-disciplinary economics (behavioral/psychological, agent-based computational/complexity, sociological, anthropological) for a fair while longer than that.

    Both perspectives on the hamermesh teaching seem to offer some valuable insights into PD games generally (as tokeneconomist points out, I wouldn’t expect a single-play PD to be used to emphasize the existence of multiple equilibria – a Stag Hunt may be more appropriate) and into how economics students may come to perform exceptionally in public goods games, although I think also some rather lofty expectations are made of the students in a basic game theory class (we may want to explain Nash equilibrium solidly before engaging classes over what a game’s payoffs represent, whether von Neumann-Morgenstern utilities, dollars, utility derived specifically from satisfying social norms, etc.!).

    Unfortunately, I think the primary message in this set of replies is that genuine animosity exists between economists and sociologists – and, in fact, that belittling treatment of folks working in the other discipline has become so commonplace that it can even be seen as a non-threatening and acceptable tone to adopt. I would be remiss in laying this claim at the feet of sociologists alone (although I think tokeneconomist was quite civil and engaging, his/her open attitude to other disciplinary perspectives is rather alien in my experience of economists), but I think this issue has more depth to it than a simple difference of perspective; I think tokeneconomist was picking up on a very real, rather generally belligerent tone in sociological critiques of economics/economists, and it would be a mistake to write this incident off as simply different interpretations of the original post.


  8. “I would be remiss in laying this claim at the feet of sociologists alone.”

    Right. I did *not* make a belligerent critique of the *discipline* of economics. I did make a joke that referenced a famous paper that, at least among the now-professional economists I know from my days taking grad classes in econ at Wisconsin is one that they find kind of funny/intriguing (yes, a sociologist who knows some actual formal micro!). But I didn’t just make a joke. I used it to attack a position that is taken by *some* economists (and some non-economists). And I attacked it because it is a position that has political implications — implications that manifest, for example, in the final chapter of superfreakonomics with its presumption that collective action is all but impossible as a justification for geo-engineering — and because I thought Hamermesh was claiming empirical support for that position that was not in fact justified by his result. It is just flat out wrong to ascribe what I wrote to disciplinary warfare — unless of course you want to kick Robert Murphy out of economics.

    Murphy posted a couple of days after me, with essentially the same reasoning as me, and highlighted exactly the same passages as me. This is not because Murphy is plagiarizing me (or Silas Barta, who posted some hours after me with again similar reasoning). I seriously doubt he chanced across scatterplot. It is because I took a position that is both reasonable and entirely consistent with understandings of game theory held by many economists. I did give my post a stronger sociological spin around norms and their enforcement — which is a disciplinary contribution — but the main point here is that Murphy (econ), Barta (mechanical engineering I think), and Whitford (sociology) all have the same basic problem with Hamermesh’s take on his experiment. The fact is, Hamermesh either reasoned badly, or he moralized through the back door (or both).

    If Hamermesh had framed his story about how to get students to understand nash equilibrium, I would not have said boo. But then he would have been posting on an econ pedagogy blog. Instead he posted on a blog that is supposedly about the “science of everyday life” and he argued that the science of everyday life says something that it does not say. Yes, I know, somebody said something wrong on the internet! Oh dear. But still, part of what blogs do is to point out when that happens in a very public place. And on scatterplot, presumably one does so especially when sociology has something to say on the matter. So that’s what I did. I neither kicked off nor contributed to the disciplinary war. Or at least I didn’t among those with a sense of humor.


  9. Hi, Rabid, if your comment was aimed at me, I actually agree with you about cross-disciplinary hostility. I wrote my comment quickly at the same time jdw was writing, so my response was written before, not after, that one, although it got posted after. I don’t think differences of perspective are “simple” and did not use or imply that word, as you’ll see if you look again at what I wrote. If you interpreted the brevity of my remark to mean that I think differences of perspective are something superficial we can all work out in 10 minutes over a beer, you are wrong.


  10. jdw –

    I agree that your critique was substantive (hence that Murphy’s is, similarly), and that Hamermesh could certainly dig a few more layers off the onion of Nash ea quilibrium in PD games if he had been more nuanced in proclaiming the one defector was alone in understanding the game.

