American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to American Economic
Journal: Economic Policy.
http://www.jstor.org
American Economic Association
Altruism and the Child Cycle of Alumni Donations
Author(s): Jonathan Meer and Harvey S. Rosen
Source: American Economic Journal: Economic Policy, Vol. 1, No. 1 (February 2009), pp. 258-286
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/25760034
Accessed: 18-08-2015 16:54 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
American Economic Journal: Economic Policy 2009, 1:1, 258-286
http://www.aeaweb. org/articles.php?doi=10.1257/pol. 1.1.258
Altruism and theChild Cycle ofAlumni Donations1^
By Jonathan Meer and Harvey S. Rosen*
We study alumni contributions to an anonymous research univer
sity. If alumni believe donations will increase the likelihood of their
child's admission, and if this belief helps motivate their giving, then
thepattern of giving should vary systematically with the ages of their
children, whether the children ultimately apply to the university, and
theadmissions outcome. We call thispattern thechild cycle of alumni
giving. The evidence is consistent with the child-cycle pattern. Thus,
while altruism drives some giving, the hope for a reciprocal benefit
also plays a role. We compute rough estimates of theproportion of
givingdue toselfishmotives. (JELD91, D64,121)
fn an essay on the economics of altruism, Paul A. Samuelson (1993, 143) writes:
Mesmerized by Homo economicus, who acts solely
on
egoism, econo
mists shy away from altruism almost comically. Caught in a shameful
act of heroism, theyaver: "Shucks, itwas only enlightened self interest."
Sometimes it is.At other times itmay be only rationalization... "If I res
cue
somebody's son, someone will rescue mine."
Samuelson concludes that such arguments render economists guilty of "face sav
ing tautologies." Theodore C. Bergstrom and Oded Stark (1993, 149) similarly criti
cize economists' stance toward altruistic behavior: "Why are economists convinced
thatHomo economicus is selfish? No doubt we find considerable support for this
hypothesis in the behavior of our colleagues."
The notion thatmainstream economics categorically rejects the existence of altru
ism is essentially false, however. As noted below, a substantial theoretical literature
explicitly allows for the possibility that human behavior is unselfish and draws out
the implications of altruism in a variety of contexts. Contrary to Samuelson, itwould
be more accurate to characterize economists' view of the importance of altruism as
agnostic rather than skeptical. They are willing to contemplate the possibility that
altruism is an important motivator of behavior but, at the same time, do not rule
out selfishness. Charles T. Clotfelter's (1985) important volume on charitable giving
*Meer: Department of Economics, Stanford University, Stanford, CA 94305 (e-mail: jmeer@stanford.edu);
Rosen: Department of Economics, Princeton University, Princeton, NJ 08544 (e-mail: HSR@princeton.edu). We
are grateful toAlan Auerbach, B. Douglas Bernheim, William Bowen, Charles Clotfelter, Ronald Ehrenberg, Jean
Grossman, Bo Honore, Donna Lawrence, Brian McDonald, Ashley Miller, JohnMorgan, Sriniketh Nagavarapu,
Andres Santos, Julie Shadle, Burton Weisbrod, and two anonymous referees for useful suggestions. Brent
Newhouse and Zhihao Zhang provided excellent research assistance. This research was supported in part by
Princeton University's Center forEconomic Policy Studies and in part by the Stanford Institute forEconomic
Policy Research.
f To comment on this article in the online discussion forum visit the articles page at:
http://ww w.aeaweb.org/articles.php?doi=10.1257/pol. 1.1.258
258
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. I MEER AND ROSEN: ALTRUISM AND THE CHILD CYCLE OF ALUMNI DONATIONS 259
is typical in this regard. In an introductory section, he provides a list of possible
motivations for charity. Some involve narrow self-interest, such as the expectation
that donors and their families will receive services or favorable publicity for their
businesses in return for their contribution. But the list gives equal footing to altruism
associated with social norms or a sense of duty or commitment. Clotfelter takes no
position on the relative importance of the various motivations. He merely observes
that they could all be operative.1
Of course, saying that both selfishness and altruism can be present does not tell
us that both motivations actually guide behavior. This is an empirical question, but
empirical work using observational data is rare in this area. Perhaps the primary
reason is the difficulty inmeasuring the benefits that donors expect to receive. For
example, news accounts of donations to hospitals indicate that one reason for giving
is the hope that, if individuals find themselves in the hospital in the future, theywill
receive particularly attentive treatment.2 But how does one measure themagnitude
of this benefit?Without quantifiable indicators of the potential selfish benefits, one
cannot estimate how responsive giving is to their existence.
This paper uses a unique data set that allows us to assess whether donors' contri
butions to a nonprofit institution are affected by the expectation of a reciprocal ben
efit.We study alumni contributions to an anonymous selective research university,
henceforth referred to as Anon U. The proprietary data provided by Anon U contain
detailed information about donations made by alumni as well as a variety of their
economic and demographic characteristics. The data also include information on
the ages of the children of the alumni, whether they applied for admission toAnon
U, and, if so, whether theywere accepted. The premise of our analysis is simple. If
alumni believe donations increase the likelihood of their children being accepted
toAnon U, and if this belief helps motivate their giving, then the pattern of giving
should vary systematically with the age of their children, whether the children ulti
mately apply toAnon U, and the outcome of the admissions process. Specifically, if
reciprocity influences the behavior of donors, one would expect that, ceteris paribus,
the presence of children increases the propensity to give, that giving drops off after
the admissions decision is rendered, and that the decline is greater when the child is
rejected. We refer to this pattern as the child cycle of alumni giving.
An interesting feature of this phenomenon is that the institutionmakes no prom
ise of reciprocity. True, children of Anon U alumni have a higher rate of accep
tance than other students,3 but this does not prove that having a parent who made
donations in the past increases a child's likelihood of admission. Nevertheless, the
view that reciprocity exists is widespread. As one account of the college admis
sions process stated, "Traditionally, universities have relied on gifts from alumni,
1
In the same way, Dugan et al. (2000) ascribe a number ofmotivations to university donors. The list includes
"avoidance of social stigma, tax incentives, recognition for generosity, a response to past or deterrence to future
solicitation, and quid pro quo for services rendered indirectly such as access to elite social circles or business
contacts." However, at the top of their list is "pure altruism."
2
Julie Bick, "The Hospital Worked Wonders. Can you Return the Favor?" New York Times, August 5, 2007.
3
According to public information, children of alumni at Anon U are accepted at roughly three times the rate
of other applicants. William G. Bowen et al. (2005) document the importance of correcting fordifferences in the
characteristics of applicant pools when assessing the importance of legacy preferences, however.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
260 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
who are rewarded with 'legacy' preferences for their children."4 In 1976, a federal
judge held inRosenstock v.Board ofGovernors of theUniversity ofNorth Carolina
that admission preferences for the children of out-of-state alumni were not uncon
stitutional, since "alumni provide monetary support for theUniversity." The judge
viewed this as "a reasonable basis and...not constitutionally defective." Perceptions
of reciprocity may be reinforced by university administrators who link the accep
tance of alumni children to the financial support of their institutions. In a recent
interview with theWall Street Journal, the president of Princeton University was
asked, "Why does Princeton give admissions preference to alumni children...?"
Her response was, "We are deeply dependent on the generosity of our alumni each
and every year...They are extremely important to the financial well-being of this
university."5We know of no statistical evidence on whether alumni donations at any
university affect admissions probabilities for their children, and if so, how much.
For our purposes, the key insight is that generating the child cycle of alumni giving
requires only the perception of reciprocity.
Determining whether a child life cycle exists is important because of the opportu
nity itprovides to shed light on the general question ofmotivations for altruism. As
Lise Vesterlund (2006) notes, several important policy issues surround the question.
Suppose, for example, thatdonors perceive the gain from a donation to be primarily
private. In that case, if the government provides a grant to the institution, there is no
strong reason for the donor to cut back on his or her donations, and vice versa. Thus,
determining whether or not a child cycle of giving is present gives us some evidence
about the importance of "crowding out" at least in this context.6 In addition, the
extent towhich a donation for a public good ismotivated by a private gain affects
whether private provision of the public good is optimal and whether government
intervention is required to enhance efficiency. Gaining a better understanding of the
motivations for alumni giving is also of independent interest because of its impor
tance to the financing of higher education. In 2004-2005, alumni contributed $7.1
billion to higher education, about 28 percent of all voluntary support.7
In Section I, we briefly review some pertinent previous research in this area.
Section II describes the data and econometric framework. The results are presented
in Section III. The evidence is strongly consistent with the child-cycle pattern.
Alumni parents of teenage children who apply toAnon U make larger donations
than alumni whose children do not apply. Once an alumnus's child is accepted,
his donations fall off substantially. If the child is rejected, giving falls off dramati
cally. Section IV discusses the sensitivity of the results to alternative specifications
of themodel. Section V concludes with a summary and suggestions for additional
research.
4
Daniel Golden, "How Lowering the Bar Helps Colleges Prosper," Wall Street Journal, Saturday/Sunday,
September 9-10, 2006.
5
JohnHechinger, "The Tiger Roars," Wall Street Journal, July 17,2006.
6
Vesterlund (2006) notes that, relative to studies of crowd-out using survey or tax data, experimental studies
tend to find stronger evidence of a public motive for giving.7
Other sources of voluntary support include other individuals, corporations, foundations, and religious and
other organizations. See Chronicle ofHigher Education (2006).
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER AND ROSEN: ALTRUISM AND THE CHILD CYCLE OF ALUMNI DONATIONS 261
I. Previous Literature
The role of altruism in human behavior has long been of interest to economists. As
Serge-Christophe Kolm (2000, 7) notes, "all great economists have considered the
effects of positive social sentiments," including Adam Smith, John Stuart Mill, Leon
Walras, Vilfredo Pareto, and JeremyBentham. Inmore recent times, notions of altru
ism have been brought to bear in theoretical analyses of charitable giving (Gary S.
Becker 1974), rescues (William M. Landes and Richard A. Posner 1978), commercial
policy (Julio J.Rotemberg 2000), and remittances ofmigrants to theirhome countries
(Frederic Docquier and Hillel Rapoport 2000), among other important social phenom
ena. Additional discussion can be found inAmartya Sen (1977) and Kolm (2000).
