Why Do People Give? An Experimental Test of Pure and Impure Altruism - - PowerPoint PPT Presentation

Why Do People Give? An Experimental Test of Pure and Impure Altruism

Lise Vesterlund ∗ Mark Ottoni-Wilhelm † Huan Xie ‡

∗University of Pittsburgh †IUPUI and IU Lilly Family School of Philanthropy ‡Concordia University

Workshop in Multidisciplinary Philanthropic Studies

March 13, 2014

Overview

• Standard economic model of charitable giving is impure

altruism.

• Dual-motive theory: altruism and egoism/warm glow
• Cornes & Sandler (1984), Steinberg (1987), Andreoni (1989)
• Batson, Darley, & Coke (1978): “Motivation for helping may

be a mixture of altruistic and egoistic desires.”

• Lots of work over the past 25 years has been built on this

model.

• Andreoni (2006), Vesterlund (2010)
• Twenty-five Years of Impure Altruism, UCSD

Overview

• However, there has not been a test of a prediction of the

impure altruism model that was not purposefully built into the model to begin with.

• Nor a test that, for all practical purposes, the model could fail.
• In this paper we conduct a charitable giving experiment that

tests a prediction not purposefully built into the model.

• Results consistent with impure altruism – the model passes the

test, but ...

• ... egoism/warm glow motives among the participants were

weak.

• Most of the giving in the experiment was due to altruism.

Overview

• Innovations
• New test that is closely integrated with theory.
• New experimental design that carefully creates the theoretical

framework inside the lab.

• Measure individual heterogeneity in altruism and egoism/warm

glow.

• Significance
• Fundamental question about human behavior – existence of

altruism

• The strong altruistic motivation we see in the experiment

differs from previous experimental results that have suggested egoism/warm glow to be the predominant motive people give.

Theory - a giving situation

• You are each paired with a different child (between 1 and 12

years old).

• The child’s family home has been destroyed by fire.
• You will be given money which you can allocate between the

child and yourself.

• The money you allocate to the child will be given to the Red

Cross.

• The Red Cross will use the money to buy books for the child,

and will give the books to the child immediately after a fire.

Theory - a giving situation

• The Red Cross provides the family with temporary shelter,

clothing, a meal, a ”comfort bag.”

• The Red Cross has no funds for any comfort items for the

children.

• In addition to the amount of your money you allocate for the

child’s books, a foundation has agreed to donate a fixed amount of money towards the child.

• The foundation’s fixed donation is independent of your

allocation.

• Hence, the total amount the Red Cross will be able to spend
• n books for the the child is the foundation’s fixed donation

plus the allocation you make to the child.

Theory - a giving situation

• Why use the funds to buy books?
• The books provide a great bridge for Red Cross volunteers to

connect with kids and get them talking about what they’ve experienced.

Theory - a giving situation

• What will you decide to do?
• The foundation will donate $4 towards the child.
• You have $46 to allocate between the child and you.
• How much will you allocate to the child?
• Deciding . . .

Experiment

.1 .2 .3 .4 Fraction 10 20 30 40 46 Dollars allocated to the child

Source: Vesterlund, Ottoni-Wilhelm, Xie (2014).

Figure 1. Foundation = $4, Your income = $46.

Reflection

Think for a moment: Why did you pick the amount you allocated to the child? What motivated you?

Theory - Why did you pick the amount you allocated to the child?

• Egoism/warm glow motivations – intensive focus on . . .
• + . . . seek positive self-evaluation of identity (moral principles)
• – . . . avoid self-punishment for not helping (guilt)
• + . . . seek approbation (extrinsic)
• – . . . avoid shame (extrinsic)
• 3 more intrinsic:
• personal distress - alleviate the distress you feel hearing about

the child and the fire.

• seek “empathic joy”
• seek relief of negative mood state.

Theory - Why did you pick the amount you allocated to the child?

• Altruistic motivations – intensive focus on the amount of

books needed to accomplish the goal.

• What amount of books are needed to allow the volunteers to ...
• ... connect with the child and get her/him talking about what

they’ve experienced?

• ... initiate a process through which the child will begin to

recover a sense of well-being?

