Full text of DOI:10.1002/1520-6688(200022)19:3<451::AID-PAM5>3.0.CO;2-E

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 451
Public Subsidies and
Arthur C. Brooks
Charitable Giving: Crowding
out, Crowding in, or Both?
Abstract
Whether government subsidies to nonprofit organizations leverage (crowd in) pri-
vate donations, or rather crowd them out has been actively debated for some time. A
third hypothesis, explored theoretically and tested empirically in this paper, is that
the two phenomena are actually not inconsistent with one another: At low levels of
subsidies, government support may stimulate private giving, whereas at high levels
it could have just the opposite effect. The model presented is based on this idea,
which yields implications relevant to nonprofit management and public policy, and
tests it with data on symphony orchestras. The conclusion is that the maximization
of private donations and total “unearned” revenues are not compatible goals. Fur-
ther, nonprofits that suffer from short-term liquidity problems or managerial short-
sightedness may face a “subsidy trap,” in which they are forced to rely on subopti-
mal levels of subsidies in terms of maximizing the firm’s revenues. © 2000 by the
Association for Public Policy Analysis and Management.
INTRODUCTION
Considerable energy in the nonprofit and public policy communities is spent trying
to understand the relationship between public subsidies and private philanthropy:
Are they substitutes or complements? Or, in the language of nonprofit economics, is
there “crowding out,” or “crowding in”? This question is an important one for nonprofit
organizations and public agencies that support them. If public support is a substitute
for private donations, managers and public officials should know the mechanism by
which one revenue source displaces the other, such that fund-raising investment and
public funding programs are not designed and implemented in a self-defeating way.
On the other hand, if they are complements, the leveraging effect is valuable to
understand so it can be exploited optimally.
This paper departs from the existing literature on this question: It creates and tests
a model that suggests that crowding out and crowding in are not incompatible
phenomena. In so doing, it arrives at several new implications for public and private
nonprofit policy and management. First, it predicts that either crowding in or crowding
out might occur, and that a nonprofit might even alternate between the two. Second,
it suggests that the goals of the maximization of “unearned” revenues and private
Manuscript received July 1998; rejected after review November 1998; manuscript resubmitted December 1998; revise
and resubmit recommended March 1999; revision received April 1999; sent out for review May 1999; paper accepted
November 1999.
Journal of Policy Analysis and Management, Vol. 19, No. 3, 451–464 (2000)
© 2000 by the Association for Public Policy Analysis and Management
Published by John Wiley & Sons, Inc.

452 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
donations cannot be pursued simultaneously. Third, it exposes a “subsidy trap,” in
which certain nonprofit firms, due to illiquidity or managerial short-sightedness, get
“stuck” year after year relying on levels of public subsidies that correspond to total
revenues that are below the highest possible level. It also suggests ways that nonprofits
can (and cannot) escape this trap.
THE ISSUE
In the continuing polemic over the role of government funds in the finances of
nonprofit organizations, it has become common to hear that government support
provides “seed” money to nonprofit firms (National Endowment for the Arts, 1998).
Particularly at low levels of support, it is argued, public funds can get the ball rolling
for the private donations that make up much of most nonprofits’ revenues. In contrast
to this argument, a case in the public economics literature has been made that public
subsidies will not stimulate private giving to nonprofits; indeed, they may even crowd
this giving out. This falls within the family of arguments about the crowding out of
private expenditures by government provision of goods and services (Friedman, 1962;
Feldstein, 1974; West, 1975).
Theory relevant to this subject began with the result from standard neoclassical
price theory that the only donations not displaced would be those from donors with
a very high demand for the nonprofit’s product, but who were unable to manipulate
subsidy levels (Hochman and Rodgers, 1977). That is, the crowding-out phenomenon
should almost certainly be the one in effect. Subsequent theory and experimental
work has generally been consistent with this, predicting that crowding out should be
complete (Bergstrom, Blume, and Varian, 1986; Warr, 1982), or at least partial
(Andreoni, 1993; Duncan, 1999). Relaxing the assumptions of the neoclassical model,
however, a few theoretical papers have predicted that, on the contrary, there might be
crowding in instead (Rose-Ackerman, 1986; Seaman, 1980). One area of agreement
among all the theoretical treatments is that the question must be settled empirically.