Altruistic behavior within families has received particularly extensive attention
because of its implications for several important policy questions. In particular, the
efficacy of fiscal policy hinges on the extent towhich intergenerational bequests are
motivated by altruism. Becker (1974) considers a theoretical model inwhich parents
are altruistic and shows that, under certain assumptions (an important one being that
parents' utility depends only on family income) they have no incentive tobe strategic
with respect to their children. In contrast, B. Douglas Bernheim, Andrei Shleifer and
Lawrence H. Summers (1985) develop a theory inwhich bequests are motivated not
only by altruistic concern for children, but also with the hope for reciprocity in the
form of care or attention.
Theories relating to intrafamily altruism have been tested in a rich economet
ric literature. This literature has developed because there are observable variables
related to the selfish gains thatmight be obtained from seemingly altruistic behavior.
For example, one can look at the amount of contact between elderly parents and
children and how it is related to the parents' bequeathable wealth. The results are
mixed. Bernheim, Shleifer, and Summers (1985); Joseph G. Altonji, Fumio Hayashi,
and Laurence J.Kotlikoff (1992); and Alessandro Cigno and Furio C. Rosati (2000)
find evidence that gifts from parents to children have a strategic component, while
Kathleen McGarry and Robert F. Schoeni (1995), Lakshmi K. Raut and Lein H.
Tran (2000), and Yannis M. Ioannides and Kamhon Kan (2000) find that altruistic
motives predominate.
Turning to themuch different but important case of charitable donations, which
amounted to about $250 billion in 20048, researchers have more or less taken
for granted that selfish motives play a role. According to Clotfelter (1985, 38),
"Individuals may volunteer for organizations in order for their families or them
selves to consume services;' Burton A. Weisbrod (1978, 34) ismore pointed: "The
extent towhich narrow self-interest lies behind the donations ofmoney and time to
nonprofit organizations is little understood, but there can be no doubt that donors
often do benefit through themaking of business contacts and the receipt of favorable
publicity for good deeds." Similarly, ithas been suggested that selfish motives may
8
Giving USA Foundation (2005).
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
262 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
underlie donations to universities. "Donors demand attention and prestige supplied
by college fundraisers"9 (JangH. Yoo andWilliam B. Harrison 1989, 367).
Are such assertions valid? Erik Schokkaert and Luc Van Ootegem (2000) sum
marize survey evidence on reasons for giving. Also, a number of laboratory experi
ments have investigated the extent towhich contributions topublic goods aremarked
by altruism. (See, for example, James Andreoni (1993), John A. List (2007), and
the extensive survey by Andreoni, William T. Harbaugh and Vesterlund (2008).10
James C. Cox (2004) designed an experiment to determine how much of individu
als' behavior is driven by altruism or reciprocity. He found thatwhile reciprocity is
an important component ofmotivation, a large proportion of the subjects still show
some elements of pure altruism. Similarly, Gary Charness and Ernan Haruvy (2002)
found that altruism plays a role in a gift-exchange experiment.
Inmarked contrast to the literature on giving within the family, however, we have
been able to find no statistical work on motivations for charitable giving based on
observational data. The same holds for the literature on the determinants of alumni
giving to their universities. Econometric studies of alumni giving have examined a
wide array of variables such as attitudinal measures of satisfaction with the under
graduate experience, income, marital status, number of children, occupation, the
state of the stock market, marginal tax rates, gender, ethnicity, academic perfor
mance as an undergraduate, extracurricular activities including varsity athletics,
membership in fraternities or other social clubs, whether the individual received
financial aid, performance of athletic teams, and so on.11 However, we have found
no systematic attempts to assess whether self-interestmight have a role in explaining
giving behavior.
The likely reason for the dearth of such research is the absence of measurable
indicators of the benefits donors expect to receive in return for their donations. Our
study is premised on the notion that alumni believe donations enhance the probabil
ity that their children will be admitted to their alma mater. Therefore, the presence
of children, their ages, and their admissions status are measurable indicators of the
potential for reciprocal benefits generated by donations. In this view, alumni believe
that donations buy them entrance into a lottery inwhich the prize is admission for
their children. As we stressed above, whether the probability of one's child being
admitted depends on prior or expected future donations is unknown.12 However, as
long as alumni perceive that their contributions improve their children's chances of
being admitted and that greater contributions by others lessen the odds, then this
mechanism is operative.
9
This suggests that efforts by college development offices could be an important determinant of alumni giv
ing.Certainly, this iswhy colleges have such offices in the firstplace. While there is some evidence of a correla
tion between development costs and donations across institutions (Harrison, Shannon K. Mitchell, and Steven P.
Peterson 1995), it is difficult to ascribe a causal relationship because the variables are likely jointly determined.
10
For an ethnographic approach to studying motivations for charitable giving, see Teresa Odendahl (1990).
11
See, for example, Clotfelter (2003), JamesMonks (2003), James L. Shulman andWilliam G. Bowen (2001,
Chapter 10),Alton L. Taylor and Joseph C. Martin (1995), and Phanindra V. Wunnava and Michael A. Lauze
(2001).
12
In particular, our data do not allow us to explore this hypothesis forAnon U, as we have no information on
the attributes of rejected students. (See below.)
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. I MEER AND ROSEN: ALTRUISM AND THE CHILD CYCLE OF ALUMNI DONATIONS 263
II. Data and Econometric Model
A. Data
Our primary data source is the administrative archives of Anon U's develop
ment office, which contain information on all alumni donations from 1983 to 2006.
There are no issues of selectivity with respect to the sample because every alumnus
is included. The data are proprietary and sensitive, and individuals' names were
stripped from the records before being made available to us. Our unit of observation
is a yearly giving opportunity. For example, if an individual has been an alumna
for five years, she accounts for five giving opportunities in our analysis, starting in
the first fiscal year after graduation. Multiple gifts in the same year are summed
together. The development office data also include information on academic major,
extracurricular activities when the alumnus was an undergraduate, post graduate
education, occupation, residence, whether he or she ismarried to another graduate
ofAnon U, as well as information on the age and admissions status of the alumnus's
children.13 Anon U's registrar supplemented these data with information on SAT
scores, academic honors, ethnicity, type of high school, summary evaluations made
by the admissions office during the application process, and grade point average.
The registrar's data are available only as far back as the class of 1972, so we restrict
most of our analysis to this group of individuals.
We begin with 547,836 observations representing 35,556 alumni. We delete 27,992
observations because ofmissing data on the child's age, essential information forour
analysis. We deleted an additional 1,100 observations because the child withdrew
his or her application before a decision was rendered, and another 32,041 because of
missing data for other variables. Altogether, our analysis sample has 487,913 obser
vations on 32,488 alumni.
We focus on two dimensions of alumni giving. First is the probability that an
alumnus made any gift at all in a given year.14 Universities care about the propor
tion of their alumni who make donations. Anon U, for example, makes considerable
effort to contact as many alumni as possible and urge them togive something, even if
it is just a few dollars. Second, we analyze the amount donated in any given year.15
The mean and standard deviation of each of these variables are shown at the top of
Table l.16The unconditional mean gift (in 2006 dollars) is $466. The relatively large
13
The data regarding post-graduation attributes are collected through various means, including surveys, class
Web sites, and the alumni magazine. Class officers and the alumni records office also contribute to these data.
Although the approach is not entirely systematic. Anon U's development office is confident that the data are fairly
complete and accurate.
14
Pledges without an associated gift are not counted.
L>>
As is typically the case, a few relatively large gifts account fora disproportionate amount of Anon U's dona
tions. For example, in2006, the top 1percent of gifts accounted for 69.2 percent of total giving. In results avail
able upon request, we also estimate the probability that the alumnus is a "class leader" in a given year, where a
class leader isdefined as an individual who donated an amount greater than or equal to the 90th percentile of gifts
inhis or her class. The results with respect to the child cycle are qualitatively the same for the probability of being
a class leader as they are for the probability of making any gift at all, and for the amount of thegift.16
As noted above, these are summary statistics of our observations, which are not the same as summary statis
tics for the alumni themselves. In effect, thedata in the table weight alumni characteristics by the number of years
each alumnus was in the sample. Therefore, changes in the demographic structure of Anon U may not be fully
evident. However, the summary statistics in the last panel of Table 1report some variables forwhich themeans
are taken over the number of alumni rather than the number of observations.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
264 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
Table 1?Variable Definitions and Summary Statistics3
Variable Description Mean
Standard
deviation
TotalYear
LogTotalYear
Didgive
Yearssince
Yearssince2
Spouseisalum
Male
Race/ethnicity
White
Amerind
Black
Hispanic
Asian
Secondary schooling
Public
Boarding
Private
Schloth
SATmath
Total giving for year in 2006 dollars
Log of total giving for year in 2006 dollars
1 ifany donation given in year
Number of years since graduation
Number of years since graduation, squared
1 if the spouse is an alumnus
1 if the alumnus ismale
Omitted category: 1 if the alumnus iswhite
1 if the alumnus is a Native American
1 if the alumnus isblack
1 if the alumnus isHispanic
1 if the alumnus isAsian
Omitted category: 1 if the alumnus attended public school
1 if the alumnus attended boarding school
1 if the alumnus attended private school
1 if the alumnus attended another type of school
SAT math score. Scores prior to 1996 are adjusted
to reflect recentering of the scoring scale
SATverbal SAT verbal score. Scores prior to 1996 are adjusted
to reflect recentering of the scoring scale
Admissions office "nonacademic" ranking5
A Omitted category: 1 if the alumnus received the highest
nonacademic ranking from the admissions office
B 1 if the alumnus received the second highest
nonacademic ranking from the admissions office
C 1 if the alumnus received the third highest
nonacademic ranking from the admissions office
D 1 if the alumnus received the fourth highest
nonacademic ranking from the admissions office
E 1 if the alumnus received the fifthhighest
nonacademic ranking from the admissions office
Admissions office "academic" ranking
A Omitted category: 1 if the alumnus received the highest
academic ranking from the admissions office
B 1 if the alumnus received the second highest
academic ranking from the admissions office
C 1 if the alumnus received the third highest
academic ranking from the admissions office
D 1 if the alumnus received the fourth highest
academic ranking from the admissions office
E 1 ifthe alumnus received the fifthhighest
academic ranking from the admissions office
1 if the alumnus played on a varsity team
1 if the alumnus played on a club team
