Ethical concerns and the reach of markets:

A choice experiment on Americans’ views about paying kidney donors*

Julio Elias (Universidad del CEMA)

Nicola Lacetera (University of Toronto)

Mario Macis (Johns Hopkins University)†

Abstract

Regulation and public policies are often the result of competition and compromise

between different views and interests. In several cases, strongly held moral beliefs

voiced by societal groups lead lawmakers to prohibit certain transactions or to

prevent them from occurring through markets. However, there is limited evidence

about the specific nature of the general population’s opposition to using prices in

such contentious transactions. We conducted a choice experiment on a

representative sample of Americans to examine preferences for legalizing

payments to kidney donors. We found strong polarization, with many participants

in favor or against payments regardless of potential supply gains. However, about

20% of respondents would switch to supporting payments for large enough supply

gains. Preferences for compensation have strong moral foundations. Respondents

especially reject direct payments by patients, which they find would violate

principles of fairness. We corroborate the interpretation of our findings with the

analysis of a costly decision to donate money to a foundation that supports donor

compensation.

Keywords: repugnant transactions, morality, markets, preferences, kidney donation.

JEL Codes: C91, D01, D63, D64, I11.

* This is a substantively revised version of a previous paper titled “Efficiency-Morality Trade-offs in Repugnant

Transactions: A Choice Experiment” (NBER WP 22632). We benefited from conversations with Sandro Ambuehl, Lanier

Benkard, Leo Bursztyn, Jon de Quidt, Liran Einav, Christine Exley, Claudine Gartenberg, Martin Gaynor, Ori Heffetz

Alessandro Iaria, Alex Imas, Jeffrey Kahn, Nancy Kass, Judd Kessler, Matthew Mitchell, Matthew Osborne, Gerard Padro

i Miquel, Devin Pope, Iyad Rahwan, Jim Rebitzer, Alvin Roth, Heather Royer, Cass Sunstein and Justin Sydnor; and from

the comments of the attendees of presentations at several universities and conferences. We received ethics approval from

the Research Ethics Board at the University of Toronto (protocol 30238) and the Homewood Institutional Review Board at

Johns Hopkins University (protocol 00001991); the registration number at the American Economic Association’s Registry

for Randomized Controlled Trials is AEARCTR-0000732. We gratefully acknowledge the financial support of the Johns

Hopkins University “Catalyst Award”, a sub-grant from the research program “The Economics of Knowledge

Contribution and Distribution” (funded by the Sloan Foundation), a Research and Scholarly Fund grant from the

University of Toronto Mississauga, and the Dean’s Research Funds at Johns Hopkins Carey Business School. Nicola

Lacetera is thankful to the Center for Economic Studies in Munich for the hospitality during the development of part of

this study. Gilda Assi and Daphne Baldassari provided excellent research assistance. † Emails: je49@ucema.edu.ar; nicola.lacetera@utoronto.ca; mmacis@jhu.edu, respectively.

2

“We need to understand better and engage more with the phenomenon of ‘repugnant transactions,

which often serves as an important constraint on markets. As economists, we have to understand

folk ideas about what we can do in the market better than we do. They’re a big issue. And that’s

not to say that economists are the ones who are necessarily right.” Alvin Roth (2007).

1. Introduction

In 2017, about 35,000 patients received an organ transplant in the United States. Yet in the same

year over 50,000 people joined the waiting list, which currently counts about 115,000 patients.

Of these patients, 95,000 need a kidney (UNOS 2017). The average wait time is over 4.5 years,

and thousands of Americans die because they cannot find a donor. The annual cost of the kidney

shortage is $5-7 billion. Each transplant reduces medical costs by about $200,000; the benefits

reach $1.1 million per recipient if we add the value of increased life expectancy and quality

(Held et al. 2016). This health emergency concerns many other countries. Scholars and

practitioners have discussed legalizing donor compensation to enhance supply; according to

Becker and Elías (2007), for example, payments between $15,000 and $30,000 would eliminate

the waitlist within a few years. However, compensation is illegal virtually everywhere.1

Ethical concerns such as the exploitation of participants, coercion, undue influence, and unfair

allocation of organs are often indicated as main determinants of the opposition to paying donors.2

A further concern is that compensation would violate human dignity and other sacred values.3

Delmonico et al. (2002), for example, state that payments are “ethically unacceptable […]

despite the purported benefits of such a sale for both the buyer and the seller”, and that

“Fundamental truths of our society, life and liberty, should not have monetary price”. These

words characterize organ donor payments as repugnant transactions, i.e. exchanges in which the

parties want to transact, but third parties disapprove and wish to prohibit the trade (Roth 2007).

In this paper we provide, to our knowledge, the first investigation into the nature of

preferences of Americans toward paying organ donors. Although there is a lively public debate on

this issue, there is no evidence regarding the sources of aversion to legalizing donor compensation

among the general population, and on whether opposition is absolute (i.e., paying donors violates

sacred values) or, conversely, people would accept a paid-donor system if it produced large

enough organ supply gains.

1 Remuneration is illegal in all countries except in the Islamic Republic of Iran. In the U.S., the key legislative reference is

the 1984 National Organ Transplant Act (NOTA), which prohibits the transfer of human organs for “valuable

consideration”, punishing violators with fines and prison time. Certain countries contemplate some incentives to motivate

donations, such as allocation priority, kidney exchanges, and symbolic awards (Kessler and Roth 2012; Leider and Roth

2010; Niederle and Roth 2014; Roth, Sönmez and Unver et al. 2004; Stoler et al. 2016). 2 See Basu (2007); Halpern et al. (2010); Kerstein (2009); Radin (1996); Rippon (2012); Satz (2010). Ambuehl (2018) and

Ambuehl, Niederle and Roth (2015) provide experimental evidence of whether remuneration leads to undue influence. 3 See for example Council of Europe (2015); Delmonico et al. (2002); Grant (2011); Sandel (2012); WHO (2004).

3

We conducted an online choice experiment on a representative sample of 2,666 Americans

whom we recruited through a professional survey firm. Our experimental design included two

main sources of variation.

First, we randomly assigned each respondent to one hypothetical paid-donor kidney

procurement and allocation system, and asked them to view it as an alternative to the current

system. Each individual made five binary choices to indicate whether they would support the

proposed system or if they would prefer to keep the current one. The features of the alternative

system were the same in all five choice scenarios, except for the number of transplants that, in

each scenario, we asked each participant to assume the system would generate. The features that

were constant within a system were the nature of compensation (cash or non-cash, such as

contributions to college or retirement funds or tax rebates), the amount of the payment ($30,000

or $100,000), and the identity of the payer (a public agency or the kidney recipient). Having

respondents make five binary choices with different kidney supply levels allowed us to

characterize their preferences toward outcomes (number of kidneys procured) and procedural

aspects of paid-donor procurement and allocation systems. Payments by patients represent

standard market transactions, whereas if compensation is from a public agency the allocation

depends on medical need and not on purchasing power. We could therefore determine whether

respondents oppose payments per se, or if they are concerned about the distributional

consequences of private transactions; in other words, we could test for differences in attitudes

toward patients paying and donors being paid. Non-cash compensation could allay the concern

that individuals might be unduly pressured to give their kidney if they are in urgent need of cash

(Satel 2006); different payment amounts could affect perceptions of exploitation or undue

influence (Ambuehl, Niederle and Roth 2015). Our design also lets us estimate the “slope” of the

relationship between favor for a paid-donor system and the increases in transplants it is assumed

to achieve, and whether this relationship varies with procedural features of the system.

Second, we had a random half of participants express their moral views about both the

current system and the paid-donor system to which we assigned them (at each hypothesized

supply level), with reference to six principles: autonomy of choice, undue influence, exploitation

of the donor, fairness to the donor, fairness to the patient, and human dignity. With this part of

the survey we could assess whether the respondents’ attitudes toward the outcomes and

procedural features of paid-donor systems have ethical roots, and which principles are more

relevant. Because only a subset of the subjects received the ethical assessment questions, we were

also able to determine whether making ethical issues salient affects the support for a paid donor

system and the response to supply gains.

We found that approval of paid-donor systems increases with the size of the kidney supply

gains. The support at any given supply increase, however, differs across systems. In particular,

4

systems where the patient pays for the organ receive much lower acceptance than systems where

compensation comes from a public agency. The nature and amount of payments, in contrast,

have limited effects on support. Thus both process and outcomes have an impact, on average, on

the support for paid-donor systems, with some features of the process, namely the identity of the

payer, being particularly relevant.

Exploiting the within-subject features of our design, we document a polarization in attitudes,

with large shares of respondents either in favor of payments (46%) or against (22%) regardless of

the size of organ supply gains. About 20% of participants, however, would switch from opposing

to supporting payments if supply gains were sufficiently large.

Ethical considerations are a major determinant of attitudes towards using prices in this market;

these considerations, moreover, vary widely in the population. The opposition to systems that

contemplate payments by the organ recipient mainly derives from concerns about the fairness of

the resulting organ allocation. We also found that the responses to supply gains and to the

procedural features of the systems are similar with and without the ethical assessment module;

thus individuals are likely considering ethical issues when making their choices, regardless of

whether one makes them salient. Moreover, attitudes toward payments for kidney donors

correlate with broader ethical views that we assessed using Graham et al. (2011)’s “moral

foundations” assessment tools. Individuals with more deontological beliefs are more likely to

oppose payments regardless of the kidney supply gains, whereas those who place high value on

compassion, freedom, and pleasure are more likely to support the legalization of payments.

Finally, we corroborated the findings and their interpretation by adding an incentive-

compatible choice module to the survey. We gave the respondents the possibility to profit (or

incur a cost) from having the researchers donate (or not) money to a foundation that is in favor of

expanding allowable payments to organ donors. Participants in “patient pays” conditions show a

lower propensity to donate to a pro-compensation foundation and are willing to sacrifice some

compensation in order to express their opposition. Moreover, participants who opposed

payments regardless of the supply gains are less likely to donate, whereas those who switched

from being opposed to being in favor at some higher supply gain, and even more so those who

expressed support at any supply gain, are more likely to donate. We found opposite patterns of

donations to a foundation that opposes paying organ donors.

Our study provides insights to several literatures within economics. In recent years numerous

studies focused on how fairness concerns, identity, religious beliefs, political ideology, dignity,

and social status influence utility and decisions.4 Calls for economists to consider ethical forces

4 See for example Akerlof and Kranton (2000); Bénabou and Tirole (2009, 2011); Bénabou, Ticchi and Vindigni (2015);

Benjamin, Heffetz, Kimball and Rees-Jones (2012); Benjamin, Choi and Fisher (2016); Bursztyn, Callen, Ferman, Gulzar,

Hasanain and Yuchtman (2015); Kuziemko, Norton, Saez and Stantcheva (2015).

5

as guiding decisions have a long history (Smith 1822; Marshall 1890; Sen 1999). Shleifer (2004)

discusses the effect of market forces on the diffusion of certain morally censurable behaviors,

and Falk and Szech (2013) and Bartling, Weber and Yao (2015) study whether market

interactions erode ethical values and social responsibility. Evidence on whether individuals

perceive trade-offs between ethical beliefs and supply considerations in the case of repugnant

transactions is, however, virtually non-existent.

Other studies analyze whether economic returns affect the decision of individuals to adopt a

morally unacceptable behavior, such as lying or cheating on school tests.5 These investigations

focus on activities that are (plausibly) universally considered morally wrong, are illegal

everywhere and their legalization is not considered as a policy option. Our interest is about

transactions that are morally contentious but that can be (and indeed often are) contemplated as

actual policies. Moreover, we focus on individuals’ attitudes toward activities that others (and

not necessarily themselves) undertake. This is, in principle, a decision process that is different

from choosing between an illegal or universally immoral act and a private economic or social

gain. Many other morally controversial transactions have features similar to payments for organ

donors; examples include gestational surrogacy, prostitution, abortion, eating meat from certain

types of animals, and so on. Some of these activities are legal in certain countries and not in

others, and opinions about their morality vary widely (Healy and Krawiec 2017; Satz 2010).