The empirical nonprofit literature has produced no unambiguous answer on this
score. A number of studies have found evidence of crowding out (e.g., Payne, 1998;
Kingma, 1989; Steinberg, 1985). Others, however, have found no significant
relationship between the revenue sources (e.g., Brooks, 1999; Khanna, Posnett, and
Sandler, 1995), or even some slight leveraging effect (e.g., Schiff, 1985).
This paper suggests that these seemingly contradictory findings may be reconcilable,
beyond just attributing them to structural differences between firms within the nonprofit
sector. The nature of the relationship between public subsidies and private giving to
nonprofit organizations may depend on their magnitude relative to one another.
DO PUBLIC SUBSIDIES LEVERAGE OR CROWD OUT PRIVATE GIVING? MAYBE BOTH
While the literature has traditionally tested the question of whether leveraging or
crowding out occurs, a compelling argument can be made that it might be the case
that the two effects are not mutually exclusive. Imagine a typical nonprofit firm, one
that can receive “unearned” revenues from both governmental (G) and private sector
(P) sources.1 Assume that private donors react in their giving to government subsidies
as well as to a number of other variables (X). Also assume that if the firm were to
receive no funding from the government, its unearned revenues would be P0>0 dollars
of private support.
1 While not satisfying from an economic standpoint for the obvious reasons, the term “unearned revenue”
is common parlance in the nonprofit sector; for this reason, it is used here to denote contributions that are
not direct compensation for the sale of a product.

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 453
As government subsidies become positive, one can see how the leveraging argument
might be the relevant one. Low levels of public funds may be effective in “priming the
pump” for private donations. This idea would seem consistent with the finding that
smaller organizations—those presumably with relatively undeveloped philanthropic
bases—see a large return to their (small) fund-raising initiatives, such as matching
funds campaigns (Brooks, 1997). In addition, the organization functioning with no
government support might be entirely unknown or just starting out, and as such might
benefit from the publicity and credibility, however small, associated with a grant.
When government funds increase beyond a certain point (call it G*>0) for this
firm, the crowding-out arguments begin to sound more plausible. Organizations that
are largely funded by the government begin to appear as quasi-public agencies, and
just as few citizens make private contributions (voluntarily) to their governments,
fewer and fewer people may be inclined to give to these organizations. In addition,
heavily subsidized firms often suffer from the image of requiring public funds to stay
afloat, and thus might inspire little confidence among private contributors. The top
half of Figure 1 graphically depicts this model, as described so far.
The strict concavity of the curve drawn in Figure 1 may at first seem a strong
assumption to make. However, it actually allows the fewest possible additional
assumptions to be made on the relationship between private and public funding.
Specifically, in the absence of a compelling argument for doing so, no acceleration of
the leveraging effect before the critical point is assumed, nor is any deceleration of
the crowding-out effect beyond this point. Thus, strict concavity is assumed to maintain
the greatest possible generality.
The relationship in the top of Figure 1 can be expressed mathematically as follows.
P = P(G,X)
(1)
n
where P ∈ ℜ , X ∈ℜ+ , and G ∈[0,G ], where Gmax ∈[G*,∞) is the point at which the
organization is entirely publicly fmuanx ded and all private gifts are crowded out. In
addition, P′(G*,X)=0 and P″(G,X)<0 for all values of G>0, by the strict concavity
assumed above. This indicates the presence of a unique global maximum level of P,
P*, corresponding to the level of public subsidies G*.
+
This is an important assumption: There is no simultaneity between P and G. While
G influences P, the reverse does not occur. But this assumption has empirical
justifications, as will be shown in the next section.