1 if the alumnus graduated magna, summa, or cum laude
1 if the alumnus was a member of a fraternity or sorority
466.14
2.425
0.5563
12.05
206.5
0.1302
0.6507
0.8195
0.00363
0.06929
0.03798
0.06958
0.5792
0.1395
0.2622
0.01916
703.1
Varsity
Clubsport
Honors
Greek
Major
Molbio
Small social sciences
English
Economics
Public policy
Political science
Psychology
History
MAE
EE/CS
701.9
0.03220
0.4660
0.4188
0.07897
0.00401
0.1536
0.4242
0.2708
0.1435
0.0079
0.3892
0.1728
0.4532
0.6949
49,512
2.425
0.4968
7.826
231.3
0.3365
0.4768
0.3846
0.06012
0.2539
0.1911
0.2544
0.4937
0.3464
0.4398
0.1371
75.95
75.77
Omitted Category: 1 if the alumnus majored
inmolecular biology
1 if the alumnus majored in anthropology, urban studies,
or Sociology
1 if the alumnus majored inEnglish
1 if the alumnus majored in economics
1 if the alumnus majored inpublic policy
1 if the alumnus majored inpolitical science
1 if the alumnus majored in psychology
1 if the alumnus majored in history
1 if the alumnus majored inmechanical/aerospace
engineering
1 if the alumnus majored in electrical engineering
or computer science
0.1073
0.07949
0.05841
0.08796
0.04890
0.1182
0.03534
0.1765
0.4988
0.4934
0.2697
0.06319
0.3605
0.4942
0.4444
0.3506
0.08858
0.4876
0.3780
0.4978
0.4604
0.02940 0.1689
0.3095
0.2705
0.2345
0.2832
0.2157
0.3229
0.1846
0.05533 0.2286
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. I MEER AND ROSEN:ALTRUISMANDTHECHILD CYCLEOF ALUMNIDONATIONS 265
Table 1?Variable Definitions and Summary Statistics (Continued)
Variable Description Mean
Standard
deviation
Major
Arch & Civ
Small humanities
Small engineering
Small sciences
Minor
No minor
African/African
American studies
American studies
Latin
Finance
Theater
Public policy
Other engineering
Other sciences
Other humanities
Teaching
Reunion
Post baccalaureate education
NoPostAB
PhD
Masters
JD
MD/DDS
MBA
Alumni-based sample
EverGave0
Children0
AnyAcceptedc
PropApplied0
PropAcceptedc
I if the alumnus majored in architecture or 0.07040 0.2558
civil engineering
I if the alumnus majored inart, art history, 0.11800.3226
classics, East Asian studies, linguistics, music,
Near Eastern studies, philosophy, religion, or
languages and literature
1 if the alumnus majored in "engineering," 0.03132 0.1742
operations research and financial engineering,
or chemical engineering
1 if the alumnus majored inapplied mathematics, 0.1375 0.3444
astrophysics, biochemistry, biology, chemistry,
ecology and evolutionary biology, geology,
mathematics, physics, or statistics
Omitted category: 1 if the alumnus received no minor 0.7673 0.4226
1 if the alumnus received a minor inAfrican or 0.02303 0.1500
African American studies
1 ifthe alumnus received a minor inAmerican studies 0.02324 0.1507
1 ifthe alumnus received a minor inLatin 0.00186 0.04305
1 ifthe alumnus received a minor in finance 0.00324 0.05683
1 if the alumnus received a minor in theater 0.0129 0.1130
1 if the alumnus received a minor in public policy 0.05023 0.2185
1 ifthe alumnus received a minor in architecture, 0.0184 0.1344
basic engineering, bioengineering, electrical
engineering, geological engineering, management,
materials sciences, or robotics
1 ifthe alumnus received a minor in applied and 0.0273 0.1630
computational mathematics, biophysics.
cognitive studies, environmental studies, science
inhuman affairs, or neuroscience
1 ifthealumnus received aminor ina humanities field 0.0526 0.2233
1 if the alumnus received a teaching certificate 0.01377 0.1165
1 if the year after graduation is some multiple of 5 0.1795 0.3838
Omitted category: 1 if the alumnus has no 0.6053 0.4888
advanced degree
1 ifthe alumnus has a PhD or equivalent degree 0.0674 0.2508
1 if the alumnus has a masters 0.1381 0.3450
1 ifthe alumnus has a JD 0.1004 0.3006
1 ifthe alumnus has a medical degree 0.06173 0.2407
1 ifthe alumnus has an MBA 0.0899 0.2860
Proportion of alumni who gave at least once 0.8828 0.3216
since graduation
Proportion of alumni with at least one child 0.2347 0.4238
Proportion of all alumni who had at least one 0.0212 0.144
child admitted toAnon U
Conditional on the eldest child reaching age 17, 0.5870 0.4924
proportion of alumni who had at least one child
apply toAnon U (n
-
2.840)
Conditional on at least one child applying, 0.4107 0.4921
proportion of alumni who had at least one child
admitted toAnon U (//
=
1,667)
a
Based on 487,913 observations on gift giving from 1983 to 2006. Represented here are 32,488 alumni who
graduated from 1972 to2005. Unless otherwise indicated, these are summary statistics of our observations, which
are not the same as summary statistics for the alumni themselves. In effect, the data in the table weight alumni
characteristics by the number of years each alumnus was in the sample.
bThe nonacademic ranking isbased on attributes such as musical talent, athletic ability, volunteer work, etc.
c
Based on 32,488 observations on alumni who graduated from 1972 to 2005, observed in2006. Unless other
wise noted, these are summary statistics for the alumni themselves. These variables do not enter the regression
models.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
266 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
standard deviation, $49,512, reflects thepresence of enormous outliers. To reduce the
likelihood that outliers drive our results, we take the log of the amount given.17 In
addition, we also estimate our models without the top 1percent of the observations,
and find that the results are essentially unchanged. With respect to the probability
of giving, Table 1 shows that about 55.6 percent of the giving opportunities result
in a donation to the university. Relative to other schools, this is a high participation
rate. Indeed, Anon U is at the edge of the right tailwith respect to the proportion of
alumni who make contributions.
Most of the explanatory variables in the table are dichotomous. For each set of
dichotomous variables, the "omitted category" is the variable that is excluded from
the regressions. About 65.1 percent of our observations are associated with male
alumni. Historically, Anon U was an all-male institution and did not confer degrees
on women until the 1970s.Whites comprise 81.9 percent of our observations. A total
of 57.9 percent of the observations are associated with secondary education at a pub
lic school; almost 39 percent with participation in undergraduate varsity athletics;18
and 45.3 percent with individuals who receive honors when they graduate. About 40
percent receive a post baccalaureate degree.
Unfortunately, the data include no direct information on income, an important
determinant of giving (Shulman and Bowen 2001, 404). We address this issue in
twoways. First, for a large subset of our alumni, we have information that is closely
related to permanent income, occupation and field.19 Table 2 shows the occupations
and fields for the 344,342 observations, representing 20,039 alumni, forwhich we
have this information.20 The fields of education, finance, health care, and law are
highly represented. We re-estimate our basic models with this subsample including
the occupation and field data in order to see whether our substantive results are sen
sitive to their inclusion. As shown below, they are not. Second, ifwe are willing to
think of an alumnus's permanent income as an unchanging attribute (at least during
our sample period), thenwe can model itas a fixed effect.We show below that our
substantive results are unchanged with fixed-effects estimation.
B. Characterizing theChild Cycle
We characterize the child cycle by a vector of dichotomous variables indicating
whether the alumnus has a child, and if so, his or her age and admissions status.
We discuss the treatment of families with several children below. Recall that in our
framework, alumni with children may believe that a gift toAnon U will some day
17
A logarithmic transformation presents problems for observations that take a value of zero. As noted below,
we set 320 gifts that are greater than zero but less than or equal to $1 equal to $1.01. Therefore, observations for
which there is no giving are associated with $1, forwhich the logarithm is zero.
18
Varsity athletes are defined as those who participated in a varsity-level sport, not necessarily receiving a
varsity letter.Club sports are defined as those thatdo not confer a varsity letter.
19
In this context, it is important to note that a number of the variables in our basic specification are also cor
related with income including gender, ethnicity, college major and grade point average, advanced degrees, years
since graduation, and location. Moreover, Brendan M. Cunningham and Carlena K. Cochi-Ficano (2002) point
out that SAT scores are related closely to family socioeconomic status as well.
20
Due to lack of reliable data regarding the start and stop dates of occupation and field, these variables indicate
whether the alumnus was ever involved in that field or occupation, rather thanwhether they are involved during
the particular year of observation.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. 1 MEERAND ROSEN:ALTRUISMAND THECHILD CYCLEOF ALUMNIDONATIONS 267
Table 2?Position and Field Definitions and Summary Statistics0
Variable Description Mean
Standard
deviation
Field
Arts
Agriculture
Architecture
Pharmaceuticals
Communications
Consulting
Education
Finance
Health care
(Business/Industry)
Hospitality
Information technology
Law
Manufacturing
Retail
Transportation
Federal government
State Government
Foreign government
Nongovernmental
Organization
Religion
Other
Multilateral organization
Military
Occupation
Government worker
Miscellaneous worker
Physician/dentist
White collar
Attorney
Executive
ifthe alumnus ever worked in the arts field
ifthe alumnus ever worked in the agriculture field
if the alumnus ever worked in the architecture field
if the alumnus ever worked in the pharmaceuticals field
ifthe alumnus ever worked in the communications field
if the alumnus ever worked in the consulting field
ifthe alumnus ever worked in the education field
if the alumnus ever worked in the finance field
ifthe alumnus ever worked in the health care field
ifthe alumnus ever worked in the hospitality field
if the alumnus ever worked in the IT field
if the alumnus ever worked in the law field
if the alumnus ever worked in themanufacturing field
if the alumnus ever worked in the retail field
ifthe alumnus ever worked in the transportation field
ifthe alumnus ever worked for the federal government
if the alumnus ever worked for a state government
ifthe alumnus ever worked for a foreign government
if the alumnus ever worked in theNGO field
if the alumnus ever worked in the religion field
if the alumnus ever worked in another field
if the alumnus ever worked in themultilateral
organization field
if the alumnus ever worked for themilitary
if the alumnus ever worked as a government worker
ifthe alumnus everworked in some
miscellaneous occupation
ifthe alumnus ever worked as a physician or dentist
ifthe alumnus ever worked ina white collar occupation
if the alumnus ever worked as an attorney
ifthe alumnus ever worked as an executive
0.06254
0.00187
0.02516
0.02336
0.09619
0.1009
0.1222
0.1947
0.1700
0.00457
0.1150
0.1883
0.07509
0.02251
0.01014
0.04406
0.02515
0.00275
0.02832
0.01052
0.27108
0.00191
0.00747
0.01007
0.08177
0.1339
0.3079
0.2673
0.4863
0.2421
0.04324
0.1566
0.1511
0.2949
0.3011
0.3275
0.3960
0.3756
0.06743
0.3190
0.3909
0.2635
0.1483
0.1002
0.2052
0.1566
0.05234
0.1659
0.1020
0.4445
0.04371
0.08612
0.09982
0.2740
0.3405
0.4616
0.4426
0.4998
a
Based on 344,342 observations on gift giving from 1983 to 2006 for individuals forwhom data on field and
position are reported. A total of 20.039 alumni who graduated from 1972 to 2005 are represented.
generate a reciprocal benefit, and therefore the presence of a child should increase
the probability ofmaking a gift. Perhaps, though, having a child is correlated with
unobserved variables that also drive giving decisions. For example, individuals who
become parents might care about young people in general, and hence be particularly
willing to support higher education. But if so, there would be no reason to expect
giving to decline just when the child exceeds the age at which college admissions
decisions are made. In contrast, the child-cycle framework implies that once the
child is beyond that age, giving will drop off, because the admissions lottery is
over. To examine how giving varies with the age of the child, we include a series of
dichotomous variables, CHILDh which take a value of one if the alumnus has a child
of age /and zero otherwise. The range of / is from zero (less than one year old) to
26 years old and older.
