The theory of crowding-out predicts an increase in government expenditure will decrease private propensity to donate. The National/Canadian Survey of Giving, Participating, and Volunteering for 1997, 2000, and 2004 is used to test whether Canadian private donations to health, education, environment, and social welfare are affected by changes in government transfers to the equivalent targets. Empirical evidence from this study suggests there is no support for the theory of crowding-out. Government expenditure appears to have no significant effect on private donations across all 4 categories. Other specifications suggest that the crowding out (or in) effect may depend on individual awareness of government actions.
Top Rated Pune Call Girls Shikrapur ⟟ 6297143586 ⟟ Call Me For Genuine Sex S...
Crowding Out Giving? The Effect of Government Transfers on Private Donations in Canada
1. Crowding-Out Giving?
The Effect of Government Transfers on
Private Donations in Canada
Mai Bui
Dr. Lemieux- ECON 594 - UBC
Aug. 1st 2008
The theory of crowding-out predicts an increase in
government expenditure will decrease private propensity
to donate. The National/Canadian Survey of Giving,
Participating, and Volunteering for 1997, 2000, and 2004
is used to test whether Canadian private donations to
health, education, environment, and social welfare are
affected by changes in government transfers to the
equivalent targets. Empirical evidence from this study
suggests there is no support for the theory of crowding-
out. Government expenditure appears to have no
significant effect on private donations across all 4
categories. Other specifications suggest that the crowding
out (or in) effect may depend on individual awareness of
government actions.
We believe that one of the greatest costs of our present welfare system is that it ...poisons the springs of private
charitable activity.
1
Milton and Rose Friedman
1
Friedman, M., and Rose D. Friedman, Free to Choose: A Personal Statement (New York: Harcourt Brace Jovanovich, 1980) 123.
2. Page |2
INTRODUCTION
The behavioural effect of governmental transfers on private propensity to donate has
been a source of interest, for the last couple of decades, to public economists looking to
extricate the actual outcome from the intended outcome of governmental policy. Spending
policies rely on an accurate picture of how donors react to public application of their tax
money, and optimal government spending differs depending on whether private donations are
complements, substitutes, or neural with respect to government expenditure. Furthermore,
understanding the effect of public expenditure on private philanthropy would shed some light
on how donating enters utility functions and the determinants of willingness to donate.
Theories predict a dollar of government transfers will crowd out a dollar of private
donations, thereby neutralizing the purported effect of government spending policy and
suggesting a suboptimal use of tax dollars. And yet, subsequent models have made room for an
incomplete crowd out, and others, still, have responded with models allowing for a crowd-in of
donations. Much empirical research has been conducted on the subject, as well, with the
results varying widely. While there have been some studies done in the UK, most are on the US
and none currently exist on the effect of Canadian government expenditure on propensity to
donate. However, results may not be validly extrapolated from one particular country and time
period and applied to another. Whether crowding out happens depends on donor awareness
of government policy and actions, and whether or not information is accurate and donors tend
3. Page |3
to follow current events are characteristics dependent on the particular national demographic
and time period.
The goal of this paper is to add to the literature by focusing on Canadian donors’
reaction to their respective provincial governments’ spending. It provides an update using
recent microdata from the National/Canadian Survey of Giving, Volunteering, and Participating
for the years 1997, 2000, and 2004. Figure 1 shows provincial government expenditure per
capita from 1995 to 2005 for four important subsectors. Even though there is an upward trend,
spending varied significantly among the subsectors throughout the decade. Instead of looking
only at social welfare, I focus on a wider range of public goods: education, healthcare, social
welfare, and the environment. Regression analysis shows that there is no support for crowding
out or in for any of the public goods; that is, government transfers appear to have no significant
effect on donor behaviour across the four subsectors.
The rest of the paper is organized as follows. After a summary of the theoretical work
on the crowding-out hypothesis is a survey of the literature. Next is an overview of the
empirical approach used, followed by a discussion of the results.
4. Page |4
Figure 1
Source: Statistics Canada, Table 385-0002 - Federal, provincial and territorial general government revenue and
expenditures.
THEORIES OF CROWDING-OUT
The theory of crowding-out predicts that when transfers to a public good increase,
private contributions to that good falls. The mechanics can be divided into two familiar effects:
a substitution and an income effect. With the substitution effect, an increase in government
transfer lowers the need for charitable funding, thereby decreasing the marginal benefit of
donating and increasing the marginal benefit of private consumption. Under the income effect,
an increase in government transfers lowers disposable income through taxes and reduces the
5. Page |5
financial ability of taxpayers to make donations2. Both effects operate in the same direction to
crowd out private donations, with the income effect being the less-significant of the two3.