Theory - a DIFFERENT giving situation

• What will you decide to do?
• The foundation will donate $10 towards the child.
• You have $40 to allocate between the child and you.
• How much will you allocate to the child?
• Deciding . . .

Experiment: situations 1 and 2

.1 .2 .3 .4 Fraction 10 20 30 40 46 Dollars allocated to a child Fdtn = $10, You = $40 Fdtn = $4, You = $46

Source: Vesterlund, Ottoni-Wilhelm, Xie (2014).

Figure 3. Two decisions.

Reflection again

Did you change the amount you allocated to the child? Why?

Theory - Pure Altruism

• If your goal was to get $15 to the Red Cross because that is

enough books to help the volunteers connect with the child, then the two budgets you were just faced with—($4, $46) and ($10, $40)—are equivalent.

• Faced with ( $4, $46), you give $11 (books = $4 + your $11).
• Faced with ($10, $40), you give $5 (books = $10 + your $5).
• The total amount of money in play was the same

($50)...“balanced-budget.”

• When the foundation went from $4 → $10 (+ $6), you went

from $11 → $5 (− $6).

• Crowd-out (balanced-budget) = −$6

+$6 = −1

Theory - Pure Altruism - Income effect

• What if we held the foundation’s donation fixed at $4 and

increase your income from $40 to $46?

• Faced with ( $4, $40), say you give $9 (books = $4 + your $9

= $13).

• Faced with ( $4, $46), you give $11 as before (books = $4 +

your $11 = $15).

• Income effect = $15−$13

$46−$40 = +$2 +$6 = +.33 (from perspective of

books)

• Income effect =

$11−$9 $46−$40 = +$2 +$6 = +.33 (from perspective of

your contribution)

Theory - Pure Altruism - Unfunded crowd-out

• What if when the foundation’s donation went up from $4 →

$10 we held your income fixed at $40?

(the foundation’s increase is not funded from your income...“unfunded”)

• Faced with ( $4, $40), you give $9 as before (books = $4 +

your $9 = $13).

• Faced with ($10, $40), you give $5 as before (books = $10 +

your $5 = $15).

• The total amount of money in play has gone up from $44 →

$50.

• When the foundation went from $4 → $10 (+ $6), you went

from $9 → $5 (− $4).

• Income effect = $15−$13

$50−$44 = +$2 +$6 = +.33 (from perspective of

books)

• Crowd-out (unfunded) =

$5−$9 $10−$4 = −$4 +$6 = −.67 (from

perspective of your contribution)

• Crowd-out (unfunded) = −1 + Income effect = −1 + .33

Theory - Pure Altruism - In-kind transfers

• Standard intermediate microeconomics course – food stamps

example.

• Balanced-budget
• Person gets $4 per week in food stamps and has $46 in cash

income, say they spend %15 week on food ($4 food stamps + $11 cash) and $35 on other goods (cash).

• If $10 per week in food stamps and has $40 in cash income,

would still spend %15 week on food ($10 food stamps + $5 cash) and $35 on other goods (cash).

• Unfunded
• Person gets $4 per week in food stamps and has $40 in cash

income, they spend $13 on food ($4 food stamps + $9) and $31 on other goods.

• If $10 per week in food stamps and $40 in cash income, would

spend %15 week on food ($10 food stamps + $5 cash) and $35 on other goods (cash).

• Some of the $6 increase in food stamps was spent on food

(.33 – the income effect on food) and the rest was spend on

• ther goods (.67 – the income effect on other goods).

Theory - Pure Altruism

• Increasing cash income and increasing amount of in-kind

(food stamps, books–charity) is the same.

• Income effect on books–charity is called q1
• In the previous example q1 = .33

Theory - EGOISM/Warm Glow - Unfunded crowd-out

• If you said “You know what, what happened to that kid is

terrible, and it’s my duty to help out. I’ll give $20 out of my $40) income.”

• Faced with ( $4, $40), you give $20 as before (books = $4 +

your $20 = $24).

• Faced with ($10, $40)—“still my duty to give $20”—(books =

$10 + $20 = $30).

• The total amount of money in play has gone up from $44 →

$50.

• When the foundation went from $4 → $10 (+ $6), you

remained at $20.