The top of Figure 1 suggests a natural extension of the model from the relationship
between G and P to that between G and total unearned revenues (TR). Defining
TR = P(G,X) + G
(2)
we can see that TR is a concave function of subsidies, since TR″=P″<0. Further,
maximizing TR with respect to G yields TR′=P′(G)+1=0, so
P′(G**) = –1
(3)
From the concavity of P, we can conclude that G**>G*. Note that at zero public
subsidies, TR=P0 since all unearned revenue comes from private donations. Similarly,
at the maximum level of public subsidies (where the firm has become entirely public),
TR=Gmax. This is depicted in the lower half of Figure 1.

454 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
Figure 1. The relationship between total unearned revenues and public subsidies.
There is a point beyond which total unearned revenues actually fall when public
subsidies increase. This relationship has two implications that will motivate the
discussion in the following sections. First, Figure 1 suggests that there is a range of
public subsidies that is too high from the standpoint of maximizing unearned revenues.
Specifically, any level above G** will drive TR below its maximum of TR**. Second, if
TR were at its maximum at TR**, there would tend to be crowding out between
public subsidies and private donations. In Figure 1, this is seen in the fact that G** is
to the right of G*, and consequently sits on the downward-sloping portion of the P(G)
curve: As subsidies rise, donations fall.
In an ideal world, nonprofits would optimize by accepting only G** in subsidies
while attracting P** in donations. In reality, however, charities may not be able to
attract or control the right amount of subsidies, or may not understand the underlying
relationship at work. As a result, firms may sit on any part of the P(G) curve. Between
G=0 and G*, the firms experience crowding in; between G* and Gmax, crowding out
occurs. These two possibilities may explain why the empirical studies on this issue
have seen such mixed results when measuring the phenomenon linearly. Indeed, the

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 455
empirical section of this paper will reveal that both of these firm types are probably
present among symphony orchestras.
EMPIRICAL EVIDENCE AND INTERPRETATION
Do the existing data on the nonprofit sector support the hypothesis of nonlinear
crowding out? In general, this is a difficult question to answer because data are limited.
Most specific nonprofit industries have very little variance in their data with respect
to the level of public funding. (For example, it is rare that within a particular industry,
one firm receives almost no public funding while another receives most of its revenues
from the government.)
One exception to this lack of variance is the case of the arts nonprofit sector, owing
to the fact that arts organizations tend to be organized and financed in a variety of
ways, from volunteer ensembles to semiprofessional groups to full-time professional
organizations. This variety translates into a wide variance in private philanthropy
and public subvention, allowing testing of the hypothesis in this paper.
The data used to estimate the model represent 253 American symphony orchestras
over eight concert seasons from 1984 through 1991. These data were collected by the
American Symphony Orchestra League (ASOL) and aggregated into averages over
groupings of orchestras of similar size in the Wolf Organization’s The Financial
Condition of Symphony Orchestras (1992). The groupings are as follows:
•
•
•
•
•
19 orchestras with 1991 budgets of $8.5 million to $38.7 million
22 orchestras with 1991 budgets of $5 million to $8.5 million
48 orchestras with 1991 budgets of $1.8 million to $5 million
68 orchestras with 1991 budgets of $.63 million to $1.8 million
96 orchestras with 1991 budgets of up to $630,000
Hence, the dataset consists of 40 observations: one observation for each of the five
groups over eight years.2
The variance in these data can be seen in Table 1, which contains the descriptive
statistics for the levels of government support, private giving, earned income, and the
average number of concerts performed per year.
Table 1. Descriptive statistics on the orchestra data.
Government
support
Private
support
Earned
income
Concerts
per year
Mean
$537,251
$262,351
$558,421
$1,842,618
$17,874
$1,598,774
$799,434
$2,034,711
$783,501
127
150
70
Median
Standard deviation
Range
$1,808,347
$5,396,768
$61,492
$2,683,876
$7,877,434
$38,373
211
14
Minimum
Maximum
$1,860,492
$5,458,260
$7,915,807
225
2 Obviously, data on the individual orchestras would be better than the averages. However, this is the only
publicly available version of this proprietary dataset. And in spite of the limitation this presents, it will be
shown that it is still useful.