The existence of the substitution effect is subject to some assumptions about the way
giving enters a donor’s utility function. In the Ultra-Rational case, a donor’s decision to give is
made by equating marginal cost (to the donor) and marginal benefit (to the recipient). A donor
will give until the marginal cost of giving equals the marginal benefit. Thus, donations are made
only to raise welfare, and the government is seen as a direct agent in transferring income from
the contributor (through taxes) to the recipient (through transfers). Donations and government
transfers are perfectly substitutable, implying a complete crowd-out. In the case of
Interdependent Utility Functions, the donor’s marginal benefit from contributing is affected by
the recipient’s marginal benefit from receiving. Government transfers lower the marginal
utility of donations for the recipient, and thus for the donor as well, although the effect is
incomplete and implies an imperfect (fractional) crowding-out. The Warm Glow Hypothesis4
(Andreoni, 1990) acknowledges that the act of giving itself has intrinsic benefits to the donor
independent of the effect it may have on the recipient or to the total supply of the public good.
In other words, the contributor’s utility function is independent from the utility of recipient,
2
Information on the substitution and income effects, Interdependent Utility Functions and Ultra-Rational is summarized using
Burton Abrams and Mark Schmitz, “The ‘Crowding-Out’ Effect of Governmental Transfers on Private Charitable Contributions”,
Public Choice, 33(1): 1978, 31
3
If a government wants to increase hospital funding, they are likely to reallocate tax money from, say, financing a museum
renovation rather than directly increasing taxes in order to do so. Thus, a taxpayer partial to health care improvement will find
her disposable income (ability to donate) might be unchanged, but she might decide to substitute away from donating to health
care now that it is being publicly taken care of.
4
James Andreoni, “Impure Altruism and Donations to Public Goods: A Theory of Warm-Glow Giving”, The Economic Journal,
100(401): 1990, 464-477
6. Page |6
and donating enters the donor’s utility function positively by increasing the ‘warm glow’. In the
case of complete independence, with warm-glow being the only motivation for giving,
government spending does not crowd out private donations.
Fewer theories seek to explain the crowding-in phenomenon, which suggests that
potential donors somehow see government transfers as complements to their own giving.
Brooks (2000) and Khanna et al (2000) note that charities can use the fact that they are able to
attract public funding as a signal to potential donors that their money would be put to good
use. Furthermore, government attention stimulates the interest of private donors who might
otherwise have overlooked the cause and organization5. Matching grants, where the
government pledges an amount for every X dollars of private donations a charity can fundraise,
are another (albeit, trivial) way for crowding-in to take place6. One might also speculate that
crowding-in happens because of the bystander-apathy effect7. Experiments in psychology have
shown that a person witnessing an emergency, such as someone else having a seizure, is less
likely to help when she is one of many apathetic bystanders than if she were the only witness.
Being one of many in the same situation diffuses the sense of responsibility amongst all
witnesses. If we see that others are not reacting, then we will not react, either, because we
suspect the blame for not reacting is negligible when many are responsible, or because we look
to each other’s reactions to discern whether or not the situation requires action. In the case of
5
A. C. Brooks, “Is there a Dark Side to Government Support for Non-Profits?” Public Administration Review, 6(3): 2000.,212
6
R. D. Hood et al, “Economic Determinants of Individual Charitable Donations in Canada”, Canadian Journal of
Economics, 10( 4): 1977), 667.
7
Darley, J. M. and B. Latané. “Bystander Intervention in Emergencies: the Diffusion of Responsibility”. Journal of Personality
and Social Psychology. 8(4): 1968. pp. 377-83
7. Page |7
a crowding-in, a lack of ‘reaction’ from the government justifies donor apathy and discourages
action. On the other hand, a move by the government to fund carbon-emission reduction
projects, for example, could trigger similar gestures from citizens. Although bystander apathy is
seldom used to explain economic phenomena, it could provide insight into why some studies
have shown that private donations vary positively with government transfers.
EVIDENCE FROM THE LITERATURE
There is no consensus in the empirical literature on whether government expenditure
crowds in, out, or is independent of, private donations. Noticing a drop in private charitable
contributions right around the time government expenditures increased, Abrams & Schmitz
(1978) focus on the crowding out of donations to social welfare programs in the US. Using US
census and tax returns for the period 1948-1972, they create a quasi-panel to determine the
extent to which state and federal social welfare transfers crowd out private donations.
Charitable donations are grouped by AGI class, while federal transfers are in per capita terms.
After controlling for average disposable income (before contributions) and the price of
donating, p=1-t, where t is the marginal tax rate on the first dollar contributed in class and year,
they find a small but significant crowd out effect. Their estimates suggest that a 1% increase in
government transfers to social welfare programs is met by a 0.2% fall in private donations.