• Income effect = $30−$24

$50−$44 = +$6 +$6 = +1.00 (from perspective of

books)

• Crowd-out (unfunded) = $20−$20

$10−$4 = $0 +$6 = 0.00 (from

perspective of your contribution)

• Crowd-out (unfunded) = −1 + (Income effect = +1) = 0.00

Theory - Egoism/Warm Glow - Income effects (plural)

• You mean to tell me that if that person’s own income went

from $40 → $46 they would give ALL of it to buy more books?

• Not likely. Say they would give an additional $3.60
• Her/his q1 = $3.60

$6

= .60

• Then when the foundation increases its donation of books

from $4 → $10, it is as if the person has a second, “extra” income effect = .40 so that she/he spends all the $6 additional income on books, and none of it on other goods.

• This second “extra” income effect is called q2.
• q2 is evoked only by the foundation’s $6 increase in donation.
• Pure Egoism/Warm glow ⇔ q1 + q2 = 1, and of course

q2 > 0.

Theory - Impure Altruism

• In between the two extremes of Pure Altruism and Pure Warm

Glow.

• Income effects
• Pure altruism : q2 = 0
• Impure altruism: q1 + q2 < 1, and q2 > 0
• Pure Warm Glow : q1 + q2 = 1, and q2 > 0
• Crowd-out (balanced-budget / unfunded)
• Pure altruism : 0 , and −1 + q1
• Impure altruism: −1 + q2, and −1 + q1 + q2
• Pure Warm Glow : −1 + q2, and 0

Background

• Pure altruism came first, in economic work
• Edgeworth (1881), Boulding (1962), Hochman & Rodgers

(1969), Becker (1974).

• No one even thinking about Pure Warm Glow
• “Utility theory will mislead us if we conclude from it that the

motivation for philanthropy is no different from that for other forms of expenditure” (Boulding 1962).

• Pure altruism theory crashes into real world evidence
• Econometric studies of unfunded crowd-out (κ) between -.05

and -.35 (Steinberg 1991, Andreoni 2006, Vesterlund 2010).

• Even if κ = -.35, q1 would have to be +.65 !!!

Background

• Impure altruism = altruism + warm glow
• Cornes & Sandler (1984), Steinberg (1987), Andreoni (1989)
• “Motivation for helping may be a mixture of altruistic and

egoistic desires; it need not be solely or even primarily altruistic to have an altruistic component” (Batson, Darley, & Coke 1978).

• Although there likely are other sources of utility from giving:

seek approbation, avoid shame, positive self-evaluation (duty), etc...

• ...“fairly obvious that neutrality will be unlikely to hold with

warm-glow giving” and will produce a model “consistent with empirical observations” (Andreoni 1989).

• In economics, warm glow was added to the Pure Altruism

model to fix the crowd-out prediction so that it aligned with the econometric evidence.

Experiments

• Previous work has tested a Pure Altruism null hypothesis
• Bolton & Katok (1998) DG-student: |κBal| = .74 (q2 = .26)–

moderate warm glow

• Eckel, Grossman, & Johnston (2005): DG-charity: |κBal| = 0

(q2 = 0) – Pure Warm Glow

• Voluntary contribution mechanism experiments
• |κBal| from .975 (q2 ≈ 0) to .672 (q2 = .328)
• Andreoni (1993), Chan, Godby, Mestelman, & Muller (2002),

Gronberg, Luccasen, Turocy, & Van Huyck (2012), Sutter & Weck-Hannemann (2004) – it depends

• Much recent work has emphasized the practical importance of

warm glow

• e.g., Eckel, Grossman, & Johnston (2005), Crumpler &

Grossman (2008), DellaVigna, List, & Malmendier (2009), Tonin & Vlassopoulos (2010, 2013)

Predictions

• A balanced-budget crowd-out test can serve as a test of pure

altruism — H0 : κBal

A

= −1 ⇔ q2 = 0.

• A balanced-budget crowd-out test cannot serve as a test of

impure altruism — H0 : |κBal

Im | < 1 ⇔ q2 > 0

• Pure warm glow is also consistent with q2 > 0.
• Practically, even if the true model is pure altruism, any

confidence interval around ˆ κIm ≈ −1 will be consistent with some q2 > 0, however small.

