More Money, More Problems?\\ On the Repugnance of Material Incentives Sandro Ambuehl,\; Muriel Niederle, \; Alvin E. Roth Keywords: \\ JEL Codes: Introduction A model of moral assessments of voluntary transactions Motivation and Disclaimer This is a purely descriptive model, not a normative one. The model is non-circular. It reduces the question of whether a transaction is repugnant to a question of technological feasibility. This model is not intended to be an encompassing model of how people make moral assessments. Instead it is to clarify the forces behind the one single moral idea that high material incentives can be coercive, and to derive the implications of these forces. Other considerations may very well play a role too. There are two other moral arguments that this model explicitly abstracts from: (i) Crowding out (of intrinsic incentives) (ii) issues related to inequality. The statements in my model apply even in a world with a perfectly equitable, but possibly heterogenous population (in which, e.g., everyone pays the same small fee for a kidney fund, which is used to induce people with the smallest preference for holding on to their kidney to sell it at a price.) Domains the model does capture Organ sales Participation in medical experiments Prostitution Indentured servitude Dwarf tossing Gray area Surrogate motherhood Selling dead body to medical institute. This depends on the view whether the dead body should be assessed similarly to a living body (for those people, the model applies) or whether the dead body is just a lifeless piece of material (for those people, the model does not apply). Domains the model does not capture Things where the structure ``$A$ accepts, $B$ thinks $A$ should never have been offered the transaction'' is not given. (e.g. futures market in terrorist events, life insurance on third parties ) Setup There are two goods, money and health. Each agent $i$ makes tradeoffs according to U(m,x) = _i u_x(x) + u_m(m) The distribution of $ _i$ in the population follows C.D.F. $F$. Each agent $i$ has initial endowment $(x_i, m_i)$. Both $u_x$ and $u_m$ are strictly increasing, but we explicitly note that $u_m$ may converge towards an upper bound. Economically, it may be impossible to compensate sufficiently large health losses with money. (Note: This is what the replaceability assumption captures.) A buyer looks for one seller of the good. The buyer makes a take-it-or-leave-it offer; he will pay $m_t$ for $x_t$ units of the good. The seller observes an unbiased noisy signal of his preference parameter. We write $d_i = _i - _i$ for the difference between sellers' best estimate of his preference parameter (determines behavior), and his true preference parameter (determines beliefs). Justification for a imperfect information about the noisy signal about the preference parameter derives from the fact that the seller decisions is determined by his beliefs about how he will feel upon taking part in the transaction, and these beliefs may be less than perfectly accurate. The parameter $ $ captures how noisy the signal about the preferences is. For the moment, we set $ = 0$. An observer observes the terms of the transaction and the excess demand for the offer made. If there is positive excess demand, assignment is random. From his observation, he forms beliefs $ _s$ about the seller's preferences. For simplicity we assume that the observer knows his preferences perfectly, although the model would not be substantially changed by assuming otherwise. (Note: selection + noise already gives a mechanism for welfare-decreasing choices. Hence, I wouldn't even have to exogenously assume partial perspective taking. In this case, however, it wouldn't be possible to address the IRB argument ``even when able to consent'') The observer judges how moral it is to offer the transaction $(x_t, m_t)$ to a seller with preferences $ _s$ and endowment $(x_s, m_s)$ . The observer's own endowment is $(x_o, m_o)$. m(x_t, m_t; s) &=& a( ) u_x (x( ) - x_t ) + u_m (m( ) + m_t ) where $ (0,1)$ is the perspective-taking parameter, and $a( ) = a_s + (1- ) a_o$, $x( ) = x_s + (1- ) x_o $, $m( ) = m_s + (1- ) m_o$. When the observer judges offering the transaction to a pool of potential sellers, he judges the morality of the transaction by $M(x_t,m_t) = E $ where $w$ is strictly increasing and weakly concave (allowing the seller to disproportionately weight the less well-off sellers). We call a transaction repugnant if $M(x_t, m_t) - M(0,0) < 0$. Importantly, rather than inferring the seller's welfare from his actions, the observer considers the outcome of the transaction (the several dimensions) and uses (partially) his own endowment / preferences to assess whether this transaction makes the seller better or worse off. There are several possible reasons for partial perspective taking: (i) Paternalism (the observer