456 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
The literature cited in the introduction on the crowding-out debate in general has
tested empirical models of crowding out in a variety of ways (Steinberg, 1993, p.
103). The general idea in all of them is to regress a measure of private philanthropy
on a measure of government support (and other relevant variables), and to do so in
such a way that endogeneity between philanthropy and government subsidies does
not bias the results unduly. Means to achieve this have included the use of instrumental
variables (Brooks, 1999; Kingma, 1989), seemingly unrelated regression (Lindsey and
Steinberg, 1990), and the assumption of an information lag in government support,
in the case that this is theoretically supportable (Brooks, 2000). The last assumption
is used here, as current-year public sector commitments to orchestras and other arts
organizations nearly always represent government funding decisions made in a prior
period (meaning that public support paid this year can affect private philanthropy,
but not vice versa).
Models will be fitted to these panel data, which have been adjusted for the divergent
scale of operations of the different-sized orchestras by dividing each observation
through by the number of concerts presented per year by the mean orchestra in the
group. All dollar-denominated variables are in constant 1996 dollars. Consistent with
the literature already cited, the variables included in the regressions are:
•
•
•
•
PRIVATEit = average private donations for orchestras in group i in year t, divided
by the average number of concerts performed.
PUBLICit= average government subsidies for orchestras in group i in year t,
divided by the average number of concerts performed.
DEVELOPMENTit = average fundraising expenditures for orchestras in group i
in year t, divided by the average number of concerts performed.
EARNEDit = average earned income for orchestras in group i in year t, divided
by the average number of concerts performed.
•
•
GROUP i (where i =1,2,3,4) = dummies for four of the five groups.3
TIME = trend variable.
In addition, to test the central hypothesis, a squared term for government support is
added:
•
PUBLIC-SQit = squared average government subsidies (divided by the average
number of concerts) for orchestras in group i, in year t.
The models estimated regress this period’s philanthropy on a lagged value of itself
(to capture “habitual” behavior in giving), this period’s earned revenues and
development expenditures (to net out the effects of the orchestra’s concert income
and “advertising” expenses for donations), the time trend, and the group dummies.4
Support of the hypothesis that leveraging is followed by crowding out of philanthropy
as government support increases would be seen in the significance, sign, and
magnitudes of the regression coefficients for PUBLIC and PUBLIC-SQ. Specifically,
the hypothesis is supported if the coefficient on PUBLIC is positive and significant,
the coefficient on PUBLIC-SQ is negative and significant, and the resulting knee in
the curve occurs reasonably close to the range of the data on PRIVATE.
Two models are fitted to the data. First, the coefficients are estimated with ordinary
least squares (OLS). Perhaps not surprisingly, diagnostics indicate problems with
3 The effect for the fifth group is caught in the regression’s constant term.
4 Lagged values of PUBLIC and PUBLIC-SQ were not found to be either significant or to contribute mate-
rially to the explanatory power of the model. They are thus omitted to conserve scarce degrees of freedom.

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 457
Table 2. Regression estimates of the nonlinear crowding-out model.
Independent variable
Coefficients
OLS Model
GLS Model
PUBLICt
2.0286 (0.8012)**
2.4956 (0.2852)***
-0.00015 (0.00002)***
-0.0357 (0.0710)
3.0353 (0.6113)***
-0.1500 (0.0985)
4957.0 (1528.0)***
-700.19 (799.3)
PUBLIC-SQt
PRIVATEt-1
DEVELOPMENTt
EARNEDt
GROUP 1
GROUP 2
GROUP 3
GROUP 4
TIME
-0.00013 (0.00005)**
-0.11034 (0.1182)
2.0774 (0.9780)**
0.12007 (0.2081)
3497.0 (3303.0)
-238.67 (1774.0)
-803.42 (424.9)*
-497.32 (508.5)
138.02 (119.5)
-572.51 (233.5)***
-125.14 (219.8)
198.10 (36.87)***
-56.369 (562.7)
CONSTANT
825.86 (1386.0)
Adjusted R2 = .99
Log of likelihood
function=-270.57
BUSE R2 = .99
Log of likelihood
function=-250.85
Dependent variable: PRIVATEt
N=35 (5 panels, 7 periods each)
Notes: Standard errors in parentheses.