Even if giving increases as admissions time approaches (at approximately 18
years of age) and falls thereafter, hopes for reciprocity need not be atwork. Perhaps,
for example, a child of college age reawakens fond memories of an alumnus's
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
268 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
undergraduate days, or inspires thoughts of experiences that the alumnus and his
child might share during future parents' weekends. This could lead to an increase in
an alumnus's propensity to give. To investigate this possibility, we take advantage of
information on whether the child ultimately applies for admission. Suppose that by
the time the child is a teenager, an alumnus can reasonably estimate the probability
thathis orher childwill ultimatelyapply.Such an estimatecouldbe based on the
child's expressed preferences for type of college, academic performance, and so on.
If so, the perceived payoff to the admissions lottery should be higher for alumni
whose children ultimately apply than for those who do not, and so should their dona
tions.We therefore include a set of interaction terms, CHILD {Appl, which multiply
CHILD {by a dichotomous variable that equals one if a child of age /eventually
applied toAnon U and zero otherwise. We assume thatparents can form reasonably
accurate expectations about whether their children will apply only when the children
are in their teens, so thatCHILD {Appl is defined only forvalues of ifrom 14 through
17 years of age. Under the joint hypothesis that alumni can predict with some accu
racy whether their children will apply and that expected reciprocity is a motivation
for giving, these interaction terms should have positive coefficients.
Similarly, we define a series of dichotomous variables, CHILD{NoAppl, which
equals one if the child ultimately did not apply and zero otherwise, with /running
from 14 to 17years of age. If expected reciprocity is present, the coefficients on these
variables will be smaller than those associated with theCHILD\Appl variables, but
still positive. They remain positive because presumably some parents in this group
believe that their children will apply, so their giving should be higher than that of
members of the omitted category, who have no children at all. A third series of
dichotomous variables, CHILDjYoung, equal one if the child was not old enough to
have applied by the end of our sample in 2006, with /running from 14 to 16 years
of age.
Turning now to the outcome of the admissions decision, we expect it to have no
impact on giving if altruism is the only motivation. On the other hand, to the extent
that giving ismotivated by expected reciprocity, we expect parents of admitted chil
dren to reduce giving, as there is no longer an expected gain. This effect will be
attenuated if these alumni perceive thatAnon U has "held up its side of thebargain,"
and reciprocate by continuing to give. Below, we examine some other reasons why
admittance of one's child might not lead to a dramatic decrease in giving. Turning
now to the parents of rejected children, not only is the prospect of an expected gain
gone, but the alumnus may perceive that the university has not reciprocated properly,
and therefore retaliates by reducing donations even further.
To examine these conjectures about the impact of the admissions decision, we cre
ate a set of dichotomous variables: CHILDtAcc, which equal one if the child applied
toAnon U and was accepted and zero otherwise; and CHILDtRej, which equal one
if the child applied toAnon U and was rejected and zero otherwise.21 For these vari
ables, /runs from 18 through 26 years old and older.
21
The data allow us todistinguish between those who are accepted but do not attend and those who do attend.
However, nearly all ofAnon U's alumni children who are accepted toAnon U choose to attend. Therefore, sample
sizes are too small tomeasure accurately any differences in the two populations.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. I MEER ANDROSEN:ALTRUISMAND THECHILD CYCLE OF ALUMNIDONATIONS 269
A complication arises when families have multiple children. Which one should
be used for characterizing the child cycle? In our basic results reported below, all
the child-cycle variables are defined in terms of the first child. This makes sense
because the giving decision surrounding the first child is unaffected by any prior
personal experience of reciprocity. However, some of the giving that occurs during
the first child's cycle might be affected by the presence of younger children. For
example, alumni might continue tomake large donations after their first child is
accepted out of concern about the admissions prospects of younger children. Given
the large number of child-cycle variables, it is infeasible to include the cycles for
multiple children and their interactions inone model. Therefore, we simply estimate
the child cycle based on the last child in the family. As shown below, the substantive
results are essentially the same as those'based on the first child.
Itwould be cumbersome and uninformative to report summary statistics for each
of the large number of variables that characterize the child cycle. To provide some
basic information, we note that in2006, the last year of our sample, 23.5 percent of
the alumni had at least one child. Of those who had a child, themean age was 13.6
years old. Conditional on reaching age 17, 1,501 alumni children, representing 52.9
percent of that subsample, had applied toAnon U, and 37.2 percent were accepted.
Other summary statistics that relate to the child cycle are reported in the last panel
of Table 1 under the heading "Alumni-based sample."
C. Econometric Model
We model the decision tomake a gift toAnon U with a probit model:
Pr(Gjt)
$[a
+ CYCLER +
Xjty+ YEARt(32+
LOCJt(33
+
CLASS,-j34],
wherePr(G^) is theprobabilitythatalumnusj makes a gift inyear t\<&[ ] is the
cumulative normal distribution function; CYCLEy,
is the vector of variables charac
terizing the child cycle as discussed above; Xjt
is a vector of the alumnus's personal
characteristics; YEARt is a set of time effects; LOCjt
is a set of location effects (state
or foreign country of residence); and
CLASS)
is a set of class effects (equal to one if
the alumnus graduated in a given year and zero otherwise). The time effects account
for the impacts of the state of the business cycle, the stockmarket, and so on.22 The
state effects allow for the possibility that alumni who live closer toAnon U might
be more likely to donate, and their children might be more likely to attend. The
class effects control for common influences on alumni in the same class, such as the
political milieu when theywere undergraduates, the presence of certain professors
or administrators, and so on.
As noted above, we have more than one observation per alumnus. Because the
errors for the observations for a given alumnus are likely to be correlated, the stan
dard errors are adjusted for clustering within individuals.
22
Ralph Bristol (1991) emphasizes the role of the stock market and Ronald G. Ehrenberg and Christopher
L. Smith (2003) document the importance of macroeconomic conditions. Time effects also take into account
changes in the value of the university's endowment (Sharon M. Oster 2001).
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
270 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
When we turn to the actual amount of the gift,we face two issues that arise in all
studies of donative behavior. First, a substantial number of the observations are zero.
Second, there are a few very large outliers. For example, the three largest gifts in our
sample are $3.1 million, $6 million, and $31.1 million. To address the first issue, we
use the Tobit estimator, which explicitly takes censoring into account. The second
problem suggests thatwe transform the data to reduce the influence of outliers. We
take logarithms. Because the logarithm of zero is not defined, we set the 320 positive
gifts thatwere less than or equal to $1 equal to $1.01.23 In effect,we have censoring
at the point where the logarithm of the gift is equal to zero, and can then apply Tobit
straightforwardly. There is, of course, some arbitrariness to this procedure. To assess
its robustness, we also estimate the model in levels, firstwith the entire sample
and then eliminating the top one percent of the observations in order to reduce the
impact of outliers. The substantive results with respect to the child-cycle variables
are not affected. As with the probit estimates, we correct for correlation among the
error terms for any given individual by using a clustering procedure.
We assume that the determinants of the amount of giving are the same as those
that affect the probability of giving. Under this condition, themarginal effects gen
erated by the probit model and theTobit model are the same up to a constant of pro
portionality. Because the quantitative magnitudes of the child-cycle variables on the
probability of giving and the amount of giving are both of interest,we report both
sets ofmarginal effects below.
III. Results
A. ProbabilityofMaking a Gift
Given the large number of child-cycle variables, themost convenient way topresent
the child-cycle coefficients is in a graph.24 In Figure 1, the horizontal axis measures
the child's age, and the vertical axis shows the incremental effect on the probability
ofmaking a gift (relative to having no children). The dashed lines indicate 95 per
cent confidence intervals. Note that overlapping confidence intervals do not imply
necessarily thatone cannot reject thehypothesis that the difference between the two
associated coefficients is zero. We return to this issue below.
The graph starts when the child is born and shows that having a child increases
the probability of giving by about 13 percentage points. The incremental effect of
the child's presence generally increases with the child's age, reaching about 17 per
centage points by the time he or she is 13 years old. These coefficients are estimated
precisely. When children reach the age of 14, our model distinguishes between those
who eventually apply to Anon U and those who do not. This is reflected by the
fact that the line divides at age 14.Comparing the two sets of coefficients, we see
that at every age from 14 to 17 years old, the incremental probability of giving is
greater for alumni whose children ultimately applied. The differences at each age
23
Dropping these observations, which represent only 0.12 percent of all gifts, leaves our results essentially
unchanged.