• More fundamentally, the second income effect q2 > 0 was

intentionally built in to yield |κBal

Im | < 1 in the first place.

• An unfunded crowd-out test can serve as a test of pure warm

glow — H0 : κGlow = 0 ⇔ q1 + q2 = 1.

Limits to impure altruism

• As Giving by others (e.g., the foundation)

→ ∞ ⇒ q1 + q2 → 1 ⇔ κ → 0.

• Under fairly weak conditions on preferences.
• Ribar & Wilhelm (2002).
• New test:
• Increase Giving by others (e.g., the foundation) from a low

level to a high level and test H0 : |κG Low

−i | ≥ |κG High −i |.

• This prediction was not intentionally built into the impure

altruism model.

Limits to impure altruism: Cobb-Douglas illustration

a = .40 b = .10 q1 + q2 |k| .2 .4 .6 .8 1 q1 + q2, |k| 20 40 60 80 100 Giving by others (G-i) q1 + q2 Crowd-out

Notes: The Cobb-Douglas parameters are a = .40 and b = .10. U = a log(G) + b log(gi) + (1 - a - b) log(xi). Income is held constant at wi = $40.

Figure 4A. Cobb-Douglas q1 + q2 as giving by others (G-i) increases.

New test

• Increase giving by others from a low level to a high level and

test H0 : |κG Low

−i | ≥ |κG High −i |.

• Intuition
• When giving by others is Low, motivated by need to provide

the charitable good (books): q1 > 0 and q2 ≈ 0.

• As giving by others ↑, need for more books ↓ but warm glow

motivation ↑: q2 ↑

• Eventually, when giving by others ↑ to a high enough level,

q2 ↑ enough so that q1 + q2 ≈ 1 ⇔ κ ≈ 0.

• Any giving still being done is motivated by warm glow (at the

margin).

• This prediction was not intentionally built into the impure

altruism model.

New test - Experimental design

• Must hold need being addressed by the charity fixed as giving

by others ↑.

• Careful control over giving by others, including giving by
• thers outside the experiment intended to address the same

need.

• “Individualized charity” design.

Predictions - Five tests

Table: 1b. Five tests

Test Model being tested H0 1 Pure Altruism |κBal| ≥ 1 2 Pure Warm glow κ ≥ 0 3 Impure altruism - decreasing crowd-out |κG Low

−i | ≥ |κG High −i |

4 Impure altruism - increasing q2 qLow

2

≤ qHigh

2

5 Impure altruism - at G High

−i

qHigh

2

> 0 and qHigh

1

+ qHigh

2

< 1

Individualized charity

• Created a new public good through collaboration with the

American Red Cross in Southwestern Pennsylvania.

• In the event of a fire in SW PA, the Red Cross helps the

affected families find temporary shelter, provides them with clothing, a meal, and gives them essential toiletries.

• Prior to study no items given to the children affected by the

fire.

• Each participant in the experiment is paired with a different

child whose family home has suffered extensive fire damage, and ...

• ... can give some money to the Red Cross to use to buy books

for the child.

• The books provide a bridge for Red Cross volunteers to

connect with the child.

• Neither the Red Cross, nor other donors, provide

”bridge-books” for this child.

Instructions 1

• You will be asked to make six decisions.
• One of your six decisions will be randomly selected for

payment.

• Your payment will result from your decision, plus $5 for

showing up to the study.

• A research foundation provided the funds.

Instructions 2

• You are paired with a child (between 1 and 12 years old).
• The child’s family home has been destroyed by fire.
• In each of your six decisions you will be given money which

you can allocate between the child and yourself.

• The money you allocate to the child will be given to the Red

Cross.

• The Red Cross will use the money to buy books for the child,

and will give the books to the child immediately after a fire.

Instructions 3

• The Red Cross provides the family with temporary shelter,

clothing, a meal, a ”comfort bag.”

• Unfortunately, the Red Cross has no funds for any comfort

items for the children.

• We have joined the Red Cross to collect funds to buy books

for the affected children.

• In each of your six decisions you will be given money which

you can allocate between the child you are paired with and yourself.

Instructions 4

• In addition the foundation has agreed to donate a fixed

amount of money towards the child, independent of your allocation.

• The total amount to be spent on the child is the sum of the

foundation’s fixed donation and the allocation you make to the child.