    My thinking there is an underlying hostility that tokeneconomist rightly picked up on is not meant to cheapen your insights; to the contrary, I meant – and perhaps I failed here – to express a sense of mourning that the various quite productive points delivered by both sides seemed to take a back seat to blame and aggression.

    OW –

    What I meant to express is that I don’t think the dispute is a matter of perspective; I think there is an underlying hostility in the delivery of the original post and the first few replies, and that tokeneconomist correctly detected that hostility. I also think that underlying tone is one rather mild example of a more general, intense phenomenon between soc and econ (even more fiery in anthro circles, actually!), at least as played out in the blogosphere, and that it’s something worth lamenting.

    Honestly, I’m a bit surprised that this is the first such incident I’ve encountered on the soc/anthro blogs/listservs I read.


  11. p.s.

    I was actually rather excited to see jdw provoke this reply from tokeneconomist earlier:

    “Whether collusion or defection is socially optimal is not as simple as you claim, either. It depends on the objective function…”

    But disappointed that the conversation did not return to it. How much information can safely be stowed away in behaviorally unclear utility (objective) functions is a topic of much interest to me; certain classical formulations of game theory actually *define* utility as something which humans through their behavior maximize (under which interpretation it could not be the case that Nash equilibria observed behaviorally in PD games being played are suboptimal!), but it is clear we lose much in the way of testability by adopting that approach, and it seems to me we avoid the hard problem of figuring out how to measure phenomena like satisfaction/happiness/”social optimality” in adopting such an approach.

    This issue bears itself out in tokeneconomist’s later claim that “The things outside the game that influence the game do so by changing the payoffs. In other words, the payoff isn’t actually (20,0) versus some equal split.”

    i.e., if we do find that players do not defect, we may either conclude that: A) they weren’t *really* playing a Prisoner’s Dilemma, or B) they weren’t pay-off (utility) maximizing, and a common choice in game theory is to adopt position A. I think it’s a rather fascinating question which of these choices is the better one for purposes of improving theory.


    1. To build on that post a bit, here’s a problem I think defending classical game theory by claiming that, if people cooperative, then they are clearly not actually playing a Prisoner’s Dilemma (once you modify their payoffs to reflect payoffs from social norms):

      One of very important uses of the utility-maximizing assumption in classical economics is that it appears as if utility-maximization is *good*. That is, if we can improve everyone’s utility over their current levels, or improve some folks’ without decreasing others’, then we’ve effected a desirable outcome. So, if more utility is good, then socially desirable outcomes can be identified, in a weak way, as those for which at least some people improve their utility accounting with no one sacrificing utility.

      But, if we go as far as to adopt utility-maximization as an axiom – as is implicit in interpreting rampant cooperation in an apparent Prisoner’s Dilemma to mean that a Prisoner’s Dilemma was not actually being played – then it is no longer clear whether utility-maximization is at all a good thing. This obstacle seems like a large one, to me, in approaches to understanding observations in experimental game theory.


  12. Um, Rabid, the main thread is assumptions in models, which is cool. But on the side thread, I don’t know why you would imagine that “perspective” and “hostility’ are somehow mutually exclusive sets. I confess to feeling my hostility rise at what seems to me to be a case of willful obtuseness, although it is probably just your blindness due to your perspective. As someone who has published rational action models in sociology, I am well aware of the overt hostility such work evokes, and have personally been the target of what have seemed to me to be insane attacks that had nothing to do with anything I actually said. At the same time, a lot of the hostility directed by sociologists toward economists seems to me to be justified by the behavior of many economists.

    And if it has never occurred to you that perspectives involve hostility, maybe you need to read more sociology.

    Edit: The point is that I felt insulted by your comment, and you don’t even know why I thought it was insulting. That’s what perspective is about.l


    1. I admit that I do not know what about the way I expressed my disagreement was offensive. I could understand certainly how the content of my message might offend, but I’m not sure there’s much to be done about that – altering it to be inoffensive would, I think, mean misrepresenting my position.