24
The coefficients and standard errors themselves are available upon request.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER ANDROSEN:ALTRUISMAND THECHILD CYCLEOF ALUMNIDONATIONS 271
Figure 1. Incremental Effect of the Child Cycle on the Probability of Making a Gift
Notes: This figure graphs the child-cycle coefficients from a probit model of the probability ofmaking any dona
tion in a given year, estimated using 487,913 observations. The vertical axis shows the incremental effect (relative
to having no children) on the probability ofmaking a gift in a given year as a function of the first child's age and
admissions status. The dashed lines are 95 percent confidence intervals.
are statistically significant from each other.Moreover, the joint test that each pair of
coefficients is different is highly significant (the chi-squared testwith four degrees
of freedom is 31.78, which is associated with p
=
0.0000). This differential is con
sistent with the hypothesis that alumni reasonably can predict the likelihood that
their children will apply toAnon U and that reciprocity in the form of admission is
expected.25
At age 18, the graph splits again, this time between applicants who were rejected
by and those who were accepted atAnon U.26 The graph indicates that, conditional
on applying, the probability of giving increases by about 34 percentage points for
an alumnus whose 18 year old child is accepted. The incremental probability falls
after acceptance (by age 20 it is down to about 27 percentage points), but remains
elevated into the child's mid-20s. The parents of unsuccessful applicants behave
very differently.As the figure indicates, the incremental probability ofmaking a gift
falls off substantially at age 18, and at age 19 and older, it is essentially zero. At each
age, the differences between probabilities for alumni whose children were accepted
and those who were not are statistically significant except for those whose children
25
Note also that this finding is inconsistent with the notion that the child-cycle pattern isdue to the fundraising
office focusing on alumni with children approaching college age. To explain this finding, one would have to argue
that the fundraisers have perfect foresight with respect to the future behavior of alumni children.
26
Because our unit of observation is based on an entire year, there is some ambiguity in precisely when the
admissions decision becomes known. The year inwhich the child turns 18 years old is a sensible choice.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
272 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
were 26 years old or older. In addition, the joint hypothesis that the coefficients are
pairwise equal is rejected easily (the chi-squared testwith 18 degrees of freedom
is 260.32, which is associated with p
=
0.00). Interestingly, having a child rejected
lowers the probability of giving to the level of alumni who have no children. Indeed,
formost ages one cannot reject the hypothesis that the coefficients for individuals
without children and parents of rejected children are the same.
One concern with our interpretation of these findings is that the likelihood of giv
ing and the likelihood that a child applies toAnon U are both driven by some unob
served third variable, perhaps the extent towhich a parent feels an affinity toAnon
U. To investigate this possibility, we estimate amodel inwhich we allow the impact
of the child's age, at all ages, to depend on his or her eventual application status.
That is,CHILDfAppl andCHILD{NoAppl arealso includedforages 0 through14.If
our results are driven by affinity forAnon U, thenwe would expect the differences
between these variables at young ages to be as important as when the children are
teenagers. However, this is not the case. When the dependent variable is the prob
abilityofmaking a gift,thejointhypothesisthatthecoefficientson CHILD\Appl
and CHILDtNoAppl are equal for ages 0 through 13 cannot be rejected (p
=
0.1415).
These findings increase our confidence that the child-cycle results are not being
driven by the alumnus's unobservable affinity forAnon U. Further evidence along
these lines is presented in Section IV.
Another possible problem with our interpretation of Figure 1 is that reduced
giving after admissions might be driven by income effects associated with tuition
payments. If tuition were the important factor, then we would also expect to see
decreases in giving among alumni whose children did not apply to Anon U but
instead attended other institutions. As Figure 1demonstrates, however, these alumni
do not exhibit anything like the substantial decreases in giving thatwe see for the
alumni of accepted children.27
In short, the patterns inFigure 1 cannot be explained readily by the public good
provision, by a "warm glow" from giving, or by income effects associated with
tuition payments. In contrast, the results fitwell in the child-cycle framework.
Other variables.?The coefficients on the other variables in the basic model are
of some interest, because they allow us to see whether the determinants of giving
at Anon U are similar to those that have been found in previous studies of alumni
giving. Since they are not of central importance to documenting the existence of a
child cycle to alumni donations, however, we discuss them in theAppendix to this
paper. The Appendix shows that, taken together, our results are very much in line
with those from previous studies. While no school is "typical," Anon U appears not
tobe idiosyncratic with respect to the determinants of the donation decision. It is not
unreasonable to expect, therefore, that the child-cycle results would also generalize
to other selective institutions.
27
Another argument along the same lines is thatwhen one's child is accepted at another institution, new
opportunities for charitable giving open at that institution. Again, though, if thiswere the case, we would expect
thebehavior of theparents of rejected children and theparents of children who never applied tobe about the same,
in contrast to Figure 2.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. I MEER AND ROSEN: ALTRUISM AND THE CHILD CYCLE OF ALUMNI DONATIONS 273
3
-0.5-.?.-.-.-.-.-.-.-.-.-.?.-\
-1 -I
Age
Figure 2. Incremental Effect of the Child Cycle on the Amount Given
{LastChild)
Notes: This figure graphs the child-cycle coefficients from theTobit model of the amount of donations in any
given year, estimated using 487,913 observations. The vertical axis shows the incremental effect (relative to hav
ing no children) on the log of amount donated in a given year as a function of the firstchild's age and admissions
status. The dashed lines are 95 percent confidence intervals.
B. Amount ofGiving
Figure 2 graphs the child-cycle coefficients from the Tobit model of the amount
of donations. As noted above, the probit and Tobit coefficients are the same up to a
constant of proportionality, so thefigure does not provide any trulynew information.
The magnitudes are of some interest, however. In this context, it is important to note
that,because the dependent variable is the logarithm of giving, small coefficients are
approximately percentage changes. However, this approximation is not very good
for large coefficients, so caution is required in their interpretation.
A challenge to the child-cycle interpretation of Figure 2 is that total giving over
an alumnus's lifemay not be affected much by having an eligible child, just the
timing of donations. To investigate this possibility, we estimate a cross sectional
regression inwhich the left-hand-side variable is lifetime giving as of 2006,28 and
the right- hand-side variable includes the basic demographic variables in Table 1
augmented with a series of dichotomous variables for number of children and con
tinuous variables for the age of each child. We find that lifetime giving is 109 percent
higher for alumni with one child and an additional 58 percent higher with a second
28
Specifically, this is computed as the sum of giving in constant dollars over all years that the alumnus has
been in the sample.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
274 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
child. Further, lifetime giving increases by 5.1 percent for each year of the first
child's age and 2 percent for each year of the second child's age. In short, lifetime
giving is affected by the presence and age of children. The child cycle does not arise
simply because alumni are shifting donations over time.
C. The Role ofDirected Giving
Giving by parents of accepted children remains high after the admissions deci
sion has been rendered, a result that is not necessarily wholly consistent with the
child-cycle model. Perhaps the alumnus is showing tangible evidence of warm feel
ings engendered by the acceptance of his or her child.29 Without ruling out this
explanation, we note that another forcemay be operative. Certain donations made by
alumni with children on campus could be less public goods than relatively targeted
benefits for their progeny. An example is a donation earmarked for a child's varsity
team.
To explore this possibility, we estimate the probability that an alumnus makes
a directed gift in a particular year as a function of the variables in Table l.30We
find that, conditional on making any gift, alumni with 17-year-old children who
are applying toAnon U are 3.4 percentage points more likely tomake a directed
gift than alumni who have no children, and the difference is statistically signifi
cant. After admission, this figure increases substantially. Alumni with 18 year old
children who have been accepted are 6.1 percentage points more likely tomake a
directed gift,while those whose accepted children are 19-years-old are 12.8 percent
age points more likely tomake a directed gift. The impact of having a child accepted
at Anon U on directed giving peaks at 16.5 percentage points when the accepted
child is 21 years old. It remains statistically significant through age 25, but drops to
an insignificant ?0.22 percentage points for children aged 26 and older. In contrast,
conditional on giving, the parents of rejected children are not significantly more
likely tomake a directed giftwhen their children are of college age.31
In short, during the time an alumnus's child is on campus, the probability ofmak
ing a gift aimed at specific purposes, conditional on making a gift at all, is elevated.
This phenomenon might be due to the fact that prior to thematriculation of their
children, parents know little about the activities of certain campus organizations, or
even of their existence. But if that is the case, it is hard to explain why the relative
likelihood of directed giving drops after the child graduates. We conjecture that at
least part of the explanation is that the specific purposes directly benefit the child.
Therefore, elevated giving after the admission of one's child may be due in part to
nonaltruistic motivations.
29
Another possibility is that the alumnus is concerned about admissions prospects for younger children.
However, as shown below, the same tendency exists when we look at the child cycle for the last child in the
family.
30
A directed gift is defined as one notmade through the general annual appeal. It is possible that individu
als could support a particular student organization without going through the official channel of alumni giving.
However, in this case, the donation is unlikely to be tax deductible, hence, there are strong incentives to give
through the university.
31
The full set of results is available upon request.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. 1 MEER AND ROSEN: ALTRUISM AND THE CHILD CYCLE OF ALUMNI DONATIONS 275
D. Basic Model: Summary
The child-cycle pattern comes through clearly in our estimates. This is not to say
that altruism is unimportant?people without any children give substantial amounts
ofmoney, after all. For the top 1percent of all gifts, unconditional on class or year,
2,875 gifts came from alumni with children, while 2,003 came from alumni without
children. Among the top 1percent of lifetime (cumulative to 2006) givers, 212 had
children and 110 did not. However, it is hard to explain the patterns found inFigures
1 and 2 on the basis of altruism alone.
Ifwe are willing tomake some strong assumptions, we can be more precise about
the relative roles played by altruistic and selfish motivations. Specifically, suppose
that: 1)Giving by childless alumni is done without the expectation of receiving any
reciprocal benefit. Of course, other motivations, such as public recognition or donat
ing to research projects that could be useful to one's business, may also be present.
To the extent that they are, our estimate of the proportion of giving due to altruism
may be considered an upper bound. 2) The additional giving by alumni with chil
dren who do not ultimately apply toAnon U is unselfish as well. As conjectured
above, it is generated for one reason or another by warm feelings toward Anon U
thatare associated with having children.32 3) The additional giving by alumni whose
children do apply is motivated by self-interest. Under these assumptions, we can
use our estimates of the difference in giving associated with a child who did not
apply relative to a child who did apply to estimate the self-interested component
of giving. A complication arises because this differential depends on the child's
age. Seventeen years old seems a sensible choice because at this age the application
choice has generally been made, so that alumni whose children do not apply are not
making any precautionary donations. At the same time, these children have not yet
been accepted, so parents do not have an incentive tomake directed donations as
discussed in Section IIIC. For the same reason, neither can donations be influenced
by warm feelings due to the acceptance of one's child.
Under these assumptions, and using the estimated coefficients on CHILDl7Appl
and CHILDl7NoAppL we calculate that about 52 percent of giving by alumni whose
children apply toAnon U is due to altruism and the remaining 48 percent is due to
self-interest.33 To the best of our knowledge, this is the firstattempt to use observa
tional data to decompose donative behavior into altruistic and self-interested com
ponents, and it suggests that the two motivations are of about equal importance, at
least in this context.