• The amount of money that you can allocate between the child

and you, as well as the foundation’s fixed donation to the child, will vary across the six decisions.

Instructions 5

• If you choose not to allocate any funds to the child, then the

money to be spent on the child will be limited to the research foundation’s fixed donation.

• Only you have the opportunity to allocate additional funds to

the child.

• Neither the Red Cross nor any other donors provide books to

the child.

Instructions 6 — Why use the funds to buy books?

Children’s needs are often overlooked in the immediate aftermath

• f a disaster because everyone is concerned primarily with putting

the fire out, reaching safety, and finding shelter, food and clothing...just the basics of life. So many times, I’ve seen children just sitting on the curb with no one to talk to about what’s happening... for this reason I’ve found trauma recovery experts in the community to work with us to train our volunteer responders in how to address children’s needs at the scene of a disaster....... being able to give the children fun and distracting books will provide a great bridge for our volunteers to connect with kids and get them talking about what they’ve experienced.

Decisions 1-6 as seen by a participant

Test 1: Pure altruism

H0 : κBal

A

= −1 ⇔ q2 = 0

p = .255 p = .002 .2 .4 .6 .8 1 1.2 Crowd-out (Balanced-budget) Low $4-10 High $28-34

Notes: Vesterlund, Ottoni-Wilhelm, & Xie (2014) Table 2, columns 1 and 2. Tests are for complete crowd-out. Estimates not adjusted for corner decisions.

At low and high giving by others

Table 2. Balanced-budget crowd out:

Result: Pure altruism not rejected at G Low

−i , but is rejected at

G High

−i

.

Test 2: Pure warm glow

H0 : κ ≥ 0

p < .001 p < .001 .2 .4 .6 .8 1 Crowd-out (Unfunded) Low $4-10 High $28-34 Change

Notes: Vesterlund, Ottoni-Wilhelm, & Xie (2014) Table 4, columns 1--3. Tests are for zero crowd-out (bars 1 and 2), and for zero change in crowd-out (bar 3). Estimates not adjusted for corner decisions.

From a low to a high level of giving by others

Table 4a. Change in unfunded crowd out:

Result: Pure warm glow is rejected.

Test 3: Impure altruism

H0 : |κG Low

−i | ≥ |κG High −i |

p < .001 p < .001 p = .013 .2 .4 .6 .8 1 Crowd-out (Unfunded) Low $4-10 High $28-34 Change

Notes: Vesterlund, Ottoni-Wilhelm, & Xie (2014) Table 4, columns 1--3. Tests are for zero crowd-out (bars 1 and 2), and for zero change in crowd-out (bar 3). Estimates not adjusted for corner decisions.

From a low to a high level of giving by others

Table 4b. Change in unfunded crowd out:

Result: H0 : Crowd-out stayed the same (or got bigger) is rejected ⇒ Results consistent with Impure Altruism.

Test 4: Impure altruism - intuition

H0 : qLow

2

≤ qHigh

2

Table 5. Additional evidence supporting the impure altruism model. Account for corner decisions Change in unfunded crowd-out (restricted sample)a Change in balanced-budget crowd-out (full sample) Unfunded crowd-out and own income effect Giving by others Low/ High (1) Low/ High (2) High (3) Giving by others (G-i)

• 0.53
• 0.99
• 0.41 d

(0.07) (0.09) (0.07) Giving by others interacted 0.12 b 0.18 c ‒ with a dummy indicator that giving by others is high (0.08) (0.12) Income (wi) ‒ ‒ 0.40 d (0.07) Budgets 1, 2, 3, 4 2, 5, 4, 6 3, 4, 6 n 70 85 85

Test 5: Impure altruism - at a high level of giving by others

H0 : q2 > 0 and q1 + q2 < 1

Table 5. Additional evidence supporting the impure altruism model. Account for corner decisions Change in unfunded crowd-out (restricted sample)a Change in balanced-budget crowd-out (full sample) Unfunded crowd-out and own income effect Giving by others Low/ High (1) Low/ High (2) High (3) Giving by others (G-i)