      I’m certainly open to claims that hostility between the disciplines is justified, or – more to the point, in my mind – productive. However, I took “this exchange is a good illustration of how people with different sensibilities can see…either a reasonable comment on an issue or outrageous/hurtful…attacks” to mean you thought there was no underlying, objective reality to the claim that the original post/few replies were insulting – which is of course a very different argument from acknowledging that hostility and defending it as justified.

      As I disagree with that argument, I expressed as much; it’s certainly possible that I misunderstood your original post, and would be happy to apologize if so, but I am not certain whether that’s the case.


  13. It’s interesting to see the way this conversation has developed.

    First, some technical points. Just as much as social sanctions from peers could be part of the game not included in the monetary payoffs defined by Hammermesh, so could sanctions or rewards from the professor. If you assume that the seven students who chose to collude understood that their peers would sanction them for defecting, thus reducing the value of defecting below the monetary payoff, it’s also reasonable to consider that maybe they also understood the professor would reward students for behaving as economists predict. The idea that giving teachers the answer they want is rewarded is probably something students have encountered before they get to college! If the students understood that the payoff to either decision had three components (money, sanction from peers, and sanction from professor), then the seven students who colluded might have done so EITHER because they valued some of the components differently than the others — for example, cared more about their reputations with peers or less about their reputation with the professor — OR because they didn’t understand which rewards and sanctions accompanied which strategies. The students could have come to different decisions even fully understanding that there would be sanctions from their peers and their faculty if they valued those payoffs differently, or if they understood the game differently. There is nothing about the situation that Hammermesh describes that makes it obvious the students didn’t understand all of these potential components of the payoffs in advance.

    There’s also a much more charitable interpretation of the situation than you allow. Perhaps the eight participant — the girl who did not collude — was a particularly serious student, maybe a social outcast, a “nerd.” Maybe Hammermesh saw the class tide turn against this student after she was revealed to have screwed her classmates, so he tried to lessen this peer sanction by publicly supporting this student. We’ve all seen elementary school teachers encourage their students to clap for a classmate who is usually excluded. Maybe that’s what Hammermesh was doing. I don’t know. But you don’t either.

    As I’ve said, I find many of the posts I receive on Scatterplot. But I don’t buy the claim that you didn’t deliberately contribute to a disciplinary war by making generalizations about all economists with statements like “Freakeconomists do know that it is possible to change human behavior. And they do understand the power of social sanctions. They just don’t like sharing.”


  14. PS. I would have been much more sympathetic to your point if you’d made it without such righteous indignation and without attributing such negative intentions to Hammermesh. Further, it was your reply at 11:11 on Oct. 23 that I found most objectionable, and why I commented. You may not think that your comment contributed to “disciplinary wars” (again, your words, not mine), but I certainly read it as piling on to economists.


    1. I don’t understand the objection to that particular comment. Perhaps I’m being obtuse. The point was that: (1) situations are embedded. So it doesn’t make sense to think of 1-off games. They might be useful on a theoretical level, but on an empirical one they’re basically useless (as there are no real world conditions where they make sense). Your own reply (#14) begins with this very point. (2) Hammermesh seeks to *produce* an economic world, not reveal the world to be understood by these principles. By rewarding the behavior that is most consistent with the model – rather, than, say, what most people do – the exercise is a performative one. “This is how you should act.” As the reference to Marwell and Ames suggests, there is empirical evidence to suggest that just is what happens with economists. This hardly seems hostile.

      I think part of the problem is that the economic point of view and the sociological one argue on different terrains. The economic responses are theoretical: sociologists don’t get and are hostile to their theoretical models. The sociological responses are empirical: these models don’t explain real world behavior. I must confess to be hostile in this later instance. Data matter. This is the point of the thread and wider critiques of superfreakanomics. And of the free rider (see Marwell and Ames: there is little evidence for this phenomena, except for economists). And as the consequences are often very real and important, being critical isn’t just disciplinary quibbling. It’s important to do.


      1. The theoretical models are often not meant to literally explain or predict real world behavior. They are simplified and stylized — that’s the point of a model! They provide a benchmark against which real world behavior can be measured. “Relative to risk-neutral agents with perfect information and access to complete credit markets, people in this experiment are _____, ______, and ______.”