*indicates statistic is significant at α=.10.
** indicates statistic is significant at α=.05.
*** indicates statistic is significant at α=.01.
ρ in the GLS model was estimated using Parks’ pooling procedure for panel data.
The constant term captures the effect of the fifth group of orchestras.
heteroskedasticity and autocorrelation within the panels. A method designed to deal
with these specific problems suggested by Parks (1967) is appropriate for these data
and is used to re-estimate the coefficients.5 The resulting generalized least squares
(GLS) regression effectively neutralizes the problems with heteroskedasticity and
autocorrelation. The results of both the OLS and GLS regressions are presented in
Table 2.
The extraordinarily high R-squared measures do not reflect multicollinearity, but
rather an overall lack of variance in the sample resulting from the fact that each
observation is itself an average across the “raw” data.
The two models are quite consistent with one another, and in terms of the hypothesis
being tested, the results are unambiguous. Both this period’s public funds term and
its square are significant at high levels, and they have the predicted signs. Furthermore,
the knee in the curve they imply is within the range of the data used. In other words,
the empirical tests strongly support the nonlinearity hypothesis set forth earlier. Other
significant variables are development expenditures, two of the fixed effects, and the
time trend. The finding that development expenditures positively affect private giving
is not surprising and is consistent with other work done on symphony orchestras
(Brooks, 1997). The significant positive trend indicates autonomous increases in
5 This method employs a set of assumptions on the disturbance covariance matrix that is cross-sectionally
heteroskedastic and time-wise autoregressive. For more details, see Kmenta (1986, pp. 616–625).

458 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
philanthropy over time. This last finding is interesting in that it is not consistent with
some recent National Endowment of the Arts (NEA) research (Cobb, 1995) that claims
a negative trend in private giving to the arts.
The relationship between government support and private philanthropy measured
in the GLS model is charted in Figure 2. Here, the middle group of orchestras is
selected (48 orchestras), and all values besides public subsidies and private giving are
held at their means. According to this model, the average group 3 orchestra will,
ceteris paribus, tend to experience a leveraging effect up to about $8,200 in government
support per concert; after this point, private support will begin to be crowded out.
The relation between Figure 2 and the actual data for each orchestra group is
summarized in Figure 3. The figure emphasizes that Figure 2 is not an unreasonable
extrapolation from the data. In addition to the fact that the curve does not use values
far outside the range of data, we see that both crowding out (group 1) and crowding
in (groups 2–5) are apparent within the orchestra groups. To demonstrate this, the
data in each panel are compared to their values fitted to PUBLIC and the mean values
of the other significant variables in Table 2.
It should be noted that the empirical treatment here is not intended to represent
the only model of the hypothesized relationship. The approach here was employed
primarily to balance simplicity with strength of results. The purpose was to test the
central hypothesis, which these models do in a transparent, convincing manner.
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
3,000
6,000
9,000
12,000
15,000
Figure 2. The observed relationship between private donations and public subsidies
per concert (fitted values for group 3 orchestras).

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 459
Public
Public
Public
Public
Public
Figure 3. Group 1 orchestras see crowding out, while the other groups experience
crowding in.
Note: Individual observations are indicated with dots, while the lines give their predicted values. Pre-
dicted values are made from regressions within individual groups using the mean values of the significant
variables in Table 2.