32
One concern is thatgiving by alumni whose children do not apply may be motivated by thedesire to enhance
the prospects of younger children. However, as shown below, this portion of the child cycle for the last child in a
family is very similar to the child cycle for the firstchild.
33
The calculation is done as follows. We assume, without loss of generality, that giving associated with no
children (the baseline) is 1.We then exponentiate the coefficient on CHILDl7Appl, which gives us a figure of 6.06,
the amount donated by alumni whose 17-year-old children applied toAnon U, relative to the baseline. Next, we
exponentiate the coefficient on CHILDxlNoApl to obtain the relative amount given by those whose 17-year-old
children did not apply, which is 3.14. Under our assumptions, the proportion due to altruism has two components:
baseline altruism, and the increment associated with having children. The proportion of giving from baseline
altruism is 1/6.06 (= 0.165), while the proportion due towarm feelings associated with having children is (3.14
l)/6.06
(= 0.353), and the proportion associated with selfish reasons is (6.06
-
3.14)76.06 (= 0.482).
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
276 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
It isworth noting, however, that the self-interested component is likely to differ by
the type of charitable giving (Vesterlund 2008). A gift to one's local religious con
gregation entitles one to counseling and fellowship, while the rewards for donating
to international relief efforts are less tangible. Given thatwe cannot observe what
rewards a donation toAnon U garners, one might expect that our estimate of the
self-interested component of donations to universities is a lower bound.
IV. Alternative Specifications
In this section we present some alternative specifications of our model in order to
assess the robustness of the basic results.
A. Subsequent Children
So farwe have characterized the child cycle in terms of the first child in the fam
ily.A possibledrawbackis thatgivingalong thefirstchild'scyclecouldbe affected
by concerns about younger children's admissions prospects. Therefore, estimating
the cycle for the last child might allow a cleaner test of themodel.34 We began by
re-estimating our basic model using information on the last child rather than the
first,which, in effect, ignores any possible impact of older siblings. All alumni with
complete data, including those who only have one child, are included in this sample.
After deleting observations with missing data on the child's age (27,882 observa
tions), this sample contains 488,297 observations.
For brevity, we present only the graphical representation of the child cycle for
amounts given (see Figure 3). If anything, the child-cycle pattern is even stronger
than for the first child. Note that giving by those whose last child was rejected is
lower than by those with no children at all, and significantly so at ages 22- and
23-years-old. This is a sharper relative decrease thanwe observed for first children.
Within our framework, this suggests that parents of rejected first children still give
some amount with an eye toward enhancing their younger children's prospects. But
with the last child, thismotivation disappears.35
A natural question is whether the character of the child cycle is affected by a
previous child's outcome. Does acceptance of a first child toAnon U reinforce the
notion thatdonations generate a reciprocal benefit? One way to answer this question
would be to interact the child-cycle variables for the second child with indicators for
the first child's outcome. Estimating such amodel is infeasible, however, as itwould
involve hundreds of right-hand-side variables, many of which are all or nearly all
zeros. Instead, we augment the specification for the second child with three indicator
34
For families with only one child, the firstchild is also considered to be the last. Future children may not yet
be born, but it is reasonable to expect that at least for the children's age group that is our primary interest, 14years
and older, most families are unlikely to have another child. In addition, for some families, previous children may
be sufficiently old that alumni are concerned about the admissions prospects for grandchildren.
35
This raises the question of whether other family relationships might be associated with expectations of
reciprocal benefits to giving. For example, grandparents might make donations hoping to enhance the likelihood
of admissions for theirgrandchildren. Unfortunately, inour data set,we are only able to reliably link grandparents
and grandchildren when themembers of the intermediate generation also attended Anon U. This would leave us
with a small and very unrepresentative sample of grandparent-grandchild pairs.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. INO. 1 MEER ANDROSEN:ALTRUISMAND THECHILD CYCLEOF ALUMNIDONATIONS 277
Figure 3. Incremental Effect of the Child Cycle on the Amount Given
(LastChild)
Notes: This figure graphs the analogs to the child-cycle coefficients from theTobit model, using 482,760 obser
vations, when the basic model is re-estimated using the last child's age and admissions status rather than the
first child's. The vertical axis shows the incremental effect (relative to having no children) on the log of amount
donated in a given year as a function of the firstchild's age and admissions status. The dashed lines are 95 per
cent confidence intervals.
variables relating to the status of the first child: whether the first child was rejected
by Anon U, whether the first child was accepted, and whether the first child did not
apply toAnon U at all. Note that these variables are not constant over time. Their
values change as the first child grows older and his or her admissions status becomes
known.36
We find that the shape of the giving cycle associated with the second child is unaf
fected by the inclusion of these variables. However, if the firstchild isknown to have
been accepted, the entire child cycle shifts upward. The probability ofmaking a gift
increases by 11.6 percentage points. If the first child is rejected, the probability of
making a gift falls by 6.1 percentage points. (All of these figures are statistically sig
nificant.) If the firstchild did not apply, there is no significant change in the probabil
ityof making a gift. These results have a straightforward interpretation within the
child-cycle framework. Rejection of the first child indicates to the alumnus that his
or her expectations regarding reciprocity were to some extent incorrect, and giving
36
In particular, the variables for first child acceptance and rejection can take on a value of one only for those
years inwhich the firstchild is 18-years-old or older, while the indicator for firstchild nonapplication can be one
only for those years inwhich the first child is 14-years-old and older.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
278 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
behavior is adjusted accordingly. In the same way, acceptance of the firstchild rein
forces the perception that reciprocity is present.
As an additional test,we estimate the cycle for third children.37 The shape of the
cycle associated with the third child is very similar to that of both the firstand sec
ond children. The oldest child's rejection reduces the amount given by 14.5 percent
(imprecisely estimated), while the second child's rejection reduces the amount given
by a statistically significant 39 percent. The firstand second child's acceptances
increase the amount given by 134 percent and 118 percent, respectively, and both
figures are statistically significant. Again, information that reinforces or weakens
the perception of a reciprocal benefit affects giving.
B. Occupation and Field
As noted earlier, a drawback of our data set is lack of information on income or
wealth. However, for a subset of observations, we have detailed information on the
alumnus's occupational field and position. We know whether the individual ever
worked in a number of fields, including consulting, finance, information technol
ogy, health care, education, and so on. From the position data, we can classify the
alumnus as an executive, government worker, academic, attorney, physician, white
collar worker, or some other occupation. We believe that this information, together
with the variables inour basic model, do a reasonable job of proxying forpermanent
income. Using the field and position data reduces our sample size substantially, from
487,913 to344,342givingopportunities,which iswhywe didnot includethesevari
ables in our basic model.
To establish a baseline, we begin by estimating our basic model (that is, themodel
without the field and position variables) with the smaller sample. The child-cycle
graphs are virtually identical to their counterparts in Figures 1 and 2. When we
augment thismodel with the field and position variables, the estimated child cycle
is essentially unchanged. We present a graphical summary of the child cycle for the
amount given in Figure 4. The same tendencies thatwe saw in Figure 2 are clearly
present. Hence, the existence of a child cycle is not sensitive to the inclusion of a rich
set of variables relating to the alumnus's permanent income.38
C. Permanent Income and Fixed-Effects Estimation
Another approach to dealing with missing income data begins with the hypothesis
that giving depends on the alumnus's permanent income. If so, then a sensible alter
native is fixed-effects estimation, which controls for any attributes of an alumnus that
37
For this specification, there are 493,854 observations representing 32,817 alumni with 1,687 third children.
38
Although the child-cycle coefficients do not substantially change, some of the other coefficients do. For
example, inour basic model, being an economics major increases the amount of giving by about 85 percent. Once
we take occupation into account, however, this figure drops to 37 percent. In part, the coefficient in the basic
model reflects the fact thatAnon U's economics majors are particularly likely to go into the field of finance which,
by itself, increases the amount of giving by about 75 percent, ceteris paribus. Interestingly, the coefficients on the
race variables do not change substantially. For instance, the independent effect of being black is?56.8 percent in
the basic model and -59.4 percent when we augment themodel with occupation and field. Other race variables
are similarly unaffected. Detailed estimates for these models are available upon request.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER ANDROSEN:ALTRUISMAND THECHILD CYCLE OF ALUMNIDONATIONS 279
3
-1 J-1
Age
Figure 4. Incremental Effect of the Child Cycle on the Amount Given
(Controlling for Position and Field)
Notes: This figure graphs the analogs to the child-cycle coefficients from theTobit model, using 344,342 observa
tions, when the basic model is re-estimated including information on alumni's fields and positions. The vertical
axis shows the incremental effect (relative tohaving no children) on the log of amount donated ina given year as a
function of the firstchild's age and admissions status. The dashed lines are 95 percent confidence intervals.
do not change over time (or at least over the length of our sample period). Indeed,
a fixed-effects model takes into account any time-invariant unobservable variables
thatmight drive both giving behavior and the admissions status of an alumnus's
child, and hence confound the child-cycle interpretation of our findings. Such unob
servables include affinity toAnon U, generosity, quality of undergraduate experi
ence, and so on. Estimating fixed effects in nonlinear models is difficult at best, due
to the incidental parameters problem (JeffreyM. Wooldridge 2002, 484). Therefore,
we use ordinary least squares.
To establish a baseline, we first estimate the equations with OLS but without fixed
effects. The qualitative outcomes are very similar to those seen above. Next, we
include fixed effects. Figures 5 and 6 show the resulting child cycles for the prob
ability of making a gift and the amount given, respectively. If our results were in
fact being driven by permanent income or other time-invariant unobservables, then
the child cycle would be less pronounced than itwas previously, or perhaps disap
pear altogether. However, ifanything, the fixed-effects estimates aremore consistent
with the child-cycle framework. The probability that an alumnus with an accepted
22-year-old child makes a gift is 18.5 percentage points higher than for an alumnus
without children. But this differential trends downward to 8.5 percentage points by
the time the child reaches age 25. The incremental probability of an alumnus with
a rejectedchildmaking a giftdrops to?0.044 percentagepointswhen thechild is
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
280 AMERICANECONOMIC JOURNAL-ECONOMIC POLICY FEBRUARY2009
Figure 5. Incremental Effect of the Child Cycle on the Probability of Making a Gift
(Fixed-Effects Estimates)
Notes: This figure graphs the analogs to the child-cycle coefficients when the basic probit model is re-estimated
with OLS and fixed effects. The vertical axis shows the incremental effect (relative to having no children) on the
probability ofmaking a gift ina given year as a function of thefirst child's age and admissions status. The dashed
lines are 95 percent confidence intervals.