intentionally says that the seller's decision is bad for him, even though the seller thinks otherwise) (ii) Inability to fully appreciate the seller's perspective. Implications Indifferent sellers We first consider the case in which the transaction is such that the seller is just indifferent between accepting the offer and rejecting it. The observer judges this by seeing that there is no excess demand for the transaction. The observer judges a transaction as Repugnant if $M(x_t, m_t) - M(0,0) < 0$ Coercive if it is repugnant and $m_t>0$. (Immediate implications) An observer will never judge a transaction as repugnant if he would accept it himself An observer may judge a transaction as not repugnant even if he would reject it himself CONJECTURE: If the observer is able to fully take perspective regarding $ $, but not regarding money, then a transaction $(x_t, m_t)$ is judged as coercive only if the transaction $(x_t, 0)$ would be rejected. Immediate. (Comparative statics regarding the seller's and observer's material endowments.) Conditional on the assumption that a single seller accepts the transaction and there is no excess demand: Richer observers are more likely to judge the transaction as repugnant: $M$ is decreasing in $m_o$. If $u''_m / u'_m$ is non-increasing (e.g. CARA or CRRA), then the observer judges incentivizing a poorer seller as more repugnant than incentivizing a richer seller: $M$ is increasing in $m_s$. If healthier observers are more (less) likely to judge the transaction as repugnant ($M$ is de(in)creasing in $x_o$), then the observer judges incentivizing a more (less) healthy seller as more repugnant ($M$ is de(in)creasing in $x_s$). (Comparative statics regarding the properties of the transaction.) Conditional on the assumption that a single seller accepts the transaction and there is no excess demand, the larger $m_t$, the more likely the transaction is judged as repugnant. The above comparative statics is stronger, the richer the observer. In-kind transactions are judged as least repugnant. Conditional on a seller having accepted the transaction, increasing payment makes the transaction less likely to be judged as repugnant. The larger the excess demand for a transaction, the less likely it is to be judged as repugnant. (i) and (ii) capture the intuition that giving money to a seller is not per se repugnant, it is incentivizing the transaction that is repugnant. Example: Seller is indifferent between taking part or not in medical experiment $A$ for \$1,000, and also indifferent between taking part or not in medical experiment $B$ for \$10,000. Which of these is more repugnant? Intuition suggests the second, as predicted by the model. The result on in-kind transactions relates to the World Medical Association's Helsinki Declaration (1964) (very similar to current U.S. IRB): ``Medical research with a vulnerable group is only justified if the research is responsive to the health needs or priorities of this group and the research cannot be carried out in a non-vulnerable group. In addition, this group should stand to benefit from the knowledge, practices or interventions that result from the research.'' (Population-dependent comparative statics) Suppose there are two sellers with $x_s^1 = x_s^2 = x_s$, $ _s^1 = _s^2 = _s$, but $m_1^s < m_2^s$. Consider two payment schemes: (A) Offer the lowest uniform amount such that both sellers accept, (B) Offer the lowest amount that is proportional to the endowment such that both sellers accept. Payment scheme (B) is more repugnant than payment scheme (A). Fix the minimal transaction price $m_t$ that is required to induce both sellers to accept the transaction under payment scheme (A). Then maximally lower the payment for seller 1 such that he still accepts the transaction. Since this change does not affect incentives (both sellers still accept the transaction), but make seller 1 worse off, payment scheme (B) is more repugnant. (Ability to predict the consequences of the decision.) Suppose one seller is indifferent between accepting and rejecting the transaction, and that there is no excess demand. Now decrease $ _s$, and simultaneously decrease $ _ $ such that $ _ _s + (1 - _ ) _s$ stays constant. This change makes the transaction more repugnant. (Interpretation: The seller has the same ``true'' $ _s$ before and after the change, but after the change, he selected into the transaction by mistake) d &=& u_m' (m_s + m_t ) d \\ d &=& - u_x(x_s - x_t) - u_x(x_s) < 0 (The terms involving health remain unchanged, since the change in $ _ $ is chosen to balance out any change in $ $ in the observer's assessment.) This is negative due to $u_x'>0$. (Extensions) Intensive vs. extensive margin Ability to predict the consequences of the decision Under the assumption that bargaining brings each seller closer to his indifference point (whereas a fixed price leaves