Motivation

I’m interested in whether Kahan’s account of reciprocity can be scaled up to account for all compliance or non-compliance with authoritative laws generally, rather than just criminal prohibitions on street crime and tax law. Particularly, I’m interested in whether it might provide a public choice theoretic grounding of Joseph Raz’s account of authority. Raz’s Normal Justification Thesis states that “the normal way to establish that a person has authority over another person involves showing that the alleged subject is likely better to comply with reasons which apply to him (other than the alleged authoritative directives) if he accepts the directives of the alleged authority as authoritatively binding and tries to follow them, rather than by trying to follow the reasons which apply to him directly” (Raz, quoted in the Stanford Encyclopedia of Philosophy, “Authority”). As long as a citizen usually does a better job of acting according to the balance of (first-order) reasons (that is, accomplishing the best outcomes, all things considered) by following authoritative commands (e.g., laws) than by evaluating all of the reasons for herself, then she should follow that kind of command.

Let h0 be: Commands from institutions (legitimate or not) will add to or subtract from the balance of reasons, but will not be authoritative. (For instance, a penalty is a reason not to do something prohibited, to the extent that you anticipate being made to pay it, because you don’t want to pay it. Likewise, a rule in a public goods dilemma that everyone should contribute y=m of their endowment where m is some constant to which all players, or most players, agree might be a reason not to contribute y less than m, if the rule creates an [extra-legal] norm that everyone should contribute m.)

Let h1 be: In communities characterized by high social organization and high social influence, commands from legitimate institutions will be authoritative. (The first three terms are Kahan’s terminology, the last is Raz’s.) Otherwise, commands from institutions will simply add to or subtract from the balance of reasons.

[Note that there are a number of possible variations on this hypothesis: Kahan’s factors could be necessary but not sufficient conditions for authority, or a smaller number of them could be sufficient (e.g. legitimacy itself might be enough for authority), or factors could also be added “trust,” etc. Also, this hypothesis looks at the Kahanian factors as on/off, rather than as continuums, with might better match real world reactions to lawful commands.]

I think that h1 is plausible, and I’ll just very briefly sketch the reasons why I think this is:

High social organization means that community members develop norms and agreements that enable them to share baseline reasons (those that authority might override) that overlap to a large degree, meaning that most people who will be subject to authoritative commands have similar interests in the first place.

Legitimacy (in the Kahanian sense) means that the institutions issuing commands will, to a certain, unspecified extent mirror the consensus norms created (in part) through high social organization, meaning that the commands of these institutions should do a good job capturing the reasons that apply to subjects.

High social influence means that there will be strong non-legal reasons to follow community norms anyway, which reduces the likelihood of (purportedly) authoritative commands asking a subject to do something radically different from the action that a consideration of the balance of reasons in a particular case is likely to call for. (Given restrictions on authority that Raz is willing to accept [see the SEP, §3.1], this may not affect the authoritative status of particular commands as mean that more commands from an institution are authoritative.]

If these three considerations are necessary conditions for authoritative commands/laws, then it seems to follow that: (1) authoritative laws generally (maybe always) have low enforcement costs because of community buy-in, (2) laws are authoritative only when they do a good job mirroring what people really think (including their propensity for reciprocity), and (3) the penalties associated with authoritative laws will rarely need to be applied, and probably don’t need to be particularly big. This may also mean that authoritative commands only exist in highly developed legal systems.

If any of this is too technical to follow or opaque, I’m happy to elaborate further, though I could be wrong about all of this, in which case further elaboration may not make it any clearer.

Jonathan

In discussing the limitations of game experiments, Camerer and Fehr state: “Of course, games are reductions of social phenomena to something extremely simple, but they can always be made more complex. A painter who first sketches a line drawing on a blank canvas has reduced a complex image to two dimensions of space and color. But the line drawing is also a platform on which more complex images can be restored” (Camerer and Fehr, p. 85). I wonder if this is true.

I do not doubt that we can reduce a complex social phenomenon to a single transaction that can be modelled, create a simple language with which to explain this transaction, and later re-introduce complexity. What I am skeptical about is the suggestion that, at the end of all this, we can arrive at a more accurate account of the social phenomenon that we originally sought to understand. My doubt arises from two very different types of considerations.