19yearsold and staysinsignificantlydifferentfromtheprobabilityfora childless
alumnus. The results for amount given are similarly pronounced. We conclude that
our results are not likely affected by time-invariant unobservable variables.
D. Augmented Sample
Our analysis sample is based on alumni who graduated between 1972 and 2005.
We have additional data on alumni who graduated before 1972. This sample, which
includes alumni from classes as early as 1914, is far larger, containing 939,671 giving
opportunities between 1983 and 2006. Ithas many more alumni children who are at
college age and beyond, a group that is essential to estimating the child cycle. Some
5,096 children applied since 1983 (41.9 percent of whom were accepted), compared
to 1,501 in our basic sample. The tradeoff is a less rich set of explanatory variables,
because we lack data on SAT scores, admissions rating, race, grade point average,
secondary school type, and honors for the pre-1972 classes.
When we estimate themodel using the augmented sample, the results are nearly
identical to those from the basic sample. Graphs of the coefficients are available
upon request. Thus, our main results continue to hold using a data set with much
more information on a critical group of alumni, those with children old enough to
have gone through the admissions process.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER ANDROSEN:ALTRUISMAND THE CHILD CYCLE OF ALUMNIDONATIONS 281
Figure 6. Incremental Effect of the Child Cycle on the Amount Given
(Fixed-Effects Estimates)
Notes: This figure graphs the analogs to the child-cycle coefficients when the basic Tobit model is re-estimated
with OLS and fixed effects. The vertical axis shows the incremental effect (relative to having no children) on the
logarithm of the amount of giving in a given year as a function of the first child's age and admissions status. The
dashed lines are 95 percent confidence intervals.
V. Conclusions
Our starting point is an old question in economics. To what extent does philan
thropy stem from altruism rather than the expectation of receiving some reciprocal
benefit? Research on this topic using observational data is rare because quantify
ing a reciprocal benefit is difficult. To address the problem, we analyze a unique
data set that allows us to estimate how alumni contributions to a university relate
to a perceived benefit?an improvement in the likelihood that their children will be
admitted.
We find that the presence of children increases an alumnus's giving, that giving
drops off after the admissions decision, and that the decline is far greater when
the child is rejected. In short, alumni giving varies systematically with the age and
admissions status of the alumni's children. This child cycle of alumni giving is con
sistentwith the hypothesis that some donations are made in the hope of a reciprocal
benefit. The result is robust to choice of estimation method and alternative specifica
tions of themodel, and does not appear to be due to unobservable variables such as
underlying affinity to the university.
Our results do not imply that self-interest is the only motivation behind donative
behavior. As we document in the text,many alumni with no apparent reason to
expect a reciprocal benefit, at least in terms of a higher admissions probability for
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
282 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
their children, are extraordinarily generous. In the context of the larger debate over
themotivations for altruism, our analysis shows that both selflessness and giving
with the hope of reciprocity are present.
We do not know whether these results generalize to other selective universities.
However, we are encouraged by the fact that other institutions seem quite similar
to the university we study with respect to other variables that affect giving. Hence,
behavior with respect to the child cycle might be similar as well. That said, itwould
be useful to investigate whether the child cycle ispresent at other selective schools.39
Similarly, itwould be informative to study the trajectory of alumni giving at non
selective schools. At such schools, the child cycle is not operative, and we would
expect to see a path of alumni giving, with respect to their children's age, that is less
steep, and that exhibits less of a falloff after the child's admissions decision.
Appendix: Other Covariates in the Basic Model
Our basic model includes variables that characterize the child cycle of giving as
well as a set of covariates to control forother alumni characteristics thatmight affect
their giving decisions. This Appendix reports and discusses the estimated impacts of
the latter set of variables on the probability ofmaking a gift.
The results are reported inTable Al. The coefficients on the linear and quadratic
terms for years since graduation imply that the probability of making a gift falls
for about the first20 years after graduation, and then turns upward.40 With respect
to gender, men are 4.6 percentage points less likely to donate in a given year, cet
eris paribus.41 Whites are more likely to contribute thanAmerican Indians, African
Americans, Hispanics, and Asians. The gap is largest with African Americans, who
are 16 percentage points less likely tomake a gift than are whites. These gender and
ethnic/racial differentials are similar to those reported in previous studies (Monks
1993)42 Alumni who attended boarding or private schools are somewhat more likely
to contribute than those who attended public schools. There is no discernible impact
of home or alternative schooling on the probability of giving relative topublic school
attendees.
As noted above, the admissions office produces summary evaluations of appli
cants on the basis of both academic and nonacademic criteria, such as musical talent,
athletic ability, volunteer work, and so on. An A is the highest score and an E is the
lowest score. Alumni who received the lowest nonacademic ratings at the time of
admissions are 6.9 percentage points less likely tomake donations. On the other
39
By definition, the child cycle can be operative only at selective schools.
40
The impact of years since graduation could embody several different effects, including changes in income,
changes in attitudes toward the institution as one ages, and the extent towhich solicitations of alumni vary over
time.
41
We also interacted the gender variable with the child-cycle variables to see if the child-cycle pattern is
different formen and women. The only branch of the child cycle forwhich there is a statistically significant dif
ference is for the accepted children. For this group, there is amuch steeper fall in giving after acceptance on the
part ofmale alumni than there is for female alumni. Further, the fall in giving after graduation ismuch steeper
formale than female alumni.
42
Some of these differentials may be due to the fact that income and wealth differ across ethnic groups. As
noted in the text,when we re-estimate themodel for a subsample of our data, which includes some reasonable
measures of permanent income, the differentials do not disappear.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER ANDROSEN:ALTRUISMAND THECHILD CYCLEOF ALUMNIDONATIONS 283
Table Al?Other Variables in the Basic Probit Modelu
Didgive Didgive
Yearssince
Yearssince2
Spouseisalum
Male
Race/ethnicity
Amerind
Black
Hispanic
Asian
Secondary schooling
Boarding
Private
Schloth
SATmath
SATverbal
Admissions office "nonacademic" ranking1
B
C
D
E
Admissions office "academic" ranking
B
C
D
E
Varsity
Club sport
Honors
Greek
GPA?Second quartile
GPA?Third quartile
GPA?Bottom quartile
-0.01455
(0.00085)
0.00034
(0.00002)
0.1401
(0.005968)
-0.04591
(0.00482)
-0.09796
(0.03579)
-0.1622
(0.00979)
-0.1112
(0.01146)
-0.07193
(0.00823)
0.01963
(0.00656)
0.01021
(0.00501)
-0.02547
(0.01764)
-0.00010
(0.00009)
0.000002
(0.00010)
0.00123
(0.01326)
0.00505
(0.01365)
-0.01293
(0.01563)
0.06919
(0.03214)
0.01503
(0.00724)
0.02461
(0.00939)
0.00710
(0.01231)
0.02640
(0.02365)
0.04912
(0.00494)
0.00860
(0.00598)
0.00105
(0.00628)
0.1285
(0.00501)
0.00313
(0.00662)
-0.02053
(0.00812)
-0.05714
(0.00946)
Major
Small social sciences
English
Economics
Public policy
Political science
Psychology
History
MAE
EE/CS
Arch & Civ
Small humanities
Small engineering
Small sciences
Minor
African/African American studies
American studies
Latin
Finance
Theater
Public policy
Other engineering
Other sciences
Other humanities
Teaching
Reunion
Post-baccalaureate education
PhD
Masters
J.D.
M.D./D.D.S.
M.B.A.
-0.03655
(0.01782)
-0.02557
(0.01426)
0.08084
(0.01408)
0.04885
(0.02069)
0.02286
(0.01428)
-0.02501
(0.01583)
0.03214
(0.01388)
0.06214
(0.01650)
0.06090
(0.01520)
0.06510
(0.01504)
-0.03285
(0.01411)
0.09559
(0.01666)
-0.00471
(0.01377)
-0.03727
(0.01555)
0.09132
(0.01337)
-0.00103
(0.04341)
0.09217
(0.02246)
-0.07022
(0.01876)
-0.00890
(0.01896)
0.02029
(0.01739)
0.00304
(0.01282)
-0.00678
(0.00936)
-0.00768
(0.01930)
0.06318
(0.00141)
0.04909
(0.00955)
0.08972
(0.00670)
0.1483
(0.00721)
0.1258
(0.00901)
0.1873
(0.00703)
a
This table shows the coefficients on variables other than those that characterize the child cycle. (The
child-cycle variables themselves are summarized graphically in Figure 1).Each figure shows the incremental
effect on theprobability ofmaking a gift ina given year, as estimated by a probitmodel. The model is estimated
using the basic sample of 487,913 observations. The figures inparentheses are standard errors. Figures in italics
are statistically significant at the 5 percent level. Standard errors are adjusted for clustering based on individuals.
In addition to the variables listed, the regression includes location effects, year effects and class effects, which
are suppressed forbrevity.
hThe nonacademic ranking isbased on attributes such as musical talent, athletic ability, volunteer work, etc.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
284 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
hand, students in the highest academic category are somewhat less likely tomake
donations than those with lower ratings. SAT scores do not appear to have any sta
tisticallysignificantimpacton theprobabilityofgiving.
We now turn from variables that are known before matriculation at Anon U to
those that reflect the alumnus's undergraduate experiences. Involvement in a varsity
sport increases the probability of giving by about 5 percentage points, and member
ship in one of Anon U's fraternities or sororities increases giving by 13 percentage
points.43 These results are consistent with previous findings that students who were
actively engaged in extracurricular activities as undergraduates are more likely to
make donations as alumni (Dugan et al. 1999). With respect to academic perfor
mance, receiving honors has no effect on the probability of giving. However, the
probability of giving increases with grade point average (GPA). Those in the bottom
quartile of theGPA distribution were 5.7 percentage points less likely tomake a gift,
while those in the third quartile were 2.1 percentage points less likely. There is no
significant difference in giving between the second and top quartiles.
Consistent with earlier studies, giving patterns differ substantially by course of
study (Dugan, et al. 1999,Monks 2003). Alumni who majored in engineering, eco
nomics,andpublicpolicyhave relativelyhighprobabilitiesofmaking a giftlaterin
life. Those who majored in the small social sciences (such as sociology) and small
humanities departments (such as linguistics) tend to have relatively low probabilities
ofmaking a gift later in life. Students with minors in finance aremore likely tomake
subsequent gifts (by about 9 percentage points), while those with minors in theater
are less likely (by about 7 percentage points).