THE SHORT-RUN TRADEOFF AND A “SUBSIDY TRAP”
Having established a convincing case for the hypothesis of nonlinear crowding out,
issues of the short- vs. long-run may now be examined. When time and liquidity are
considered in the nonprofit firm’s (or the government’s) problem, an entirely different
conclusion may emerge, as a result of the fact that private donors’ reactions to a
change in the level of subsidies may take time to materialize completely. To show
this, the idea of a short-run version of the curve in Figure 1 is introduced, and depicted
in Figure 4.
Curve SR is a hypothesized short-run tradeoff curve; LR is the long-run curve in
Figure 1. SR is the curve that firms see within a given period, after public subsidies
have been received, but before private donors have sufficient time to react fully to
this level of subsidies. SR intersects LR at the SR maximum because LR is made up
of all of the “best” short-run possibilities. The crowding out reaction of donors being
muted before the passage of time, SR will lie outside of LR on either side of its
maximum.

460 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
Figure 4. A short-run tradeoff leads to a long-run trap.
Testing the proposition that SR has the shape it does—notably, that it lies off the
LR curve at all points besides its maximum—is more difficult than testing the
hypothesis in the last section. This is due to the fact that data in most of the nonprofit
sector are simply not collected frequently enough to gauge very short-run effects of
government spending on private giving. However, a tentative case for this hypothesis
can still be made using the symphony orchestra data. This can be done by comparing
private support and successive lags in public support. Support of the proposition
here would be seen in a finding that the correlation between (the absolute value of)
changes in public subsidies and private giving rises somewhat after the initial period,
as reaction to subsidies has time to fully materialize (and then later drops off, as
would be expected). This is far from perfect evidence of the phenomenon in question,
but could be considered suggestive nonetheless.
Table 3 gives the correlation matrix for changes in private giving this period and in
public subsidies from this period as well as two prior periods. The first column reflects
the pattern in correlations consistent with the short-run tradeoff proposition.
Given this relationship, why would a nonprofit organization or government treat
SR as the relevant curve, as opposed to LR?6 There are two possible reasons: First,
the firm could have short-term liquidity problems such that moving up in total revenue
over the long run would mean an unaffordable short-run revenue sacrifice. In Figure
6 Note that SR is only defined as the curve drawn in Figure 4 in the event that the firm starts at subsidy level
G***; otherwise, it would be facing a different short-run tradeoff curve corresponding to a different start-
ing subsidy level.

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 461
Table 3. Correlations between the absolute value of changes in private giving and
government support to orchestras.
∆P
| |
∆G t
∆G t-1
∆G t-2
t
|
|
|
|
|
|
∆P
| |
t
1
0.01
1
0.61
0.37
1
0.26
0.55
0.37
1
∆G t
|
|
|
|
0.01
0.61
0.26
∆G t-1
|
0.37
0.55
∆G t-2
|
0.37
4, note that movement from G*** (the short-run maximum) down to G** would lead
to increased total unearned revenue over time, but lower revenues in the present
period. If the firm were insufficiently liquid to absorb the short-run loss, it would be
precluded from moving up LR to higher revenues and remain “stuck” below TR**,
period after period.
Second, managerial and bureaucratic short-sightedness could play a role in keeping
public subsidies at a suboptimally high level. If a decisionmaker’s personal time horizon
did not extend to the period corresponding to LR, there would be no reason to make
the short-term revenue sacrifice it represented. Similarly, it could be mistakenly
believed that SR represented the only set of tradeoffs to be had.
SUMMARY AND IMPLICATIONS FOR POLICY AND MANAGEMENT
The theoretical and empirical literature on the crowding-out question has until now
assumed linearity in the relationship between the variables of interest. A puzzle that
has emerged in this literature is the inconsistency of findings: In some estimations,
government subsidies crowd-in private giving, while in others, donations are crowded
out. This paper argues that relaxing the linearity assumption may resolve this puzzle.
At low levels of government funding, philanthropy might be encouraged, but beyond
a certain point crowding out begins.
Two important conclusions follow from the analysis in this paper. The first involves
the incompatibility of a total unearned revenue objective with a private donations
objective. The second concerns the existence of the short-run subsidy trap.