      2. One-off games are probably of less relative importance than repeated games in real life, but I’m not sure claims that they do not exist hold up under scrutiny, at least not without doing some rather serious gymnastics to produce iterated effects.

        Example from real life: choosing whether to pick up a hitch-hiker. Ignoring the improbable, you have little expectation of ever seeing the hitch-hiker again, and may conduct your choices in this situation as if you will never play this or any other game with this other player again, ever. What about using a one-off model in this situation is contrived?


      3. A bit more in defense of models: I really think it’s naive to object to models because they don’t predict real-world behavior, or to think that economists blindly expect them to! This, after all, is exactly the point of the first year of graduate training in economics. Budding economists learn the assumptions underlying the standard models, where the assumptions come from, how the assumptions drive the results of the models, which results go through when the assumptions are violated and which do not, how sensitive the results are to the assumptions, and how to predict or at least bound results with different or fewer assumptions. And then, a very large fraction of those economists go on to do empirical work, interpreting their results as providing information for or against those models and assumptions. There is a huge amount of empirical work in economics; for evidence of that, take a look at the articles in the most recent AER (

        A model helps us understand if X causes Y, or if Y actually causes X, or if some Z causes X and Y. It’s not a substitute for empirical work; it’s a guide in how to interpret that empirical work. Data matter to economists as well, and our use of models does not undermine that.


      4. On the hitch-hiker: choosing to pick up a hitch-hiker isn’t a good example of a PD, as the model is for a moment of interaction, wherein both actors have a simultaneous choice of how to act.* With the hitch-hiker scenario, the question is simply on the level of what the driver will do (granted, the hitchhiker can ask to be picked up or not — but if the hitchhiker doesn’t ask to be picked up, there is no choice for the driver to make). Put differently, it’s not clear what the payoffs in the interaction of “cooperate” vs. “defect” would be (or that they would be different — that if you both cooperated you’d both be better off). But to the direction question: why isn’t a one-off game? Because things outside the situation influence the decision. The assumption of the one-off game is that the decision can be explained as just a product of that interaction. And I would assert that it can’t.

        * This point about the simultaneous quality of the interaction is an important one — and I’m curious if it’s been taken up. Anyone know? It matters if the interaction is not simultaneous, because then you know how the other actor has acted before you do (or, you have to act before the other does).


      5. On models: I’m going to write a longer post on this. But the cliff notes version is this:

        “[Models] provide a benchmark against which real world behavior can be measured.” — This is the key difference. For me, the real behavior is the benchmark against with models are evaluated. This difference is not trivial.

        “There is a huge amount of empirical work in economics.” I agree. But drawing on the above, the place of that empirical work is different. My claim is not economists don’t care about the real world. It’s how the position above suggests care about it: in relationship to the model. And I care about it as a thing in itself. Models aren’t a benchmark for behavior; behavior is that which allows us to adjudicate competing models.


      6. Shakha –

        I didn’t mean the hitch-hiker example to illustrate a Prisoner’s Dilemma specifically. But, I think it certainly constitutes a game; it could be viewed as a trivial game in which the hitch-hiker plays no real role (such trivial games do exist, e.g. the dictator game —, or as a game in which the hitch-hiker’s actions are various forms of advertisement (leg out-stretched, colorful sign, thumb pointing down the road), though the appropriate payoffs to assign those various choices may not be particularly obvious. The driver’s available strategies would, of course, be to pick the hitch-hiker up (or not). The trivial game would be completely non-iterative (or “one-off”), as best I can tell; the game where the hitch-hikers options are recognized would be a game with a two-step structure – not quite one-off, but not iterated for any significant length either.

        Concerning simultaneity – what you express here:

        “It matters if the interaction is not simultaneous, because then you know how the other actor has acted before you do (or, you have to act before the other does).”