Turning to schooling afterAnon U, alumni who continue their education aremore
likely tomake donations than those who do not, a finding consistent with previous
studies (Dugan et al., 1999, Monks 2003). Finally, we note that, consistent with
previous research (James H. Grant and David L. Lindauer 1986, Olsen, Smith, and
Wunnava 1989) the likelihood of giving increases substantially during reunion years,
with the probability increasing by 6.3 percentage points.
Taken together, our results are verymuch in line with those from previous studies.
While no school is "typical," Anon U appears not to be idiosyncratic with respect to
the determinants of the donation decision.
REFERENCES
Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff. 1992. "Is the Extended Family Altru
istically Linked? Direct Tests Using Micro Data." American Economic Review, 82(5): 1177?98.
Andreoni, James. 1993. "An Experimental Test of the Public-Goods Crowding-out Hypothesis."
American Economic Review, 83(5): 1317-27.
Andreoni, James, William T. Harbaugh, and Lise Vesterlund. 2008. "Altruism in Experiments." In
The New Palgrave Dictionary ofEconomics. 2nded., ed. StephenN. Durlauf and Lawrence E. Blume.
New York: Palgrave Macmillan. http://www.dictionaryofeconomics.com/article?id=pde2008_
A000240>doi:10.1057/9780230226203.0035.
Becker, Gary S. 1974. "A Theory of Social Interactions." Journal of Political Economy, 82(6): 1063-93.
43
Several organizations did not provide membership information for the class of 2001 and beyond. Interacting
indicators for those classes with the Greek indicator did not change the estimate of the main coefficient
substantially.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
VOL. 1NO. 1 MEER AND ROSEN:ALTRUISMAND THECHILD CYCLEOF ALUMNIDONATIONS 285
Bergstrom, Theodore C, and Oded Stark. 1993. "How Altruism Can Prevail in an Evolutionary
Environment." American Economic Review, 83(2): 149-55.
Bernheim, B. Douglas, Andrei Shleifer, and Lawrence H. Summers. 1985. "The Strategic Bequest
Motive." Journal of Political Economy, 93(6): 1045-76.
Bowen, William G., Martin A. Kurzweil, Eugene M. Tobin, and Susanne C. Pichler. 2005. Equity and
Excellence inAmerican Higher Education. Charlottesville, VA: University of Virginia Press.
Bristol, Ralph. 1991-1992. "How Much Will Alumni Give in the Future?" Planning for Higher Edu
cation, 20(2): 1-12.
Charness, Gary, and Ernan Haruvy. 2002. "Altruism, Equity, and Reciprocity in a Gift-Exchange
Experiment: An Encompassing Approach." Games and Economic Behavior, 40(2): 203-31.
Chronicle ofHigher Education Almanac. 2006. 53(1): 30. http:///www.aafrc.org/press_releases/index.
cfm?pg=trustreleases/tsunamigifts.html.
Cigno, Alessandro, and Furio C. Rosati. 2000. "Mutual Interest, Self-enforcing Constitutions and
Apparent Generosity." In The Economics of Reciprocity, Giving and Altruism, ed. L. A. Gerard
Varet, S. C. Kolm, and J.Mercier Ythier, 226-47. London: Macmillan Press.
Clotfelter, Charles T. 1985. Federal Tax Policy and Charitable Giving. Chicago: University of Chicago.
Clotfelter, Charles T. 2003. "Alumni Giving to Elite Private Colleges and Universities." Economics
of Education Review, 22(2): 109-20.
Cox, James C. 2004. "How to Identify Trust and Reciprocity." Games and Economic Behavior, 46(2):
260-81.
Cunningham, Brendan M., and Carlena K. Cochi-Ficano. 2002. "The Determinants of Donative
Revenue Flows from Alumni of Higher Education: An Empirical Inquiry." Journal of Human
Resources, 37(3): 540-69.
Docquier, Frederic, and Hillel Rapoport. 2000. "Strategic and Altruistic Remittances." In The Eco
nomics of Reciprocity, Giving and Altruism, ed. L. A. Gerard-Varet, S. C Kolm, and J.Mercier
Ythier, 285-97. London: Macmillan Press.
Dugan, Kelly, Charles H. Mullin, and John J. Siegfried. 2000. "Undergraduate Financial Aid and
Subsequent Alumni Giving Behavior." Vanderbilt University Working Paper No. 00-W40.
Ehrenberg, Ronald G., and Christopher L. Smith. 2003. "The Sources and Uses of Annual Giving
at Selective Private Research Universities and Liberal Arts Colleges." Economics of Education
Review, 22(3): 223-35.
Gerard-Varet, L. A., S. C. Kolm, and J.Mercier Ythier, eds. 2000. The Economics of Reciprocity,
Giving and Altruism. London: Macmillan Press.
Giving USA Foundation. 2005. "CharitableGiving Rises 5 Percent toNearly $250 Billion in2004."
http://www.aafrc.org/press_releases/index.cfm?pg=trustreleases/tsunamigifts.html.
Grant, James H., and David L. Lindauer. 1986. "The Economics of Charity: Life-Cycle Patterns of
Alumnae Contributions." Eastern Economic Journal, 12(2): 129-41.
Harrison, William B., Shannon K. Mitchell, and Steven P. Peterson. 1995. "Alumni Donations and
Colleges' Development Expenditures: Does Spending Matter?" American Journal of Economics
and Sociology, 54(4): 397-412.
Ioannides, Yannis M., and Kamhon Kan. 2000. "The Nature of Two-Directional Intergenerational
Transfers ofMoney and Time: An Empirical Analysis." In The Economics of Reciprocity, Giving
and Altruism, ed. L. A. Gerard-Varet, S. C. Kolm, and J.Mercier Ythier, 314-31. London: Mac
millan Press.
Kolm, Serge-Christophe. "The Theory of Reciprocity." In The Economics of Reciprocity, Giving and
Altruism, eds. L. A. Gerard-Varet, S. C. Kolm, and J.Mercier Ythier, 314-31. London: Macmillan
Press.
Landes, William M., and Richard A. Posner. 1978. "Salvors, Finders, Good Samaritans and Other
Rescuers: An Economic Study of Law and Altruism." Journal of Legal Studies, 7(1): 83-128.
List, John A. 2007. "On the Interpretation of Giving inDictator Games." Journal of Political Econ
omy, 115(3):482-493.
Marr, Kelly A., Charles H. Mullin, and John J. Siegfried. "Undergraduate Financial Aid and Subse
quent Alumni Giving Behavior." Quarterly Review of Economics and Finance, 45(1): 123-43.
McGarry, Kathleen, and Robert F. Schoeni. 1995. "Transfer Behavior in the Health and Retirement
Study: Measurement and the Redistribution of Resources within the Family." Journal of Human
Resources, 30: SI84-226.
Monks, James. 2003. "Patterns of Giving toOne's Alma Mater among Young Graduates from Selec
tive Institutions." Economics of Education Review, 22(2): 121-30.
Odendahl, Teresa. 1990. Charity Begins at Home: Generosity and Self-interest among the Philan
thropic Elite. New York: Basic Books.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions
286 AMERICANECONOMIC JOURNAL:ECONOMIC POLICY FEBRUARY2009
Olsen, Katherine, Amy L. Smith, and Phanindra V. Wunnava. 1989. "An Empirical Study of the Life
Cycle Hypothesis with Respect toAlumni Donations." American Economist, 33(2): 60-63.
Oster, Sharon M. 2003. "Is There a Dark Side to Endowment Growth?" New Directions for Institu
tional Research, 119: 81-90.
Raut, Lakshmi K., and Lien H. Tran. 2000 "Reciprocity with Two-Sided Altruism in Intergenera
tional Transfers: Evidence from Indonesian Family Life Survey Data." In The Economics of Reci
procity, Giving and Altruism, ed. L. A. Gerard-Varet, S. C. Kolm, and J.Mercier Ythier, 314-31.
London: Macmillan Press.
Rosenstock v. Board of Governors of the University of North Carolina, 423 F. Supp. 1321.
Rotemberg, Julio J. 2003. "Commercial Policy with Altruistic Voters." Journal of Political Economy,
111(1): 174-201.
Samuelson, Paul A. 1993. "Altruism as a Problem Involving Group Versus Individual Selection in
Economics and Biology." American Economic Review, 83(2): 143-48.
Schokkaert, Erik, and Luc Van Ootegem. 2000. "Preference Variation and Private Donations." In The
Economics of Reciprocity, Giving and Altruism, ed. L. A. Gerard-Varet, S. C. Kolm, and J.Mercier
Ythier, 78-95. London: Macmillan Press.
Sen, Amartya K. 1977. "Rational Fools: A Critique of the Behavioral Foundations of Economic The
ory."Philosophy and Public Affairs,6(4): 317-44.
Shulman, James L., and William G. Bowen. 2001. The Game of Life: College Sports and Educational
Values. Princeton, NJ: Princeton University Press.
Taylor, Alton L., and Joseph C. Martin. 1995. "Characteristics of Alumni Donors and Nondonors at
a Research I, Public University." Research inHigher Education, 36(3): 283-302.
Vesterlund, Lise. 2006. "Why Do People Give?" In The Nonprofit Sector: A Research Handbook. 2nd
ed., Walter W. Powell and Richard Steinberg, 568-90. New Haven, CT: Yale University Press.
Weisbrod, Burton A. 1978. "The Forgotten Economic Sector: Private but Non-Profit." Challenge,
21(4): 32-36.
Willemain, Thomas R., Anil Goyal, Mark Van Deven, and Inderpreet S. Thukral. 1994. "Alumni
Giving: The Influences of Reunion, Class, and Year." Research in Higher Education, 35(50),
609-29.
Wooldridge, Jeffrey M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: MIT Press.
Wunnava, Phanindra V., and Michael A. Lauze. 2001. "Alumni Giving at a Small Liberal Arts Col
lege: Evidence from Consistent and Occasional Donors." Economics of Education Review, 20(6):
533-43.
Yoo, Jang H., and William B. Harrison. "Altruism in The "Market" For Giving and Receiving: A
Case of Higher Education." Economics of Education Review, 8(4): 367-76.
This content downloaded from 147.251.189.14 on Tue, 18 Aug 2015 16:54:42 UTC
All use subject to JSTOR Terms and Conditions