The analysis here shows that, under the initial assumptions (which are supported
by empirical testing) about the behavior of private donations at different levels of
public support for nonprofit firms, the level of subsidies that maximizes donations
cannot be coincident with the level of subsidies that maximizes total unearned
revenues. To be more specific, the former level will be lower than the latter. As a
result of this, if total revenues are at their maximum, the firm will experience crowding
out of donations by subsidies.
It is common in policy problems that objectives be in competition, of course
(although it is a point often forgotten). However, it is notable when two common
managerial objectives cannot be attained even coincidentally. Development directors
of nonprofits, who often consider both of the measures discussed here to be targets
(however difficult to attain, since government subsidies cannot be directly
manipulated), might take special note of this.

462 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
This finding has potentially interesting social welfare implications as well. A
“benevolent social planner” might choose the total revenue-maximizing level of
subsidies to maximize the production of the goods and services produced by the
organization, which are arguably generally quasi-public. On the other hand, the
planner might prefer the level of subsidies that maximizes private giving, as this
effectively devolves decisions about funding these quasi-public goods to individuals
and hence is arguably more democratic.
The possibility of a short-run subsidy trap, suggested by the discrepancy between
the long-run tradeoff and its short-run corollary, has strong policy and management
implications. As has been shown, trouble with liquidity or a misunderstanding of the
subsidy–donation tradeoff could lead to the consumption of a suboptimally high level
of subsidies by some firms in the long run, which in turn could trap them at levels of
total unearned revenues below the maximum. Thus, the inability or unwillingness to
forego a relatively small amount of public funds in the current period (in order to
stimulate even more in private donations) could represent a significant long-term
sacrifice.
Two other policy points bear making. First, even if crowding out is occurring at the
margin for orchestras in the aggregate, it still may be that subsidies stimulate giving
for individual organizations. Second, government would do well to try and ascertain
the degree of crowding in or out across firms and subsectors. In a world of scarce
public resources, this information would allow policymakers to reallocate subsidies
across the nonprofit sector in a way that uses these funds most efficiently.
Certain types of nonprofit firms should be especially prone to the subsidy trap.
First, there is the firm that has relatively high levels of government subsidization,
and year-to-year trouble with liquidity, such that a small decrease in these subsidies
would jeopardize the firm’s immediate ability to cover its operating costs. Second,
there is the government-subsidized firm that has traditionally paid little attention to
private donations.
How might a firm be lifted out of a short-run subsidy trap? In light of the analysis
reported in this paper, the answer to this question is not more government aid. If a
larger subsidy were used to increase the nonprofit’s short-term liquidity, the firm’s
total unearned revenues—both short-term and long-term—would fall. Therefore, the
solution to this trap lies either in an exogenous increase in philanthropy or increased
funds from another source, such as earned revenues (which were not found in the
empirical model to significantly affect private giving).
In the case of a firm in a short-run subsidy trap, cognizance of the futility of greater
public subsidies would be very valuable for managers and policymakers. For instance,
limited resources for promotion and development might not be squandered on
lobbying as opposed to, say, commercial expansion. And the existence of the trap as
well as the futility of greater government funding might be useful to convey to potential
donors. An explicit appeal to donors to “invest” in a firm in the present to allow it the
liquidity to accept a (lower) level of government funding (associated in the long run
with higher private giving) might be an effective fund-raising strategy.
For their helpful suggestions, I am grateful to Al Slivinski, Jorge Martínez, John Thomas,
Günther Schulze, Bruce Seaman, Bill Luksetich, and three anonymous referees. The usual
caveat applies.
ARTHUR C. BROOKS is Assistant Professor of Public Administration and Economics,
Andrew Young School of Policy Studies, Georgia State University, Atlanta.

Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both? / 463
REFERENCES
Andreoni, J. (1993). An experimental test of the public-goods crowding-out hypothesis. Ameri-
can Economic Review, 83(5), 1317–1327.