Some of the concepts that we used and that influenced our research design, finally, come

from outside of economics, and in particular from studies in moral foundation theory and

experimental ethics.6 We see great value in bridging across disciplines to provide a deeper

understanding of the ethical constraints to the reach or markets.

In Section 2 we outline the simple framework that guided our empirical investigation.

Section 3 describes the research design and the subject pool. We then describe the findings in

Section 4, and discuss their implications in the concluding Section 5.

5 Among studies of preferences for truthfulness and economic incentives, see for example Gibson, Tanner and Wagner

(2013) and Gneezy (2005). Jacob and Levitt (2003) and Martinelli et al. (2018) show that monetary rewards induce

teachers and students to cheat on tests. 6 Birnbacher (1999); Bonnefon, Shariff and Rahwan (2016); Doris and Stich (2005); Graham et al. (2013); Grant (2011);

Haidt (2007); Knobe et al. (2012); Molewijk et al. (2004); Tanner, Medin and Iliev (2008); Tetelock et al. (2000).

6

2. Motivating Framework

We describe below a basic framework that guided our design of the choice experiment. Consider

an individual who evaluates an alternative way for society to organize a particular transaction,

and compares her utility from introducing this policy to the utility that she receives from the

current system. There are two sets of factors that affect utility: the effect that the alternative

system has on the supply of the particular good or service that is the object of the transaction,

and the details of the organization of the exchanges. In our survey we test whether both outcome

and procedural features of a transaction affect individual utility. Our specific transaction of

interest is the procurement and allocation of kidneys for transplants, therefore from now on we

refer to this case specifically. We normalize the utility from the current system to be equal to

zero, and express the utility that an individual “voter” i derives from an alternative procurement

and allocation system s as:

𝑈𝑖𝑠 = 𝑓𝑖(Π𝑠, Δ𝑄𝑠),

where Δ𝑄𝑠 represents the change in the number of transplants with respect to the system

currently in place (what we refer to as the main outcome of the system); and Π𝑠 includes

characteristics that define the procurement and allocation rules of a particular system (the

process). An individual supports the introduction of an alternative system s with a set of process

features Π𝑠 and an expected increase in supply Δ𝑄𝑠 if 𝑈𝑖𝑠 > 0.

An explicit functional form for 𝑈𝑖𝑠 helps us identify some cases of interest without loss of

generality. Suppose that we can express:

𝑈𝑖𝑠 = Π𝑖𝑠 + 𝜚𝑖𝑠Δ𝑄𝑠.

The coefficient 𝜚𝑖𝑠 (possibly individual- and system-specific) indicates how utility responds to

increases in supply. Π𝑖𝑠 represents the reaction to the specific characteristics of a system

(regardless of supply) relative to the current system. Again, the value of Π𝑖𝑠 can be positive or

negative: an individual may derive disutility from the procedural features of a system, whereas

others may find those same features welfare enhancing.

If both Π𝑖𝑠 and 𝜚𝑖𝑠 are positive, then an individual will vote in favor of the alternative system

regardless of the size of the gains in transplants that the system will produce. If, instead, the

individual reacts positively to supply increases, but Π𝑖𝑠 < 0, then we may observe two choice

patterns. In one case, the absolute value of Π𝑖𝑠 is large enough that the individual will not prefer

the alternative system for any increase in supply within the relevant range; in a second case, for

lower absolute values of Π𝑖𝑠 (i.e. for only moderate aversion to the procedural aspects of a

system), there will be a level S1 of Δ𝑄𝑠, within the reveal range of feasible increases, that will

make the individual shift to supporting an alternative system for any Δ𝑄𝑠 > 𝑆1. An individual

with aversion to the procedural aspects of a system, and also a negative marginal utility to

7

increases in transplants (for example because her disutility increases if more and more

individuals engage in what she considers an undesirable transaction), would always oppose an

alternative system as compared to the current one. Finally, a decision maker may have a

preference for the procedural features of an alternative system, but also value increases in supply

negatively; in such a case within the relevant range of supply increases these individuals may

support an alternative system up to an increase in supply of S2, and oppose that alternative

system for any Δ𝑄𝑠 > 𝑆2; for example, “excessive” supply may raise concerns about what drives

certain people to donate, and these concerns may outweigh the utility from supply increases.

We designed our choice experiment to characterize the respondents’ preferences for paid-

donor kidney procurement and allocation systems as highlighted in the basic framework

described above. Specifically, we investigate the impact of procedural features (Π𝑖𝑠) and kidney

supply increases (𝜚𝑖𝑠 ). Moreover, we explore whether the respondents’ different reactions to the

systems’ outcomes and processes have roots in moral values. The next section provides the

details of our experimental design.

3. Research Design and Subject Pool

We designed an online randomized survey and administered it through a professional survey

firm (Qualtrics) to a sample of adult respondents, representative of the U.S. population. The

survey ran from November 15 to December 7, 2017.7

3.1 Experimental design

In the introductory part of the survey, we informed the participants that we would collect their

opinions regarding alternative kidney procurement and distribution systems, that we would

inform U.S. Congress members as well as the Secretary of Health and Human Services about the

results of the study, and that their answers would be anonymous. We also told the respondents

that the sample in the study was representative of the U.S. population. In including these

features, our objective was to increase the perceived consequentiality of the study (that is, that

the study can potentially influence policy) while reassuring the participants that their individual

answers would be completely anonymous (Vossler, Doyon and Rondeau 2012).

In the second section we provided a description of several aspects related to the procurement

and allocation of kidneys in the United States: what is kidney failure, the various types of kidney

donations (i.e., deceased and living donors, directed and undirected donors), the living kidney

donation process (including some information regarding the surgery and the associated risks and

7 Survey materials (including texts and snapshots from the actual survey) are in the Appendix. The survey itself is

available at this web address: https://jhubusiness.qualtrics.com/jfe/form/SV_1NypHI8IT20GYap

8

recovery for the donor), and the characteristics of the current U.S. organ procurement and

distribution system, including the size of the kidney shortage, and the fact that the 1984 National

Organ Transplant Act (NOTA) prohibits compensation to organ donors. This part of the survey

was somewhat lengthy, but we wanted to ensure that all participants had the same initial

information about the topic. Giving details about the context of interest is frequent, and in fact

encouraged, in contingent valuation studies, for example in the valuation of natural resources, in

order to enhance the reliability of the respondents’ expression of their willingness to pay in

absence of market information (Carson 2012).

We ended this second part by informing the participants that in the next section we would ask

them to express their opinions about an alternative organ procurement and distribution system.

3.1.1 Assignment to different alternative kidney procurement and allocation systems

We then randomly assigned participants to one of eight alternative procurement and allocation

systems. Each system was a combination of the following attributes (Table 1 provides a

summary):

(1) Nature of the payment: cash or non-cash. In the “cash” systems, donors would receive cash

compensation (a deposit to their bank account), whereas in the “non-cash” systems donors could

choose between “tax credits, tuition vouchers, loan repayment, or contributions to a tax-free

retirement account”;

(2) Identity of the payer: the patient or a public agency. In the “public agency” systems, donors

would receive compensation from a public agency coordinated by the U.S. Department of Health

and Human Services, and kidneys from paid donors would be allocated to patients on the waiting

list according to priority rules based on medical urgency, blood and tissue match with the donor,

time on the waiting list, age and distance to the donor. In the “patient pays” systems, donors

would receive compensation directly from the kidney recipient; however, we also specified that a

public agency, coordinated by the U.S. Department of Health and Human Services, would

regulate and oversee the process.

(3) Size of the payment: low ($30,000) or high ($100,000).

9

Table 1: Characteristics of the kidney procurement and allocation systems randomly assigned to the study participants

nature of

compensation amount payer

System 1 Cash $30K Public agency System 2 Cash $100K Public agency System 3 Cash $30K Patient System 4 Cash $100K Patient System 5 non-cash $30K Public agency System 6 non-cash $100K Public agency System 7 non-cash $30K Patient System 8 non-cash $100K Patient

All systems thus included compensation to kidney donors. The “patient pays” systems

represent standard market transactions, whereas in the “public agency” systems, payments are

from a third party, and the allocation is based on priority rules, not purchasing power. This

variation allows determining whether the aversion is to payments per se, or if it is related to the

identity of the payer and the resulting distributional consequences. Some proponents of

compensation, for example, argue that payments are more ethically acceptable if they come from

a public agency and they are not in the form of direct cash. Non-cash forms of compensation, in

particular, could lessen the concern that vulnerable individuals might be induced to give away

their kidney because they are in immediate need of cash (Satel 2006). However, in its strong

form the opposition to payments appeals to deeper reasons (e.g., violation of human dignity) that

make any form of payments unacceptable, irrespective of regulation and public intervention

(Delmonico et al. 2002; Sandel 2012). Finally, our design also allows testing for whether

attitudes towards compensating donors depend on the amount of compensation. One could argue,

for example, that concerns such as exploitation of the participants would be less relevant if the

donors received a relatively large sum. Conversely, large sums may increase worries about

undue influence (Ambuehl, Niederle and Roth 2015). We chose the sums of $30,000 and

$100,000 to be sufficiently different from each other while being within the range of amounts

that the public policy debate actually discussed, and that would also keep the systems cost-

effective (Becker and Elias 2007; Held et al. 2016).

3.1.2 Supply increases and support for alternative systems

After describing the alternative organ procurement system assigned to each respondent, we asked

them to indicate whether they would support the proposed system or if they would prefer to keep

the current one. We did so under five scenarios, where each scenario asked the respondent to

assume that the alternative system would result in a certain number of kidney transplants per

year (and the corresponding fraction of the annual demand for kidneys), ranging from 19,000

10

(roughly the number of kidney transplants currently performed in the US, or about 50% of the

annual demand)8 to 38,000 (corresponding to roughly 100% of the annual demand according to

estimates from Held et al. 2016). For each of the five choices, which we presented in ascending

order of supply gains in separate pages, we provided a table summarizing the characteristics of

the alternative system, together with the features of the current system. We told the respondents

that we were asking them to consider each of the five scenarios separately, i.e. that they were to

take each level as the best available estimate of the number of kidney transplants performed

annually. We specified that a “Yes” on each choice question would indicate that they would

prefer the alternative system, whereas a “No” would indicate that they would prefer the current

system. Choices were thus binary “referenda” between the alternative and current system

(Vossler et al. 2012).

Our design therefore combines between-subject and within-subject variation. We assigned

each participant to only one alternative, out of the eight combinations that we described above,

and each participant made multiple Yes-No choices on their assigned system for different kidney

supply outcomes. The within-subject component of the design lets us characterize each

individual’s preferences and potentially identify “types” in the population. For example, as we

noted in Section 2, some individuals may not be in favor of paid-donor systems regardless of

how many more transplants it would generate; others, in contrast, may support paying donors at

any level of kidney supply; finally, some participants may be willing to vote in favor of the paid-

donor system only if it provides a sufficiently high number of additional kidneys.