First of all, it seems to me that this type of project can transform the subject it seeks to explain. In reducing a person to a game-playing agent (e.g., one who does not communicate with or have knowledge of others, and performs tasks divorced from the social context in which they are normally integrated), we might not reach a blank frame on which to build a realistic model of human agency. Rather, in creating a new baseline against which human action is measured and understood, we might end up with a model that does not support the re-introduction of the complexity that we originally stripped away.

At another level, I wonder if rules/norms actually govern human action. It seems to me that this assumption is a driving force behind the game-theory project. Ostrom, for example, suggests that learning heuristics, norms, and rules should occupy an important role in second-generation models of rationality. While I agree in part, I have doubts about whether human behavior is explained, or merely described, by rules—about whether rules are the causes, or merely reasons that we give, for what we do. If rules are best understood as descriptive, game experiments may not give rise to a “deeper” picture of the social world.

Neither of the two points I have briefly touched on here are criticisms of the use of game theory in the social sciences. I raise them merely as things to consider when thinking about the fascinating behavior that games uncover.

I am trying to square the invitation to reciprocate in, for example, the Gift Exchange Game described in Camerer and Fehr at p. 63, with the Crowding-Out theory we discussed last week. The basic question is when and how much especially generous wage offers to supply a given level of effort can crowd out natural (altruistic) motivations to supply that effort. I’ve thought of a variation on Jolls’s babysitter example to test one instinct.

Here is the arrangement: Parent offers babysitter wage w, which is roughly the lowest wage babysitter would accept to do the promised work. Additionally, parent offers to pay premium z upon return if babysitter agrees not to shirk his responsibilities. Parent conditions payment of z on reporting: parent will ask babysitter at the end of the night whether he shirked his responsibilities. If babysitter claims not to have shirked, then he will receive w + z, but if he admits to shirking, he will receive only w.

This arrangement is very similar to Jolls’s example, which David helpfully describes below. In both examples, the parent offers a “fair” premium for nothing more than the services sought and available at w. In both cases, parents rely on reciprocity norms and trust where monitoring is infeasible.

Does it make any difference that in my hypothetical the premium is candidly described as payment for doing precisely what the babysitter would already do for w? Does the end-of-night inquisition change things? Perhaps Frey and Jegen would say yes: Babysitter could shirk at w or w + z without penalty, and thus the decision to do the promised level of work is, in homo economicus’s, mind, altruistic and irrational. Offering w + z signals trust and invites reciprocation for fairness sake, but perhaps offering premium z only after inquisition crowds-out whatever motivation might come from fairness concerns. In my example, z may be like payment for donating blood: some (most?) babysitters are naturally inclined to provide the promised level of service regardless of payment.

If crowding-out is likely in my hypothetical, what stops it from occurring in Jolls’s? Perhaps Deci and Ryan can provide some guidance–Is it the sense of competence and autonomy that comes with a higher wage and no monitoring? Does trust fulfill our basic need for a sense relatedness, which is injured by the inquisition?

This isn’t meant to be an argument, since I’m not sure whether my hypothetical changes things at all. But the relationship between premium wages for services and crowding-out is at least worth pondering.

Kahan argues that, because of the multipile equilibria and hetergoneity of actors in collective action problems, maximum cooperation “probably requires that reciprocity dynamics be supplemented with appropriately tailored incentives, most likely in the form of penalties aimed specifically at persistent free-riders.” (p. 9).

However, some sticky policy issues may have so many intertwined moving parts that it is impossible to segregate and target “dedicated free-riders.” Any external incentives put into play then risk also hitting more reciprocal-minded actors and crowding out reciprocal norms.

So, what to do when no single policy prescription is “smart” enough to hit only one set of actors within a given collective action disposition but not the others? I suggest two alternatives. One is the default option: offer no external incentive and let the reciprocity norms settle on an equilibrium. The second is to shoot for the middle – i.e., design an incentive to push “neutral reciprocators” to cooperate. They represent the middle, and perhaps largest, block in Kahan’s model of heterogeneity. (Figure 3, p. 8). Incentives designed for this group will not be strong to push towards 100% cooperation because of more dedicated free-riders, and they may crowd out some of the reciprocal motivations of more natural cooperators. But overall, an effective design likely will settle on an equilibrium north of 50% — perhaps 60-80% cooperation.

Ultimately, choosing between these two alternatives will depend on the natural default’s equilibrium point and a host of other factors, such as the degree of monitoring available, as in Jolls’s examples.