Bergstrom, T.C., Blume, L.E., & Varian, H.R. (1986). On the private provision of public goods.
Journal of Public Economics, 29, 25–49.
Brooks, A.C. (1997). Improving the orchestra’s revenue position: practical tactics and general
strategies. Symphony Orchestra Institute research studies series 1. Evanston, IL: Symphony
Orchestra Institute.
Brooks, A.C. (1999). Do public subsidies leverage private philanthropy for the arts? Empirical
evidence on symphony orchestras. Nonprofit and Voluntary Sector Quarterly, 28(1), 32–45.
Brooks, A.C. (2000). Is there a dark side to government support for nonprofits? Public Admin-
istration Review, 60(3), 211–218.
Cobb, N.K. (1995). Looking ahead: Private sector giving to the arts and humanities. Washing-
ton, DC: President’s Committee on the Arts & Humanities.
Duncan, B. (1999). Modeling charitable contributions of time and money. Journal of Public
Economics, 72, 213–242.
Feldstein, M.S. (1974). Social security, induced retirement, and aggregate capital accumula-
tion. Journal of Political Economy, 82, 905–926.
Friedman, M. (1962). Capitalism and freedom. Chicago: University of Chicago Press, 80–81.
Hochman, H., & Rodgers, J. (1977). The optimal tax treatment of charitable contributions.
National Tax Journal, 30, 1–19.
Khanna, J., Posnett, J., & Sandler, T. (1995). Charity donations in U.K.: New evidence based on
panel data. Journal of Public Economics, 56, 257–272.
Kingma, B.R. (1989). An accurate measure of the crowd-out effect, income effect, and price
effect for charitable contributions. Journal of Political Economy, 97, 1197–1207.
Kmenta, J. (1986). Elements of econometrics (2nd ed.) New York: Macmillan.
Lindsey, L., & Steinberg, R. (1990). Joint crowdout: An empirical study of the impact of federal
grants on state government expenditures and charitable donations. (National Bureau of Eco-
nomic Research Working Paper #3226). Cambridge.
National Endowment for the Arts. (1998). Did You Know...? http://arts.endow.gov/Guide/Facts/
DidYa2.html.
Parks, R.W. (1967). Efficient estimation of a system of regression equations when disturbances
are both serially and contemporaneously correlated. Journal of the American Statistical As-
sociation, 62, 500–509.
Payne, A.A. (1998). Does the government crowd-out private donations? New evidence from a
sample of nonprofit firms. Journal of Public Economics, 69, 323–345.
Rose-Ackerman, S. (1986). Do government grants to charity reduce private donations? In Su-
san Rose-Ackerman (Ed.). The economics of nonprofit institutions (pp.313–329). New York:
Oxford University Press.
Schiff, J. (1985). Does government spending crowd out charitable contributions? National Tax
Journal, 38, 535–546.
Seaman, B.A. (1980). Economic models and support for the arts. In W.S. Hendon, J.L. Shanahan,
& A.J. MacDonald (Eds.), Economic policy for the arts (pp.80–95). Cambridge, MA: Abt
Books.
Steinberg, R. (1985). Empirical relations between government spending and charitable dona-
tions? Journal of Voluntary Action Research, 14, 54–64.

464 / Public Subsidies and Charitable Giving: Crowding out, Crowding in, or Both?
Steinberg, R. (1993). Does government spending crowd out donations? Interpreting the evi-
dence. In A. Ben-Ner & B. Gui (Eds.). The nonprofit sector in the mixed economy (pp. 99–
125). Ann Arbor, University of Michigan Press.
Warr, P.G. (1982). Pareto optimal redistribution and private charity. Journal of Public Eco-
nomics, 19, 131–138.
West, E.G. (1975). Educational slowdown and public intervention in 19th-century England: A
study in the economics of bureaucracy. Explorations in Economic History, 12, 61–87.
The Wolf Organization. (1992). The financial condition of symphony orchestras. Washington,
DC: American Symphony Orchestra League.