producer surplus for many suppliers) bargaining over the price is more repugnant than having an exogenously fixed price. Formally suppose that bargaining power is with the buyer, and that prices are just high enough that 10 sellers supply, at heterogenous marginal values of money (all of which are higher than judged by the observer). For bargaining, every seller is at the indifference point, and thus gets a deal that is judged to make him worse off. For the posted price, some sellers have a strictly positive consumer surplus, and hence are not judged as being made worse off (only the marginal suppliers are). Once we make this assumption, how does technological feasibility still matter? Replaceability matters because if the good sold is replaceable, the seller can replace the loss. Since the observer uses different weights to assess the transaction, this means that the maximal size of the loss the seller takes, once he realizes his mistake, is small. What the model so far doesn't explain Not paying at all is better than paying a petty amount Paying \$1,000 is better than paying \$50 Non-monotonicity Ideas Note: The comparative statics of the model crucially depend on whether increasing the payment is associated by an increase in the health supplied that keeps the seller just indifferent, or whether the increase is conditional on having accepted. The experiment does in no way control for that, and hence, the model can be made to fit the data by making the right assumptions on the observers' beliefs. Very low payment gets the really desperate people With the current design, we just ask about the invitation. We don't say the subject is on the fence / indifferent. Hence, we have two effects going on: (i) selection of people (ii) compensation of selected people. If the seller were indifferent in each conditions, then we should get the decrease in repugnance (effect (i)). If we allow people to have erroneous assessments about their preferences, then the really low payment gets only the most desperate subjects / those who most underestimate the value of health, and the higher payments get better calibrated people (but there would be excess demand). Very low payment is insulting (cannot be in model) Eating Olives etc. Suppose I don't like olives, but observe somebody eating olives. Would that be judged as repugnant? Observer's posterior expectation about the `sellers' preferences after eating olives: $E[ | 0]$. Observer's taste for Olives is negative: $ _o < 0$. The Observer will not judge eating Olives as repugnant if $ E[ | 0] + (1 - ) _o 0$, i.e. if $ $ is large enough. Suppose that and observer randomly drawn from the population observes the `seller' consume good $x$. Suppose $ $ is the same for all observers. Then The less mass the distribution $F$ has on non-negative $ $, the more likely an observer is to judge consuming good $x$ as repugnant (even conditional on $ _o < 0$). Intuition: Much mass on positive $ $: ``Many people like Olives. That's a normal preference even if I don't happen to have it. Little mass on positive $ $: ``Nobody else likes doing that. There must be something wrong with that guy. He shouldn't be doing it.'' Florian's comment: Why isn't a situation such as incentivizing someone to do work not equally explained by the model? Related to Olives: In some domains, observers might be more able or more willing to take the perspective of the other person. Alternative explanation for survey findings: Rich people think poor people make bad decisions, and are lured in by money; poor people don't think so. Connecting the model to the data Control tightly: Excess demand (i.e. is the wage changed given incentives, or is the wage changed to incentivize?) Income of the observer Income of the person who accepts the transaction For which goods are material incentives for provision repugnant? Conjecture: Incentivizing transactions is repugnant for goods for which money can compensate, but not replace a loss (necessary condition) decision to sell is permanent, cannot be undone (necessary condition) good sold is large, not gradual Example: Wage labor not repugnant indentured servitude dwarf tossing ? prostitution ? More repugnant to pay someone a lot of money to buy his house to tear it down and build a large development than to buy it and live in it. (Here, respect probably matters too. Similar to buying a kidney and using it as a soccer ball.) Older stuff What the model clarifies Makes sense of what philosophers and the IRB mean when they say that a large amount of money may be coercive. Larger benefits make choice less voluntary: If I observe somebody doing something without a lot of money, I infer that his loss on the relevant dimension must be small. Why is it viewed as less repugnant to accept a job in an iron ore mine than to sell a lobe of the lung? One answer: Selling a part of the lung has more