This is the first time I have made an Internet blog post, so proceed with caution:

Professor Jolls does a great job of discussing what she describes as the “fairness dynamic” and its relation to the way employers and employees decide wages. At the beginning of her article, she tells a story of how one of her acquaintances chose a pay level for a babysitter. As soon as I read this, I couldn’t help thinking of analyzing the rest of the contrast in light of the (admittedly small amount of) traditional economics knowledge that I have.

Professor Jolls’ friend paid more than the reservation wage because “the sitter [was] going to be in [her] house alone caring for [her family’s] children for a significant amount of time, and [they] just didn’t feel [they] should negotiate for a lower level of pay even though [they] believe [they] could have gotten one.” Professor Jolls’ article suggests that the babysitter was paid “fairly” “in order to encourage diligence and hard work on the employees’ part.” Yet if employee-employer interactions either are repeat games or involve some level of negotiation, traditional economics alone would tell us that the marginal increase above the reservation wage is a cost of higher service.

Professor Jolls nicely shows us that the situation is a bit more complicated. A parent may be willing to pay a babysitter beyond her reservation wage for a variety of reasons:
(A) the babysitter is offering two levels of service: basic babysitting services offered at w, the reservation wage rate; and premium babysitting services offered at w+z, where z is the marginal rate above the reservation wage that will induce the babysitter to offer the higher quality care. Offering the services at a rate of w is certainly more enjoyable for the babysitter, as she will probably spend more of her time watching TV, talking on the phone, or engaging in some other activity instead of caring for the baby. The babysitter is willing to avoid engaging in these activities if the price is right — if she is paid an extra z per hour.
(B) The parent believes that a higher wage will increase the morale of the babysitter, and this in turn will increase her performance. See 77 N.Y.U.L. Rev. 47, 52-3.

Babysitters would be a perfect experimental study group to see how behavior changes at different levels of pay and monitoring. I would be curious to see how the quality of babysitting services differs among the following situations (if possible, legally and otherwise, the quality of the services provided by the babysitters would be monitored in all situations without them knowing):
(1) the babysitter is getting paid at the reservation wage w, and nobody is ever home to monitor the babysitter’s quality of work
(2) the babysitter is getting paid at the reservation wage w, and one of the parents is always home when the babysitter is taking care of the infant. The parent, to one’s dismay, does not help raise the child even though he or she is home.
(3) the babysitter is getting paid at the reservation wage w+z, and nobody is ever home to monitor the babysitter’s quality of work
(4) the babysitter is getting paid at the reservation wage w+z, and the same parent as in (2) monitors the babysitter’s quality of work
(5) the babysitter is getting paid at the wage rate w, and both parents stop by the house at various times of the day for different lengths of time. Although there are times when the parents are not home for hours at a time, the babysitter is unable to accurately predict when the parents will be home. This would be an attempt to “maintain[] the right level of terror”, a practice opposed by Professor Jolls for child care work. 77 N.Y.U.L. Rev. 47, 64.
(6) the babysitter is getting paid at the wage rate w+z, and both parents stop by the house at various times of the day for different lengths of time. Although there are times when the parents are not home for hours at a time, the babysitter is unable to accurately predict when the parents will be home. This would be an attempt to “maintain[] the right level of terror”, a practice opposed by Professor Jolls for child care work,

And most importantly,

(7) The babysitter gets a base wage rate of w every week. At the beginning of each week, a coin is flipped determining whether the babysitter gets a bump in salary. If the coin is heads, the babysitters will get the additional wage 2z, and if the coin is tails, the babysitter will not get any wage above the reservation rate z. The numbers are designed so as to maintain an expected value of w+z for the salary. Nobody is at home while the babysitter performs her job.

(8) The babysitter gets a base wage rate of w every week. At the beginning of each week, a coin is flipped determining whether the babysitter gets a bump in salary. If the coin is heads, the babysitters will get the additional wage 2z, and if the coin is tails, the babysitter will not get any wage above the reservation rate z. The numbers are designed so as to maintain an expected value of w+z for the salary. One of the parents is always home while the babysitter performs her job.

I’d be fascinated to know the results of these experiments, especially those of (7) and (8)

This is the blog for the seminar on “Motivation” offered at Harvard Law School in 2008. We’ve never used this learning technique (or this particular software) before, so there undoubtedly be some experimentation and adjustment as we go along. Feel free to make substantive comments on any of the course readings or to contribute reactions to any of the comments made in class. It’s important to be honest and concise — and, of course, to treat one another with respect.

Terry