• 0.53
• 0.99
• 0.41 d

(0.07) (0.09) (0.07) Giving by others interacted 0.12 b 0.18 c ‒ with a dummy indicator that giving by others is high (0.08) (0.12) Income (wi) ‒ ‒ 0.40 d (0.07) Budgets 1, 2, 3, 4 2, 5, 4, 6 3, 4, 6 n 70 85 85

Structural model

Ui = αlog(G) + βlog(gi) + (1 − α − β)log(xi)

Table: Cobb-Douglas

All Except always at a corner (N = 85) (N = 78) Altruism (α) .594 .569 (.025) (.025) Warm glow (β) .021 .026 (.009) (.009) ρ .902 .829 (within-person corr.) (.016) (.025) Notes: Non-linear random effects Tobit.

Crowd-out in the representative model

a = .594 b = .021 q1 + q2 |k| .2 .4 .6 .8 1 q1 + q2, |k| 20 40 60 80 100 Giving by others (G-i) q1 + q2 Crowd-out

Notes: The Cobb-Douglas parameters are a = .594 and b = .021. U = a log(G) + b log(gi) + (1 - a - b) log(xi). Income is held constant at wi = $40. These are the a, it:b} estimates from the representative Cobb-Douglas unility function. (N = 85).

Figure 9. Cobb-Douglas q1 + q2 as giving by others (G-i) increases.

Counterfactual: if there had been no altruism

• α = 0 and β = .021
• 95 % of the giving would be gone.

Individual heterogeneity in altruism and warm glow

Ui = αilog(G) + βilog(gi) + (1 − αi − βi)log(xi)

.2 .4 .6 .8 1 Altruism (a) .2 .4 .6 .8 1 Warm glow (b)

Notes: Points on the y-axis are pure altruism participants. Points on the x-axis are pure warm glow participants. n = 78.

Figure 8: Cobb-Douglas Altruism and Warm Glow Parameters

Individual heterogeneity in altruism and warm glow

Ui = αilog(G) + βilog(gi) + (1 − αi − βi)log(xi)

.2 .4 .6 .8 1 Altruism (a) .2 .4 .6 .8 1 Warm glow (b) Women Men

Notes: Points on the y-axis are pure altruism participants. Points on the x-axis are pure warm glow participants. n = 78.

Figure 8b: Altruism and Warm Glow: Women and Men

Conclusion - Significance

• First test of a prediction of the impure altruism model, not

designed into the model from the beginning.

• Strength of warm glow detected in an experiment is sensitive

to the level of exogenous provision of the public good at which crowd-out is being measured, and ...

• ... other characteristics of the context, such as the purpose of

the non-profit organization to which participants can give.

• In the experimental context we established, participant motives

differ from previous experimental results that have suggested warm glow to be the predominant motive people give to charity.

Conclusion - Significance

• Practical implications
• How fundraising communication takes into consideration ”Why

might potential donors give?”

• How donors reflect on the question ”Why do I give?”
• Evidence for the existence of altruism in charitable giving.

Thank you

• Kong Wah Lai, Michael Menietti, and Linnea Warren who

helped conduct the experiments.

• Sandi Wraith and the American Red Cross of South Western

Pennsylvania.

• IU Lilly Family School of Philanthropy Research Fund Grant

54-921-10.

What is q2 ?

• Mechanical answer: Impure altruism places no restrictions on

the size of q2.

• q2 is what it needs to be in order to generate a prediction that

is consistent with the available crowd-out evidence:

• ˆ

κIm = − .05, if q1 = .02, would be rationalized by q2 = .93

• a “cosmological constant” for charity
• Economics answer: q2 is a difference between two income

effects:

• dG ∗

dG−i − dG ∗ dwi = [q1 + q2] − q1

• dx∗

i

dwi − dx∗

i

dG−i = [1 − q1] − [1 − (q1 + q2)]

• Psychological answer: q2 arises because there are 7 specific

types of warm glow/egoism, none of which are satisfied when G−i ↑:

• 2 extrinsic: avoid shame, seek approbation.
• 5 intrinsic: avoid guilt, seek positive self-evaluation of identity,

seek “empathic joy,” seek relief of negative mood state, reduce personal distress.

What is q2 ?

• Pop culture answer . . .

If this is q1 . . .

. . . then this is q2

. . . warm glow indeed!