irreplaceability - there's no way of getting it back. On the other hand, I can always quit the job in the ore mine. (Needs to be more specific.) Why offering a very large payment to somebody to leave a house that he has an emotional attachment to is less bad than offering the same large payment to somebody to sell his kidney. Open questions What are the dimensions? Possible (cheap) way out: As models of salience etc., say the model applies once dimensions are given. (Even the standard general equilibrium model doesn't say what one good is.) This is enough to usefully apply to model to predict which kinds of transactions that affect only a defined set of dimensions will be judged as more or less repugnant. Why do the moral considerations apply to some dimensions but not to others? I.e. why do I care about someone's level of health vs. money, but not about the number of someone's Rolex watches vs. the number of his Mercedes cars? Because the latter are replaceable, even if I may not choose to do so. (So the same outcome is not repugnant if it is chosen but replaceable, but is repugnant if it is chosen but irreplaceable. To that extent, the model seems to contain some reduced-form assumption about wanting to change your mind later. Also, a transaction that makes someone accept a big financial wealth risk may be judged as repugnant if the amount lost cannot be earned back. BUT: then, all is within the money domain!) Examples of morally reprehensible incentives: Sandel Cash for sterilization; cast for intrauterine devices for HIV positive women Sandel: Bribery Objection: Corruption consists in buying and selling something that should not be up for sale. [...] We corrupt a good, an activity, or a social practice whenever we treat it according to a lower norm than is appropriate to it. [...] Should we regard our bodies as possessions that we own and can use and dispose of as we please, or do some uses of our bodies amount to self-degradation? My model captures and generalizes the self-degradation objection. Portrayal of economists: ``In all domains of life, human behavior can be explained by assuming that people decide what to do by weighing the costs and benefits of the options before them, and choosing the one they believe will give them the greatest welfare, or utility. If this idea is right, then everything has its price.'' Of Sandel's points, my model does not capture (but these things are mostly about crowding out): Paying kids for good grades Paying people to take healthy behaviors Tradable procreation permits This is mainly about inequality But: ``parents who want an extra child must induce or entice other prospective parents to sell off their right to have a child.'' This is captured by the model (assuming that those who sell their right to a child can, only with difficulty, earn back that right). Also: ``Central to the norm of parental love is the idea that one�s children are inalienable; it is unthinkable to put them up for sale.'' This is precisely about the compensation vs. replacement argument. Life insurance on third parties (this is outside the domain of $A$ accepts a transaction, and $B$ thinks he should never have been offered the transaction.) Literature Review Medical Research Guidelines India: p.59 EU: Article 12 Australia: 2.2.9 and 2.2.10 Kenya: Point 8) on p. 12 of the guidelines South Africa: p.89 Repugnance Roth Nussbaum on Objectification Nussbaum on Prostitution Krawiec papers Satz book Baron Spranca 1997, ``protected values'' Andre, 1992, ``blocked exchanges'' (cited in Baron Spranca) ``What others consider fair'' literature Kahnemann, Knetsch, Thaler AER Dual entitlement theory: Transactors have entitlement to terms of reference transaction and firms are entitled to reference profits Reference transaction = (reference price or wage, reference profit for the firm) basis for judgment because it's normal, not because it's just firm not entitled to arbitrarily threaten the transactor's reference may not decrease wage because unemployment higher entitlement does not carry over to new transaction if old employee leaves, can hire new at lower wage if line of business changed, can change wage for existing employees rules of fairness permit a firm not to share in the losses that it imposes on its transactors, without imposing on it a duty to share its gains when reference point of firm is threatened, it may set new terms to protect reference profit at cost of buyer when profits increase, firm is not required to pass that on to consumer Coding of outcomes action by firm more likely judged unfair if it causes loss than if it cancels possible gain action by firm more likely judged unfair if it causes gain to firm than if it averts loss costs are assigned to specific goods. Hence unfair to increase price of old stock when the price of new stock increases. Occasions for price changes profit reductions firm can pass entire cost increase to consumers (no sharing of pain needed!) profit increases firm not required to pass cost savings to consumers increases in market power when blizzard increases surplus from snow shovel, firm cannot take any part of that surplus, since that would violate the consumer's reference price unfair for firm to take advantage of increased monopoly power action that deliberately exploits the special dependence of a particular individual is exceptionally offensive. introduction of explicit auction would also enable the firm to gain at the expense of its transactors, and is consequently unfair. (exception: proceeds to charity) goods for which active resale market exists can be auctioned. (reason: increase in price offset by increase in resale price if plan is to resell) Kahnemann, Knetsch, Thaler, Journal of Business relies on same data and essentially replicates AER paper Knobe et al. Experimental Philosophy Agent is more likely to be judged as the cause of an event when the event is previously judged to be bad. Examples Chairman example Professor and student taking pens example Theories Distortion theories Competence theories Objectivism vs. relativism Method: Two people disagree, respondent selects who's right, or "it depends" Children are objectivist, 20-30 relativist, old people objectivist if people with disagreement in the vignette come from different cultures, more people choose "it depends" Money Vohs et al., 2006: Money prime makes people more self-sufficient. Reduces giving and asking for help, makes people prefer to work and play alone, makes people put more physical distance between self and new acquaintance Heyman, Ariely: two markets that determine the relation between effort and payment, monetary and social. Former are sensitive to the magnitude of compensation, latter are not. (This is about the less relevant crowding-out hypothesis.) McGraw, Tetlock (2005): Four different ways in which social relationships are structured: 1. Communal sharing (e.g. within close relationship) 2. Equality matching (tit-for-tat relationships) 3. Authority ranking (e.g. military) 4. Market pricing. Studies: (i) Pen for sale. Pen was either bought on market, received as a gift to signify friendship (or two other treatments). Selling pen is repugnant in second condition. (ii) Sell watch to either regular customer or close family friend (or two other treatments). People think sale should occur at lower price in the latter case. (iii) Exchange with roommate for not having to bring out trash. Paying \$15 per month is regarded as weird. Paying the electrical bill of \$15 each month is acceptable. (iv) President either sells night in Lincoln bedroom for \$250,000, or reciprocates to large campaign backers with letting them stay in the Lincoln bedroom for a night. Former is more repugnant (but possibly not only because it involves money, but because the terms of trade are much more explicit than in the latter scenario). Autonomy literature Bartling, Herz, Fehr, EMA forthcoming Subjects intrinsically value having decision rights (above the material benefits thereof) Fehr, Herz, Wilkening, Lure of Authority Social preferences literature Coffman: Bartling, Fischbacher, ``Delegation and Responsibility": Dictator game with two possible choices, can delegate choice to dictator with same incentives, third party can punish. Responsibility is shifted, and dictators delegate in order to avoid responsibility. Eisenkopf, Fischbacher, ``Doing well by doing good'': Trust game with two trustors, for which giving has different multiplication rates. Trustor 1 can delegate to trustor 2. Finding: Trustees reward the trustor who actually made the transfer more handsomely. Delegation only pays for trustor 1 if trustor 2 has a much higher multiplication rate. Paharia et al.: Similar to the delegation papers, but with psychology methodology Fehr, Fischbacher: Dictator game with costly third-party punishment. Third parties punish, and do so the more the fairness norm is violated. Chavez, Bicchieri, 2013: Less closely related Falk's mouse killing paper Killing of mouse is material externality on mouse Bartling, Weber ``Do markets erode social responsibility?''. In markets Lying Ariely, Mazar, ``Dishonesty of Honest People" People lie a little, don't maximize profits they could get by lying Erat, Gneezy, ``White Lies'' Altruistic (benefits receiver, hurts sender) vs. Pareto (benefits everyone) white lies. Many people don't even tell Pareto white lie. Gneezy, AER, 2005 Lying increases in benefit to self, decreases in cost to other Eisenkopf et al, ``Size matters'' When deciding how much to punish a liar, comparative statics similar to Gneezy 2005 hold Brooks article in NYT: Ariely study: taxi riders take longer route less often with blind than sighted person, because they would feel bad cheating a blind person.