3.1.3 Eliciting moral views about procurement and allocation systems

Within each of the eight systems, we further randomized participants, with equal probability, to

also express (or not) opinions about some ethical features of the system. We asked those who

were in the conditions that included the morality judgments to report their assessment, on a scale

from -10 to +10, of whether the system (1) benefits or exploits the donors, (2) respects or limits

individual autonomy, (3) allows individuals to make fully informed choices or exerts undue

influence, (4) is fair or unfair to the patients, (5) is fair or unfair to the donors, and whether (6) it

promotes or violates human dignity. In selecting these principles we followed the literature in

philosophy and bioethics.9 Respondents could choose any integer number on a slider in the [-10,

+10] interval. To guide the interpretation of the different numbers, we added seven verbal

8 According to UNOS, 17,878 kidney transplants were performed in 2015, 19,060 in 2016, and 19,851 in 2017. See

https://unos.org/data/transplant-trends/. 9 See for example Council of Europe (2015), Nuffield Council on Bioethics (1995), Radin (1996), Satz (2010), United

States Task Force on Organ Transplantation (1986), and World Health Organization (2004).

11

expressions above the sliders; we wrote, for example, “very unfair to donors” above the -10

mark, “neutral” above the 0 mark, and “very fair to donors” over the +10 mark.

Participants provided their moral assessments prior to expressing their choice in each of the

five kidney supply scenarios. We therefore have within-subject variation that allows us to

determine whether the morality valuations are purely assessments of the process defined by the

system, or whether also the outcomes, namely the assumed increases in kidney transplants,

influenced the ethical views of the respondents. Before showing the respondents their assigned

alternative system, we also asked those in “ethics assessment” subgroups to evaluate the current

organ procurement and distribution system on the same features and using the same scale. This

provided us a baseline set of morality assessments that will later allow us to construct, for each

individual, morality valuations of the alternative system relative to the current system. Including

the ethics assessment thus allows us to determine whether and to what extent ethical

considerations explain respondents’ expressed favor for the alternative systems; moreover,

including these questions only for a subset of the respondents enables us to assess whether

prompting the respondents to think about ethical issues affects the support for a system or the

responses to supply gains.

3.1.4 Quality of responses

Because the law does not currently allow the policy options that we consider, one cannot run an

actual-choice experiment. We thus have to proceed with hypothetical scenarios.10 We adopted a

few strategies to enhance the reliability of the responses and the ensuing quality of our data.

At the beginning of the survey, we asked the respondents to commit to provide truthful

answers. This was of course a soft commitment; however, prior research has shown that these

prompts help to motivate survey participants to give complete and accurate responses (Cibelli

2017). We then introduced survey modules that would increase the perceived consequentiality of

the answers, measured whether the participants perceived the topics of the survey as important,

and collected their beliefs on whether their responses will have some impact on policy. Finally,

we added an incentive-compatible choice to assess the respondents’ preferences related to

supporting payments to organ donors. We describe these modules below.

Consequentiality. A critique to hypothetical survey analyses is that the opinions that

participants express might not represent their “true” preferences. Before prompting the

respondents to make their choices, we reminded them that we would send a letter to US

10 This challenge is common to other studies, such as Benjamin, Kimball Heffetz and Szembrot (2014) on the analysis of

subjective well-being and the relationship between happiness and choice; Kuziemko et al. (2015) on preferences for

redistribution; Andreoni and Sprenger (2012) on time and risk preferences; and Kessler and Roth (2014) on priority

rules and organ donor registration.

12

Congress Representatives and the Secretary of Health and Human Services reporting the

distribution of the study respondents’ preferences with respect to that scenario, and we included

an example of a letter (see the Appendix).11 We then included questions to gather information

about the key requirements that the literature on contingent valuation (e.g. Carson and Groves

2007 and Vossler et al. 2012) identifies to assess the respondents’ perceived consequentiality in

hypothetical surveys: first, individuals should regard the topic as important; second, the

respondents should perceive their responses as having a potential effect on actual policy choices.

Immediately following the set of five Yes-No choices regarding the alternative organ

procurement system, we asked the respondents to rate how strongly they felt about the choices

they had just made. They could choose one of four answers: “very confident” [about my

answers], “somewhat confident”, “somewhat unsure”, or “very unsure”. Then, to measure the

extent to which participants cared about the topic and perceived their answers to have some

chance of influencing policy, we asked them whether they believed that public authorities should

take their answers into consideration, and whether they believed the authorities will take their

answers into consideration (we randomized the order of these two questions). For these

questions, the answers from which to choose were “not at all”, “very little”, “little”,

“somewhat”, and “very much”. Finally, we asked the respondents to express their beliefs about

the probability that Congress would pass legislation allowing various types of organ donor

compensation. Taken together, these elements of our survey provide complementary ways to

assess whether respondents considered the topic as being policy-relevant, and whether they

expected that the survey could have some impact on policy. We also assessed whether these

beliefs affected the results of our analyses.

Social influence. To assess whether participants were concerned about other people’s

opinions about the issues, we asked whether their choices were affected by considerations about

how others might be voting. We also asked what share of people in the United States they

believed would be in favor of various types of compensation to organ donors.12 We used the

responses to these questions to assess whether the participants’ responses were affected by the

perceived popularity of payments to organ donors in the general population.

Incentive-compatibility. Finally, to obtain an incentive-compatible indicator of our

respondents’ attitudes towards compensating organ donors, we offered them an extra $1 cash

reward if they authorized us to make a $1 donation on their behalf to a foundation that supports

the expansion of allowable forms of organ donor compensation. This gave the respondent the

11 Kuziemko et al. (2015) adopted a similar approach; they included the respondents’ willingness to send a letter to their

Congresspersons as an outcome variable in their study of attitudes toward redistribution. 12 We included cash payments, reimbursement of lost wages and other expenses related to the donation process, health

insurance for organ donors, and tax credits. Details are in the Appendix.

13

opportunity to profit from allowing us to donate to a pro-compensation organization or,

conversely, to express a costly message of opposition to organ donor payments if they chose not

to authorize us to make the donation. Following Bursztyn, Egorov and Fiorin (2017), we told

participants that they would be randomly assigned to one among two different organizations, one

that favored payments to organ donors and one that opposed such payments. Doing so ensured

that the participants would not make any inference about the researchers’ own preferences.13 In

the analysis, we correlate the respondents’ donation decision with the characteristics of the

system to which we assigned them, as well as with their attitudes towards compensating donors

from the analysis of their choices in the hypothetical choice experiment.

3.1.5 Moral Foundations

The survey included a module that collected information on the participants’ “moral

foundations”. We used a set of questions from Graham et al. (2011) that would measure the

importance that each respondent placed on the following values: equality, freedom, spirituality,

justice, tradition, approval by others, compassion, giving, pleasure, purity, and pragmatism.14

The module also included a vignette, also from Graham et al. (2011), describing a moral

dilemma that would measure whether a respondent is characterized by deontological or

consequentialist preferences. The answers to these questions allow us to assess any correlation

between the participants’ choices in our survey experiment and their more general moral beliefs

as assessed by a set of validated questions used in moral psychology.

3.1.6 Socio-demographics

The final module of the survey included socio-demographic questions (gender, age, income,

education, religious beliefs, political orientation on social and economic matters, relationship

status, and if the respondents had children) and questions on whether the respondents made

donations or volunteered in the recent past, had a blood transfusion or knew anyone who did, had

an organ transplant, were waiting for a transplant, or knew anyone in those conditions.

3.2 Subject Pool

3.2.1 Socio-demographic characteristics of the respondents

Our sample consists of 2,666 respondents. Column (1) of Table 2 below shows characteristics of

the survey participants. The survey firm constructed the sample to be representative of the adult

13 Again following Bursztyn et al. (2017), we assigned respondents to the pro-compensation organization (the American

Transplant Foundation) with 80% probability, and to the organization opposing payments (the National Kidney

Foundation) with 20% probability. Doing so maximized our statistical power while not deceiving the participants. 14 We obtained the questions at this website: http://www.yourmorals.org/. Details are in Section A of the Appendix.

14

US population in terms of composition by gender, age, education, and ethnicity. The statistics in

column (2) confirm that we achieved representativeness on these features; other characteristics of

the respondents (including marital status, labor market status, and income) are also very similar

to those of the US population. About 51% of the respondents were women. The average age was

49 years. 63.5% of the subjects were non-Hispanic whites, 12.6% black, and 14.1% Hispanics.

About 59% were married, 23% had a college degree, 60% were employed and 4% unemployed,

and 34.5% reported an annual household income lower than $50,000.

Figure 1 shows that we achieved balance of the sample across the sixteen experimental

conditions (eight systems, further divided by whether participants received the ethics assessment

module or not).

Table 2: Characteristics of the respondents and comparison with American Community Survey Data

Share of:

Qualtrics sample (N=2,666)

(1)

ACS 2016

(2)

Women 50.8% 51.4% Age 18-34 24.2% 30.2% Age 35-54 36.4% 33.5% Age 55+ 39.5% 36.3% White (non-Hispanic) 63.5% 61.3% Black 12.6% 13.3% Hispanic 14.1% 17.8% Asian 6.3% 5.7% Other race/ethnicity 3.4% 1.9% Less than HS diploma 7.5% 12.6% HS diploma/GED 24.4% 27.7% Some college or Associate degree 30.3% 31.0% Bachelor’s degree 22.9% 18.3% Graduate degree 14.8% 10.5% Married 59.4% 51.6% With children 36.3% 23.6% In labor force 64.4% 63.1% Employed (full or part time) 60.4% 59.1% Unemployed 4.1% 5.8% Income <$50,000 34.5% 43.7% $50,000 <= Income < $100,000 35.6% 30.0% Income >=100,000 29.9% 26.3%

Notes: The table shows summary statistics from our Qualtrics sample in column (1) as well as the corresponding statistics from the American Community Survey (ACS) for 2016 (https://www.census.gov/acs/www/data/data-tables-and-tools/subject-tables/). The ACS employment statistics are for population 16 years and over, and marital status is for population 20 years and over. The remaining statistics are all for population 18 years and over.

15

Figure 1: Balance checks

Notes: This figure shows the mean and confidence intervals of several individual characteristics in the sixteen conditions of the survey experiment, where a condition is characterized by one of the eight assigned systems (as described in Table 1 above) and by whether respondents received the moral principles module or not. The horizontal line in each graph represents the average for a given characteristic among the respondents in condition 1, and the solid lines in correspondence to each number in the x axis between 2 and 16 report the averages for each of the conditions 2 through 16. The dotted lines are the confidence intervals of the coefficient estimates from a linear regression of a binary indicator for a given features (e.g. being a woman) on indicators for each condition 2 through 16, centered around the average of the particular features for the respondents in each condition.

16

3.2.2 Reliability of the survey responses

To assess the participants’ perceived consequentiality of their survey responses, following

Carson and Groves (2007) and Vossler et al. (2012) we determined whether our respondents

cared about the policies that we asked them to consider, and believed that the results of this

survey would have at least some influence on policy. About 80% of respondents stated that their

responses, once communicated to policymakers, should be “somewhat” or “very much” taken

into consideration; only 3.3% said that policymakers should not take their responses into

account. Moreover, only about 13% of participants stated that there is no chance that

policymakers will take their answers into consideration, and only between 3.0% and 4.2% of

respondents attributed zero probability to the event that Congress will legalize some form of

compensation to donors. When we asked respondents how strongly they felt about their answers,

55.6% declared to be “very confident” and 37.9% “somewhat confident”.

Thus, the vast majority of respondents appear to believe that the topic is important enough to

merit the attention of policymakers and that their responses have some chance of affecting

policy; they also felt confident in their answers. Of course, also the answers to these questions

could suffer from different biases. However, the overall consistency of the respondents’ answers

to our core survey modules and, especially, the results from our analysis of an incentive-

compatible outcome (a monetary donation to a foundation that favors expanding allowable

payments to organ donors; see Section 4 below), make us confident about the quality and

reliability of our survey data.

4. Findings

We organize the description of the findings in seven parts. First, we report raw statistics and

regression estimates of the overall relationship between support for a paid-donor system and its

procedural aspects and hypothesized kidney supply gains. We then explore the heterogeneity in

preferences and the polarization of views about donor compensation, thus identifying “types” in

the population of respondents. In the third part we move to studying how ethical views vary in

the population and relate to the features of the various paid-donor systems. Next, we analyze the

joint role of supply, procedural aspects, and moral views on preferences for legalizing payments.

In the fifth subsection we explore more in depth the moral foundations of the attitudes toward

payments. The sixth set of analyses includes the results of the incentivized monetary donation

experiment and their relationship with the survey experiment findings. Finally, we describe some

robustness tests that we performed to evaluate whether the participants’ perceived relevance of

the topic, confidence in their answers, consequentiality of the survey, and beliefs about the

popularity of payments to organ donors in the US population, affected our main results.

17

4.1 Support for alternative systems increases with kidney supply gains, and depends

strongly on the identity of the payer

Figure 2 shows the percentage of respondents who expressed support for legalizing the

alternative system to which they were assigned, at each of the five hypothesized supply levels. In

the picture, we expressed the supply levels in terms of gains (in thousands of transplants per

year) over the current system. In our data, support for a paid-donor system increases with the

level of assumed increases in transplants. The increase in support is roughly linear, and the slope

is similar across all eight systems that we considered. The level of support at each supply level

differs across systems, however. In particular, the identity of the payer causes a significant shift

in support: systems where a public agency compensates donors received significantly more

acceptance than systems where the patient pays the kidney donor. For example, the raw means

show that to receive the same support rate that a system with $30,000 cash payments from a

public agency obtains even without any supply gains over the current system (about 63%), a

system with $30,000 cash payments directly from the kidney recipient would require an increase

of 19,000 transplants (i.e., it should be able to cover 100% of the annual demand for kidneys).

Figure 2: Support for alternative organ procurement systems over the current system, by level of transplant gains.

Notes: Each line represents the percentage of participants assigned to one of the eight alternative systems who stated that they would support the adoption of that system as compared to the existing one. In the legend, $30K and $100K indicate the payment amounts of the alternative system (USD30,000 or USD 100,000); “cash” and “non-cash” indicate whether the system included cash payments or non-cash compensation (of the same dollar amount); PA and PVT indicate that the payer was a public agency or the kidney recipient, respectively.

18

Table 3 reports estimates from linear probability regressions. The outcome variable is an

indicator equal to 100 if a respondent selected their assigned alternative system at a given supply

level, and 0 otherwise; regressors include the supply level, expressed as percentage point gains

over the current number of transplants per year,17 and indicators for each system in some

specifications, or binary indicators for the three key features of each system: the level of payment

(1 for $100,000, 0 for $30,000), the type of payment (1 for cash, 0 for non-cash), and the identity

of the payer (1 for private payments from the recipient, 0 for payments from a public agency).

We also added interaction terms between these indicators and the measure of supply increases.18

The estimates confirm the evidence from the raw averages in Figure 2. Supply gains

significantly increase support for a system; the estimates imply that, on average, a one-

percentage point increase in transplants leads to a 0.26 percentage-point increase in the support

for a paid-donor system (columns 1 and 2). Systems with patient payments receive a support that

is 15 percentage points lower than systems with payments by a public agency. This is a large

effect; for comparison, the effect of payer identity on the support for a system corresponds to the

effect of a 15/.26=57 percentage-point increase in transplants. The type and amount of payments,

in contrast, have small and statistically insignificant effects on support. The effect of supply

gains is similar between the different systems. Therefore both process and outcomes have an

impact, on average, on the support for procurement and allocation systems that include payments

to kidney donors. One feature of the process, namely the identity of the payer, is particularly

relevant in determining the level of support for a system. However, the systems’ features do not

seem to affect the marginal response to supply increases (thus, overall the parameter ϱ from the

framework in section 2 does not appear to be system-specific).

17 We chose this specification because the relationship between support and supply level that inform the raw data (see

Figure 2) is roughly linear. Regressions with binary indicators for different supply levels provided very similar results (see

Table B1 in the Appendix). 18 Because each respondent was assigned to only one alternative organ procurement system, and the five supply levels

were the same for all respondents, these regressions are equivalent to individual fixed-effects analyses.

19

Table 3: The effect of supply gains and procedural features on support for paid-donor systems

Notes: The table reports the coefficient estimates from linear regressions of the support for a system, on the hypothesized supply increase and binary indicators for each of the eight systems or binary indicators for the three features of each system: the level of payment (1 for $100,000, 0 for $30,000), the type of payment (1 for cash, 0 for non-cash), and the identity of the payer (1 for private payments from the recipient (PVT), 0 for payments from a public agency). Standard errors, clustered at the respondent level (the regressions include 2,666 participants), are in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

Outcome variable:

Regressors: (1) (2) (3) (4)

Supply increase (%pts.) 0.264*** 0.264*** 0.247*** 0.265***

(0.018) (0.018) (0.045) (0.036)

$100K cash, PA -1.645 -1.043

(2.964) (3.550)

$30K cash, PVT -14.461*** -15.639***

(3.075) (3.611)

$100K cash, PVT -13.086*** -14.263***

(3.125) (3.658)

$30K non-cash, PA 5.462** 4.894

(2.690) (3.279)

$100K non-cash, PA -0.186 0.088

(2.915) (3.518)

$30K non-cash, PVT -15.327*** -17.459***

(3.070) (3.625)

$100K non-cash, PVT -13.269*** -12.533***

(3.172) (3.696)

Cash -1.591 -1.580

(1.535) (1.819)

PVT -15.026*** -16.046***

(1.542) (1.825)

$100K -1.067 -0.000

(1.538) (1.822)

$100K cash, PA x Supply increase -0.024

(0.068)

$30K cash, PVT x Supply increase 0.046

(0.071)

$100K cash, PVT x Supply increase 0.046

(0.070)

$30K non-cash, PA x Supply increase 0.022

(0.066)

$100K non-cash, PA x Supply increase -0.011

(0.070)

$30K non-cash, PVT x Supply increase 0.084

(0.073)

$100K non-cash, PVT x Supply increase -0.029

(0.069)

Cash x Supply increase -0.000

(0.037)

PVT x Supply increase 0.040

(0.037)

$100K x Supply increase -0.042

(0.037)

Constant 64.356*** 66.659*** 64.782*** 66.630***

(2.112) (1.481) (2.435) (1.692)

Observations 13,330 13,330 13,330 13,330

R-squared 0.036 0.035 0.036 0.035

Favor for alternative system (=100 if in favor, 0 if opposed)

20

4.2 Participants are heterogeneous in their responses to kidney supply increases

Figure 2 suggests a further pattern in the data. A sizeable share of the respondents (with some

differences across systems) expressed preference for a paid-donor system over the current one

even when the alternative system would hypothetically yield no gains in the number of

transplants. Moreover, between 20% and 40% of the respondents (again, depending on the

specific system) supported the current no-payment system even when the hypothesized supply

gains from the paid-donor system would allow to cover the entire annual need for transplants.

These findings are consistent with the presence of individuals with very different preferences,

including some with strong opposition to payments regardless of the supply gains, and

individuals who, in contrast, support compensation irrespective of any effects on transplants. The

positive slopes of the curves in Figure 2 and the sign of the estimated coefficient on supply gains

that we reported in Table 3 derive from individuals who responded to hypothetical supply gains

by changing their position in favor or against donor compensation. Overall we characterize five

distinct “types” among the respondents. About 22% of participants expressed opposition to the

paid-donor system, regardless of the hypothesized increase in transplants that the system would

yield (hereinafter we refer to this group as the “always opposed”). Conversely, 46% stated favor

for payment systems at all supply levels, including the case where we assumed that a payment

systems would produce the same number of transplants as the current system (“always in

favor”).19 About 18% of participants expressed opposition to payments in case of zero gains in

transplants, but switched at some higher supply gains (“from opposed to in favor”). In contrast, a

small subset of 4.5% of respondents showed the opposite pattern in their choices, supporting

payments for no or small supply increases, but expressing opposition for higher supply levels

(“from favor to opposed”). Finally, 10% of participants showed non-systematic patterns of

choice with no correlation between support for the alternative system and kidney supply levels

(“Others”).20 Among the respondents who went from opposing to supporting the alternative

system, almost 40% of them switch at the first opportunity (i.e., when the paid-donor system is

assumed to increase transplants by 14 percentage points compared to the current system)

whereas 15% of this subset of respondents become willing to support the paid-donor system only

when it is hypothesized to cover the entire demand for kidney transplants.21 As we show in the

Appendix (Table B2), the distribution of these types differed across systems. Notably, for the

19 Note that the lack of a relationship between support for a paid-donor system and supply gains for these two categories

does not imply that these respondents do not place any value on supply gains; following the framework that we laid out in

Section 2, these choice patterns are consistent with very strong views about the desirability of the procurement and

allocation rules of a given paid-donor system. 20 This behavior could derive from a lack of interest or attention on the one hand, or an explicit desire to randomize

(Agranov and Ortoleva 2016). The small share of this fifth group, as well as some additional evidence that we report

below, are consistent with these individuals not paying sufficient attention or not showing interest in the survey. 21 Figure B1 in the Appendix shows details on the switchers separately for each of the eight paid-donor systems.

21

systems where the patient pays we observe the largest proportions of “always opposed”, and,

conversely, the lowest proportions of “always in favor” and “from opposed to in favor” (the

proportions of “from in favor to opposed” and “others”, instead, appear to be distributed

uniformly across the eight experimental conditions).

4.3 Ethical concerns for alternative systems vary widely in the population

Our next step is to describe the respondents’ ethical assessment of their assigned alternative

organ procurement system. Recall that we included the morality module for only a random half

of respondents (1,276 participants). In section 4.4 below we describe the impact of the ethics

assessment module on the degree of favor for the alternative systems; in this section, we

document vast heterogeneity in moral concerns across participants, and identify the primary

drivers of this heterogeneity.

Figure 3 reports the distribution of ratings for each principle for all eight paid-donor systems

and five kidney supply levels combined, as well as the distribution of the ratings for the

corresponding principles for the current system. We grouped the 21 possible discrete scores

(from -10 to 10) into seven groups in order to smooth the curves and make them more legible.

Recall that positive numbers express consistency with a moral principle (e.g. autonomy of choice

or respect of human dignity) whereas negative numbers indicate violation of those principles

(e.g., coercion of choice or harm to human dignity). The main insight from Figure 3 is the wide

variance in the morality assessment scores. The majority of ratings is on the positive side of the

spectrum; for example, in most cases respondents saw at least some moderate benefit for donors,

moderate respect of autonomy or dignity, and some fairness toward patients and donors.

However, a substantial share of the ratings is on the negative side, thus indicating a perception of

violation of certain moral principles.

22

Figure 3: Distribution of ethical principles scores

A. Benefit to the donor (vs. exploitation)

B. Respect (vs. limit) of individual autonomy

C. Fully informed choice (vs. undue influence)

D. Fairness (vs. unfairness) to donors

E. Fairness (vs. unfairness) to patients

F. Promotion (vs. violation) of dignity

Notes: The graphs report the distribution of ratings for each of the six ethical principles that we asked the participants to consider. We aggregate all eight paid-donor systems and five kidney supply levels, and included the distribution of the corresponding scores for the current system. We grouped the 21 possible scores (from -10 to 10) into seven groups.

There is variation also for the current, unpaid-donor system. For example, about 30% of the

respondents indicated that the current system at least moderately exploits donors (score of -1 or

lower), and a similar share saw it as at least moderately unfair to patients. Moreover, there is a

23

high positive correlation between the ratings of the six principles both for the alternative systems

and for the current system. However, the individual-level correlation between the paid-donor and

current system ratings of a given principle is close to zero for all six moral features.

Table 4 and Figure 4 show the relationship between moral views and three factors: the supply

gains from the paid-donor system over the current system, the features of the eight alternative

systems, and the respondent types that we identified in Section 4.2 above. Our objective here is

to begin to assess whether a person’s overall reaction to a paid-donor system depends on moral

concerns, and the extent to which the respondents’ different moral views relate to the number of

transplants that the system makes possible (i.e., its outcome), to the system’s features (i.e., the

process through which organs for transplantation are obtained and distributed), or to both. For

these analyses, we constructed a relative measure of moral judgment, that is, a measure of moral

concerns toward paid-donor systems as compared to the current one. From each of the morality

scores attributed by the respondents to their assigned paid-donor system at each kidney supply

level, we subtracted the corresponding rating that the subject gave to the current system

(respondents rated the current system only at the current kidney supply level), and multiplied this

difference by -1. Thus, positive (negative) values of these relative moral concern scores indicate

that a respondent considered their assigned paid-donor system at a given supply level as being

less (more) consistent with a particular ethical principle than the current system. Because of the

high correlation, at the individual-supply level, between the six differences (or measures of

relative moral concerns), in some analyses we consider their average.22

Table 4 reports estimates from regressions of the relative concerns for each of the six moral

principles (and, in the last column, of their average) on supply increases (again in linear forms

and expressed as percentage points), as well as on indicators for the identity of the payer, the

type of payment, and the amount of the payment. Overall, supply increases have a small,

negative correlation with moral concerns. Cash payments increase all six moral concerns,

although not significantly in the case of undue influence. Payments by the kidney recipients

significantly increase concerns that choices may not be fully free, that the system would be

unfair to patients, cause undue influence, and that it would be disrespectful of human dignity;

and the large payment amount significantly increases concerns related to the system’s fairness to

the kidney recipients. In fact, the features of the system have an especially strong effect on

concerns for fairness toward the kidney recipient: private payments in particular increase this

concern, but also payments in cash and of higher amount correlate strongly with higher concerns

for fairness.

22 A principal component analysis identified one factor as subsuming the six measures, and the values of these factors

exhibit a correlation with the average of over 0.99.

24

Table 4: The effect of outcomes and procedural features on moral concerns

Notes: The table reports coefficient estimates from linear regressions of moral concerns on the hypothesized supply increase and binary indicators for the three features of each system: level of payment (1 for $100,000, 0 for $30,000), type of payment (1 for cash, 0 for non-cash), and identity of the payer (1 for private payments from the recipient (PVT), 0 for payments from a public agency). Standard errors, clustered at the respondent level (the regressions include the 1,276 participants who received the ethical principles module), are in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

In Figure 4, finally, each group of five columns reports the respondents’ average moral

concerns at each level of hypothesized kidney supply increases, according to a respondent’s type

as determined by their pattern of support for a paid-donor system. Moral concerns are very

different for individuals who are opposed to payments regardless of kidneys procured, those in

favor of payments at all kidney supply levels, and for individuals who increased or decreased

their support with increasing supply gains. For individuals with strong preferences (always in

favor or against), moral views correlate only weakly with the supply levels, suggesting that these

individuals are in favor or against a paid-donor system for reasons related to their procedural

features. For respondents who started out opposing the paid-donor system but switched to being

in favor at higher supply levels, moral concerns are substantially lower at higher supply levels,

whereas the opposite pattern characterizes the participants who decreased their support for

payments as the kidney supply increased.23 We also performed analyses of variance for this

summary measure of moral concerns, first by system and supply level, and then by system,

supply level, and type of respondent. A model with the eight paid-donor systems and supply

levels explains about 2.4% of the variance in moral concerns, with the systems accounting for

93% of the explained variation and organ supply only 7%. Adding indicators for the respondent

23 The averages in the graph aggregate the eight different systems. We report separate analyses for each different system in

the Appendix (Figure B3); the aggregate findings in Figure 4 here reproduce qualitatively for each system.

Outcome variable:Concerns for

exploitation

Concerns for lack of

autonomous choice

Concerns for

undue

influence

Concerns for

fairness to

donors

Concerns for

fairness to

patients

Concerns for

harm to human

dignity

Avg. rate of moral

concerns

Regressors: (1) (2) (3) (4) (5) (6) (7)

Supply increase (%pts.) -0.003 -0.014*** -0.013*** -0.003 -0.024*** -0.016*** -0.012***

(0.002) (0.002) (0.002) (0.002) (0.003) (0.002) (0.002)

Cash 0.832** 0.783** 0.618 0.906** 1.089*** 0.983*** 0.868***

(0.366) (0.354) (0.379) (0.392) (0.367) (0.355) (0.291)

PVT 0.401 1.167*** 1.300*** 0.571 3.235*** 0.958*** 1.272***

(0.367) (0.354) (0.380) (0.393) (0.371) (0.356) (0.291)

$100K -0.507 0.159 0.231 -0.051 0.784** 0.094 0.118

(0.364) (0.353) (0.378) (0.390) (0.365) (0.353) (0.289)

Constant -3.246*** -1.462*** -0.071 -3.383*** -2.973*** 0.395 -1.790***

(0.349) (0.342) (0.356) (0.380) (0.341) (0.333) (0.274)

Observations 6,380 6,380 6,380 6,380 6,380 6,380 6,380

R-squared 0.006 0.012 0.011 0.005 0.062 0.012 0.021

25

types increases the share of explained variance to 32%, with these types accounting for over 92%

of the explained variance.24

Figure 4: Average moral concerns by type of respondent and supply increase

Notes: Each group of five columns reports the respondents’ average ethical concerns at each level of hypothesized kidney supply increases, according to a respondent’s type as determined by their pattern of support for the paid-donor system. The transplant increases are expressed in thousands. “Always opposed” indicates individuals who did not support the alternative system at any supply level. The “always in favor” participants expressed support for the alternative systems for all five supply increases. The respondents in the “from opposed to favor” group are those who opposed the alternative system at lower level of hypothesized supply, and then switched to supporting it. The “from in favor to opposed” group includes the individuals who supported the alternative systems at lower supply levels, and switched to opposing it at higher levels. The measure of ethical concerns is the difference between the rating that a respondent, at a given supply level, gave to a particular principle with reference to the paid-donor system, and the rating of the same principle for the current system. Both ratings could vary between -10 and +10, with negative scores indicating violation of moral principles positive scores indicating consistence with moral principles. After multiplying the differences by -1, we obtained scores that represent increased concerns with regards to a particular issue (e.g. exploitation or unfairness). The graph in this figure reports, on the vertical axis, the mean of concerns, i.e. the average of the differences for the six principles.

4.4 Moral concerns are major determinants of the choice to support paid-donor systems

Figure 5 shows the likelihood of supporting the paid-donor system at each supply level, in three

different cases, according to whether the average moral concerns for the paid-donor system was

(1) greater than 1 (“higher moral concerns”), (2) less than -1 (“lower moral concerns”), and

24 The R-squared of the regressions in Table 4 is also small (the largest is 0.062, obtained in the regression of concerns for

fairness to the patient). Individual-specific determinants as expressed by the “preference type” are much stronger

predictors of one’s moral concerns than the features of a system per se.

26

(between -1 and 1 (“similar moral concerns”).25 The graph shows that moral views strongly

correlate with preferences for a paid donor system. Conditional on moral views, larger supply

gains also increase support for paid-donor systems, although the relatively large shares of

respondents who approve or disapprove of the paid-donor system regardless of supply flatten

these gradients.26

Figure 5: Support for a paid-donor system, moral concerns and supply level

Notes: The columns in this graphs represent the percentage of choices in favor of a paid-donor system, for each combination of hypothesized supply gain over the current system, and weather the average moral score that a respondent, for a system and a given supply gain, gave to the paid donor system was higher than (lower moral concerns), lower than (higher moral concerns) or equal to the score to the current system.

Tables 5 and 6 report estimates from linear probability regressions where, again, the outcome

variable is an indicator for whether a respondent expressed support their assigned paid-donor

system at a given supply level.27 In Table 5, the model whose estimates are reported in column

(1) included as right-hand-side variables the supply increases and the indicators for the level and

type of payment and the identity of the payer; in column (2) we added a dummy for whether the

respondents received the morality assessment module; and in column (3) we included

25 Because we calculated the summary measure of moral concerns as the difference between two averages of six scores

(see Section 4.3 above), it is unlikely that this difference will be precisely zero; therefore we allowed for a narrow interval

(of a distribution with values between -20 and 20) to represent similar moral views that a respondent holds between a paid

donor system (at a given kidney supply level) and the current system. 26 Figure B4 in the Appendix shows the gradients for the “from opposed to in favor” and “from favor to opposed” groups. 27 Table B3 in the Appendix reports versions of the regressions from Tables 5 and 6 including controls for the respondents’

socio-demographic characteristics. The sign, magnitude, and statistical significance of our coefficients of interest are

virtually unchanged.

Higher moral concerns

Similar moral concerns

Lower moral concerns

0 %pts. 14 %pts. 25 %pts. 38 %pts. 50 %pts.

23.8% 22.5%

32.2%28.8% 29.2%

55.8% 57.5%61.6%

61.6% 64.4%

79.2%

84.8% 86.7% 86.9%89.4%

Supply increases (% points over current system)

27

interactions of the morality module dummy with the system’s features and supply level. Before

analyzing the relationship between moral views and support for a system, we wanted to assess

whether the mere presence of questions about moral principles would affect individuals’

preferences for paying donors; the fact that compensation for organ donors is a morally charged

and controversial topic suggests that making moral considerations salient may affect choices,

unless individuals already consider moral considerations even without an explicit prompt.

Table 5: Salience of ethical issues and support for paid donor systems

Notes: The table reports the coefficient estimates from linear regressions of the support for a system on the assumed supply increase, binary indicators for each of the eight systems or binary indicators for the three features of each system, and a binary indicator for whether a respondent also received the moral principles module. The estimates in column (1) are the same as in column (2) of Table 3 above, where we only included supply increases and the three system feature dummies. In column (2), the estimates are from a model that included the indicator for receiving the moral principles module, and the specification whose estimates are in column (3) also included interaction terms between the moral principle module indicators and both the supply level and the system feature dummies. Standard errors, clustered at the respondent level (the regressions include 2,666 participants), are in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

The figures in Table 5 indicate that the participants’ response to supply gains and to the

systems’ procedural features were broadly similar with and without the morality assessment

Outcome variable:

Regressors: (1) (2) (3)

Supply increase (%pts.) 0.264*** 0.264*** 0.316***

(0.018) (0.018) (0.026)

Cash -1.591 -1.670 -1.193

(1.535) (1.534) (2.071)

PVT -15.026*** -15.114*** -15.862***

(1.542) (1.541) (2.077)

$100K -1.067 -1.047 -1.204

(1.538) (1.536) (2.074)

Morality module -3.911** -1.608

(1.538) (2.973)

Supply increase (%pts.) x Morality module -0.108***

(0.037)

Cash x Morality module -0.955

(3.079)

PVT x Morality module 1.547

(3.094)

$100K x Morality module 0.362

(3.082)

Constant 66.659*** 68.603*** 67.490***

(1.481) (1.640) (2.000)

Observations 13,330 13,330 13,330R-squared 0.035 0.036 0.037

Favor for alternative system (=100 if in favor, 0 if opposed)

28

module. However, prompting the respondents to express their opinions about the systems’

morality made them, on average, 3.9 percentage-point less likely to express favor for a paid-

donor system (column 2); moreover, it also made them less sensitive to supply changes, as

shown by the negative and significant estimate of the coefficient on the interaction term between

the supply variable and the indicator for the moral principles module (column 3).28 Thus,

although the results are broadly consistent with a pre-existing familiarity of the respondents with

the terms in which the public debate typically frames these and other, similarly controversial

issues (e.g., gestational surrogacy, compensation for plasma donors and so on), making morality

considerations salient does have some independent effect. In the light of these results, and

because the actual debate about the legalization of payments for kidney donors focuses on both

moral arguments and supply effects, we believe that making both sets of issues salient to the

respondents is an appropriate choice.

Table 6 reports estimates for the sample of 1,276 participants who received the morality

assessment module (more generally, most of the analyses below focus on this subsample). The

regressors in this case include the relative moral concerns for each of the six principles (columns

(1) through (4)), or the average of the six (columns (5) and (6)). The inclusion of the ethical

judgments among the covariates substantially increases the explanatory power of the regression

models as compared to those whose estimates are in Table 5 above. Supply gains still show a

strong positive effect on the support for a paid-donor system; however, the marginal effect, as

compared with column (3) of Table 5, declines from about 0.21 to 0.14-0.16, suggesting that, in

part, moral considerations drive responses to supply gains. There still is a large, negative impact

of the identity of the payer on the support for a system, but the point estimate is, in absolute

value, only about half as that from the specification without the moral views measures (compare

column 3 in Table 5 with column 1 in Table 6). This is a strong indication that the opposition to

direct payments by kidney recipients, the strongest feature affecting individuals’ choices, has

moral foundations. In particular, the estimates (columns 3 and 4 of Table 6) indicate that

respondents worry that a private payment system would be unfair to kidney recipients, consistent

with the findings on the determinants of moral concerns that we reported in section 4.4 above.

Because each respondent expressed their own ethical assessments at each of the five supply

levels, we can also perform individual fixed-effects regressions. Columns (2) and (6) of Table 6

show that the introduction of fixed effects does not lead to substantial changes in the estimates:

the sign, magnitude, and statistical significance of the estimated coefficients are in most cases

28 Appendix Figure B2 shows that the subset of respondents who received the ethics assessment module were a few

percentage points more likely to always oppose payments (23.4% vs. 19.3%) and correspondingly less likely to be willing

to switch from opposed to in favor (15.4% vs. 19.6%). The fractions of always in favor of payments, instead, were similar

among respondents with and without the morality assessment module (45.1% vs. 46.4%).

29

similar to those from specifications without fixed effects. Regressions that include individual

fixed effects also explain a much larger share of the variance in the support for paid-donor

systems. This, again, is consistent with attitudes toward compensation being highly

heterogeneous in the population.29 We further explore this aspect below.

Table 6: The impact of supply and moral considerations on the support of a paid-donor system

Notes: The table reports the coefficient estimates from linear regressions of the support for a system on the hypothesized supply increase, binary indicators for each of the eight systems or binary indicators for the three features of each system, and measures for the relative moral concerns for the assigned paid donor system. The specifications whose estimates are in columns (5) and (6) include the average relative moral concerns over the six principles. Standard errors, clustered at the respondent level (the regressions include 1,276 participants), are in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

29 Note again that we can identify the coefficient on supply gains because about 25 percent of respondents switched

support at different levels of these gains. These estimates are therefore an indication of the overall shift in the share of

individuals who would favor payments, including those who are always in favor regardless of the assumed supply level.

Outcome variable:

Regressors: (1) (2) (3) (4) (5) (6)

Supply increase (%pts.) 0.146*** 0.136*** 0.145*** 0.160*** 0.153*** 0.147***

(0.024) (0.027) (0.025) (0.025) (0.024) (0.027)

Cash 1.956 0.676 1.548 1.817

(1.862) (2.072) (1.879) (1.874)

PVT -8.054*** -5.927*** -10.378*** -8.509***

(1.900) (2.115) (1.887) (1.888)

$100K -0.083 1.191 -0.736 -0.302

(1.855) (2.064) (1.868) (1.865)

Concerns for exploitation -0.575*** -0.407* -0.572***

(0.164) (0.227) (0.167)

Concerns fo lack of autonomous choice -0.381** -0.304 -0.519***

(0.165) (0.247) (0.167)

Concerns for undue influence -0.802*** -0.974*** -0.845***

(0.162) (0.239) (0.164)

Concerns for fairness to donors -0.587*** -0.846*** -0.734***

(0.177) (0.271) (0.174)

Concerns for fairness to patients -0.905*** -1.528*** -2.593***

(0.144) (0.237) (0.115)

Concerns for harm to human dignity -1.338*** -0.913*** -1.655***

(0.180) (0.282) (0.170)

Avg. rate of moral concerns -4.565*** -5.033***

(0.132) (0.430)

Constant 59.256*** 55.447*** 58.174*** 61.378*** 57.711*** 54.108***

(2.028) (1.313) (2.104) (2.004) (1.884) (0.772)

Individual fixed effects No Yes No No No Yes

Observations 6,380 6,380 6,380 6,380 6,380 6,380

R-squared 0.289 0.758 0.167 0.278 0.283 0.756

Favor for alternative system (=100 if in favor, 0 if opposed)

30

We can use the estimates in column (6) to compare the overall effect of supply

considerations and moral concerns, with the caveat that, because of the (ever slight) correlation

between supply gains and moral views, these comparisons should be taken with caution. For

example, an increase in average moral concerns by 1.5 points, i.e. roughly the difference

between a system with $30,000 cash payments by a public agency and one with the same amount

and type of payment from the organ recipient, would reduce overall support for payments by

about 5*1.5=7.5 percentage points; in absolute terms, one could obtain an equivalent increase in

support rates with a supply gain of 7.5/0.15 = roughly 50 percentage points, similar to the

amount that we saw in the raw data from Figure 2.

4.5 Attitudes toward payments for kidney donors correlate with broader moral views

To investigate the determinants of the divergent views that we documented, and to better

understand the role of ethical considerations, in this section we explore whether and how

attitudes toward paid-donor systems correlate with the respondents’ “moral foundations”, which

we assessed through a set of modules from Graham et al. (2011), as well as with their political

and religious beliefs.

Figure 6 reports graphs with the distribution of respondents across the five patterns of

choices that we defined above, according to their religious and political views, as well as their

responses to questions in the moral foundations module. With reference to religious views, we

separate the respondents who stated that they were agnostic or atheists (about 12.9% of the

sample) from those who reported being religious (pooling together different religious

allegiances). We then analyze the distribution across the five types for the respondents who

reported being liberal or being conservative on economic and social issues. Attitudes toward

religion arguably relate to an individual’s ethical views; moreover, recent studies showed a link

between religious views and economic decisions and outcomes (Bénabou, Ticchi and Vindigni

2016; Benjamin, Choi and Fisher 2016). Also, views on the organization of the economy and

society may relate to opinions about the reach of market transactions. Regarding more direct

measures of moral views, we considered a subset of the moral foundation questions that we

asked the participants (we decided to ask a larger set of questions to obfuscate the issues in

which we were especially interested). First, we administered a “moral foundations” module with

eleven values, and focused on compassion, freedom, pleasure, pragmatism, giving, and tradition.

Individuals who value compassion highly may be more responsive to increases in kidney supply

and as such be less opposed to paying donors; this attitude may also characterize respondents for

whom freedom is a particularly important value, as well as those who hold values such as

pleasure (potentially related to utilitarian views) and pragmatism in high regard. To the extent

that people who value tradition also have a preference for the status quo, then one might expect

31

these individuals to be more opposed to payments. Finally, a high consideration of giving as a

moral value may lead to disapproving of payments if giving is interpreted as a gratuitous activity,

but also to supporting payments if they lead to more giving. Following Graham et al. (2011), we

used a 9-point scale (from -1 = opposed to my values, to 7 = of supreme importance) for

questions about moral values in the survey. In order to have subsamples of meaningful size, for

each value we grouped individuals according to whether they gave a high score to a value (6 or

7) or medium/low (5 and below). In general, most individuals scored the values relatively high

(the modal category in all cases is 5 or higher) and low scores were rare, thus we wanted to

isolate those who scored a value at the extreme. The share of respondents who gave a score of 6

or 7 ranges from 19.2% (for pleasure) to 56.8% (for compassion). Next, we used a vignette (from

the same source as the moral foundations questions) describing a moral dilemma that would

measure whether a respondent is characterized by deontological or consequentialist/utilitarian

preferences. The dilemma consisted in deciding whether to open a hatch in a sunken submarine

that would result in certain death for one crew member but that would save the rest of the crew.

Stating that one should not sacrifice a life even if doing so would save several other lives is an

indication of a deontological view, whereas agreeing that the right choice is to sacrifice that

single person is an indication of a more consequentialist or utilitarian view. In our context,

respondents who gave a deontological answer to the vignette would more likely be opposed to

payments to organ donors regardless of how many additional lives may be saved as a result.30

Respondents, whether they are religious or not, do not differ much in their overall attitudes

toward a paid donor system.31 However, individuals with conservative views about society and

the economy more frequently oppose payments than those with liberal economic and social

views. A stark difference emerges in relation to the response to the ethical dilemma in the

“submarine” vignette. The distribution of the individuals who gave the deontological response

shows a substantially higher share of participants who are opposed to payments regardless of the

supply gains. A high importance that a respondent gives to such values as compassion, freedom,

giving, and pleasure correlates with a greater likelihood that the individual is strongly in favor of

payments to kidney donors.

30 In section A of the Appendix we report details on the moral foundations questions and the vignette. 31 In the graphs we report the distribution of types after excluding the fifth groups (“others”). Because of the lack of

discernible patterns in the choices of these participants, we concluded that it was not insightful to consider their underlying

moral views for the purposes of the analysis here.

32

Figure 6: Distribution of paid-donor support types by religiosity, political views, and moral foundations

Religiosity

Economic policy views

Social policy views

Vignette response

Importance of Pragmatism

Importance of Tradition

Importance of Freedom

Importance of Giving

Importance of Pleasure

Importance of Compassion

33

We obtain additional evidence that the attitudes toward legalizing payments, and the role that

supply gains play in determining these attitudes, correlate with broader moral views from the

analysis of the time that respondents took to fill our survey, and from an analysis of the open

comments that respondents left at the end of the survey. Previous research shows that

deontological judgments tend to be faster than consequentialist ones (Sunstein 2014). Consistent

with this idea, we find that the respondents who oppose payments regardless of supply gains

were indeed faster in completing the survey than the other two largest groups or types of

respondents (those always in favor and those who went from opposing to supporting

compensation for higher supply gains). We find comparable differences when we focus on the

time participants took to respond to the “submarine” vignette, with the individuals always

opposed to payments being faster to complete this part of the survey, and individuals

recommending to not kill the character (an answer that the literature interprets as deontological)

having a shorter response time. We report graphical representations of these findings in the

Appendix (Figures B5 and B6).

In the Appendix (Tables B4 and B5) we also report an analysis of the open comments that

the survey participants could leave at the end of the questions. Of the 2,666 survey respondents,

330 left some comment. We asked three independent raters to classify this feedback into a set of

categories, including expressions of opposition to paying donors (73 comments), support for

donor payments (32), appreciation for the topic and the survey (111 comments), personal

experience (15), or other (99).32 We find that the nature of the comments correlates strongly with

the participants’ attitudes toward paid-donor systems; respondents who left a comment

expressing opposition to organ donor payments are much less likely to be in favor of

compensating donors, and less likely to donate money to the pro-compensation foundation.

Moreover, participants who were always opposed to organ donor payments were 5.9 percentage

points more likely to leave a comment (corresponding to a 51% increase over the baseline). This

is consistent with the idea that individuals who feel more strongly about an issue are also likely

to be more vocal. In contrast, the features of the paid-donor system assigned to the respondents,

and the prompt to think about ethical issues do not correlate with the likelihood that the

respondents left a comment.

Taken together, the evidence that we just presented strongly suggests that the respondents’

attitudes towards paying organ donors as we measured them relate to a broad set of moral values

as commonly assessed in moral and social psychology.

32 591 respondents wrote something in the space provided; however, 236 wrote “no comment” (or equivalent expressions)

and 29 typed some random characters. We found strong concordance across the three raters’ classifications and used a

majority rule (i.e., two out of three) to assign comments to categories. We assigned comments without a majority to the

“Other comments” category.

34

4.6 Stated preferences are correlated with the incentive-compatible choice to donate to an

organization that that supports the expansion of allowable payments to organ donors

In our last set of regression analyses we use the respondents’ choice in our incentivized donation

experiment as the dependent variable. The estimates in Table 7 below corroborate our

interpretation of the findings from the analysis above; in particular, there is strong consistency

between the overall stated attitudes toward paid-donor systems, and the respondents’ incentive-

compatible donation behavior.

Columns (1) through (4) report linear regression estimates of the choice to donate to the

American Transplant Foundation (ATF), which the respondents knew supported an expansion of

allowable forms of organ donor compensation. The regressors include the features of the paid-

donor system that the donors assessed in the stated support experiment, an indicator for whether

a respondent received the moral principles module, the respondent’s overall relative moral

concern, and in a separate regression (column 4), indicators for the respondent’s “type” as

expressed by their pattern of support for the paid donor system across the five kidney supply

levels. Column (1) shows that respondents who were assigned to a system with payments by the

kidney recipient have a lower propensity to donate to ATF. This confirms the crucial role of a

feature that, as we showed above, respondents did not prefer, and that prompted greater aversion

toward payments. The estimates in column (2) reveal that the exposure to questions about the

morality of a paid-donor system does not correlate with the decision to donate to the ATF,

consistent with our findings in Section 4.3 above that the morality module only had a limited

effect on respondents’ attitudes toward payments. We then limited the sample to the participants

who received the morality assessment module, and included in the regression a measure of the

respondents’ overall moral concerns towards the paid-donor system; as column (3) shows, the

estimated coefficient is negative and statistically significant, and the estimated coefficient on the

private payment indicator loses statistical significance (and is smaller). This is consistent with

moral concerns influencing attitudes, and with the aversion to private payments depending on

moral considerations. The estimates in column (4) simply represent average donation rates for

the five types of respondents that we identified.33 Participants who opposed payments regardless

of the supply gains are 21.5 percentage points less likely to donate to ATF, whereas those who

switched from being opposed to being in favor at some non-zero supply gain, and even more so

those who expressed support at any supply gain, are 13.2 and 20.3 percentage points more likely

to donate to ATF, respectively.

33 The unclassified “others” type is the omitted category (their donation rate was close to the overall average of 51%).

35

Table 7: Donation behavior

Notes: The table reports the estimates from linear regressions of the choice to donate to the American Transplant Foundation (ATF) or to the National Kidney Foundation (NKF), expressed as binary (0-1) indicator, on the following covariates: indicators for the features of the paid-donor system that respondents considered in the survey; an indicator for whether a respondents received also the morality principles module; a summary measure of the moral concerns the participants expressed for the paid-donor system (relative moral concerns averaged over the six principles and five supply levels); and overall respondent attitudes toward a paid-donor system as expressed by their pattern of stated support for different supply levels. *** p<0.01, ** p<0.05, * p<0.1.

Columns (5) through (8) report the estimates with the decision to donate to the National

Kidney Foundation (NKF), an organization that opposes payments to donors, as the outcome

variable. Although the smaller sample size makes it difficult to advance clear inferences, there

are some noteworthy differences with respect to the correlates of donation to ATF. First, higher

moral concerns for paid-donor systems correlate positively with the likelihood to donate to the

NKF. Second, respondents opposed to donor payments regardless of the supply gains, as well as

those who switched from supporting payments at low supply gains to opposing them at higher

supply gains are more likely to donate to the NKF than the other groups of participants.34

These findings are consistent with those from our survey, and with the interpretation that we

gave to those results. We see this as an important corroboration for our stated preferences survey.

34 Appendix Table B6 reports regressions that include controls for the respondents’ ethical assessments of the paid-donor

systems, their moral foundations, and the vignette. Individuals with deontological preferences are significantly less likely

to donate to ATF, individuals who place high importance on the principle of compassion are more likely to donate to ATF

and NKF, and, perhaps surprisingly, respondents with a high value for tradition are less likely to donate to the NKF.

Outcome variable:

Regressors: (1) (2) (3) (4) (5) (6) (7) (8)

Cash -0.036* -0.037* -0.025 0.024 0.021 -0.040

(0.022) (0.022) (0.031) (0.043) (0.043) (0.062)

PVT -0.049** -0.049** -0.033 -0.024 -0.021 0.022

(0.022) (0.022) (0.031) (0.043) (0.043) (0.062)

$100K 0.003 0.002 -0.014 0.034 0.030 0.063

(0.022) (0.022) (0.030) (0.043) (0.043) (0.061)

Morality module -0.015 0.056

(0.022) (0.043)

Mean rate of moral concerns -0.024*** 0.019***

(over all supply levels) (0.003) (0.006)

Always opposed -0.215*** 0.102

(0.038) (0.075)

From opposed to favor 0.132*** 0.044

(0.042) (0.077)

Always in favor 0.203*** 0.035

(0.037) (0.067)

From favor to opposed 0.056 0.292**

(0.060) (0.122)

Constant 0.542*** 0.550*** 0.503*** 0.429*** 0.390*** 0.367*** 0.415*** 0.358***

(0.021) (0.024) (0.030) (0.033) (0.043) (0.048) (0.066) (0.059)

Observations 2,130 2,130 1,018 2,130 498 498 237 536

R-squared 0.004 0.004 0.068 0.106 0.002 0.005 0.058 0.013

Donation to ATF Donation to NKF

36

4.7 The results are robust to controlling for perceived consequentiality and beliefs about

the popularity of organ donor payments in the population

The issue of incentive compatibility was one of the concerns that we discussed in Section 3.1.4

above and addressed with the monetary donation experiment that we just described. In Section

3.1.4 we also considered two other topics related to the reliability and interpretation of the data:

the perceived consequentiality of the responses, and the potential role of social pressure,

conformity and “strategic” responses. Table B7 in the Appendix reports regression estimates

from the same model as in Table 6 above, where we (a) add to the regressors our measures of

participants’ perceived relevance of the topic, confidence in their answers, and consequentiality

of the survey, and (b) limit the sample to the respondents who perceived the topic as important,

were confident of their responses, and attributed at least some level of consequentiality to the

survey. Our main estimates of interest, namely the change in support at different supply gains

and for different levels of moral concerns, and the effect of the systems’ procedural features, are

very similar to the ones reported in Table 6. In general, individuals who found the topic more

important, were more confident in their responses, and perceived the survey as consequential see

the legalization of donor compensation more favorably. The estimates reported in Table B8 in

the Appendix come from regressions where we added controls for the participants’ beliefs about

the popularity of payments to organ donors in the US population, and the importance that

individuals attributed to social recognition as a guiding value in their life. Again, the inclusion of

these variables (or restricting the sample according to how the respondents answered these

questions) does not meaningfully change any of the main estimates of interest.

5. Discussion

Cultural and moral beliefs contribute to the identity and cohesion of a society (Durkheim 1893)

and, as such, have significant impact on economic decisions and outcomes (Alesina and Giuliano

2015; Bisin and Verdier 2011; Guiso, Sapienza and Zingales 2006). Studying the nature of the

aversion or support to certain market transactions, and in particular the role that cultural and

moral beliefs play in determining these views, provides insights into how to address policy-

relevant problems, and into whether policymakers can alleviate ethical concerns via institutional

design. Although proper policy design can address some ethical concerns, others, such as the

perception that a transaction violates human dignity per se, are less amenable to be addressed; if a

population feels strongly about them, this may explain why societies prohibit certain activities

even if these prohibitions reduce individuals’ freedom and private welfare.

In this paper we studied the nature of preferences of Americans toward paying organ donors.

We wanted to assess the sources of aversion to legalizing donor compensation among the general

37

population, and whether opposition was absolute or conditional on material gains such as

increases in the number of kidney for transplants.

Our analysis offers four main insights. The first is the strong polarization of attitudes: large

proportions of respondents are either in favor or against paying organ donors, irrespective of the

size of hypothesized kidney supply gains. We also confirmed the evidence from stated

preferences with our analysis of incentive-compatible monetary donation choices. Strong

polarization of opinions is a recurring feature in topics that are ethically contentious.35

The second main insight is that moral considerations are a major determinant of these

opposing views. The broad similarity of (stated and revealed) preferences about paying organ

donors among subjects whom we prompted to express their moral views and those who were not

indicates that concerns about the potential violation of ethical principles were prominent in the

respondents’ minds regardless of the salience that we (the researchers) gave to them. Our

interpretation is consistent with the additional finding that attitudes and ethical views toward

donor compensation correlate with broader moral values that respondents reported being

important to them.

Third, despite this polarization, we found that the attitudes of about one-fifth of respondents

do vary depending on how many additional transplants a paid-donor system would make

possible. A higher kidney supply increases support for legalizing organ donor compensation for

these individuals, and also reduces their ethical concerns. Thus, although our evidence shows

that moral concerns do pose a constraint to introducing a price mechanism in this context, at the

population level positive supply effects of paying organ donors may significantly change societal

support for legalizing these payments. An implication of this finding for policy is that pilot

studies of compensation to organ donors would be useful to produce evidence on the potential

effects on the number of transplants. Without this evidence, a large share of Americans would

lack a crucial element to guide their preferences. The previous findings, however, suggest that

pilots should also evaluate whether paying organ donors violates ethical principles, for example

by exerting undue influence on the decision makers (Ambuehl 2018).

A fourth key finding is that attitudes toward paid-donor systems depend on certain

procedural features. In particular, there is a difference between the aversion to paying donors and

the aversion to having a recipient pay: a significant portion of the opposition to legalizing

compensation derives from whether payments come from the kidney recipient or from a third

party such as a public agency. In the former case, concerns for fairness to patients were much

more severe than in the latter. In fact, the respect of fairness in the allocation of organs was the

most relevant ethical principle for respondents among the six that we considered. This finding

35 See for example Mouw and Sobel (2011) on abortion.

38

offers an important insight to policymakers because it indicates that appropriate institutional

design can allay a major ethical concern.

The joint consideration, in our study, of the role of supply gains as well as of procedural

aspects of the organization of organ procurement and allocation, together with the investigation

of the ethical roots of attitudes toward these features, allowed for a detailed exploration of how

people perceive the moral limits of markets, how deep their differences in attitudes are, and

whether and how individuals make tradeoffs between potentially competing values. The

conceptual framework and experimental design that we adopted derived also from an attempt to

integrate an economic approach to these issues with insights from several other disciplines, such

as work in social psychology on moral foundations and studies in experimental philosophy. We

see this as a fruitful confluence, one that would enhance our understanding of the nature of

preferences in repugnant transactions and the reasons why societies may keep certain activities

and transactions out of the marketplace. In fact, we expect our methodology to be applicable to

other morally controversial transactions. For example, there is evidence that legalizing indoor

prostitution enhances social welfare by reducing violence and STD incidence (Cunningham and

Shah 2017). Similarly, abortion tends to be safer in countries where it is legal (Faúndes and Shah

2015). Our approach could be used to assess whether these or other welfare gains are sufficient

to induce a majority of voters to legalize a transaction in spite of ethical concerns.

39

References

Agranov, M. & Ortoleva, P. (2017). Stochastic choice and preferences for randomization. Journal of

Political Economy, 125(1), pp.40-68.

Akerlof, G. A., & Kranton, R. E. (2000). Economics and identity. Quarterly Journal of Economics, 715-

753.

Alesina, A. & Giuliano, P. (2015). Culture and institutions. Journal of Economic Literature, 53(4), 898-

944.

Alesina, A., Stantcheva, S., & Teso, E. (2018). Intergenerational mobility and preferences for

redistribution. American Economic Review, 108(2), 521-54.

Ambuehl, S. (2018). An Offer You Can't Refuse? Incentives change how we think, Working paper,

University of Toronto.

Ambuehl, S., Niederle, M., & Roth, A. E. (2015). More money, more problems? Can high pay be

coercive and repugnant? American Economic Review Papers and Proceedings, 105(5), 357-360.

Andreoni, J., & Sprenger, C. (2012). Risk preferences are not time preferences. American Economic

Review, 102(7), 3357-3376.

Bartling, B., Weber, R.A. & Yao, L. (2014). Do markets erode social responsibility? The Quarterly

Journal of Economics, 130(1), 219-266.

Basu, K. (2007). Coercion, Contract and the Limits of the Market. Social Choice and Welfare, 29 (4),

559-579.

Becker, G. S., & Elias, J. J. (2007). Introducing incentives in the market for live and cadaveric organ

donations. Journal of Economic Perspectives, 21(3), 3-24.

Bénabou, R., & Tirole, J. (2009). Over my dead body: Bargaining and the price of dignity. American

Economic Review, 99(2), 459-465.

Bénabou, R., & Tirole, J. (2011). Identity, morals, and taboos: Beliefs as assets. Quarterly Journal of

Economics, 126(2), 805-855.

Bénabou, R., Ticchi, D., & Vindigni, A. (2015). Forbidden fruits: the political economy of science,

religion, and growth. National Bureau of Economic Research working paper 21105.

Benjamin, D.J., Choi, J.J. & Fisher, G. (2016). Religious identity and economic behavior. Review of

Economics and Statistics, 98(4), 617-637.

Benjamin, D. J., Heffetz, O., Kimball, M. S., & Rees-Jones, A. (2012). What do you think would make

you happier? What do you think you would choose? American Economic Review, 102(5), 2083.

Benjamin, D. J., Heffetz, O., Kimball, M. S., & Rees-Jones, A. (2014). Can marginal rates of substitution

be inferred from happiness data? Evidence from residency choices. American Economic Review,

104(11), 3498-3528.

Benjamin, D. J., Kimball, M. S., Heffetz, O., & Szembrot, N. (2014). Beyond Happiness and Satisfaction:

Toward Well-Being Indices Based on Stated Preference. American Economic Review, 104(9), 2698-

2735.

Birnbacher, D. (1999). Ethics and social science: Which kind of co-operation? Ethical theory and moral

practice, 2(4), 319-336.

40

Bisin, A. & Verdier, T. (2011). The economics of cultural transmission and socialization. In Handbook of

social economics (Vol. 1, pp. 339-416). North-Holland.

Bonnefon, J.F., Shariff, A. & Rahwan, I., 2016. The social dilemma of autonomous vehicles. Science,

352(6293), pp.1573-1576.

Bursztyn, L., Callen, M., Ferman, B., Hasanain, A., & Yuchtman, N. (2015). Identifying Ideology:

Experimental Evidence on Anti-Americanism in Pakistan, Working paper.

Bursztyn, L., Egorov, G. & Fiorin, S., 2017. From extreme to mainstream: How social norms unravel.

National Bureau of Economic Research working paper 23415.

Carson, R. T. (2012). Contingent valuation: A practical alternative when prices aren't available. Journal

of Economic Perspectives, 26(4), 27-42.

Carson, R.T. & Groves, T. (2007). Incentive and informational properties of preference questions.

Environmental and resource economics, 37(1), 181-210.

Cibelli, K., 2017. The Effects of Respondent Commitment and Feedback on Response Quality in Online

Surveys, doctoral dissertation, University of Michigan.

Council of Europe (2015). Council of Europe Convention against Trafficking in Human Organs.

Cunningham, S. & Shah, M. (2017). Decriminalizing indoor prostitution: Implications for sexual violence

and public health. Review of Economic Studies, forthcoming.

Spital, A., Delmonico, F.L., Arnold, R. & Youngner, S.J. (2002). Ethical incentives--not payment--for

organ donation. New England Journal of Medicine, 347(17), 1382-1384..

Doris, J. & Stich, S., 2005. As a matter of fact: Empirical perspectives on ethics. The Oxford handbook of

contemporary philosophy, pp.114-152.

Durkheim, E., 1893. The Division of Labor in Society (Anthony Giddens ed., W.D. Halls, trans., 1984).

Falk, A., & Szech, N. (2013). Morals and markets. Science, 340(6133), 707-711.

Faúndes, A., & Shah, I. H. (2015). Evidence supporting broader access to safe legal

abortion. International Journal of Gynecology & Obstetrics, 131(S1).

Gibson, R., Tanner, C., & Wagner, A. F. (2013). Preferences for truthfulness: Heterogeneity among and

within individuals. American Economic Review, 103(1), 532-548.

Gneezy, U. (2005). Deception: The role of consequences. American Economic Review, 95(1), 384-394.

Graham, J., Nosek, B. A., Haidt, J., Iyer, R., Koleva, S. & Ditto, P. H. (2011). Mapping the moral

domain. Journal of Personality and Social Psychology, 101, 366-385.

Graham, J., Haidt, J., Koleva, S., Motyl, M., Iyer, R., Wojcik, S.P. & Ditto, P.H. (2013). Moral

foundations theory: The pragmatic validity of moral pluralism. Advances in experimental social

psychology 47, 55-130.

Grant, R. W. (2011). Strings attached: Untangling the ethics of incentives, Princeton University Press

Guiso, L., Sapienza, P. & Zingales, L. (2006). Does culture affect economic outcomes?. Journal of

Economic perspectives, 20(2), 23-48.

Haidt, J. (2007). The new synthesis in moral psychology. Science, 316(5827), 998-1002.

Halpern, S.D., Raz, A., Kohn, R., Rey, M., Asch, D.A. & Reese, P. (2010). Regulated payments for living

kidney donation: an empirical assessment of the ethical concerns. Annals of internal medicine,

152(6), 358-365.

41

Healy, K. & Krawiec, K.D., 2017. Repugnance Management and Transactions in the Body. American

Economic Review, 107(5), pp.86-90.

Held, P. J., McCormick, F., Ojo, A., & Roberts, J. P. (2016). A Cost‐ Benefit Analysis of Government

Compensation of Kidney Donors. American Journal of Transplantation, 16(3), 877–885.

Jacob, B.A. & Levitt, S.D., 2003. Rotten apples: An investigation of the prevalence and predictors of

teacher cheating. The Quarterly Journal of Economics, 118(3), pp.843-877.

Kerstein, S.J. (2009). Kantian condemnation of commerce in organs. Kennedy Institute of Ethics Journal,

19(2), 147-169.

Kessler, J. B., & Roth, A. E. (2012). Organ Allocation Policy and the Decision to Donate. American

Economic Review, 102(5), 2018-2047.

Kessler, J.B. & Roth, A.E., 2014. Don't take 'no' for an answer: An experiment with actual organ donor

registrations. National Bureau of Economic Research working paper 20378.

Knobe, J., Buckwalter, W., Nichols, S., Robbins, P., Sarkissian, H., & Sommers, T. (2012). Experimental

philosophy. Annual Review of Psychology, 63, 81-99.

Kuziemko, I., Norton, M. I., Saez, E., & Stantcheva, S. (2015). How elastic are preferences for

redistribution? Evidence from randomized survey experiments. American Economic Review, 105(4),

1478-1508.

Leider, S., & Roth, A. E. (2010). Kidneys for sale: Who disapproves, and why?. American Journal of

Transplantation, 10(5), 1221-1227.

Marshall, A. (1890). 1920. Principles of economics. London: Mac-Millan.

Martinelli, C., Parker, S.W., Pérez-Gea, A.C. & Rodrigo, R., 2018. Cheating and incentives: Learning

from a policy experiment. American Economic Journal: Economic Policy, 10(1), pp.298-325.

Molewijk, B., Stiggelbout, A.M., Otten, W., Dupuis, H.M. & Kievit, J. (2004). Empirical data and moral

theory. A plea for integrated empirical ethics. Medicine, Health Care and Philosophy, 7(1), 55-69.

Mouw, T. & Sobel, M.E., 2001. Culture wars and opinion polarization: the case of abortion. American

Journal of Sociology, 106(4), pp.913-943.

Niederle, M., & Roth, A. E. (2014). Philanthropically Funded Heroism Awards for Kidney Donors. Law

& Contemp. Probs., 77, 131.

Nuffield Council on Bioethics (1995). Human Tissue: Ethical and Legal Issues, London.

Radin, M. J. (1996). Contested commodities: the trouble with trade in sex, children, body parts, and other

things. Cambridge, MA: Harvard University Press.

Rippon, S., 2012. Imposing options on people in poverty: the harm of a live donor organ market. Journal

of medical ethics, pp.medethics-2011.

Roth, A. E. (2007). Repugnance as a Constraint on Markets. Journal of Economic Perspectives, 21(3),

37-58.

Roth, A., Sönmez, T., & Unver, U. (2004). Kidney Exchange. Quarterly Journal of Economics, 119(2),

457-488.

Sandel, M. J. (2012). What money can't buy: the moral limits of markets. Macmillan.

Satel, S. (2006). Organs for sale. The American.

42

Satz, D. (2008). The Moral Limits of Markets: The Case of Human Kidneys. In Proceedings of the

Aristotelian Society, 108(1), 269-288.

Satz, D., 2010. Why some things should not be for sale: The moral limits of markets. Oxford University

Press.

Sen, A. (1999). On ethics and economics. OUP Catalogue.

Shleifer, A. (2004). Does Competition Destroy Ethical Behavior? American Economic Review, 94(2),

414-418.

Stoler, A., Kessler, J. B., Ashkenazi, T., Roth, A. E., & Lavee, J. (2017). Incentivizing organ donor

registrations with organ allocation priority. Health Economics, 26(4), 500-510.

Sunstein, C.R., 2014. Is deontology a heuristic? On psychology, neuroscience, ethics, and law. The

Jerusalem Philosophical Quarterly, 63, 83-101.

Tanner, C., Medin, D. L., & Iliev, R. (2008). Influence of deontological versus consequentialist

orientations on act choices and framing effects: When principles are more important than

consequences. European Journal of Social Psychology, 38(5), 757-769.

Tetlock, P. E., Kristel, O. V., Elson, S. B., Green, M. C., & Lerner, J. S. (2000). The psychology of the

unthinkable: taboo trade-offs, forbidden base rates, and heretical counterfactuals. Journal of

Personality and Social Psychology, 78(5), 853.

United States Task Force on Organ Transplantation (1986). Issues and Recommendations. Task Force on

Organ Transplantation, US Department of Health and Human Services. Public Health Service, Health

Resources and Services Administration.

UNOS. 2017. https://unos.org/data/ Accessed 3/14/2018.

Vossler, C.A., Doyon, M. & Rondeau, D., 2012. Truth in consequentiality: theory and field evidence on

discrete choice experiments. American Economic Journal: Microeconomics, 4(4), pp.145-71.

World Health Organization (2004). World Health Assembly Resolution 57.18, Human organ and tissue

transplantation, 22 May 2004, http://www.who.int/gb/ebwha/pdf_files/WHA57/A57_R18-en.pdf.