        Actually captures rather well what I understand to be the treatment of simultaneity in games; an emphasis on temporal sequence isn’t really built into game theory – what is incorporated is, as you put it, whether “you know how the other actor has acted before you do.” Various multi-step games with appropriate limitations on knowledge of other players’ past actions reduce themselves to “simultaneous” games for exactly that reason; iirc, a series of Yale lectures in game theory I worked through last year focused on this point.


      7. (I shouldn’t say “focused on”, really – they touched on that issue rather lightly, providing the explanation I’ve tried to paraphrase there)


      8. Ah – I should reply to this idea, too:

        “But to the direction question: why isn’t a one-off game? Because things outside the situation influence the decision. The assumption of the one-off game is that the decision can be explained as just a product of that interaction. And I would assert that it can’t.”

        One-off models do allow for history to be encoded in their structure; the payoffs and their sources and the rules of the game (and so the number of players, who plays, the strategies available to each player, etc.) both can be modeled to acknowledge relevant past experience.

        What is essential to the one-off model is not its failure to acknowledge the past so much as its arguing that a specific interaction can be analyzed without reference to expectations about future interactions between the parties in question. The sense of future is removed, but the sense of past can be preserved and incorporated in various ways; hence through payoffs/rules there is room for “embeddedness” in one-off models, and there appear to be real-life situations (e.g. the hitch-hiker I proposed before) where one-off (or very small) games are more appropriate than lengthy or indefinitely iterated/nested ones.


      9. You say, “For me, the real behavior is the benchmark against with models are evaluated.” I don’t think that makes a model that does not accurately predict real behavior a useless or bad model. We learn from the differences between the observed behavior and the predicted behavior. The point is not that people who behave differently than predicted by models are making “wrong” or “bad” choices. The point is that they are behaving differently than agents following a simplified decision making process with a precisely defined objective and limited, clearly stated constraints. We can learn — about objectives, constraints, and ways to alleviate those constraints — by studying the differences between predicted and observed behavior.


    2. “There’s also a much more charitable interpretation of the situation than you allow.”

      Yes, there is a much more charitable interpretation, which is that I was, you know, using a *joke* — you have heard of those, right? — to make a point in a context in which I was writing a follow-on to a post where I’d objected to the climate chapter in Superfreakonomics for misinterpreting the story of Cedars-sinai precisely because they seemed to willfully ignore the fact that compliance was ultimately gotten. And I made it “personal” around the character of Hamermesh because he was, for purposes of my narrative, the autonomous actor “getting” the class to applaud, and “blessing” the heart of the defector. The intimation is that she was the hard-headed one, the others the suckers. Mine is not a crazy view. Bob Murphy basically agrees with me — and arrived there independently. You can kick him to the sociologists “side” if you want. Note also that I wrote *freakonomists* [not economists] don’t like sharing. I didn’t take a stance in a disciplinary war. I took a stance in a war that cuts across, not between, disciplines. That I freely admit.

      I think it is simply wrong to say Hamermesh did not use his experiment to tell the world about how hard it is to get cooperation (I could care less about the class; the class for me, like for him, was a vehicle to a larger point). That was the intention I really ascribed to him. In a context in which there has been relentless snark *within* economics about Elinor Ostrom getting a nobel prize, and in which superfreakonomics has taken a misleading/dangerous position on multiply specious grounds that include not understanding how cooperation happens, I think it is important to make the point that his own experiment suggests that *most* people are cooperators (or can become cooperators), and to make the further point that he reacted in a way that arguably makes it less likely that they will cooperate again the next time around. Sure, cooperation might be bad in some cases, but let’s think about how/when it happens.

      re: other stuff. I’m not going to get into it. No time. Nobody disputes that one can interpret all the bits and pieces in some sort of giant supergame [intentionally intended redundancy]. You still have the problem that the actual result was *not* a nash equilibrium. You can do with that what you will (you might want to read Murphy; he is pretty funny thereon). What I do with it is what a lot of game theorists do, which is that I think of the PD and other games as useful metaphors that sometimes illuminate aspects of reality; but I don’t mistake them for reality. Experiments that take place in class, on the other hand, are bits of reality. And they should be interpreted as such.

      re: spelling. It’s Hamermesh, not Hammermesh.


Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: