EMPATHETIC CONCERN, ALTRUISM, AND THE PURSUIT OF DISTRIBUTIVE JUSTICE _____________________________________ A Thesis Presented to the Faculty of California State University, Fullerton _____________________________________ In Partial Fulfillment of the Requirements for the Degree Master of Arts in Economics _____________________________________ By Garret Ridinger Approved by:
iii
i
ABSTRACT
This paper explores the idea that empathetic concern has the potential to influence preferences for distributive justice. A theoretical model is proposed in order to examine conditions in which empathetic concern might influence individual preferences. These hypothetical results are tested empirically using data from the General Social Survey. During the years 2002 and 2004, the survey contained information on variables designed to measure empathetic concern. Using this measure, this paper examines the influence of empathetic concern on national issues like the distribution of wealth, Social Security, welfare, aid to the poor, and foreign aid as well as more local issues like giving food or money to the homeless, volunteering for charity, and giving money to charity. Overall, a general pattern emerges suggesting that empathetic concern has a statistically and economically significant influence on preferences for distributive justice. However, empathetic concern does not seem to have much of an influence on preferences for welfare or foreign aid.
ii
TABLE OF CONTENTS ABSTRACT...................................................................................................................
ii
LIST OF TABLES..........................................................................................................
v
LIST OF FIGURES........................................................................................................ vii ACKNOWLEDGMENTS.............................................................................................. viii Chapter 1. INTRODUCTION.................................................................................................
1
Justice.................................................................................................................... 3 Empathy................................................................................................................. 6 Altruism................................................................................................................. 11 2.
THEORY............................................................................................................... 16 Empathy and Altruism........................................................................................... Theoretical Model.................................................................................................. Preferences for Redistribution and Charitable Giving................................... Influence of General Beliefs...........................................................................
3.
DATA.................................................................................................................... 30 Measurement of Empathetic Concern................................................................... Descriptive Statistics............................................................................................. Independent Variables.................................................................................... Beliefs............................................................................................................. National Issues................................................................................................ Local Issues....................................................................................................
4.
16 18 22 28
31 34 37 40 41 47
ECONOMETRIC METHODS.............................................................................. 50 National Issues....................................................................................................... 50 Local Issues........................................................................................................... 51 Strategy and Potential Problems............................................................................ 52
5.
RESULTS AND DISCUSSION............................................................................ 57 iii
6.
Results for National Issues.................................................................................... Income Redistribution.................................................................................... Social Security................................................................................................ Foreign Aid..................................................................................................... Welfare........................................................................................................... Helping the Poor............................................................................................. Results for Beliefs.......................................................................................... Fairness........................................................................................................... Luck................................................................................................................ Trust................................................................................................................
57 61 64 66 68 70 73 73 77 81
Results for Local Issues......................................................................................... Giving to the Homeless.................................................................................. Giving to Charity............................................................................................ Volunteer........................................................................................................ Results for Beliefs.......................................................................................... Fairness........................................................................................................... Luck................................................................................................................ Trust................................................................................................................
85 88 90 91 92 93 96 97
CONCLUSION..................................................................................................... 99
REFERENCES............................................................................................................... 103
iv
LIST OF TABLES Table
Page
1.
Prisoner's Dilemma with Kin Selection.........................................................
15
2.
Description of Empathy Variables and Factor Analysis Results...................
31
3.
Description of Variables................................................................................
35
4.
Summary Statistics for Independent Variables..............................................
39
5.
Summary Statistics for Beliefs......................................................................
41
6.
Summary Statistics for Dependent Variables................................................
42
7A. Frequency Data for Income Redistribution...................................................
43
7B. Frequency Data for Social Security...............................................................
44
7C. Frequency Data for Foreign Aid....................................................................
45
7D. Frequency Data for Welfare..........................................................................
46
7E. Frequency Data for Helping the Poor............................................................
47
7F. Frequency Data for Giving to Homeless.......................................................
48
7G. Frequency Data for Giving to Charity...........................................................
49
7H. Frequency Data for Volunteer.......................................................................
49
8.
Ordered Probit Results: Coefficient Estimates for the National Issues.........
59
9.
Marginal Effects at the Average for Income Redistribution..........................
64
10.
Marginal Effects at the Average for Social Security.....................................
66
11.
Marginal Effects at the Average for Foreign Aid..........................................
68
v
12.
Marginal Effects at the Average for Welfare.................................................
70
13.
Marginal Effects at the Average for Helping the Poor..................................
72
14.
Ordered Probit Results: Coefficient Estimates Given Beliefs about Fairness ........................................................................................................
74
Ordered Probit Results: Coefficient Estimates Given Beliefs about Luck and Hard Work.....................................................................................
78
Ordered Probit Results: Coefficient Estimates Given Beliefs about Trust...............................................................................................................
83
Marginal Effects at the Average for Empathy on Income Redistribution Givin Beliefs..................................................................................................
84
Marginal Effects at the Average for Empathy on Social Security Given Beliefs............................................................................................................
84
Marginal Effects at the Average for Empathy on Foreign Aid Given Beliefs............................................................................................................
84
20.
Marginal Effects at the Average for Empathy on Welfare Given Beliefs.....
85
21.
Marginal Effects at the Average for Empathy on Helping the Poor Given Beliefs............................................................................................................
85
22.
Overall Probit Results: Coefficient Estimates for Local Issues.....................
87
23.
Marginal Effects at the Average for Probit Results.......................................
89
24.
Probit Results: Coefficient Estimates Given Beliefs about Fairness ........................................................................................................
94
Probit Results: Coefficient Estimates Given Beliefs about Luck and Hard Work.....................................................................................
96
Probit Results: Coefficient Estimates Given Beliefs about Trust...............................................................................................................
97
Marginal Effects at the average for Empathy on Giving to Homeless, Givetocharity, and Volunteer Given their Beliefs.........................................
98
15. 16. 17. 18 19.
25. 26. 27.
vi
LIST OF FIGURES Figure 1.
Page
Graph of the Sample Distribution of the Derived Factor for Empathetic Concern..........................................................................................................
vii
34
ACKNOWLEDGMENTS I would like to thank my committee for all their hard work and support. Professor Kristin Kleinjans, Professor Andrew Gill, and Professor Robert Michaels have given me great advice and suggestions that have improved this thesis immeasurably. I would like to give special thanks to Professor Kleinjans and Professor Gill. Professor Kleinjans has been instrumental in guiding me through the thesis process. She took the time to read through my extremely rough drafts and provide excellent ways for me to make improvements. In addition to her tireless help with my thesis, she has helped me grow throughout my time in the program. Whether showing me an interesting article to read or giving me useful advice about my academic future, her guidance has helped me in an uncountable number of ways. Since the first day I met Professor Gill, he has always been extremely helpful. Whether giving general advice or patiently explaining to me an econometric concept, his excellent teaching ability has made me a better researcher. Both Professor Kleinjans and Professor Gill have helped me reach my goal of continuing my education at the next level. For this and everything else they have done, I am extremely grateful. None of this would be possible without my family. Especially my parents, Mark and Cindy Hollingworth, whose support and guidance have given me the opportunity to pursue my dreams.
viii
CHAPTER 1 INTRODUCTION The sentiment of justice is yet fundamental to the human mind, and whatever dispute arouses the passions of men, the conflict is sure to rage, not so much as to the question "Is it wise?" as to the question "Is it right?" —Henry George, Progress and Poverty (1912) The problem of distributive justice comes down to the unfortunate fact that if we add up everything people want it would far exceed what society can actually produce. This issue is not new and has been something that societies have struggled with throughout history. Preferences for just allocations permeate throughout human life. Bloody wars are fought over pieces of land. People quit their jobs because they feel underpaid. Business partnerships dissolve when someone believes they are not getting their fair share. While conflict is evident, cooperation can emerge from the struggle. People donate time and money to help others in need. After hours an unpaid teacher works with a child who struggles to read. Neighbors band together to pay for their friend's medical operation. All of these situations are influenced by what people believe is just. Many theories have been proposed about the “best” way to allocate resources but ultimately any theory must deal with what people believe is fair. This is because distributive justice is at its core a practical enterprise. Justice theories are concerned with how goods should be allocated, but if the theory does not work in practice its adherence
1
2 might come into question. What people believe matters. Not because they are right, but due to the fact that they must live in any theory of justice. The choices and decisions they make have the potential to make that theory successful or relegate it to the footnotes of history. Adding to our understanding of individual preferences for justice and what influences these preferences is the goal of this paper. This paper explores the idea that empathetic concern might influence a person's preferences for distributive justice. Empathetic concern is the feeling of compassion or concern about the welfare of another person and the desire to help them.1,2 Are people who exhibit higher degrees of empathetic concern more likely to support policies that help the people they are concerned about? Will more empathetic people be more likely to support government aid to the poor? Are they more likely to donate to charity? Answers to these questions can help us understand why people differ in their opinions about just allocations. Using data from the General Social Survey, this paper attempts to examine if empathetic concern influences preferences for distributive justice. I will examine national issues like the distribution of wealth, Social Security, welfare, aid to the poor, and foreign aid as well as more local issues like giving food or money to the homeless, volunteering for charity, and giving money to charity. Looking at these different issues can help us paint a contextual picture of how and when empathetic concern influences individual preferences for justice. 1
Frans B.M de Waal, “Putting the Altruism Back Into Altruism: The Evolution of Empathy,” Annual Review of Psychology 59 (January 2008): 279-300. 2
Tania Singer and Nikolaus Steinbeis, “Differential Roles of Fairness- and Compassion-Based Motivations For Cooperation, Defection, and Punishment,” Annals of the New York Academy of Sciences 1167 (June 2009): 41-50.
3 Justice Recent research on distributive justice has focused on what people actually believe is fair. As a result of this research, a group of fairness principles has emerged. Not everyone adheres to each principle but there does seem to be some broad agreement on what people consider fair. Although there are many more, I will focus on the most commonly cited principles of need, efficiency, merit, and equality. Need is a concern for the least advantaged or worst off. Survey research suggests that a large percentage of people are willing to sacrifice efficiency in order to help people in need. In one study, German economic and business students were given the choice between educating one intelligent child or one handicapped child.3 They were subsequently asked to choose between educating two, three, and four intelligent children versus only one handicapped child. In the first year of the study, 72% of the individuals preferred to give the handicapped child the funding in all scenarios.4 Over time the percentage of people choosing to give the funding to the handicapped child in all situations decreased, but still remained the majority choice.5 This suggests that concerns for need are important factors in people's judgments about fair allocations. This does not mean that efficiency is not important to people. In one paid experiment, participants from a handful of different nations voted on a choice between a
3
Wulf Gaertner, A Primer in Social Choice Theory: Revised Edition (Lse Perspectives in Economic Analysis), Rev. ed. (London: Oxford University Press, USA, 2009), 163-79. 4
Ibid.
5
Ibid.
4 Pareto inferior and Pareto superior distribution.6 People were more likely to support Pareto improvements if they did not know what allocation they would end up receiving.7 When people did know what allocation they would be receiving, approximately 72% of people voted for the Pareto improvement when the person benefiting from the improvement was already receiving a larger allocation. That is, the majority of people were willing to vote for the more efficient outcome even if it only helped the rich get richer. Of course, this means that about 28% of the people did not support a Pareto improvement if the person who would benefit from it was already receiving a higher allocation than they were.8 That is, some of the people were willing to vote for the less efficient outcome to prevent the rich from getting richer. This suggests that for some people envy can possibly influences people's preferences for allocations. While this is a large percentage of people, the majority of people did choose the Pareto superior distribution. Merit is the idea that fair allocations should be based on individual behavior. It captures the idea that people should get what they deserve. For example, it would be unfair if one worker works harder than another but both receive the same pay. In one survey conducted by Konow, people were told about identical twins who were the same in "terms of physical and mental abilities," but one works harder than the other.9 As a 6
Ibid.
7
Ibid.
8
Ibid.
9
James Konow, “Which Is the Fairest One of All? A Positive Analysis of Justice Theories,” Journal of Economic Literature 41, no. 4 (December 2003): 1197-08.
5 result the twin that worked harder earned more money. 10 When asked if this was fair, 99% of the people thought it was.11 Equality is the idea that everyone should get the same share. Evidence suggests that people tend to choose equality when there is no easy way to find out who deserves more.12 A recent experiment run by Fehr, Bernhard, and Rochenbach, had children play a sharing game for candy.13 One child could choose to get one candy and give one candy to another child or could keep two and give the other nothing.14 For ages 3 to 4, most children chose selfishly with 8.7% choosing to share, but for ages 7 to 8, 50% of the children chose to share.15 The authors also found a strong preference in children to give more equally with people in their own social group.16 The results from these studies all point to the idea that what people feel is fair or just depends on the situation. This is important for this paper, because my suspicion is that empathetic concern might not have the same effect on all the issues I will look at.
10
Ibid.
11
Ibid.
12
Ibid.
13
Ernst Fehr, Helen Bernhard and Bettina Rochenbach, “Egalitarianism in Young Children,” Nature 454 (August 2008): 1079-83. 14
Ibid.
15
Ibid.
16
Ibid.
6 Empathy The idea that empathy might motivate or influence prosocial behavior is not new, as the following quote by Adam Smith in The Theory of Moral Sentiments can attest: How selfish soever man may be supposed, there are evidently some principles in his nature, which interest him in the fortune of others, and render their happiness necessary to him, though he derives nothing from it except the pleasure of seeing it. Of this kind is pity or compassion, the emotion which we feel for the misery of others, when we either see it, or are made to conceive it in a very lively manner. That we often derive sorrow from the sorrow of others, is a matter of fact too obvious to require any instances to prove it; for this sentiment, like all the other original passions of human nature, is by no means confined to the virtuous and humane, though they perhaps may feel it with the most exquisite sensibility. The greatest ruffian, the most hardened violator of the laws of society, is not altogether without it.17 Empathy is essential in order for us to understand others. There are many different ideas that have all been described as empathy. 18 Although the main interest of this paper is in empathetic concern, a brief description of related processes is necessary as many empathetic processes are interrelated. One of the most basic forms of empathy is emotional contagion. Emotional contagion is the ability to catch the feelings or emotions of another person.19 People do not need to be aware that they caught an emotion from another person. One example is that infants will often cry when they hear another infant cry.20 This process is somewhat related to the idea of mimicry. Mimicry is the idea that 17
Adam Smith, The Theory of Moral Sentiments (London: A. Millar, 1790), http://www.econlib.org/library/Smith/smMS.html (accessed April 1, 2011). 18
Daniel Batson, "These Things Called Empathy: Eight Related but Distinct Phenomenon." In The Social Neuroscience of Empathy, ed. Jean Decety and William Ickes (Cambridge: MIT Press, 2011), 3-15. 19
Tania Singer and Claus Lamm, “The Social Neuroscience of Empathy,” Annals of the New York Academy of Sciences 1156 (March 2009): 81-96. 20
Ibid.
7 people "automatically synchronize affective expressions, vocalizations, postures, and movements with another person."21 An everyday observation of mimicry is the spread of yawning. Often when a person yawns in front of other people the yawning spreads, leading others to yawn as well. One argument related to empathy is that mimicry allows a person to automatically feel what others are feeling.22 This idea is central to the perception-action model for empathy. 23 The neuroscience-based model suggests that watching or imagining the emotion another person is feeling automatically activates a similar state in the observer.24 This hypothesis has been tested in research on pain. Evidence suggests that in regards to pain, the same areas of the brain that activate when a person is in pain are also activated when someone observes that another is in pain.25 The most common definition of empathy is feeling what another person feels.26 In this case, a person feels either exactly the same or very close to how another person feels. When another person feels sad, you feel sad as well. This type of empathy does not necessarily lead to prosocial behavior. Tania Singer gives an excellent example of torturers who use
21
Ibid.
22
Marianne Sonnby–Borgström, “Automatic Mimicry Reactions as Related to Differences in Emotional Empathy,” Scandinavian Journal of Psychology 43, no. 45 (December 2002): 433-43. 23
Stephanie D. Preston and Frans B. M. de Waal, “Empathy: Its Ultimate and Proximate Bases,” Behavioral and Brain Sciences 25, no. 1 (January 2003): 1-20. 24
Ibid.
25
Tania Singer and Claus Lamm, “The Social Neuroscience of Empathy,” 81-96.
26
Daniel Batson, "These Things Called Empathy: Eight Related but Distinct Phenomenon," 3-15.
8 empathy in order to understand how to increase the pain of their victims.27 Another idea that is related to empathy is the idea of perspective taking. Perspective taking has been described as both the ability to imagine or understand how another person is thinking or feeling, and the ability to "put yourself in another person's shoes."28 The two descriptions differ slightly. When a person imagines how another person feels, they are trying to gauge the other person's state. "Putting yourself in another person's shoes" is imagining how you would feel if you were in that person's situation. This is similar to theory of the mind, which allows a person to know that other people have thoughts, beliefs, and feelings.29 Evidence suggests that people are not born with theory of the mind. Instead, theory of the mind develops as children age and is thought to be lacking in people who struggle with autism.30 Sometimes people are distressed by the situation of others. The idea of personal distress is that people feel "anxiety" or "unease" when witnessing the circumstances of another.31 Personal distress is a feeling that comes as a result of what someone else is experiencing but is not feeling "for" another.32 People who experience personal distress might help the person who is causing them to feel bad, but only because 27
Tania Singer, "Understanding Others: Brain Mechanisms of Theory of Mind and Empathy." In Neuroeconomics: Decision Making and the Brain, ed. Paul W. Glimcher, Colin Camerer, Russell Alan Poldrack and Ernst Fehr (London: Academic Press, 2009), 251-65. 28
Daniel Batson, ed., "These Things Called Empathy: Eight Related but Distinct Phenomenon," 3-15.
29
Tania Singer, "Understanding Others: Brain Mechanisms of Theory of Mind and Empathy", 251-65.
30
Ibid.
31
Daniel Batson, "These Things Called Empathy: Eight Related but Distinct Phenomenon," 3-15.
32
Ibid.
9 they want to stop feeling bad. Personal distress presents a problem because it makes it difficult to distinguish whether a person chooses to act due to personal distress or empathetic concern. Experiments have attempted to separate the two ideas by allowing people to "escape" from the distress.33 The old adage "out of sight, out of mind," has been recreated in these experiments in order to see if these two concepts differ.34 Overall, the results suggest that people who are experiencing personal distress will not help as often when "escape" is easy compared to people experiencing empathetic concern.35 All of these concepts of empathy are interrelated. For example, the ability to "put yourself in another person's shoes" might lead a person to feel more compassion for another person. The perception that someone else is in need and the concern for the welfare of that person has been called empathetic concern.36 Sometimes labeled as sympathy, empathetic concern can potentially motivate prosocial behavior because a person values the welfare of the other person or group. The idea that a person values the welfare of the other person is what is unique about empathetic concern. In order to test this, Batson and Eklund conducted an experiment that manipulated the value a person places on the welfare of another by having people read about the other person as being
33
Daniel Batson, Jim Fultz and Patricia A. Schoenrade, “Distress and Empathy: Two Qualitatively Distinct Vicarious Emotions with Different Motivational Consequences,” Journal of Personality 50, no. 1 (March 1987): 19-39. 34
Ibid.
35
Ibid.
36
Ibid.
10 kind or being selfish.37 When people read about the other person being kind, they had increased empathetic concern for that person and were more likely to help them.38 In another experiment, four people played a collective good game for raffle tickets.39 Each person was given two sets of 8 raffle tickets that they could choose to keep, give to another player, or give to the group.40 If a person gave one set of tickets to the group, the number of tickets would increase to 12.41 Those 12 tickets were divided among the group. If a person kept both sets themselves they would have 16 tickets.42 If everyone donated to the group, then each person would receive 24 tickets. Due to the small amount of players, getting more raffle tickets greatly increased a person's odds of winning money.43 Some participants were induced to feel empathy for another person by reading about how sad a fellow participant felt over a recent breakup.44 The people who
37
Daniel Batson and Jakob Håkansson Eklund, “An Additional Antecedent of Empathic Concern: Valuing the Welfare of the Person in Need,” Journal of Personality and Social Psychology 93, no. 1 (July 2007): 65-74. 38
Ibid.
39
Daniel Batson et al., “Empathy and the Collective Good: Caring For One of the Others in a Social Dilemma,” Journal of Personality and Social Psychology 68, no. 4 (April 1995): 619-31. 40
Ibid.
41
Ibid.
42
Ibid.
43
Ibid.
44
Ibid.
11 read about the person's situation where much more likely to sacrifice their own and/or the overall groups welfare by directly giving more tickets to the person they read about.45 Altruism Empathetic concern seems to have the potential to motivate altruism, but how can altruism have survived evolution? Evolutionary biologists have struggled with this question for a long time. An illustration of this issue is demonstrated among squirrels. When squirrels sense a predator coming they sometimes call out a warning to others.46 This warning increases the chances that the other squirrels will survive, but the squirrel that sounds the alarm is now in much more danger of being eaten.47 How can this seemingly altruistic behavior have survived natural selection? This conflict between self and group interest is best captured by a prisoner's dilemma. The dominant strategy is not to cooperate but to defect. This suggests that cooperators will lose out to the selfinterested. However, following this logic tells that our squirrel will not warn others. The issue the squirrel faces is similar to problems that humans must deal with as well. How could people who often play a seemingly dominated strategy have survived evolution? An attempt at a formal answer was brought about by W. D. Hamilton. 48 The idea is that
45
Ibid.
46
Richard McElreath and Robert Boyd, Mathematical Models of Social Evolution: A Guide for the Perplexed (Chicago: University Of Chicago Press, 2007), 71-76. 47 48
Ibid.
W.D. Hamilton, “The Genetical Evolution of Social Behaviour. I,” Journal of Theoretical Biology 7, no. 1 (July 1964): 1-16.
12 altruism could evolve through what is called kin selection or inclusive fitness.49 This suggests that people might engage in altruism with people that are related to them. The idea is captured in what is known as Hamilton’s rule: r >c / b, where r is the "coefficient of relatedness."50 This rule says that the degree that someone is related to another person must exceed the cost benefit ratio of engaging in an altruistic act.51 Table 1 gives the potential payoffs to two players, where player 1 will choose to cooperate if Hamilton's rule holds. What this rule has done is transform the payoffs so that cooperating is now the dominant strategy. This is an important point because altruism can only survive evolution if altruism leads to a higher payoff. However, kin selection cannot explain any cooperation that occurs with people that are not kin. An additional way altruism can come about is through the idea of reciprocal altruism.52 Here people are essentially playing a repeated prisoner's dilemma. Cooperators can be better off if interactions with another person are repeated.53 Altruism in this case can emerge as long as fellow cooperators can interact with each other.54 However, there do seem to be situations 49
Richard McElreath and Robert Boyd, Mathematical Models of Social Evolution: A Guide for the Perplexed, 78-80. 50
Ibid.
51
Martin Nowak, “Five Rules For the Evolution of Cooperation,” Science 314, no. 580 (December 2006): 1560-63. 52
Richard McElreath and Robert Boyd, Mathematical Models of Social Evolution: A Guide for the Perplexed, 123-126. 53
Martin Nowak, “Five Rules For the Evolution of Cooperation,” 1560-63.
54
Ibid.
13 where people cooperate where it is unclear that they are engaging in reciprocal altruism. An individual might help an elderly stranger cross the street with no expectation of ever seeing that person again to reap the rewards. This type of occurrence led researchers to turn to the idea of indirect reciprocity. 55 Here people might help others in order to give themselves a better reputation.56 The reason people might care about their reputation is that it might be rewarding. Some evidence suggests that people who have reputations for being helpful tend to receive more help from others.57 Recent research in social networks has explored the idea that altruists might be more likely to interact with other altruists.58 A recent study showed that people who had higher baseline altruism have friends with higher baseline altruism.59 This suggests that cooperation can emerge in altruism "clusters."60 While altruism can theoretically evolve in natural selection, less is known about what motivates a person to be altruistic. Research in humans suggests that altruism is "not a single homogenous trait."61 Instead, altruism seems to be very context
55
Ibid.
56
Ibid.
57
Ibid.
58
Ibid.
59
Markus Mobius, Quoc-Anh Do, Stephen Leider and Rosenblat Tanya, “Directed Altruism and Enforced Reciprocity in Social Networks,” Quarterly Journal of Economics 124, no. 4 (November 2009): 1815-51. 60 61
Martin Nowak, “Five Rules For the Evolution of Cooperation,” 1560-63.
Felix Warneken and Michael Tomasello, “Varieties of Altruism in Children and Chimpanzees,” Trends in Cognitive Sciences (September 2009): 397-402.
14 dependent. It is possible that the incentive for humans to cooperate is partly driven by the idea that they care about the welfare of others. One way that economic theory has attempted to capture altruism is through the model of pure altruism.62 In this model a person cares about the welfare of another person. Typically modeled as a utility function, the utility of pure altruists increases as the utility of another person increases. An alternative model of altruism has been called impure altruism or warm glow.63 In this model, an altruist is an egoist.64 That is, people engage in altruistic acts because the act of altruism makes themselves feel good. These two models differ in their predictions about human behavior. People who are pure altruists care about the welfare of another person. So, it does not matter to them how another person's welfare increases. A pure egoist gives to another person because they feel good about giving. For example, if a pure altruist and a pure egoist observe that a homeless person needs something to eat, and they see that another person has given the homeless person some food, then it is likely that only the pure egoist would still have the same motivation to give. The pure altruist only cares about whether the homeless person is fed. How that goal is achieved is of less importance. The pure egoist does not care that the homeless person has eaten or not, only that giving makes the egoist feel good. It is likely that true human nature is a combination of these two ideas. The focus of this 62
Ulrich Mayr et. al., "Neuroeconomics of Charitable Giving and Philanthropy." In Neuroeconomics: Decision Making and the Brain, ed. Paul W. Glimcher, Colin Camerer, Russell Alan Poldrack and Ernst Fehr (London: Academic Press, 2009), 303-20. 63
Ibid.
64
Ibid.
15 paper is on pure altruism. In pure altruism models, the idea that someone cares about the welfare of another person is often assumed, but it is unclear what motivates someone to care about the welfare of another. In the next section, I will explore the idea that empathic concern can potentially motivate pure acts of altruism.
Table 1. Prisoner's Dilemma with Kin Selection65 66 Player 2 Cooperate Defect Cooperate (b - c)(1 + r), (B - C)(1 + r)
br - c, B - rC
Player 1 Defect b - rc, Br - C 0, 0 Note: Player 1's payoffs are in lower case and Player 2's payoffs are in upper case.
65 66
Martin Nowak, “Five Rules For the Evolution of Cooperation,” 1560-63.
Richard McElreath and Robert Boyd, Mathematical Models of Social Evolution: A Guide for the Perplexed , 74.
CHAPTER 2 THEORY In this section I begin by exploring the idea that empathy can potentially lead to altruism. After examining this idea, a theoretical model is presented that suggests how empathy motivated altruism might influence preferences for distributive justice. Empathy and Altruism While the replication of genes is the driving force of evolution, people who have these genes are not spending all of their time thinking about this. People often have sex not necessarily to spread their genes, but because it feels good. Sometimes the motivation is to have children, but if that were the only reason people would be having sex much less often. Instead, one of the main motivations driving sex is that people enjoy it. This is the same when people eat. People are not necessarily trying to give their body energy in order to optimize their survival so they have a better chance of spreading their genes. If this were so, people would likely eat a much healthier diet. Instead, the motivation to eat is driven by the pain and discomfort of hunger, and the release of pleasurable neurotransmitters like dopamine once food has been ingested. It is possible that altruistic acts have motivations as well. The theory of kin selection tells us that altruism can occur among relatives, but it is unlikely that the motivation a parent has to help their children is to increase the chances their genes will replicate. Instead parents often sacrifice for their children because they love them. That love provides an incentive 16
17 for a parent to care for their children, which as a result can increase the chances that their genes will spread. The reason this is important is because understanding motivations can help improve the understanding of human behavior. From an evolutionary perspective, it is difficult to understand why anyone would choose an altruistic act in complete anonymity. They gain no advantage through reputation, or future reciprocity. Nevertheless, people often donate to charities anonymously. In many experiments of the dictator game, where a person can choose to keep all of a payoff or give some to another player, many people give the other player positive amounts.67 A possible motivation for these results is the idea that people care about the welfare of other people. That is they are not engaging in altruism just to further their own fitness but partly because they generally care about the welfare of others. They might feel happy when something positive happens to someone they care about or feel bad when the person is in need. Of course, caring for others might end up increasing a person's fitness, but part of the motivation for altruistic behavior could be the concern felt for another person. Can empathetic concern motivate altruism? In one study, Kruger examined whether kin selection, reciprocal altruism, and empathetic concern influenced helping behavior.68 The author found that while the effect of empathetic concern was smaller than reciprocal altruism and preferences for kin, all three were significant in predicting helping
67
Colin F. Camerer, Behavioral Game Theory: Experiments in Strategic Interaction (Roundtable Series in Behaviorial Economics) (New York: Princeton University Press, 2003), 43-59. 68
Daniel J. Kruger, “Evolution and Altruism: Combining Psychological Mediators with Naturally Selected Tendencies,” Evolution and Human Behavior 24, no. 2 (March 2003): 118-25.
18 behavior.69 What if a person who is playing a one-shot anonymous prisoner's dilemma game knows the other person defected? Clearly, they should defect as well since it gives them the best payoff. Batson and Ahmad ran such an experiment with women.70 Participants were given the choice to cooperate or defect knowing that the other person has already defected.71 Some of the women, however, were induced to feel empathy by reading about how the other player just broke up with their boyfriend.72 Of the people given the empathy condition, 45% chose to cooperate.73 This contrasts with the nonempathy condition where only 5% cooperated.74 These experiments suggest that empathy can lead to altruistic actions. While there is experimental evidence that empathic concern can potentially lead to an increase in altruism, often experimental subjects are undergraduates and experiments take place in controlled conditions. This makes it difficult to know if these results will hold for the general population. Theoretical Model One argument for why people support redistribution or charitable giving comes from the idea of public goods.75 In order for something to be a public good it must have
69
Ibid.
70
Daniel Batson and Nadia Ahmad, “Empathy-induced altruism in a prisoner's dilemma II: what if the target of empathy has defected?” European Journal of Social Psychology 31, no. 1 (January 2001): 25-36. 71
Ibid.
72
Ibid.
73
Ibid.
74
Ibid.
19 the properties of nonrivalry and nonexlusivity.76 A good is nonrival if additional units can be consumed at zero or close to zero social marginal cost.77 A good is nonexclusive if it is impossible or close to impossible to exclude people from benefiting from the good.78 One example of a public good would be knowledge that fire is hot. Using this idea takes place at zero marginal cost and once the idea is known it would be quite hard to stop people from using it. The idea of public goods has been proposed as a reason why people might support redistribution. At first this does not seem to fit with the idea of a public good I just defined. For example, the transfer of money to the poor is not a public good, since not everyone can use the money that was given to the poor. The public good aspect in this case is the idea that poor people are being helped. The solace in knowing that people who are suffering are being taken care of is in essence a public good. This paper digs deeper into this idea by exploring why people value this type of public good. People who have empathetic concern feel compassion for others and have a desire to improve the welfare of others. They want the poor to be helped because they value the welfare of the poor. In order to explore the implications of this idea I turn to a theoretical model that attempts to explain how empathetic concern influences an individual's preferences. In the footnotes of Mathematical Psychics, Francis Edgeworth proposed a model that accounted 75
Dennis C. Mueller, Public Choice III, 3rd ed. (New York: Cambridge University Press, 2003), page
44-49. 76
Walter Nicholson and Christopher Snyder, Microeconomic Theory Basic Principles and Extensions, 10th ed. (Canada: South-Western College Pub., 2007), 679-80. 77
Ibid.
78
Ibid.
20 for the idea that people care about the welfare of others.79 Edgeworth defined a separable utility function for two people such that the first person's utility was of the form U P = P + λ Π and the second person's utility was of the form U π = Π + μ P ,
where μ, λ are measures of "effective sympathy." 80 Using this idea, let the following utility function be defined: n
U i = δ u i (πi ) + (1 − δ) ∑ E i u j (π j )
(1)
j
i
j=1
j
+
where δ∈[0,1] , E i ∈[0,1] , πi , π j∈R , and i ≠ j . Equation (1) is person i's utility function, which accounts for the utility of person i's payoff, denoted πi , plus the sum of j
the empathetic concern, E i , i feels for each person or group j times the utility of each person's payoff, denoted π j . In addition, u i (π i) is the utility person i receives from their own payoff πi . Furthermore, u ij (π j ) is the utility that person i believes that j receives from their payoff π j . Here the payoffs of πi and π j could represent many different things that people derive utility from. They could be a person's consumption set or represent a person's income. From an epistemological standpoint it is impossible for person i to truly know the utility that each person or group j receives. People nevertheless do attempt to gauge the utility of others. People often make statements like "she enjoys going to the movies way more than I do" or "10 dollars means much more to him than it does to me." In some circumstances people may assume that the other person 79
Francis Ysidro Edgeworth, Mathematical Psychics: An Essay On the Application of Mathematics to the Moral Sciences (London: C. Kegan Paul, 1881), 53. 80
Ibid.
21 receives similar utility as they do. They may also learn what others like and dislike in order to form a better understanding of the utility of others. One way people attempt to do this is through the concepts of empathy discussed in the previous section. Empathy can help people in trying to figure out how others think and feel. Since u ij (π j ) is person i's subjective belief about the utility of j, the model can account for paternalism. Person i might give food to a homeless person instead of money. Reasoning that they will use the money for nefarious means, and giving them food will make them better off. This is an important aspect of the model because it allows the idea that people might evaluate issues of distributive justice based on what they think is best for others. Just because people feel empathy for another does not necessarily mean they will act altruistically. This is because people may put a larger weight on the utility from their own personal payoff. In order to account for this, the model includes the δ term, which shows how much i weighs his or her own personal utility relative to the utility of others. If δ = 1, then a person's utility is only a function of his or her own payoff. If δ < 1, then person i's overall utility will also depend on whether or not i feels empathetic concern for the other person. If δ > 1, then the model would capture spite. However, empathetic concern is different from ideas like fairness. Ideas like altruistic punishment, and fairness can account for spiteful behavior. That is, people might lower their own payoff in order to lower the payoff of another person. The reason this is the case is that fairness dictates that people reward people who are kind to them, but punish people who hurt them.81 However, these ideas differ fundamentally from empathetic concern. A 81
Matthew Rabin, “Incorporating Fairness Into Game Theory and Economics,” American Economic Review 83, no. 5 (December 1993): 1291-2.
22 person might feel less concern for another but not the motivation to hurt them. For this reason, δ is restricted such that δ∈[0,1]. The advantage of including the δ term is that the utility function is able to account for individual heterogeneity in regards to altruism. Preferences for Redistribution and Charitable Giving The main interests are when people will want redistribution and when they will help others. While the following analysis can be extended to the more general case, assume person i only feels empathetic concern for one person j. Furthermore, take person i's income, denoted I i , as given. This can be thought of as a result of utility maximization involving income and leisure trade-offs. Person i could treat income used for consumption in the same way as income used to give to person j. In other words, the trade-off is not different when the income is used for person i's own consumption versus person j's consumption. Person i can choose to allocate this income between buying some good x at price P x and giving some s to person j at the price P s .82 P s represents the cost of giving or redistributing. For example, if person i buys food for a homeless person, then person i pays P s s , and then gives s to person j. Here P s could include a number of costs ranging from the price of the food to opportunity costs. That is, the budget constraint that person i faces is as follows: I i = P x x + P s s where P x > 0 , P s > 0 , and s, x , I i ∈R+ .
(2)
The utility function for person i is: j
i
U i = δ u i ( x) + (1 − δ)E i u j (M j + s) ,
(3)
82
Good x could also represent a basket of goods and P x could be a vector of prices for those goods.
23 where M j ∈ R+ . M j is some fixed amount of income that person j has. In addition, j i δ∈(0,1) and E i > 0 . Let u i ( x) and u j (M j + s) be increasing, continuous, and
concave functions. An important assumption to note is that the utility functions are separable. Relaxing separability could potentially lead to different results. By including M j , the model can account for the idea that a person derives utility from the income of another person. That is, a person might receive more utility when someone else has more income. More importantly, M j might influence how much s person i might give to person j. M j gives person i a reference point that could potentially determine how much s to give. An additional assumption is that person j will not change his or her behavior based on the s they receive from person i. Furthermore, person i will not change his or behavior based on person j's behavior. In addition, let M j be the true income for person j. That is, person j has not lied about M j in order to take advantage of person i's generosity. In addition, person j chooses his or her optimal income M j with no expectation of receiving any s from person i. Person i then maximizes the utility function subject to the budget constraint. (4)
j
i
L = δ u i ( x) + (1 − δ) E i u j (M j + s) + λ( I i − P x x − P s s ) ,
where (4) is the Lagrangian expression. Taking the partial derivatives with respect to x, s, and λ , and setting them equal to zero gives: (5)
∂L = δ u i' ( x) − λ P x = 0 ∂x
24 (6)
∂L j i = ( 1 − δ) E i u j ' (M j + s) − λ P s = 0 ∂s
(7)
∂L = I i − Px x − Ps s = 0 λ Setting up the Lagrangian assumes there exists some interior maximum where
* * s > 0 , and x > 0 . The main interest is when someone will choose this interior
maximum as opposed to the corner solution of s* = 0 . So when will s* > 0 ? That is, when will person i choose to give some positive amount to another person j? In order for *
s > 0 , then it must be the case that:83 j
i
(1 − δ) E i u j' (M j + s) − λ P s > 0
(8)
From (5) it is clear that λ = j
Ei >
(9)
δ u i'( x ) j . Plugging this into (6) and solving for E i gives: Px
P s δ u i ' (x) i P x (1 − δ)u j' (M j + s )
If (9) holds, then person i will choose some positive s. This suggests that higher j
values of E i will make (9) more likely to hold everything else being equal. As δ decreases, the right hand side of the equation decreases. This makes sense when looking at equation (3), since δ tells how much a person weighs his or her personal utility. Lower values of δ mean that the person places a larger weight on the utility of the other person. As P s increases, (9) could be less likely to hold. This accounts for the idea that if the cost of giving to another person is large, then a person might be less likely to give. The effect of a change in P s will depend on the utility functions. Similarly, P x shows 83
Walter Nicholson and Christopher Snyder, Microeconomic Theory Basic Principles and Extensions, 119-21.
25 that as the price of good x increases, (9) could be more likely to hold. Of course, increases in the price of x could also lead to a person to choose less s. Again, the effect of a change in P x will be influenced by the utility functions. Both u i ' ( x ) and i
u j' (M j + s) are decreasing. I have only assumed that the utility functions are increasing, continuous, and concave. This allows a great range of different functional forms for the utility functions that can potentially determine whether (9) holds. As x increases, the right hand side of (9) decreases due to diminishing marginal utility. As M j + s increases, the right-hand side of (9) increases. The value of M j is particularly interesting, because it can account for the idea that people might be less likely to give to people who are in less of a need. A person might not give a few dollars to a millionaire, but give it a person who is homeless. Now (9) only gives the condition at which s* > 0 , but does not indicate how much s person i will give. If (9) does indeed hold, then the equilibrium that maximizes the utility function from (3) subject to the budget constraint from (2) is: j
(10)
u i ' ( x) P (1 − δ) E i = x i Ps δ u j ' (M j + s)
The equilibrium condition of (10) gives the marginal rate of substitution of x for s. This condition is where it is clear how empathic concern can potentially influence a j
person's choices. As E i increases, person i is willing to trade more x for s, ceteris paribus. That is, people who have more empathic concern will want to give more to another person. In addition, as δ decreases person i will be willing to give up more x for s. When looking at P s it is clear that, as the price of s increases, person i might trade
26 less x for s. This accounts for the idea that if the cost of giving to another person is large, then a person might be less likely to give. Whether this holds will depend on the utility functions. Similarly, P x shows that as the price of good x increases person i could be willing to trade more x for s. Of course, as the price of x increases person i might end up consuming less s. The results suggest that people with higher empathetic concern will be more likely to forgo their own consumption in order to help others. Although I assumed that that j was a single person, the analysis could have assumed that j is some group. That is, person i could have felt empathic concern for a group j that represents poor people. It is possible for s to represent a direct transfer to another person or represent money redistributed through taxation. That is person i can choose to give money to another person or pay more in taxes to redistribute to another group. The results of this analysis suggest the following hypotheses: 1.
In general, people who have higher degrees of empathetic concern should be more likely to support redistribution.
2.
People who have higher degrees of empathetic concern should be more likely to give to others. A complete model has to account for the idea that people maximize their leisure
and income. When a person receives additional income this changes the optimal mix of leisure and income. In addition, the model should also account for the fact that people act strategically. That is, people will make decisions based on what other people do. Person j has an incentive to lie about his or her income in order to receive a higher s. In addition,
27 when person j receives s it is likely that they would then resolve the leisure and income maximization problem. Since person j now has a higher income, then they might prefer more leisure. In other words, person j could extract the maximum s from person i by increasing leisure and reducing income M. Now if person i knows that person j is doing this, person i might make changes as well. Knowing that person j is taking advantage of person i might make person i less empathetic towards person j. This lower empathy could lead to a smaller s going to person j or even none at all. The model presented does not account for these ideas. As a consequence, this theoretical model is only suggestive. It demonstrates how empathetic concern might potentially influence individual decision making provided that the assumptions hold. In addition to empathetic concern, people are also influenced by ideas like reciprocity and fairness. Accounting for these ideas could potentially influence the results from this analysis as well. While a complete model of human behavior still eludes researchers, the goal of this paper is to suggest an aspect of human nature that might need to be accounted for. That is, if people feel empathetic concern, then this can potentially influence human behavior. While it is possible that empathetic concern might not lead to the results from the model presented, the model nevertheless provides a starting point to think about the way that empathy can influence preferences. From (9) and (10) it is clear that many variables influence whether a person gives to another. One way to explore this idea is to look at a person's beliefs. It is likely that a person's beliefs might influence a person's level of empathetic concern, the weight they place on the utility of others, and the subjective evaluation of another person's utility.
28 These beliefs can potentially influence whether (9) holds and the possible equilibrium condition reached in (10). Influence of General Beliefs The decision to sacrifice a person's own utility is possibly influenced by a person's beliefs. Since the majority of situations people face are ones of incomplete information, people might turn to their beliefs in order to make judgments. When a person observes someone who is homeless, it is unlikely that they know how and why that person ended up that way. Lacking any concrete information people are instead forced to rely on their beliefs. For example, if a person believes that people achieve success from hard work, then they might be less likely to sacrifice their own payoff in order to help the other person. If all it takes to be successful is hard work, then the homeless person has complete control over their situation and there is no need to be concerned about it. It is possible that the empathy-altruism relationship is modulated by the perceived fairness of others. If people believe that, in general, people take advantage of other people, then they might be more likely to want to correct it. People might want to redistribute from the better off to the worse off because they might want to correct for the unfairness of others. Trust is likely to play a role as well. People who believe in general that others are untrustworthy might be less likely to engage in altruism. For example, when a person sees that another person needs help they may indeed feel empathetic concern for them, but if they believe people are untrustworthy then they might be less willing to sacrifice their own welfare in order to improve the other person's. A person who does not trust others is less likely to put themselves in vulnerable positions were they can be taken
29 advantage of. Being altruistic opens a person up to the chance that someone will take advantage of them. People who lack trust in others will be less likely to take that chance. These ideas suggest a third hypothesis: 3.
The degree to which empathetic concern motivates preferences for redistribution or giving might be modulated by general beliefs about trust, fairness, and a person's control over their own success.
CHAPTER 3 DATA The data set comes from the General Social Survey, which is nationally representative and conducted biennially on a sample of members of the United States. During the years 2002 and 2004, the survey contained an empathy and altruism module. This module contains a series of statements designed to capture empathetic concern and asks people how often they participate in certain helping behaviors. In addition to this module, the General Social Survey provides data on individual preferences for distributive justice as well as control variables such as income, race, and political views. One distinct advantage of this survey is that people have agreed to take the general survey, but did not know that empathetic questions would be asked. This potentially helps reduce the self-selection bias that would occur if people had agreed to take an empathy survey. The original sample contained 5,577 observations and the data are independently pooled cross sections. The sampling design results in not everyone receiving the same survey questions. This allows the survey to cover a much wider range of issues than if they asked each respondent every question. Not everyone received the empathy module and the observations that did not were dropped from the sample. This left 2,649 observations that received the empathy module. Subsequently, observations that had no
30
31 data on education, age, sex, race, and real income were dropped from the sample leaving 2,360 observations. If political views are included then the number of observations drops to 2,323.
Table 2. Description of Empathy Variables and Factor Analysis Results Variable Respondents rate the statements on a 5 point scale from: (1) Does not describe very well to (5) describes very well Empathy1 I often have tender, concerned feelings for people less fortunate than me Empathy2 Sometimes I don't feel very sorry for people when they are having problems
Estimated Factor Loadings* 0.5608 -0.1749
Empathy3 When I see someone being taken advantage of, I feel kind of protective towards them
0.4969
Empathy4 Other people's misfortunes do not usually disturb me a great deal
-0.1971
Empathy5 When I see someone being treated unfairly, I sometimes don't feel very much pity for them
-0.2203
Empathy6 I am often quite touched by things I see happen
0.6683
Empathy7 I would describe myself as a pretty soft-hearted person *Factor loadings were varimax rotated.
0.6615
Measurement of Empathetic Concern Since empathetic concern is the variable of interest, it is necessary to find a way to measure this concern. The General Social Survey contains a series of seven statements that participants rate on a 5-point scale, indicating how well the statement describes them. The questions come from the Interpersonal Reactivity Index and are designed to measure
32 empathetic concern.84 The seven empathy items have a Cronbach's alpha of 0.734, suggesting a good degree of internal consistency. Since all seven variables attempt to capture an overall measure of empathetic concern, factor analysis is used in order to isolate the communality in the responses. Table 2 lists the statements given to respondents as well as the estimated factor loadings. Three factors arise when running factor analysis on the seven empathy variables: factor 1 has an eigenvalue of 2.037, factor 2 has an eigenvalue of 0.47, and factor 3 has an eigenvalue of 0.01. Using the Kaiser criterion, which suggests dropping factors with eigenvalues lower than 1.0, factor 1 remains.85 The signs of the factor loadings in Table 2 suggest the relationship that is expected for empathetic concern. Empathetic concern should have positive scores on Empathy1, Empathy3, Empathy6, and Empathy7. The reason is that these statements capture feelings and desires that are thought to be related to empathetic concern.86 Individuals that feel more concerned and protective of people who are in distress are thought to exhibit greater empathetic concern. In addition, people who are more emotionally affected by others are thought to be more compassionate.87 This contrasts with the statements contained in Empathy2, Empathy4, and Empathy5. These statements are designed to capture the opposite of how people with high degrees of empathetic 84
Mark H. Davis, “Measuring Individual Differences in Empathy: Evidence For a Multidimensional Approach,” Journal of Personality and Social Psychology 44, no. 1 (January 1983): 113-26. 85
Richard J. Harris, A Primer of Multivariate Statistics, 3rd ed. (Mahwah, N.J.: Psychology Press, 2001), 408-8 86
Mark H. Davis, “Measuring Individual Differences in Empathy: Evidence For a Multidimensional Approach,” 113-26. 87
Ibid.
33 concern are likely to think and feel. People who are not affected by other's misfortunes and do not feel sympathy for others are likely to have less empathetic concern.88 The factor loadings from Table 2 suggest that this is the case. Before generating the variable, the factor is varimax rotated. The reason the factors are rotated is to maximize the variance of the squared loadings of a factor on the variables in the factor matrix.89 This procedure preserves the properties of the original factor loadings, but attempts to make the factor loadings large or small to help interpretation.90 From this the factor of interest is generated, using the factor loads as weights, into a single continuous variable for empathetic concern. Figure 1 gives the distribution for the derived factor variable for empathetic concern. The mean for the sample is approximately zero as expected. While it appears that the scores are truncated at 1, this is not actually the case. Table 4A shows that the maximum value for empathetic concern is approximately 1.0044. It is clear from the graph that a large percentage of the sample report high scores for empathetic concern.
88
Ibid.
89
Alvin C. Rencher, Methods of Multivariate Analysis, 2nd ed. (New York: Wiley-Interscience, 2002), 430-34. 90
Ibid.
0
5
Percent of Sample 10
15
34
-3
-2 -1 Scores for Empathetic Concern
0
1
Figure 1. Graph of the Sample Distribution of the Derived Factor for Empathetic Concern
Descriptive Statistics Table 3 gives descriptive information on all independent and dependent variables. It is possible that empathetic concern might influence some preferences for distributive justice more than others. In order to see this, the sample contains individual preference data for variables on national issues like the distribution of wealth, Social Security, welfare, aid to the poor, and foreign aid as well as more local issues like giving food or money to the homeless, volunteering for charity, and giving money to charity.
35 Table 3. Description of Variables Variable
Description
Scale
Income Redistribution
The government should reduce income differences between the rich and the poor (GSS Variable: EQWLTH)
1 (Govt Should) to 7 (Govt Should not)
Social Security
Social Security spending (GSS Variable: NATSOC)
1 (Too little), 2 (About right), 3 (Too much)
Foreign Aid
Foreign aid spending (GSS Variable: NATAID)
1 (Too little), 2 (About right), 3 (Too much)
Welfare
Welfare spending (GSS Variable: NATFARE)
1 (Too little), 2 (About right), 3 (Too much)
Helping the Poor
Government should improve standard of living of the poor or people should take care of themselves (GSS Variable: HELPPOOR)
1 (Govt should) to 3 (Agree with both) to 5 (People should help themselves)
Giving to Charity
Person has given money to charity at least once in the past 12 months (GSS Variable: GIVCHRTY)
1 if they have, 0 otherwise
Giving to Homeless
Person has given money or food to a homeless person at least once in the past 12 months (GSS Variable: GIVHMLSS)
1 if they have, 0 otherwise
Volunteer
Person has volunteered for a charity at least once in the past 12 months (GSS Variable: VOLCHRTY)
1 if they have, 0 otherwise
Empathy
Factor variable for empathetic concern
Continuous variable generated from factor analysis
Educ
Amount of education
Measured in years
Age
Age of person
Measured in years
Female
Sex
1 if female, 0 otherwise
Male
Sex
1 if male, 0 otherwise
Black
Race
1 if black, 0 otherwise
White
Race
1 if white, 0 otherwise
Other Race
Race
1 if a race other than white or black, 0 otherwise
2002
Year
1 if year 2002, 0 otherwise
2004
Year
1 if year 2004, 0 otherwise
36 Table 3. Description of Variables- continued Variable
Description
Scale
POLVIEWS
Respondents choose which political orientation best 7 point scale, from 1 describes them. From this the following seven dummy (Extremely Liberal) to 7 variables are generated. (Extremely Conservative)
Extremely Liberal
Measure of political orientation
1 if Extremely Liberal, 0 otherwise
Liberal
Measure of political orientation
1 if Liberal, 0 otherwise
Slightly Liberal Measure of political orientation
1 if Slightly Liberal, 0 otherwise
Moderate
Measure of political orientation
1 if Moderate, 0 otherwise
Slightly Conservative
Measure of political orientation
1 if Slightly Conservative , 0 otherwise
Conservative
Measure of political orientation
1 if Conservative, 0 otherwise
Extremely Conservative
Measure of political orientation
1 if Extremely Conservative, 0 otherwise
Log Real Income
Log of a person's real income
Continuous variable
37 Table 3. Description of Variables- continued Variable
Description
Scale
TRUST
Respondents choose the general belief about the trustworthiness of others that they most agree with. The following three dummy variables were generated from the TRUST variable.
3 point scale, 1 (Trust others), 2 (Depends), 3 (No Trust)
Trust others
In general, most people can be trusted
1 if agree, 0 otherwise
Trust Depends
In general, trustworthiness of others depends
No Trust
In general, you cannot be too careful in dealing with other people
FAIR
Respondents choose the general belief about the 3 point scale, 1 (People take fairness of others that they most agree with. The advantage), 2 (Depends), 3 following three dummy variables were generated from (People Fair) the FAIR variable.
1 if agree, 0 otherwise
Take Advantage Most people would try to take advantage of you if they 1 if agree, 0 otherwise got the chance Fair Depends
The fairness of others depends
1 if agree, 0 otherwise
People Fair
Most people would try to be fair
1 if agree, 0 otherwise
GETAHEAD
Respondents choose the general belief about how 3 point scale, 1 (Luck), 2 (Both people get ahead that they most agree with. The Luck and hard work), 3 (Hard following three dummy variables were generated from Work) the GETAHEAD variable.
Luck
People get ahead by luck or help from others
1 if agree, 0 otherwise
Both Luck and Hard Work
Luck and hard work are equally important
1 if agree, 0 otherwise
Hard Work
People get ahead by their own hard work
1 if agree, 0 otherwise
Independent Variables Table 4 lists the summary statistics for the independent variables. The empathy variable generated from factor analysis assigns a value between -3.2504 and 1.0044 for each person. Age has the potential to influence preferences for distributive justice. Priorities and circumstances can change as people age. The average age of the sample is
38 about 45 years. The sample contains no data on individuals under the age of 18, so the results will not apply to younger people. It is likely that education influences preferences for distributive justice. People who have been to college might hold different perspectives on these issues compared to people who have not been to college. In addition, years of education influence the type of jobs that people work. The average years of education is about 13.6 years. It is important to control for gender as well. Certain issues might appeal more to one gender than the other. In addition, genders are treated differently in society and as a result they might hold different preferences. There are slightly more women than men in the sample. Race is broken down from a race variable with three categories: Black, White, and Other Race. I want to control for race, because people are often treated differently based on their race. In addition, there might be cultural differences that may influence preferences for distributive justice. Year dummy variables are included in order to control for anything that might have occurred during those years that could influence individual preferences. Since the empathy module was only conducted in 2002 and 2004 the sample only contains those two years. It is likely that people's preferences for distributive justice are influenced by their political views. Many distributive justice issues involve the role of government. Including the political views helps us to control for this. The political dummy variables were created from the POLVIEWS variable that asked people to choose 1 of the 7 political views that best described them. The most common choice was Moderate at 38.14 percent. More people lean conservative, 35.95 percent, than lean liberal, 25.91 percent. Income is likely to be very important in these issues as well. This is because in
39 issues of redistribution the level of a person's income determines whether they will be on the receiving end. Following the approach by Alesina and La Ferrara in their study of preferences for redistribution, the log was applied to data on reported real income.91 This is used as an attempt to capture the diminishing marginal utility of income.
Table 4. Summary Statistics for Independent Variables Variable
Observations
Mean Standard Deviation
Min
Max
Empathy
2,360 -0.0085
0.8029 -3.2604 1.0044
Educ
2,360 13.575
2.9589
0
20
Age
2,360
16.82
18
89
Female
2,360 0.5112
0
1
Male
2,360 0.4789
0
1
Black
2,360 0.1258
0
1
White
2,360 0.8064
0
1
Other Race
2,360 0.0678
0
1
2002
2,360 0.5093
0
1
2004
2,360 0.4907
0
1
Extremely Liberal
2,323 0.0336
0
1
Liberal
2,323 0.1007
0
1
Slightly Liberal
2,323 0.1248
0
1
Moderate
2,323 0.3814
0
1
Slightly Conservative
2,323 0.1636
0
1
Conservative
2,323 0.1593
0
1
Extremely Conservative
2,323 0.0366
0
1
Log Real Income
2,360 10.022
91
45.55
1.1009
5.710 11.829
Alberto Alesina and Eliana La Ferrara, “Preferences For Redistribution in the Land of Opportunities,” Journal of Public Economics 89, no. 5 (June 2005): 897-931.
40 Beliefs In order to see if general beliefs influence the empathy altruism relationship, data on beliefs about trust, fairness, and success are used. Table 5 lists the summary statistics for the belief variables. Both Trust Others, Trust Depends, and No Trust are generated from the TRUST variable in the General Social Survey. Respondents are asked, "Generally speaking, would you say that most people can be trusted or that you can't be too careful in dealing with people?"92 Respondents could choose "most people can be trusted," "can't be too careful," or "depends."93 The different beliefs can create subsamples. That way the influence of empathetic concern can be tested between groups that hold different beliefs about trust. Both Take Advantage, Fair Depends, and People Fair are generated from the fair variable contained in the General Social Survey. People in the survey were asked, "Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?"94 Respondents could choose "would take advantage of you," "would try to be fair," or "depends."95 The responses allow testing of the idea that empathetic concern might have different influences on preferences for distributive justice depending on beliefs about fairness. The variables Luck, Both Luck and Hard Work, and Hard Work are generated from the GETAHEAD variable in 92
Tom W. Smith, Peter Marsden, Michael Hout and Jibum Kim, “General Social Surveys, 1972-2010 (Data File and Codebook),” National Opinion Research Center: The Roper Center for Public Opinion (2011) 93
Ibid.
94
Ibid.
95
Ibid.
41 the General Social Survey. People in the survey were asked, "Some people say that people get ahead by their own hard work; others say lucky breaks or help from other people are more important. Which do you think is most important?"96 Respondents can choose "hard work most important," "hard work, luck equally important," or "luck most important."97 Responses to this will allow testing of the idea that empathetic concern might have different influences on preferences for distributive justice depending on beliefs about personal control over success.
Table 5. Summary Statistics for Beliefs Variable
Observations
Mean
Min
Max
Trust Others
1,540
0.3675
0
1
Trust Depends
1,540
0.5669
0
1
No Trust
1,540
0.0656
0
1
Take Advantage
1,536
0.3828
0
1
Fair Depends
1,536
0.0951
0
1
People Fair
1,536
0.5221
0
1
Luck
1,560
0.0955
0
1
Both Luck and Hard Work
1,560
0.2564
0
1
Hard Work
1,560
0.6481
0
1
National Issues The data on the national issues comes from the General Social Survey. Five issues were examined in order to create a contextual picture of the role that empathetic concern 96
Ibid.
97
Ibid.
42 plays on these preferences. Table 6 gives the summary statistics for the dependent variables.
Table 6. Summary Statistics for Dependent Variables Variable
Observations
Mean Standard Deviation Min Max
National Issues Income Redistribution
1,531 3.7484
1.9573
1
7
Social Security
2,256 1.4403
0.5984
1
3
Foreign Aid
1,123 2.5733
0.6322
1
3
Welfare
1,119 2.9129
0.7654
1
3
Helping the Poor
1,512 2.9129
1.1751
1
5
Local Issues Give to Homeless
2,320 0.6489
0
1
Give to Charity
2,321 0.7935
0
1
Volunteer
2,322 0.4832
0
1
Table 7A shows the frequency data for the variable Income Redistribution. Respondents were given the statement "Some people think that the government in Washington ought to reduce income differences between the rich and the poor, perhaps by raising the taxes of wealthy families or by giving income assistance to the poor. Others think that the government should not concern itself with reducing this income difference between the rich and the poor."98 Respondents then chose the number that they most agreed with on a 7-point scale from 1, "government should do something to reduce income differences between rich and poor," to 7, "government should not concern 98
Ibid.
43 itself with income differences."99 Opinions on the role of government in income redistribution in the sample were divided. Approximately 18.75% of people strongly believed in government redistribution, while 13.39% strongly believed government should have no role. About 19% chose 4, which was the middle choice.
Table 7A. Frequency Data for Income Redistribution Should the Government Reduce Income Differences? Frequency
Percent of Sample
Government Reduce Differences
287
18.75
2
146
9.54
3
283
18.48
4
292
19.07
5
203
13.26
6
115
7.51
No Government Action
205
13.39
1,531
100.00
Total
Table 7B displays the frequency data for Social Security. People in the survey were asked whether the country is spending "too little," "about right," or "too much" on Social Security.100 The majority of people, 61.5%, believed too little was spent on Social Security. Only 5.49% thought too much was spent. The large amount of support for Social Security might be due to the fact that the average age of the sample is 45, which suggests that a large percentage of the sample was currently or close to receiving Social 99
Ibid.
100
Ibid.
44 Security benefits. Additionally, people receive Social Security if they have paid into the system. So, the sample might have felt that people were owed money from their previous contributions.
Table 7B. Frequency Data for Social Security Social Security Spending Frequency Percent of Sample Too little
1390
61.50
About right
742
32.83
Too much
124
5.49
2,260
100.00
Total
Table 7C gives the frequency data for Foreign Aid. People in the survey were asked whether the country is spending "too little," "about right," or "too much" on foreign aid.101 About 65% of people thought the United States was giving too much aid to foreign entities. One reason this might could occur is due to in-group bias. People tend to prefer helping their own group compared to outsiders.102 People may indeed want more aid for foreign people but nevertheless want to reduce aid to them. This could be due to people realizing that funds are limited and giving foreign aid might mean less money for their preferred program.
101 102
Ibid.
Herbert Gintis et al., “Strong Reciprocity and the Roots of Human Morality,” Social Justice Research 21, no. 2 (June 2008): 241-53.
45 Table 7C. Frequency Data for Foreign Aid Foreign Aid Spending Frequency Percent of Sample Too little
86
7.66
About right
305
27.16
Too much
732
65.18
1,123
100.00
Total
Table 7D shows the frequency data for Welfare. People in the survey were asked whether the country is spending "too little," "about right," or "too much" on welfare.103 The majority of the sample, 43.43%, think too much is spent on welfare. Other investigations into preferences for welfare find similar opposition to welfare.104 For some people, welfare might violate their beliefs about fairness.105 The fact that some people have the ability to work, but do not and still receive money might seem unfair for certain people.
103
Ibid.
104
Ibid.
105
Ibid.
46 Table 7D. Frequency Data for Welfare Welfare Spending Frequency Percent of Sample Too little
228
20.38
About right
405
36.19
Too much
486
43.43
1,119
100.00
Total
Table 7E displays the frequency data for Helping the Poor. People in the survey were asked to indicate on a 5-point scale the position they most agreed with. Respondents could choose 1, "I strongly agree the government should improve living standards of the poor," to 3, "I agree with both answers," to 5, "I strongly agree that people should take care of themselves."106 Approximately 47% of people agreed that the government should improve the standard of living for the poor and people should help themselves. This suggests that a large amount of people believed that people needed assistance but also were somewhat in control of their situation.
106
Tom W. Smith, Peter Marsden, Michael Hout and Jibum Kim, “General Social Surveys, 1972-2010 (Data File and Codebook),” National Opinion Research Center: The Roper Center for Public Opinion (2011)
47 Table 7E. Frequency Data for Helping the Poor Should the Government Improve the Standard of Living for the Poor? Frequency
Percent of Sample
Govt Action
262
17.33
2
160
10.58
Agree with both
708
46.83
4
212
14.02
People help themselves
170
11.24
1,512
100.00
Total
Local Issues The local issues are important to look at because they differ from the national issues. Decisions to give to the homeless or charity are voluntary. In addition they look at actions as opposed to opinions and can potentially add to the understanding of the role that empathetic concern plays in decision making. Table 7F shows the frequency data for Giving to Homeless. People in the survey were asked if they had given to the homeless at least once in in the past year. About 65 % gave money or food to a homeless person in the past year. Care must be taken in drawing conclusions from this however as giving to a homeless person is very much dependent on whether a person has had the opportunity to do so. Some people may have not given anything because in the area they live homeless people are quite rare. On the other hand, a person who is walking to work might give to a homeless person once in order to not look bad in front of other people passing by, but from then on may avoid
48 walking that way in the future. In addition, the amount given or how often a person in the sample gave to the homeless cannot be inferred from the data.
Table 7F: Frequency Data for Giving to Homeless Given food or money to a homeless person at least once in the past year Frequency Percent of Sample Yes No Total
1,506
64.91
814
35.09
2,320
100.00
Table 7G displays the frequency data for Giving to Charity. People in the survey were asked if they had given to charity at least once in in the past year. The majority of people, 79.41%, gave to charity at least once in the past year. This result is interesting because there are financial incentives to donating money or goods to charity in the form of tax deductions. The incentives to give to charity only typically affect people with high incomes, since people can lower their taxable incomes via itemized deductions. It is usually more beneficial for people with lower incomes to claim the standard deduction. Despite these caveats, it is possible that part of the reason for the large participation is due to tax incentives.
49 Table 7G. Frequency Data for Giving to Charity Given money to a charity at least once in the past year Frequency Percent of Sample Yes No Total
1,843
79.41
478
20.59
2,321
100.00
Table 7H shows the frequency data for Volunteer. People in the survey were asked if they had volunteered at least once in in the past year. In contrast to the charity results, 51.68 percent of people did not volunteer for charity at any time in the past year. This data does give information on how often people people volunteer or what they volunteered for.
Table 7H. Frequency Data for Volunteer Done volunteer work for charity at least once in the past year Frequency Percent of Sample Yes
1,122
48.32
No
1,200
51.68
Total
2,322
100.00
CHAPTER 4 ECONOMETRIC METHODS National Issues The dependent variables for the national issues are income redistribution, Social Security, welfare, foreign aid, and helping the poor. These dependent variables are discrete and ordered. Following the empirical strategy used by Alesina and La Ferrara, the following ordered probit model is to be estimated:107 #
Y it = α EC it + β X it + γ T + ε it ,
(4)
#
where Y it is a latent variable for individual i at time t. EC it is the measure of empathetic concern that individual i has at time t.
X it is a vector of control variables like
education and age for individual i at time t. T represents the time dummy and εit is the #
error term. Of course Y it is not observed, but Y it is observed and takes a discrete value depending on the variable in question. That is,108 #
Y it = k if μ k−1 ⩽ Y it ⩽ μ k for k = 1, 2,....., n Here μ k represents the estimated cutpoints where μ 0 = −∞ , μ n = ∞ , and n represents the discrete values Y it takes. For example, when Y it is Income Redistribution, then: 107
Alberto Alesina and Eliana La Ferrara, “Preferences For Redistribution in the Land of Opportunities,” 897-931. 108
Peter Kennedy, A Guide to Econometrics, 6th ed. (Malden, MA: Wiley-Blackwell, 2008), 241-80.
50
51 Income Redistributionit = k if μ k−1 ⩽ Income Redistribution#it ⩽ μ k for k = 1, 2,3, 4,5, 6, 7 . The ordered probit model is estimated using maximum likelihood. For simplicity, let me rewrite equation (4) as follows: #
Y it = β x it + εit ,
(5)
where x it is a vector containing the empathy measure, control variables, and time dummies. The distribution function for ordered probit is standard normal, denoted Φ( u) . The probabilities are given by:109 P( y it ∣ x it ) = Φ(μ k − β x it) − Φ(μ k−1 − β x it ) for k = 1, 2,....., n .
The marginal effects are calculated by taking the partial derivative. That is, ∂P( y it ∣ x it ) = [ϕ(μ k − βx it ) − ϕ(μ k−1 − βx it )](−β) , ∂x it where ϕ(u ) is the density function.110 Local Issues The local issues are Giving to Homeless, Giving to Charity, and Volunteer. These variables are binary. Due to this, the following probit model is to be estimated: #
Y it = α EC it + β X it + γ T + ε it ,
(6)
# where Y it is a latent variable for individual i at time t. EC it is the measure of empathetic
concern that individual i has at time t.
X it is a vector of control variables like education
and age for individual i at time t. T represents time dummies and εit is the error term. Of 109
Rainer Winkelmann and Stefan Boes, Analysis of Microdata (New York: Springer, 2006), 174-85.
110
Ibid.
52 # # course Y it is not observed, but y it is observed, where y it = 1 if y it > 0 and y it = 0 if
#
y it < 0 . For simplicity, let equation (6) be rewritten as follows: #
Y it = βx it + εit ,
(7)
where x it is a vector containing the empathy measure, control variables, and time dummies. The response probability is given by: P( y it = 1 ∣ x it ) = Φ (βx it) ,
where Φ( u) is the cumulative standard normal distribution function.111 The probit function is estimated using maximum likelihood. The marginal effects are calculated by taking the partial derivative. That is, ∂P( yit =1 ∣ x it ) = ϕ(β x it ) B , ∂xit where ϕ(u ) is the density function. Strategy and Potential Problems One area of interest is whether general beliefs can influence the degree to which empathetic concern influences preferences for distributive justice. To investigate this the sample will be restricted to only people who have a certain belief. For example, in one regression may only include people who believe that people cannot be trusted, and another might include only people who believe people can be trusted. An additional interest is whether the coefficient on empathy is different between these two groups. In order to test this, seemingly unrelated estimation will be used. This allows the use of the
111
Peter Kennedy, A Guide to Econometrics, 241-80.
53 Wald test in order to test the null that the empathy coefficients for the separate groups are the same. One worry is the potential for reverse causality. This is especially worrisome in the ordered probit model because of the dependent variables. Including political views as explanatory variables is where the problem is most likely to occur. This is because it is unclear if preferences for these national issues make people more likely to be in a certain political group or if the political group determines a person's preferences for these national issues. For example, it is plausible that being extremely liberal makes a person more likely to strongly support government redistribution of wealth. It is also plausible that believing strongly in government redistribution of wealth leads someone to become an extreme liberal. Redistribution at the national level is not necessarily voluntary, so political views on the role of government are likely to be more influential. In a perfect world, there would be some instrumental variable that could be used for political views. For example, in the early 1990's there was a large AIDS scare. During this time widespread myths and misinformation spread about AIDS. It is possible that the fear and misunderstanding of AIDS could have caused some people to become more conservative in their views about sexuality. This might have made some people more likely to identify themselves as conservative. Views about AIDS might not be correlated with these issues of distributive justice. Unfortunately, the data set does not seem to contain any suitable instruments that can be used for political views. Possible issues that might give inconsistent estimators are measurement error and omitted variable bias. Although evidence points otherwise, it is possible that the
54 measurement of the empathy variable is not truly capturing empathetic concern.112 Omitted variable bias could occur if there exists some variable that is correlated with empathy and influences the dependent variable. For example, assume that being an orphan makes a person have more empathetic concern and more likely to support income redistribution. Then not including whether a person is an orphan could present omitted variable bias. Another potential issue is that the variable generated from factor analysis is used as an explanatory variable. This creates a potential problem because any measurement error cannot be accounted for. Following the approach used by Kleinjans, Humlum, and Nielsen, standard errors are estimated using bootstrapping.113 One concern is the validity of survey data. All survey data comes with the inherent risk that people are not being truthful. People might not say how they actually feel but instead what they think others would want them to say. This however does not mean survey data has no use. Every empirical strategy has potential flaws, and the best approach would be to attempt to use as many strategies as possible to tackle an issue. That way if all the results point in a similar direction researchers can be more confident that they are on to something. That being said, it is still necessary to scrutinize the validity of any empirical approach. In contingent valuation studies, results from different surveys attempting to measure the willingness to pay for the same good often report 112
Mark Ottoni Wilhelm, “Helping Behavior, Dispositional Empathic Concern, and the Principle of Care,” Social Psychology Quarterly 73, no. 1 (March 2010): 11-32. 113
Kristin J. Kleinjans, Maria K. Humlum and Helena S. Nielsen, “An Economic Analysis of Identity and Career Choice,” Economic Inquiry (forthcoming).
55 widely different answers.114 While this is disconcerting, this might be less of an issue for the types of survey questions this paper is looking at. The reason why this could be so comes from the results of the Wason selection task. In the original task, people are asked to solve a logic problem involving four double sided cards, two with numbers and two with letters.115 Given an "if-then" rule, the person is supposed to find situations that violate the conditional statement.116 This turned out to be quite difficult for people. In most studies, people selected the correct answer less than 25% of the time.117 However, when the same logic task is changed from numbers and letters to detecting underage drinkers the results change to around 75% of people answering correctly. 118 The logic is exactly the same, but people seem to be much better at detecting rule violations in regards to people. When a person answers how much a polluted lake is worth to them, they might not really know. Due to this uncertainty they might be influenced a lot more by how the question is asked and rely on general beliefs in order to give an answer. However, in the case of moral beliefs the answers might be more reliable as people might have clearer ideas about what they think is right. The way issues are framed can create potential issues as well. This makes it difficult to be certain that results are meaningful. I hope by 114
Peter A. Diamond and Jerry A. Hausman, “Contingent Valuation: Is Some Number Better Than No Number,” Journal of Economic Perspectives 8, no. 4 (Autumn 1994): 45-64. 115
Leda Cosmides, John Tooby and Jerome Barkow, The Adapted Mind: Evolutionary Psychology and the Generation of Culture, (New York: Oxford University Press, 1995), 163-238. 116
Ibid.
117
Ibid.
118
Ibid.
56 including a number of dependent variables with different scales and different issues I can get a general idea of the role that empathetic concern plays.
CHAPTER 5 RESULTS AND DISCUSSION Results for National Issues Table 8 reports the results from the ordered probit regressions for each of the five national issues. Each column labeled (1) to (5) is a separate regression. The coefficient for empathy is significant at the 1% level for Income Redistribution, Social Security, and Helping the Poor. However, the results for Foreign Aid, and Welfare are not statistically significant. The lack of significance for Foreign Aid could be due to the large social distance between members of the U.S. to recipients of foreign aid as well as uncertainty as to whether the aid really gets there. Charness and Gneezy conducted an experiment to test whether social distance influences giving in dictator games.119 Some participants were given the name of the person they played the game with.120 People who knew the name gave significantly higher amounts to the other person.121 Due to the lack of exposure to the need of foreign people as well as in-group bias it might be harder for people to empathize with outsiders. The lack of significance for welfare could be due to people's concerns about fairness. Other studies have suggested that welfare conflicts with
119
Gary Charness and Uri Gneezy, “What's in a Name? Anonymity and Social Distance in Dictator and Ultimatum Games,” Journal of Economic Behavior and Organization 68, no. 1 (October 2008): 29-35. 120
Ibid.
121
Ibid.
57
58 people's moral principles.122 People might not perceive that others really need welfare and see them instead as free-riders. One issue that influences both foreign aid and welfare is that people vastly overestimate the degree of fraud and the amount of money spent on these programs.123 This can partly explain why there is so little support for these issues. The significance of the empathy coefficient on Income Redistribution suggests that empathetic concern might influence preferences for redistribution of wealth. This fits with the theory that people who are concerned about the needs of others might be more likely to prefer policies that benefit the people they are concerned about. A concern with the needs of the elderly could be a potential reason why the empathy coefficient for Social Security is significant. Having more empathetic concern might make people more likely to value the welfare of people who are on Social Security. In contrast to the results for Welfare, the coefficient for empathy for Helping the Poor is significant. There seems to be some difference between why empathetic concern might influence preferences for helping the poor but be less of a factor for welfare. It is possible that this is due to concerns about fairness. People seem to want to help the poor but do not seem to think welfare is the right approach. It is possible that many people simply do not think that welfare programs are worthwhile. That is they may be empathetic towards those that need welfare but think that the program does not solve the problem and hence are unwilling to trade off their own welfare to support it. 122
Herbert Gintis et al., “Strong Reciprocity and the Roots of Human Morality,” 241-53.
123
Ibid.
59 Table 8. Ordered Probit Results: Coefficient Estimates for the National issues Dependent Variable
(1) (2) Income Redistribution Social Security (-0.115*** (-0.124*** -(0.039) -(0.025) Educ --0.035*** --0.041*** -(0.010) -(0.012) Age --0.003 --0.005*** -(0.002) -(0.001) Female (-0.0904* (-0.296*** -(0.052) -(0.061) Black (-0.419*** (-0.238** -(0.105) -(0.098) Other Race (-0.328*** -0.086 -(0.101) -(0.122) 2004 --0.0172 (-0.118** -(0.052) -(0.052) Extremely Liberal (-0.489*** (-0.042 -(0.176) -(0.168) Liberal (-0.341*** (-0.012 -(0.101) -(0.102) Slightly Liberal (-0.063 (-0.009 -(0.069) -(0.080) Slightly Conservative --0.243*** --0.091 -(0.085) -(0.075) Conservative --0.432*** --0.310*** -(0.079) -(0.075) Extremely Conservative --0.654*** --0.400*** -(0.192) -(0.120) Log real income --0.115*** --0.027 -(0.030) -(0.023) cut1_cons -)0.751** -)1.203*** -(0.315) -(0.255) cut2_cons -)1.090*** -)2.558*** -(0.317) -(0.261) cut3_cons -)1.617*** -(0.314) cut4_cons -)2.143*** -(0.317) cut5_cons -)2.581*** -(0.320) cut6_cons -)2.911*** -(0.330) N -)1531 -)2256 pseudo R-sq -)0.037 -)0.038 Bootstrapped standard errors in parentheses; *p < 0.10 ** p < 0.05 *** p < 0.01. Note: White, 2002, and Moderate are the base variables. Independent Variables Empathy
60 Table 8. Ordered Probit Results: Coefficient Estimates for the National issues- continued Independent Variables Empathy Educ Age Female Black Other Race 2004 Extremely Liberal Liberal Slightly Liberal Slightly Conservative Conservative Extremely Conservative Log real income cut1_cons cut2_cons cut3_cons cut4_cons cut5_cons
Dependent Variable (3) (4) Foreign Aid Welfare -)0.013 (-0.0617 -(0.039) -(0.046) (-0.034** (-0.006 -(0.015) -(0.011) -)0.002 (-0.006*** -(0.002) -(0.002) -)0.0869 (-0.010 -(0.086) -(0.059) (-0.233* (-0.575*** -(0.119) -(0.134) (-0.500*** (-0.278** -(0.142) -(0.114) (-0.068 -)0.0119 -(0.077) -(0.077) (-0.391 (-0.889*** -(0.240) -(0.193) (-0.155 (-0.141 -(0.132) -(0.113) (-0.202* (-0.0987 -(0.109) -(0.122) -)0.218* -)0.310*** -(0.128) -(0.108) -)0.197* -)0.352*** -(0.110) -(0.116) -)0.223 -)0.423** -(0.242) -(0.213) -)0.113*** -)0.094** -(0.034) -(0.039) (-0.772** (-0.303 -(0.386) -(0.406) -)0.31 -)0.759* -(0.376) -(0.407)
(5) Helping the Poor (-0.111*** -(0.043) -)0.009 -(0.012) -)0.002 -(0.002) (-0.032 -(0.060) (-0.540*** -(0.089) (-0.328*** -(0.113) (-0.0253 -(0.050) (-0.255 -(0.232) (-0.328*** -(0.092) (-0.144* -(0.084) -)0.304*** -(0.074) -)0.381*** -(0.084) -)0.456** -(0.213) -)0.119*** -(0.029) -)0.374 -(0.258) -)0.759*** -(0.264) -)2.107*** -(0.270) -)2.689*** -(0.282)
cut6_cons N 1123 1119 1512 pseudo R-sq 0.03 0.047 0.042 Bootstrapped standard errors in parentheses; *p < 0.10 ** p < 0.05 *** p < 0.01. Note: White, 2002, and Moderate are the base variables.
61 Income Redistribution Column (1) of Table 8 gives the coefficient estimates for Income Redistribution, and Table 8 reports the marginal effects and predicted probabilities. The marginal effects for empathy are significant at the 1% level. Table 9 shows that an increase in one standard deviation for empathy leads to an approximately 2.9 percentage point increase in the probability that someone will strongly support government redistribution of wealth. Similarly, an increase in one standard deviation for empathy leads to an approximately 2.2 percentage point decrease that someone will strongly believe in no role for government redistribution. Looking at the signs for the marginal effects across all seven choices it is clear that empathetic concern tends to push the probability distribution to the left. The overall predicted probabilities for the average person report that about 16.8% of people strongly support government redistribution. Comparing this to the marginal effect, empathetic concern has a significant effect on the probability that a person will strongly support income redistribution. In addition, the predicted probabilities are calculated when empathy is at 0.5 and -0.5. Each predicted probability is statistically different from the overall category at the 5% level. This result fits with the theoretical model. People who have more empathetic concern seem to be more likely to want income redistribution. The marginal effects for education are significant at the 1% level. Looking at the marginal effects, increases in the years of education seem to make support for redistribution less likely. An increase of one year in education leads to a 0.88 percentage point decrease in the probability of strongly supporting redistribution, and a 0.68 percentage point increase in the probability of strongly supporting no role for
62 government redistribution. The race variables are both significant at the 1% level. Being Black increases the probability of strong support for government redistribution by approximately 12 percentage points compared to whites. Being in the Other Race category increases in the probability by 9.3 percentage points in strongly supporting government redistribution of wealth compared to whites. Most of the dummy variables for political views are significant at the 1% level. However, caution must be drawn in making inferences about these results as there is worry about reverse causality. Does a person's political views determine their preferences for national issues like redistribution or does preferences for redistribution lead a person to prefer a certain political ideology? An answer to this question is beyond the scope of this paper, but the political views are of concern since they have been used as control variables. Fortunately, leaving them off or including them leads to no discernible difference in the results for the coefficient on empathy which is the main interest of this paper. With the caveat that the results for political views are possibly suspect, the coefficients on Extremely Liberal, Liberal, Slightly Conservative, Conservative, and Extremely Conservative are significant at the one percent level. The base variable are those who describe themselves as Moderate. Describing oneself as Extremely Liberal suggests an increase of 14.9 percentage points in strong support for redistribution and a decrease of 7.1 percentage points in strong opposition to government redistribution relative to Moderates. The marginal effects are similar for Liberal although smaller in magnitude. Looking at people who describe themselves as Extremely Conservative, there is a decrease in probability of 11.9 percentage points in strong
63 support for redistribution and an increase of 17.4 percentage points in strongly believing there is no role for government redistribution relative to Moderates. Similar results are seen in Conservative and Slightly Conservative except the marginal effects are smaller in magnitude. The results for the income variable is significant at the 1% level. The marginal effects suggest that an increase of 100% in real income leads to a 2.89 percentage point decrease in the probability of strongly supporting government redistribution and a 2.23 percentage point increase in the probability of strongly believing that government should have no role in the redistribution of income.
64 Table 9. Marginal Effects at the Average for Income Redistribution Gov Red No Gov Red Dependent Variable 1 2 3 4 5 6 7 Empathy 0.0290*** 0.0089*** 0.0079*** 0.0039*** - 0.0104*** - 0.0091*** - 0.0224*** Educ - 0.0088*** - 0.0027*** - 0.0024*** 0.0012*** 0.0031*** 0.002*** 0.0068*** Age - 0.0007 - 0.0002 - 0.0002 0.0001 0.0002 0.0002 0.0005 Female^ 0.0227* 0.0070* 0.0062 - 0.0030* - 0.0081 - 0.0071* - 0.0176 Black^ 0.1211*** 0.0284*** 0.0162*** - -0.0259***- -0.0413***- 0.0317*** - 0.0667*** Other Race^ 0.0934*** 0.0228*** 0.0139*** - 0.0194*** - 0.0323*** - 0.0251*** - 0.0534*** 2004^ - 0.0043 - 0.0013 - 0.0012 0.0006 0.0016 0.0014 0.0033 Extremely Liberal^ 0.1485*** 0.0306*** 0.0124*** - 0.0354*** - 0.0491*** - 0.0359*** - 0.0711*** Liberal^ 0.0968*** 0.0238*** 0.0148 - 0.0198 - 0.0335 - 0.0261 - 0.0560 Slightly Liberal^ 0.0162 0.0048 0.0041 - 0.0024 - 0.0058 - 0.0050 - 0.0119 Slightly Conservative^ - 0.0562*** - 0.0193*** - 0.0197*** 0.0042*** 0.0200*** 0.0191*** 0.0519*** Conservative^ - 0.0937*** - 0.0344*** - 0.0382*** 0.0020*** 0.0322*** 0.0331*** 0.0990*** Extremely Conservative^ - 0.1190*** - 0.0511*** - 0.0679*** - 0.0168*** 0.0348*** 0.0456*** 0.1739*** Log Real Income - 0.0289*** - 0.0089*** - 0.0079*** 0.0039*** 0.0104*** 0.0091*** 0.0223*** Predicted Probabilities at variable means Overall 0.1684 0.0987 0.1955 0.2043 0.1410 0.0770 0.1150 Empathy at 0.5a 0.1837** 0.1033** 0.1992** 0.2020** 0.1355** 0.0724** 0.1040** Empathy at -0.5a 0.1547** 0.0944** 0.1914** 0.2059** 0.1459** 0.0815** 0.1263** Male 0.1565 0.0950 0.1920 0.2057 0.1453 0.0810 0.1247 Femaleb 0.1792 0.1020** 0.1982** 0.2027** 0.1315** 0.0737** 0.1071 ^dy/dx is for discrete change of dummy variable from 0 to 1; * p < 0.10 ** p < 0.05 *** p < 0.01. a statistically different compared to the Overall category b statistically different compared to the Male category
Social Security Column (2) in Table 8 shows the results for Social Security. The coefficient on empathy is significant at the 1% level. Table 10 lists the marginal effects and predicted probabilities for column (2). An increase in one standard deviation for empathy increases the probability that a person thinks too little is spent on Social Security by 4.7 percentage points. The overall predicted probability for too little is 0.6208. This result suggests that preferences for Social Security spending could be related to perceptions about the need and concerns about the welfare of older citizens. There is a common belief that Social
65 Security is progressive.124 This common belief may not be completely accurate, since people who are well off tend to live longer and receive benefits for a longer period of time compared to people who are less well off.125 A particularly interesting result is the coefficient for females which is significant at the 1% level. Being a female leads to a 11.25 percent point increase in the probability of feeling that Social Security receives too little funding relative to males. One potential explanation for this result could be that women tend to live longer compared to men. Since they live longer, they will receive benefits for more years. The dummy variable for the year 2004 is significant at the 1% level. People surveyed in the year 2004 were about 4 percentage points more likely to believe Social Security receives too little funding compared to the year 2002. This is the only regression that the year variable is significant. The race variable for Black is significant at the 1% level but the Other Race variable is not significant. Most of the political dummy variables are not significant except for people who described themselves as Conservative or Extremely Conservative. People who are extremely conservative are 15.7 percentage points less likely to believe Social Security receives too little funding and about 5 percentage points more likely to believe Social Security receives too much compared to Moderates. Again, the worry with the political views is the chance of reverse causality. The results are in line with what should be expected. People who identify themselves as conservative generally support less government spending including government programs like Social Security. The coefficient for income is not 124
Jeffrey B. Liebman, “Redistribution in the Current U.S. Social Security System,” NBER Working Paper Series w8625 (December 2001): http://ssrn.com/abstract=293238. 125
Ibid.
66 significant. It is possible that income is less of a factor in decisions about Social Security.
Table 10. Marginal Effects at the Average for Social Security Too Little About Right Empathy 0.0470*** - 0.0346*** Educ - 0.0154*** 0.0113*** Age 0.0170*** - 0.0126*** Female^ 0.1125*** - 0.0823*** Black^ 0.0800*** - 0.0616** Other Race^ - 0.0331 0.0235 2004^ 0.0447** - 0.0329** Extremely Liberal^ 0.0159 - 0.0118 Liberal^ 0.0044 - 0.0032 Slightly Liberal^ 0.0026 - 0.0019 Slightly Conservative^ - 0.0350 0.0254 Conservative^ - 0.1207*** 0.0837*** Extremely Conservative^ - 0.1572*** 0.1032*** Log Real Income - 0.0104 0.0076 Predicted Probabilities at variable means Overall 0.6208 0.3309 Empathy at 0.5a 0.6442** 0.3135** a Empathy at -0.5 0.5972** 0.3480** Male 0.5610 0.3732 Femaleb 0.6734** 0.2909** ^dy/dx is for discrete change of dummy variable from 0 to 1; *p < 0.10 **p < 0.05 ***p < 0.01. a statistically different compared to the Overall category b statistically different compared to the Male category
Too Much - 0.0124*** 0.0041*** - 0.0044*** - 0.0301*** - 0.0185** 0.0092 - 0.0118** - 0.0041 - 0.0011 - 0.0007 0.0100 0.0370*** 0.0540*** 0.0027 0.0482 0.0424** 0.0548** 0.0658 0.0356**
Foreign Aid Column (3) in Table 8 lists the results of the regression on Foreign Aid, and Table 11 gives the marginal effects and predicted probabilities. The coefficient and marginal effects for empathy are not significant. This suggests that empathetic concern does not seem to influence preferences for spending on foreign aid. In one study by Chong and
67 Gradstein, the two main factors found that influenced support for foreign aid were opinions about the performance of a person's own government and personal income.126 These factors might be strong enough to limit any influence that empathetic concern might have. It is possible that if people visited the places, learned more about the aid recipients, or viewed video and images of the people where the aid goes to, then empathetic concern might have a greater influence on preferences. Education is significant at the 5% level. Table 10 suggests that an increase of one year of education leads to an approximately 0.5 percentage point increase in beliefs that the United States spends too little on foreign aid. The coefficients for Black and Other Race are significant at the 10% and 1% level respectively. For people who are a race other than Black or White, there is a 19 percentage point decrease in the probability that they will believe that too much money is spent on foreign aid compared to whites. Income is significant at the 1% level and a 100% increase in income leads to a 1.5 percentage point decrease in the probability that a person will think too little is spent on foreign aid. This result is large since the overall predicted probability for too little is 0.0698.
126
Alberto Chong and Mark Gradstein, “Who's Afraid of Foreign Aid? The Donors' Perspective,” CESifo Working Paper Series No. 1833 183 (October 2006): SSRN: http://ssrn.com/abstract=944422.
68 Table 11. Marginal Effects at the Average for Foreign Aid Foreign Aid Spending Too Little About Right Empathy - 0.0017 - 0.0031 Educ 0.0046** 0.0080** Age - 0.0002 - 0.0003 Female^ - 0.0117 - 0.0204 Black^ 0.0356* 0.053* Other Race^ 0.0902*** 0.1041*** 2004^ 0.0091 0.0160 Extremely Liberal^ 0.0678* 0.0839*** Liberal^ 0.0228 0.0358 Slightly Liberal^ 0.0303* 0.0461* Slightly Conservative^ - 0.0262 - 0.0515 Conservative^ - 0.0238 - 0.0466* Extremely Conservative^ - 0.0256 - 0.0529 Log Real Income - 0.0151*** - 0.0266*** Predicted Probabilities at variable means Overall 0.0698 0.2765 Empathy at 0.5a 0.0689 0.2750** Empathy at -0.5a 0.0707 0.2781** Male 0.0762 0.2873 Femaleb 0.0645 0.2669** ^dy/dx is for discrete change of dummy variable from 0 to 1; *p < 0.10 **p < 0.05 ***p < 0.01. a statistically different compared to the Overall category b statistically different compared to the Male category
Too Much 0.0048 - 0.0125** 0.0005 0.0321 - 0.0886* - 0.1943*** - 0.0251 - 0.1517* - 0.0586 - 0.0764* 0.0777 0.0705* 0.0786 0.0417*** 0.6537 0.6561 0.6513 0.6366 0.6687
Welfare Column (4) in Table 8 shows the results for Welfare, and Table 12 shows the marginal effects and predicted probabilities. Similarly to foreign aid, the coefficient and marginal effects for empathy are not significant. Preferences for welfare spending do not seem to be influenced by empathetic concern. This could be due to a lack of perceived need. Other studies have suggested that people overestimate both the degree of fraud present in welfare and how much of the federal budget is spent on welfare.127 If people 127
Herbert Gintis et al., “Strong Reciprocity and the Roots of Human Morality,” Social Justice Research 21, no. 2 (June 2008): 241-53.
69 believe that the recipients are not truly in need or that the aid is not going to the people who need it, then empathetic concern is not likely to be as much of a factor. Both race variables are significant with Black at the 1% level and Other Race at the 5% level. Table 11 shows that people who identify themselves as black have an 18.4 percentage point increase in probability that they will think welfare receives too little. The results for race are quite large, especially when compared to the overall predicted probability for too little of 0.1901. Similar research by Luttmer found that support for welfare increases as the local percentage of welfare recipients increased for a person's own race, but support was lower for racially heterogeneous areas.128 These results are likely due to group loyalty. 129 That is, people might choose the policy that best helps their own group. The results for Extremely Liberal are significant at the 1% level and they have a 31 percentage point increase in probability that they will believe that welfare receives too little compared to Moderates. This is the only liberal category that is statistically significant. This contrasts with people who identify as conservative which all are significant at at least the 5% level. People who are extremely conservative are about 16.8 percentage points more likely to think that welfare receives too much money.
128
Erzo F. P. Luttmer, “Group Loyalty and the Taste for Redistribution,” Journal of Political Economy 109, no. 3 (June 2001): 500-28. 129
Ibid.
70 Table 12. Marginal Effects at the Average for Welfare Too Little About Right Empathy 0.0168 0.0075 Educ 0.0017 0.0008 Age 0.0015*** 0.0007*** Female^ 0.0026 0.0012 Black^ 0.1836*** 0.0252*** Other Race^ 0.0829*** 0.0226*** 2004^ - 0.0032 - 0.0014 Extremely Liberal^ 0.3101*** - 0.0195*** Liberal^ 0.0401 0.0144 Slightly Liberal^ 0.0277 0.0107 Slightly Conservative^ - 0.0766*** - 0.0464*** Conservative^ - 0.0850*** - 0.0543*** Extremely Conservative^ - 0.0951*** - 0.0725** Log Real Income - 0.0254*** - 0.0113*** Predicted Probabilities at variable means Overall 0.1901 0.3832 Empathy at 0.5a 0.1983 0.3867** Empathy at -0.5a 0.1816 0.3793** Male 0.1887 0.3826 Femaleb 0.1913 0.3838** ^dy/dx is for discrete change of dummy variable from 0 to 1; *p < 0.10 **p < 0.05 ***p < 0.01 a statistically different compared to the Overall category b statistically different compared to the Male category
Too Much - 0.0242 - 0.0025 - 0.0022*** - 0.0038 - 0.2089*** - 0.1055*** 0.0047 - 0.2906*** - 0.0546 - 0.0384 0.1229*** 0.1393*** 0.1676** 0.0367*** 0.4267 0.4149 0.4391 0.4287 0.4249
Helping the Poor Column (5) in Table 8 has the results for Helping the Poor, and Table 13 gives the marginal effects and predicted probabilities.. The coefficient for empathy is significant at the 1% level. An increase in one standard deviation in empathy suggests a 2.67 percentage point increase in the probability people will strongly support government action in helping the poor. Similarly, an increase in one standard deviation in empathy makes a person about 1.9 percentage points less likely to strongly believe that the poor just need to take care of themselves. In addition, the predicted probabilities are
71 calculated when empathy is at 0.5. Each predicted probability is statistically different from the overall category at the 5% level. This result makes sense as people who have empathetic concern for the poor are likely to be motivated to help them. Both race variables are significant at the 1% level. People who identify themselves as Black have a 15.5 percentage point increase in the probability that they will strongly support government action in helping the poor. All the political variables are significant at at least the 10% level. The signs on the marginal effects clearly show that liberals are more likely to support government action and conservatives are more likely to believe people should help themselves. This result fits with the ideologies of both groups. Conservatives tend to have less support for government intervention. Liberals are more likely to support government help. Income is significant at the 1% level and suggests that an increase in real income of 100% leads to a 2.85 percentage point decrease in the probability that a person will strongly support government action and a 2 percentage point increase in the likelihood that they strongly believe poor people need to help themselves.
72 Table 13. Marginal Effects at the Average for Helping the Poor Gov Action Agree w/ both People help them 1 2 3 4 5 Empathy 0.0267*** 0.0099*** - 0.0023 - 0.0153*** - 0.0190*** Educ - 0.0022 - 0.0008 0.0002 0.0013 0.0016 Age - 0.0006 - 0.0002 0.0001 0.0003 0.0004 Female^ 0.0076 0.0028 - 0.0006 - 0.0044 - 0.0054 Black^ 0.1553*** 0.0401*** - 0.0558 - 0.0699*** - 0.0698*** Other Race^ 0.0897*** 0.0264*** - 0.0260 - 0.0438*** - 0.0462*** 2004^ 0.0061 0.0022 - 0.0005 - 0.0035 - 0.0043 Extremely Liberal^ 0.0683 0.0210 - 0.0179 - 0.0343 - 0.0370*** Liberal^ 0.0887*** 0.0266*** - 0.0245 - 0.0439*** - 0.0469 Slightly Liberal^ 0.0364 0.0124 - 0.0063 - 0.0197 - 0.0229 Slightly Conservative^ - 0.0652*** - 0.0275*** - 0.0071 0.0408*** 0.0590*** Conservative^ - 0.0797*** - 0.0345*** - 0.0128 0.0507*** 0.0763*** Extremely Conservative^ - 0.0866** - 0.0413** - 0.0308 0.0584** 0.1002** Log Real Income - 0.0285*** - 0.0105*** 0.0025 0.0163*** 0.0203*** Predicted Probabilities Overall 0.1561 0.1097 0.4992 0.1391 0.0959 Empathy at 0.5a 0.1702** 0.1147** 0.4973** 0.1312** 0.0865** Empathy at -0.5a 0.1435** 0.1049** 0.4997 0.1465** 0.1055** Male 0.1521 0.1082 0.4994 0.1414 0.0988 Femaleb 0.1597 0.1110** 0.4988 0.1370** 0.0934 ^dy/dx is for discrete change of dummy variable from 0 to 1; *p < 0.10 **p < 0.05 ***p < 0.01 a statistically different compared to the Overall category b statistically different compared to the Male category
73 Results for Beliefs The main interest of this paper is to examine how empathetic concern may influence these preferences. It is possible that the ability for empathetic concern to influence people to sacrifice their own payoffs to increase other people's payoffs is affected by a person's beliefs. This hypothesis is explored by restricting the samples by certain beliefs. Doing this allows testing to see if empathetic concern is more influential given an individual's beliefs. In order to see if the coefficients for empathy are different between these groups seemingly unrelated estimation will be used. Seemingly unrelated estimation allows testing of whether the estimates for empathy given certain beliefs are statistically different from each other. Results are not reported for the middle categories for each belief as they do not add much insight to the analysis. Instead, the focus will be on the two opposite beliefs. Fairness Table 14 shows the results for the national issues given general beliefs about the fairness of other people. Looking at columns 1A and 1B it is clear that the effect of empathetic concern on Income Redistribution is larger in absolute value for people who hold the general belief that if given the chance people will take advantage of you compared to people who believe that people try to be fair. Below the estimated coefficients, the results from the Wald test show that the null can be rejected at the 10 % level that the coefficients for empathy are the same. Empathy is significant at the 1 percent level for column 1A but is not significant for column 1B. This result suggests that general beliefs about human nature may moderate the influence of empathetic
74
75 concern. People who think everyone is fair might have the same level of empathetic concern as another person, but their worldview might make them less likely to think they need to make a sacrifice to help another person. This might occur because people who believe everyone is fair probably view the world as more just than a person who believes people will take advantage of others. In an unfair world, the people who are more fair are likely to be at the bottom. This idea might make people who believe this to want to sacrifice more of their own payoff to help the people they feel empathetic concern for. Columns 2A and 2B in Table 14 display the results for preferences on national security spending. Both coefficients are significant at at least the 5% level. General beliefs about the fairness of others do not seem to change the influence of empathetic concern. Although the marginal effect for people who believe people are fair is larger than people who believe people take advantage of others, the null that the two estimates are statistically similar from each other cannot be rejected. One reason this might be the case is that Social Security is different from other redistribution programs in that people have to put money in to receive benefits. In the redistribution of wealth, people must decide if people deserve the wealth that they have and beliefs about human nature are likely to influence that decision. Since people contribute towards Social Security whether people are fair or not might have less of an effect on a person's preferences for Social Security spending. People might not worry about how fair people are, but instead whether society owes the recipients more money or not. Columns 3A and 3B in Table 14 show the results for Foreign Aid. Similarly to the overall regression foreign aid is still not significant and both estimates are not
76 statistically different from each other. This result could be due to the large social distance between people of the United States and recipients of foreign aid. People may be concerned about foreign people that need aid but have stronger empathetic concern for people closer to them. Columns 4A and 4B in Table 14 show the results of empathetic concern on preferences for welfare spending given beliefs about fairness. The coefficient on empathy for people who believe people take advantage of others is significant at the 1% level while people who believe others are fair is not significant. Table 20 shows that an increase in one standard deviation in empathy increases the probability that a person thinks welfare receives too little money by approximately 3.3 percentage points given that they believe people take advantage of others. However, the null, that the two coefficients are the same, cannot be rejected. Empathetic concern does not seem to be much of a factor in determining preferences for welfare. Due to the unpopularity of the program, it may be that people are unwilling to trade their own payoffs to support this program regardless of how much they empathize with the people receiving the aid. Columns 5A and 5B in Table 14 show the results of empathy on helping the poor given ideas of fairness. Here the difference is much clearer. Given the belief that people take advantage of others empathy is significant at the 1% level, while coefficient on empathy given the belief that people are fair is not. The null, that the two estimates are the same, can be rejected at the 1% level. The marginal effects from Table 21 suggest that an increase of one standard deviation in empathy leads to a 7.23 percentage point increase in probability that someone will strongly support government action to help the
77 poor given that they believe people take advantage of others. This contrasts with the 0.17 percentage point increase in probability for people that believe others are fair. This result suggests that beliefs about the general fairness of others might influence how much a person is willing to sacrifice their own payoffs in order to help the people who are poor. Luck Table 15 details the results of empathetic concern on the national issues given a person's beliefs about how people get ahead. It is likely that the control people have in the outcomes will influence how much an effect empathy might have on allocation preferences. One worry about this set of regressions is due to the small sample size of people who believe luck determines success. Only a small minority, 9.5%, attribute success to luck. So, when the sample is restricted to only people who believe in luck, the sample size is quite small. Columns 1A and 1B show coefficients for empathy on Income Redistribution given a person's belief that success comes from luck or hard work. Both are significant at the 5% level. With a p-value of 0.109 the null, that the two coefficients are the same, cannot be rejected at the 10 percent level. Table 17 shows that an increase of one standard deviation in empathy increases the probability that a person will strongly believe in government redistribution by 11.26 percentage points given that they believe success comes from luck. Belief in hard work has smaller marginal effects, an increase in one standard deviation in empathy leads to a 2.97 percentage point increase in the probability that a person will strongly agree with government redistribution. These results are similar to what other studies have found that have examined the role accountability plays
78
79 in people's judgments.130 People who believe luck determines success are likely to attribute differences in income to be out of a person's control. Since it is not someone's fault that they have less money and people who have more income are just lucky, it makes sense that individuals would want to help the people they empathize with more. Beliefs about hard work are different. If people reach success through hard work, then people who have lower incomes could just work harder to improve their situation. It seems unfair to punish people who worked hard and earned their success to help people who did not. This could be why people who believe hard work creates success could be less likely to swayed by empathic concern for people in need. Columns 2A and 2B, give the results for Social Security spending. The coefficient for empathy is significant at the 5% level for the hard work group. The coefficient for empathy for people that believe success comes from luck is larger in absolute value than the hard work group but is not statistically significant. I fail to reject the null that these two estimates are the same. The lack of difference between these two groups could be due to the way Social Security works. As outlined before, people who pay into the system get benefits out of the system. It seems that the beliefs about the control that people have over their success does not seem to influence the degree to which empathetic concern affects preferences for Social Security spending. Columns 3A and 3B show the results for empathy on foreign aid given beliefs about success. Both coefficients are significant at the 10% level. The null,that the coefficients are the same, can be rejected at the 5% level. What is interesting about this 130
39.
James Konow, “Which Is the Fairest One of All? A Positive Analysis of Justice Theories,” 1188-
80 result is that the coefficient on empathy for people who believe in luck is positive. Looking at Table 19 an increase in one standard deviation in empathetic concern makes a person 12.7 percentage points more likely to believe too much is spent on foreign aid. This suggests that given the belief that success comes from luck, people who have more empathetic concern want less aid to go to foreign people. It is possible that people who feel more empathic concern for people in the United States view foreign aid as taking away hep that could be going to their preferred group. If a country has only a fixed amount of money, and any money spent on one program means less for another, then people who have a high degree of empathetic concern might want to decrease the amount given to foreign people in order to increase the amount of aid given to their preferred program. The reason that the luck group might be doing this is because they might feel a strong desire to fix the inequities in the United States. Caution must be taken in drawing overall conclusions however, as the sample size for people who believe in luck is quite small. Columns 4A and 4B show the results of empathy on preferences for welfare. Both coefficients are not statistically significant and the null, that the estimates are similar, cannot be rejected. This mirrors the results from the overall regression for welfare. The support for welfare spending is low with 43% of the sample believing that the United States spends too much on welfare. It seems that beliefs about whether success comes as a result of hard work or luck do not change the degree to which empathetic concern influences preferences for welfare spending.
81 Columns 5A and 5B show the results of empathy on helping the poor given beliefs about what determines success. Empathy for the hard work group is significant at the 5% level and empathy for the luck group is not significant. The null, that the coefficients are similar, cannot be rejected. It seems that beliefs about luck or hard work are not much of a factor in determining the degree to which empathetic concern influences preferences for helping the poor. Trust Concerns about the trustworthiness of other people might change the degree to which empathetic concern influences preferences for distributive justice at the national level. Table 16 gives the results for people who believe that in general people are trustworthy or if they believe people cannot be trusted. Columns 1A and 1B show the estimates for empathy on Income Redistribution given a person's trust of others. Both coefficients for empathy are significant at the 1% level, but the null, that the estimates are similar, cannot be rejected. Looking at Table 17, the marginal effects for both the trust and no trust groups are quite similar. This results suggests that beliefs about the trustworthiness of others do not seem to influence the role of empathetic concern on preferences for redistribution of wealth. Columns 2A and 2B display the regressions for opinions on Social Security spending given beliefs about trust. Both coefficients for empathy are significant at the 5% level. The coefficient is larger in absolute value for people who believe people are trustworthy compared to people who do not. However, I fail to reject the null that the two estimates are similar. Looking at Table 18, the results show that for people who
82 believe others are trustworthy an increase in one standard deviation in empathetic concern leads to a 6.57 percentage point increase in the probability that a person will think too little is spent on Social Security. For people who believe other people cannot be trusted the percentage point increase in the probability is approximately 4.2. Columns 3A and 3B show the results for foreign aid. The coefficient on empathy for the trust group is negative and for the no trust group the coefficient is positive, but both estimates are not statistically significant. The null, that the coefficients are similar, cannot be rejected. This is in line with the previous results on Foreign Aid. Even controlling for trust, empathetic concern does not seem to have much of an influence on preferences for foreign aid. Columns 4A and 4B gives the results for welfare. While the coefficient for empathy for the trust group is larger in absolute value compared to the no trust group, the results are not significant and the null, that the estimates are similar, cannot be rejected. Even accounting for trust, empathetic concern again does not seem to have much influence on preferences for welfare. Columns 5A and 5B show the results for Helping the Poor. Both empathy coefficients are significant at the 5% level, but the null, that both estimates are the same, cannot be rejected. Table 21 shows that, given a person believes other people are trustworthy, a one standard deviation increase in empathy increases the probability that a person will strongly prefer government action to help the poor by 2.77 percentage points. Similarly, people who believe others cannot be trusted are 3.1 percentage points more likely.
83
84 Clearly, the results suggest that beliefs about the general trustworthiness of other people does not seem to change the influence of empathetic concern on preferences for the issues I have examined at the national level.
Table 17. Marginal Effects at the Average for Empathy on Income Redistribution Given Beliefs Belief Gov Red (1) 2 People take advantage 0.0632*** 0.0110*** People are fair 0.0139 0.0069 People can be trusted 0.0295*** 0.0152*** People cannot be trusted 0.0350*** 0.0080*** Success comes from luck 0.1126*** 0.0329** Success comes from hard work 0.0297** 0.0088** *p < 0.10 **p < 0.05 ***p < 0.01
3 0.0095*** 0.0067 0.0157*** 0.0060*** 0.0054 0.0109**
4 5 6 -0.0124*** -0.0216*** -0.0109*** -0.0009 -0.0058 -0.0066 -0.0007 -0.0118*** -0.0153*** -0.0075*** -0.0119*** -0.0074*** -0.0360** -0.0423** -0.0222* -0.0030 -0.0122** -0.0085**
Table 18. Marginal Effects at the Average for Empathy on Social Security Given Beliefs Belief People take advantage People are fair People can be trusted People cannot be trusted Success comes from luck Success comes from hard work *p < 0.10 **p < 0.05 ***p < 0.01
Too Little 0.0485** 0.0835*** 0.0657** 0.0420** 0.0693* 0.0459***
About Right -0.0337** -0.0623*** -0.0481** -0.0303** -0.0499* -0.0341***
Too Much -0.0148** -0.0212*** -0.0176** -0.0117** -0.0194 -0.0117**
Table 19. Marginal Effects at the Average for Empathy on Foreign Aid Given Beliefs Belief People take advantage People are fair People can be trusted People cannot be trusted Success comes from luck Success comes from hard work *p < 0.10 **p < 0.05 ***p < 0.01
Too Little - 0.0058 0.0033 0.0162 - 0.0080 - 0.0452* 0.0183*
About Right - 0.0120 0.0083 0.0304 - 0.0232 - 0.0816* 0.0293*
Too Much 0.0178 - 0.0116 - 0.0466 0.0312 0.12679* - 0.0476*
No Govt Red (7) -0.0388*** -0.0142 -0.0325*** -0.0221*** -0.0503** -0.0258**
85 Table 20. Marginal Effects at the Average for Empathy on Welfare Given their Beliefs Belief People take advantage People are fair People can be trusted People cannot be trusted Success comes from luck Success comes from hard work *p < 0.10 **p < 0.05 ***p < 0.01
Too Little 0.0414* 0.0185 0.0484 0.0116 - 0.0782 0.0213
About Right 0.0161* 0.0122 0.0327 0.0050 0.0104 0.0117
Too Much - 0.0575* - 0.0307 - 0.0811 - 0.0166 0.0677 - 0.0330
Table 21. Marginal Effects at the Average for Empathy on Helping the Poor Given Beliefs Belief Gov Action (1) 2 People take advantage 0.0723*** 0.0172*** People are fair 0.0017 0.0009 People can be trusted 0.0277** 0.0172** People cannot be trusted 0.0305** 0.0080** Success comes from luck 0.0542 0.0143 Success comes from hard work 0.0334** 0.0127** *p < 0.10 **p < 0.05 ***p < 0.01
Agree w/ both (3) 4 -0.0153*** -0.0267*** -0.0001 -0.0014 -0.0086* -0.0272** -0.0070* -0.0123** -0.0236 -0.0181 -0.0017 -0.0191**
People help themselves (5) -0.0474*** -0.0013 -0.0263** -0.0192** -0.0269 -0.0253**
Results for Local Issues Table 22 displays the results from the probit regression for the local justice variables. Each column shows the results of a separate regression for that dependent variable. These dependent variables differ from the national variables in that they are looking at actions instead of opinions. They also deal with direct transfers from one person to another person or group. An interesting observation is that the influence of education is negative for most of the national issues but for local issues is positive. People who have more education tend to be less likely to support redistribution at the national level, but seem more likely to give at the local level. One reason this could be is that education can potentially change the weight people place on justice principles.
86 People who invest more time in their education might put more weight on the justice principle of merit. That is, they might feel that they deserve to benefit from their investment in education. At the local level they are choosing how to spend their own money. The choice to give to charity is different because they are just making a decision about how to spend the money they have earned.
87 Table 22. Overall Probit Results: Coefficient Estimates for Local Issues Dependent Variable (1) (2) Giving to Homeless Giving to Charity Independent variables Empathy
-0.234*** -0.160*** (0.037) (0.041) Educ -0.022** -0.085*** (0.011) (0.011) Age -0.004*** -0.014*** (0.001) (0.002) Female -0.0264 -0.211*** (0.055) (0.066) Black -0.477*** -0.160* (0.083) (0.084) Other Race -0.545*** -0.221** (0.138) (0.103) 2004 -0.008 -0.073 (0.061) (0.062) Extremely Liberal -0.492*** -0.036 (0.168) (0.170) Liberal -0.319*** -0.023 (0.098) (0.107) Slightly Liberal -0.0241 -0.195* (0.082) (0.113) Slightly Conservative -0.120 -0.179 -(0.081) (0.117) Conservative 0.087 -0.205* -(0.083) (0.105) Extremely Conservative -0.0132 -0.173 (0.147) (0.188) Log real income -0.0954*** -0.398*** (0.028) (0.035) _cons -0.864*** -4.92*** (0.270) (0.358) N -2320 -2321 pseudo R-sq -0.047 -0.18 Bootstrapped standard errors in parentheses; *p < 0.10 **p < 0.05 ***p < 0.01. Note: White, 2002, and Moderate are the base variables
(3) Volunteer -0.181*** (0.033) -0.078*** (0.011) -0.005*** (0.002) -0.135** (0.054) -0.029 (0.090) -0.053 (0.134) -0.062 (0.054) -0.107 (0.182) -0.023 (0.098) -0.228*** (0.078) -0.264*** (0.085) -0.126 (0.086) -0.226 (0.144) -0.121*** (0.030) -2.299*** (0.329) -2322 -0.061
88 Giving to the Homeless Column (1) of Table 22 lists the results for Giving to Homeless, and Table 23 lists the marginal effects and predicted probabilities. The coefficient and marginal effects for empathy are significant at the 1% level. An increase in one standard deviation in empathy leads to a 8.1 percentage point increase in the probability that someone gave food or money to a homeless person at least once in the past year. The overall predicted probability is 0.6573. Unfortunately, this result does not detail how often people give to the homeless in a given year or if people who did not give would have if they had the opportunity. Nevertheless, this suggests that higher degrees of empathetic concern can potentially influence a persons decision as to whether or not to help someone who is homeless. A study by Goldberg looked at altruism towards panhandlers.131 The author argues that decisions to give to panhandlers cannot be accounted for by kin selection or reciprocal altruism.132 While the study found support for reputation, people still gave when no one was around.133 Combined with my results, it is likely that part of the reason people give to the homeless is because they care about the welfare of others. Education is significant at the 5% level. An increase of one year in education increases the probability that the person will give to the homeless by approximately 0.9 percentage points. The coefficients for race are both significant at the 1% level. People who 131
Tony Goldberg, “Altruism Towards Panhandlers: Who Gives?” Human Nature 6, no. 1 (March 1995): 79-89. 132
Ibid.
133
Ibid.
89 identified themselves as black or other race are approximately 16 and 17.6 percentage points more likely to give to the homeless at least once in the past year compared to whites.
For political views, only Extremely Liberal and Liberal are statistically
significant, both at the 1% level. People who are Extremely Liberal have about a 16 percentage point increase in the probability that they gave to the homeless compared to Moderates. Income is significant at the 1% level. A 100% increase in real income leads to a 3.5 percentage point increase in the probability that a person gave to the homeless in the past year.
Table 23. Marginal Effects at the Average for Probit Results Dependent Variable Give to Homeless Give to Charity Volunteer Empathy 0.0806*** 0.0394*** 0.0723*** Educ 0.0088*** 0.021*** 0.0311*** Age - 0.0016** 0.0033*** - 0.0020*** Female^ 0.0097 0.0519*** 0.0536*** Black^ 0.1596*** - 0.0394* 0.0115 Other Race^ 0.1758*** - 0.0542* - 0.0212 2004^ 0.0031 - 0.018 0.0249 Extremely Liberal^ 0.1595*** - 0.0087 0.0426 Liberal^ 0.1100*** 0.0056 - 0.0093 Slightly Liberal^ 0.0088 0.0479* 0.091*** Slightly Conservative^ 0.0434 0.0439* 0.1051*** Conservative^ 0.0314 0.0502* 0.0500** Extremely Conservative^ 0.0049 0.0425 0.0902 Log Real Income 0.0350*** 0.0977*** 0.0483*** Predicted Probabilities at variable means Overall 0.6573 0.8377 0.4807 Empathy at 0.5a 0.6995** 0.8568** 0.5172** Empathy at -0.5a 0.6137** 0.8175** 0.4450** Male 0.6522 0.8093 0.4529 Femaleb 0.6619 0.8614** 0.5064** ^dy/dx is for discrete change of dummy variable from 0 to 1; * p < 0.10 ** p < 0.05 *** p < 0.01. a statistically different compared to the Overall category b statistically different compared to the Male category
90 Giving to Charity Column (2) in Table 22 exhibits the results for Giving to Charity, and Table 23 gives the marginal effects and predicted probabilities. The coefficient and marginal effects for empathy are significant at the 1% level. An increase in one standard deviation in empathy leads to approximately a 3.9 percentage point increase in the probability that a person donated money to charity at least once in the past year. The overall predicted probability is 0.8377. Charity has long been linked to ideas of sympathy and is of no surprise that empathetic concern might make a person more likely to give.134 Education is significant at the 1% level. An increase of one year in education increases the probability that a person gave to charity by approximately 2.1 percentage points. The coefficient for age is significant at the 1% level. An increase of one year in age leads to a 0.33 percentage point increase in the probability that a person gave to charity. An intriguing result comes from the coefficient for female which is significant at the 1% level. Being a female increases the probability that a person will give to charity at least once in the past year by about 5.2 percentage points relative to men. The predicted probabilities for female are statistically different from the predicted probabilities for male at the 5% level. Research on gender differences in preferences for distributive justice suggests that women tend to weigh justice principles differently. 135 Compared to men, women seem to
134
Adam Smith, The Theory of Moral Sentiments (London: A. Millar, 1790), http://www.econlib.org/library/Smith/smMS.html (accessed April 1, 2011). 135
John T. Scott et. al., “Just Deserts; An Experimental Study of Distributive Justice Norms,” American Journal of Political Science 45, no. 4 (October 2001): 749-67.
91 put more emphasis on the principles of need and equality. 136 This could potentially explain why women are more likely to give. The variables for Black and Other Race are significant at the 10% and 5% level respectively. Here the marginal effects are in the opposite direction then they were for giving to the homeless. For example, a person who identifies themselves as Other Race has a 5.4 percentage point decrease in the probability that they donated to charity at least once in the past year. Conservative is significant at the 10% level and people who fall into that category are 5 percentage points more likely to have given to charity at least once in the past year relative to Moderates. A similar result is found for people who are Slightly Liberal. No other political views are statistically significant. Income is significant at the 1% level. An increase of 100% in real income increases the probability that someone will donate to charity by about 9.8 percentage points. Part of the reason income increases the probability of donating to charity is likely due to the incentives created from tax deductions. Although this is only likely to influence the decisions of people with high incomes. Volunteer Column (3) in Table 22 displays the results for Volunteer, and the marginal effects and predicted probabilities are listed in Table 23. The coefficient and marginal effects for empathy are significant at the 1% level. An increase of one standard deviation in empathy increases the probability by 7.23 percentage points that someone volunteered for charity at least once in the past year. The overall predicted probability is 0.4807. Of course, this does not give any indication of how often people volunteer or what they 136
Ibid.
92 volunteer for, but at least suggests that increases in empathetic concern might make a person more likely to volunteer. Most volunteering happens in a person's community or surrounding areas so it makes sense that empathetic concern might influence those choices. Education is significant at the 1% level. An increase of one year of education suggests a 3.1 percentage point increase in the probability that an individual will volunteer. Age is significant at the 1 percent level. An increase in age of one year decreases the probability that someone volunteered by 0.2 percentage points. One reason this might be the case is due to health. As people age, often their health deteriorates, which limits the activities they can engage in. The coefficient for female is significant at the 5% level. Being female increases the probability that a person will volunteer by 5.4 percentage points relative to being male. The race variables are not significant. Both Slightly Liberal and Slightly Conservative are significant at the 1% level. Relative to Moderates, identifying oneself as Slightly Liberal or Slightly Conservative increases the probability that a person volunteered by 9.1 and 10.5 percentage points respectively. Income is also significant at the 1% level. A 100% increase in real income increases the probability that someone volunteered for charity at least once in the past year by approximately 4.8 percentage points. Results for Beliefs General beliefs about human nature might potentially influence the degree to which empathetic concern influences distributive justice at the local level. However, since these beliefs are general they might not be as influential at the local level. This is
93 because deciding whether to give money to a homeless person, for instance, is often done on a case to case basis and is likely a highly context dependent situation. Fairness Table 24 gives the results for the three local variables given beliefs about fairness. Columns 1A and 1B show the results of empathy on giving to the homeless at least once in the year. Both coefficients are significant at the 1% level. The coefficient for empathy is larger for people who believe others are fair, then for people who believe others take advantage. However, the null, that these estimates are the same, cannot be rejected. Table 27 shows that for people who believe others are fair an increase of one standard deviation in empathy increases the probability that a person gave to the homeless at least once in the past year by 7.3 percentage points. Although smaller in magnitude, the marginal effects are similar for the group that believes people take advantage of others. Columns 2A and 2B shows the results for giving to charity. The coefficients for empathy for the take advantage and fair groups are statistically significant at the 5% and 1% level respectively. The null, that the two estimates are similar, cannot be rejected. It seems that general beliefs about the fairness of others does not effect the influence of empathetic concern on whether someone gave to charity in the past year. Columns 3A and 3B display the results for volunteering. For the take advantage group, empathy is significant at the 10% level, but for the fair group empathy is significant at the 1% level. The null, that the coefficients are the same, cannot be rejected.
94 Overall, at the local level general beliefs about fairness do not seem to change the influence of empathetic concern. This is likely because often these decisions to help are with people that are closer in social distance. Giving to the homeless and volunteering are likely to take place locally where general beliefs are likely to hold less weight. People at the local level also probably have more information when making their decision. They might know about what a charity does and who it provides aid to. Knowing specific details might make people less likely to fall back on general beliefs about how people act because they have more relevant information to use in their decisions.
Table 24. Probit Results: Coefficient Estimates Given Beliefs about Fairness Dependent Variable 1A 1B 2A 2B 3A 3C Giving to Giving to Giving to Giving to Volunteer Volunteer Homeless Homeless Charity Charity Belief (takeadv) (fair) (takeadv) (fair) (takeadv) (fair) empathy 0.180** 0.243*** 0.151** 0.215*** 0.117* 0.177*** (0.070) (0.063) (0.074) (0.073) (0.067) (0.062) chi2(1) 0.45 0.38 0.43 P>chi2 0.503 0.538 0.511 N 587 799 587 800 587 801 pseudo 0.084 0.059 0.166 0.2 0.065 0.064 R-sq Robust standard errors are in parentheses ; *p < 0.10 **p < 0.05 ***p < 0.01. Note: Control variables include educ, age, sex, race, year, political views, and income. The variables People Fair and Take Advantage are abbreviated by fair and takeadv, respectively. Columns A and B report results from seemingly unrelated estimation. The coefficients for empathy in Columns A and B are tested under the null hypothesis that they are the same using the Wald test.
95 Luck Table 25 gives the results for the three local variables given beliefs about success. Columns 1A and 1B show the regressions for giving to the homeless. For both groups, the coefficients on empathy are significant at the 1% level. Table 27 shows that, for people who believe success comes from luck, an increase of one standard deviation are 15.95 percentage points more likely to have given to charity at least once in the past year. Similarly, for people who believe success comes from hard work an increase in one standard deviation increases the probability that a person gave to charity by approximately 10.2 percentage points. However, the null, that the coefficients for the two regressions are the same, cannot be rejected. Columns 2A and 2B give the results for giving to charity. The coefficient for empathy is significant only for the hard work group. The null, that the coefficients for both groups are the same, cannot be rejected. Columns 3A and 3B show the results for volunteering. Only the hard work group is significant. The null, that the two empathy estimates for both groups are the same, cannot be rejected. Overall, beliefs about the determinants of success do not seem to have any effect on the degree to which empathy effects the local justice variables. These results support the idea that general beliefs about luck or hard work do not seem to modulate the influence of empathetic concern at the local level.
96 Table 25. Probit Results: Coefficient Estimates Given Beliefs about Luck and Hard Work Dependent Variable 1A 1B 2A 2B 3A 3C Giving to Giving to Giving to Giving to Volunteer Volunteer Homeless Homeless Charity Charity Belief (luck) (hard work) (luck) (hard work) (luck) (hard work) empathy 0.439*** 0.276*** -0.027 0.141** 0.105 0.185*** (0.148) (0.057) (0.125) (0.060) (0.131) (0.054) chi2(1) 1.06 1.47 0.31 P>chi2 0.303 0.225 0.575 N 148 1010 148 1009 148 1010 pseudo 0.142 0.058 0.171 0.191 0.138 0.054 R-sq Robust standard errors are in parentheses; * p < 0.10 **p < 0.05 ***p < 0.01. Note: Control variables include educ, age, sex, race, year, political views, and income. Columns A and B report results from seemingly unrelated estimation. The coefficients for empathy in Columns A and B are tested under the null hypothesis that they are the same using the Wald test.
Trust Table 26 displays the regressions for the three local variables given beliefs about the trustworthiness of others. Column 1A and 1B show the results for giving to the homeless. Both estimates for empathy are significant at the 1% level, but the null, that the two coefficients are the same, cannot be rejected. Table 27 show that an increase of one standard deviation in empathy increases the probability that a person gave to the homeless by approximately 12.23 percentage points for people that believe people can be trusted. For people that believe others are untrustworthy an increase in one standard deviation increases the probability that person will give to the homeless by approximately 6.64 percentage points.
97 Columns 2A and 2B display the results for giving to charity. Both estimates for empathy are significant at the 5% level but the null, that the coefficients are the same, cannot be rejected. Similar results are seen for columns 3A and 3B. Overall, trust does not seem to be a factor in changing the degree to which empathetic concern influences the local variables. This is probably due to the fact that at the local level general beliefs are used less often. People might fall back on stereotypes or general beliefs when they lack information but at the local level that information is more readily available.
Table 26. Probit Results: Coefficient Estimates Given Beliefs about Trust Dependent Variable 1A 1B 2A 2B 3A 3C Giving to Giving to Giving to Giving to Volunteer Volunteer Homeless Homeless Charity Charity Belief (trust) (notrust) (trust) (notrust) (trust) (notrust) empathy 0.333*** 0.190*** 0.191** 0.152** 0.181** 0.171*** (0.081) (0.057) (0.095) (0.061) (0.077) (0.057) chi2(1) 2.12 0.12 0.01 P>chi2 0.145 0.734 0.918 N 565 872 565 871 565 872 pseudo 0.081 0.058 0.207 0.162 0.075 0.062 R-sq Robust standard errors are in parentheses ; * p < 0.10 **p < 0.05 ***p < 0.01. Note: Control variables include educ, age, sex, race, year, political views, and income. Columns A and B report results from seemingly unrelated estimation. The coefficients for empathy in Columns A and B are tested under the null hypothesis that they are the same using the Wald test.
98 Table 27. Marginal Effects at the Average for Empathy on Giving to Homeless, Giving to Charity, and Volunteer Given Beliefs Belief People take advantage People are fair People can be trusted People cannot be trusted Success comes from luck Success comes from hard work *p < 0.10 **p < 0.05 ***p < 0.01
Dependent Variable Give to Homeless Give to Charity 0.0613*** 0.0488** 0.0906*** 0.0351*** 0.1223*** 0.0293** 0.0664*** 0.0454*** 0.1595*** - 0.0062 0.1020*** 0.0360***
Volunteer 0.0459*** 0.0703*** 0.0715*** 0.0678*** 0.0413 0.0734***
CHAPTER 6 CONCLUSION The results in this paper seem to support the idea that people with higher empathetic concern tend to favor more redistribution and are more likely to help others. In regards to the national variables, it seems that the specific issue matters. Empathetic concern significantly influences preferences for redistribution of wealth, Social Security, and helping the poor, but does not seem to influence opinions about welfare or foreign aid. For the local issues, empathetic concern increases the probability that people will help others in all three situations. The results are significant for giving to the homeless, giving to charity, and volunteering. Overall, I fail to find support for the hypothesis that general beliefs change the degree in which empathetic concern influences preferences for distributive justice. At the national level, there is some evidence that beliefs at how fair others are can impact the influence of empathy on the distribution of wealth and helping the poor. At the local level, general beliefs do not seem to have much of an effect. One reason why these beliefs might not have had much of an influence is that they are too general. As discussed earlier in this paper, people have pluralistic views of fairness. It is unlikely that a general belief can really capture people's views about human nature. It is possible that if I had data on more specific beliefs, then the results would improve.
99
100 The advantage of this study is that the survey was nationally representative, and dealt with actual distributive justice policies. The reason this is important is that often distributive justice researchers uses vignettes. These often give people hypothetical scenarios with complete information. While these are useful in isolating specific principles of fairness it is unclear if the results hold in the real world. For example, in the study discussed earlier by Konow people were told about identical twins who were the same in "terms of physical and mental abilities," but one works harder than the other.137 Over 99% of the people thought it was unfair that they received the same pay.138 It is unclear if that situation ever occurs in the actual world. Instead people often have incomplete and biased information on how hard others work. Looking at actual distributive justice principles is important in order to see what influences these preferences. Of course, the results from this paper rely on the validity of the survey data. While not definitive, this study adds evidence to the idea that people care about the welfare of others and make decisions taking others into account. An interesting implication of this study from a policy perspective is the wording or framing of a justice issue. Empathetic concern seems to influence preferences in regards to helping the poor, but not for welfare. How an issue is presented might determine whether empathetic concern is activated or not. Jerit and Barabas conducted an experiment regarding Social
137
James Konow, “Which Is the Fairest One of All? A Positive Analysis of Justice Theories,” 1188-39.
138
Ibid.
101 Security.139 In this experiment, participants watched a debate about Social Security and were subsequently surveyed.140 Some participants heard many misleading statements about Social Security while others did not.141 The authors found that hearing misleading statements lead more people to come to incorrect conclusions about Social Security. 142 Combined with my results, this implies that an issue can potentially be framed in a way that will activate empathetic concern and lead to increases in support for that issue. There are two main reasons why this study is important. First, if people value the welfare of others, then often their behavior will differ from models that assume people only care about their own payoff. While the results from this paper are only suggestive, it seems that empathetic concern has the potential to motivate altruistic behavior. Second, distributive justice matters to many people. Why do some people believe that the government should redistribute income and others do not? The choices people make can have a profound affect on the economy and understanding what influences these preferences is important. By looking at empathetic concern this paper adds to the growing literature attempting to understand issues of distributive justice. It is unclear if these reported attitudes will match actual behavior. People might say that they want to increase aid to the poor, but subsequently not vote for a state proposition that would do so. Future studies should see if stated preferences match with 139
Jennifer Jerit and Jason Barabas, “Bankrupt Rhetoric How Misleading Information Affects Knowledge About Social Security,” Public Opinion Quarterly 70, no. 3 (March 2010): 278-303. 140
Ibid.
141
Ibid.
142
Ibid.
102 revealed preferences. In addition, people often have misconceptions about how actual programs work. That is, individual beliefs do not always fit with reality. One interesting avenue to explore would be to see if preferences change when people learn more about how the programs actually work. Will preferences in regards to welfare change if people know the actual degree of fraud present in the program? Or will people's beliefs stay the same? In addition, future research should look at whether these results hold outside of the United States, and explore these issues using different empirical approaches.
REFERENCES Alesina, Alberto, and Eliana La Ferrara. “Preferences For Redistribution in the Land of Opportunities.” Journal of Public Economics 89, no. 5 (June 2005): 897-931. Batson, Daniel, and Nadia Ahmad. “Empathy-induced altruism in a prisoner's dilemma II: what if the target of empathy has defected?” European Journal of Social Psychology 31, no. 1 (January 2001): 25-36. Batson, Daniel, "These Things Called Empathy: Eight Related but Distinct Phenomenon." In The Social Neuroscience of Empathy. Edited by Jean Decety and William Ickes. Cambridge: MIT Press, 2011. Batson, Daniel, and Jakob Håkansson Eklund. “An Additional Antecedent of Empathic Concern: Valuing the Welfare of the Person in Need.” Journal of Personality and Social Psychology 93, no. 1 (July 2007): 65-74. Batson, Daniel, Judy G. Batson, Mathew Todd, Beverly H. Brummett, Laura L. Shaw, and Carlo M. R. Aldeguer. “Empathy and the Collective Good: Caring For One of the Others in a Social Dilemma.” Journal of Personality and Social Psychology 68, no. 4 (April 1995): 619-31. Camerer, Colin F. Behavioral Game Theory: Experiments in Strategic Interaction (Roundtable Series in Behaviorial Economics). New York: Princeton University Press, 2003. Chong, Alberto, and Mark Gradstein. “Who's Afraid of Foreign Aid? The Donors' Perspective.” CESifo Working Paper Series No. 1833 (October 2006): http://ssrn.com/abstract=944422. Cosmides, Leda, John Tooby, and Jerome Barkow. The Adapted Mind: Evolutionary Psychology and the Generation of Culture. New York: Oxford University Press, 1995. Davis, Mark H. “Measuring Individual Differences in Empathy: Evidence For a Multidimensional Approach.” Journal of Personality and Social Psychology 44, no. 1 (January 1983): 113-26. de Waal, Frans B.M. “Putting the Altruism Back Into Altruism: The Evolution of Empathy.” Annual Review of Psychology 59 (January 2008): 279-300. 103
104 Diamond, Peter A., and Jerry A. Hausman. “Contingent Valuation: Is Some Number Better Than No Number.” Journal of Economic Perspectives 8, no. 4 (Autumn 1994): 45-64. Edgeworth, Francis Ysidro. Mathematical Psychics: An Essay On the Application of Mathematics to the Moral Sciences. London: C. Kegan Paul, 1881. Fehr, Ernst, Helen Bernhard, and Bettina Rochenbach. “Egalitarianism in Young Children.” Nature 454 (August 2008): 1079-83. Gaertner, Wulf. A Primer in Social Choice Theory: Revised Edition (Lse Perspectives in Economic Analysis). Rev. ed. London: Oxford University Press, 2009. Charness, Gary, and Uri Gneezy. “What's in a Name? Anonymity and Social Distance in Dictator and Ultimatum Games.” Journal of Economic Behavior and Organization 68, no. 1 (October 2008): 29-35 George, Henry. Progress and Poverty. Garden City, NY: Doubleday, Page & Co. 1912. http://www.econlib.org/library/YPDBooks/George/grgPP27.html (accessed April 1, 2011). Gintis, Herbert, Joseph Henrich, Samuel Bowles, Robert Boyd, and Ernst Fehr. “Strong Reciprocity and the Roots of Human Morality.” Social Justice Research 21, no. 2 (June 2008): 241-53. Goldberg, Tony. “Altruism Towards Panhandlers: Who Gives?” Human Nature 6, no. 1 (March 1995): 79-89. Hamilton, W.D. “The Genetical Evolution of Social Behaviour.” Journal of Theoretical Biology 7, no. 1 (July 1964): 1-16. Harris, Richard J. A Primer of Multivariate Statistics. 3rd ed. Mahwah, New Jersey: Psychology Press, 2001 Kennedy, Peter. A Guide to Econometrics. 6th ed. Malden, MA.: Wiley-Blackwell, 2008. Kleinjans, Kristin J., Maria K. Humlum, and Helena S. Nielsen. “An Economic Analysis of Identity and Career Choice.” Economic Inquiry (forthcoming). Konow, James. “Which Is the Fairest One of All? A Positive Analysis of Justice Theories.” Journal of Economic Literature 41, no. 4 (December 2003): 1188-39.
105 Kruger, Daniel J. “Evolution and Altruism: Combining Psychological Mediators with Naturally Selected Tendencies.” Evolution and Human Behavior 24, no. 2 (March 2003): 118-25. Jerit, Jennifer, and Jason Barabas. “Bankrupt Rhetoric How Misleading Information Affects Knowledge About Social Security.” Public Opinion Quarterly 70, no. 3 (March 2010): 278-303. Liebman, Jeffrey B. “Redistribution in the Current U.S. Social Security System.” NBER Working Paper Series w8625 (December 2001): http://ssrn.com/abstract=293238. Luttmer, Erzo F. P. “Group Loyalty and the Taste for Redistribution.” Journal of Political Economy 109, no. 3 (June 2001): 500-28. Mayr, Ulrich, William T. Harbaugh, and Dharol Tankersley. "Neuroeconomics of Charitable Giving and Philanthropy." In Neuroeconomics: Decision Making and the Brain. Edited by Paul W. Glimcher, Colin Camerer, Russell Alan Poldrack and Ernst Fehr. London: Academic Press, 2009. McElreath, Richard, and Robert Boyd. Mathematical Models of Social Evolution: A Guide for the Perplexed. Chicago: University Of Chicago Press, 2007. Mobius, Markus, Quoc-Anh Do, Stephen Leider, and Rosenblat Tanya. “Directed Altruism and Enforced Reciprocity in Social Networks.” Quarterly Journal of Economics 124, no. 4 (November 2009): 1815-51. Mueller, Dennis C. Public Choice III. 3rd ed. New York: Cambridge University Press, 2003. Nicholson, Walter, and Christopher Snyder. Microeconomic Theory Basic Principles and Extensions. 10th ed. Canada: South-Western College Pub., 2007. Nowak, Martin. “Five Rules For the Evolution of Cooperation.” Science 314, no. 5805 (December 2006): 1560-63. Preston, Stephanie D., and Frans B. M. de Waal. “Empathy: Its Ultimate and Proximate Bases.” Behavioral and Brain Sciences 25, no. 1 (January 2003): 1-20. Rabin, Matthew. “Incorporating Fairness Into Game Theory and Economics.” American Economic Review 83, no. 5 (December 1993): 1291-2. Rencher, Alvin C. Methods of Multivariate Analysis. 2nd ed. New York: WileyInterscience, 2002.
106 Scott, John T., Richard E. Matland, Philip A. Michelbach, and Brian H. Bornstein. “Just Deserts; An Experimental Study of Distributive Justice Norms.” American Journal of Political Science 45, no. 4 (October 2001): 749-67. Singer, Tania, and Claus Lamm. “The Social Neuroscience of Empathy.” Annals of the New York Academy of Sciences 1156 (March 2009): 81-96. Singer, Tania. "Understanding Others: Brain Mechanisms of Theory of Mind and Empathy." In Neuroeconomics: Decision Making and the Brain. Edited by Paul W. Glimcher, Colin Camerer, Russell Alan Poldrack and Ernst Fehr. London: Academic Press, 2009. Singer, Tania, and Nikolaus Steinbeis. “Differential Roles of Fairness- and CompassionBased Motivations For Cooperation, Defection, and Punishment.” Annals of the New York Academy of Sciences 1167 (June 2009): 41-50. Smith, Adam. The Theory of Moral Sentiments. London: A. Millar, 1790. http://www.econlib.org/library/Smith/smMS.html (accessed April 1, 2011). Smith, Tom W., Peter Marsden, Michael Hout, and Jibum Kim. “General Social Surveys, 1972-2010 (Data File and Codebook).” National Opinion Research Center: The Roper Center for Public Opinion (2011) Sonnby–Borgström, Marianne. “Automatic Mimicry Reactions as Related to Differences in Emotional Empathy.” Scandinavian Journal of Psychology 43, no. 45 (December 2002): 433-43. Warneken, Felix, and Michael Tomasello. “Varieties of Altruism in Children and Chimpanzees.” Trends in Cognitive Sciences (September 2009): 397-402. Wilhelm, Mark Ottoni. “Helping Behavior, Dispositional Empathic Concern, and the Principle of Care.” Social Psychology Quarterly 73, no. 1 (March 2010): 11-32. Winkelmann, Rainer, and Stefan Boes. Analysis of Microdata. New York: Springer, 2006.