However, that analysis did not account for individual characteristics such as age,
education, and gender. They also are basing their estimations during a very particular time
8. Page |8
period, the post-WWI and Cold War era. It is possible that part of the drop in private charitable
contributions is owed to a change in sentiments particular to that time in history, for example,
an increasing sense of mistrust during the Cold War. In order to correct for strong time trends
and year-specific characteristics as well as control for some personal characteristics, their next
study (Abrams & Schmitz, 1984) uses instead a cross section of US tax returns in 1979. This
time, they focus on state level spending and include a poverty variable as well as religion and
race. Their estimation shows $1 of state transfers leads to 30 cents less of donations.
Though using tax returns has its benefits, including the ability to access detailed
information about income and applicable tax rates, it precludes accounting for donations not
claimed for tax purposes. Since only donations to registered charities are eligible to receive a
tax credit, if there are donations made to unregistered charities or donations that are
unclaimed, the results could be biased. Furthermore, depending on filing status, many US
citizens have the choice between itemizing their charitable contributions, which requires
recordkeeping, and taking a standard deduction8. Where the standard deduction is greater
than or equal to the itemized deduction, it is more advantageous to take the standard
deduction because each itemized deduction must pass under greater scrutiny. Thus, there are
quite a few donations missing from US tax data, and their absence could be correlated with tax
8 st st
“Publication 501”, IRS - US Department of Treasury, Aug. 1 2008, Aug. 1 2008
<http://www.irs.gov/publications/p501/ar02.html#d0e5569>
9. Page |9
filer status9. Instead of using tax data, Schiff (1985) uses survey data to study the effect of
social welfare spending on donations10. He presents a model in which government transfers
need not crowd out donations, and might indeed encourage donations. Using the National
Survey of Philanthropy in 1974, he finds that the effects vary between local and state transfers,
and vary between cash and welfare program transfers. Local transfers and transfers to welfare
programs crowd in, while state transfers and transfers in cash crowd out.
The discussion so far has been limited to studies of social welfare as the public good of
interest. There are many target areas that concern citizens and their governments, and it is
possible that donations to different public goods will experience varying crowding-out effects.
Brooks (2000) looks at arts and culture from 1966-1997, and welfare, health, and education
from 1955-199511. Lagged values of government spending at the state and federal level for the
period 1955-95 are compared with data of the same period on private donations to nonprofits
in those subsectors. Controlling only for GDP, year trends, and last year’s private giving, Brooks
finds donations to welfare and health care experience a small crowd out with state spending,
while education and the Arts are unaffected by any level of government spending.
9
Usually standard deductions are for those ‘financially deserving’ of deductions, (i.e. dependents, disabled) who
are likely to be the ones donating small amounts, which means lower donation amounts could be systematically
missing from the data.
10
Jerald Schiff, “Does Government Spending Crowd Out Charitable Contributions?” National Tax Journal, 38(4): 1985, 535-546.
11
A. C. Brooks, “Is there a Dark Side to Government Support for Non-Profits?” Public Administration Review, 6(3): 2000.,212
10. P a g e | 10
This present paper aims to contribute a number of things to the literature on crowding
out. Firstly, it makes an effort to fill the gap in the literature by focusing on Canada, whose tax
treatment of donations differs from that of the US. Secondly, it focuses on a wider range of
public goods, to acknowledge the extent of politically salient subsectors as well as to allow for
any “crowding” effects to vary across public goods.
EMPIRICAL APPROACH AND DATA
Based on specifications in the literature, for each of the four subsectors, empirical
estimation will take the form:
Yijt = βo + β1Gjt + β2Xijt + β3PROVj + β4YEARt + εij
where Gjt is the dollar amount of transfers per capita in province j at time t, and G denotes
healthcare, social welfare, environment, and education. Xijt represents the vector of controls
for donor characteristics. PROVj is the vector of provincial dummies, and YEARt is the vector of
survey-year dummies. Yijt is the dollar amount contributed per person i in province j at year t to
cause Y, where Y denotes healthcare, social welfare, environment, and education. Although
different variations on this model will be estimated, they will be uniform across four subsectors
in order to be consistent12.
12
A. C. Brooks, “Is there a Dark Side?”, 213.
11. P a g e | 11
A quasi-panel dataset, grouped at the provincial level, is created out of survey and
government statistics. Year dummies capture time-specific effects, and province dummies
capture group-specific effects. Because the main explanatory variable, Gjt, is invariant with
respect to individuals in the same province and year, standard errors may be biased. As such,
group effects are clustered at the province-year level where possible13. Because about a third
of the survey respondents reported donating nothing, ordinary least squares estimation should
produce biased results. The Tobin (Tobit) estimation procedure is typically used in this situation
as it corrects estimates for censoring. A more general variant, Interval Regression (Intreg), will
be used instead, as it allows for the clustering of standard errors while yielding numerically
equivalent results to the Tobit14.
Data for private contributions comes from the National Survey of Giving, Volunteering,
and Participating (NSGVP) of 1997 and 2000 and the Canadian Survey of Giving, Volunteering,
and Participating (CSGVP) of 2004. The NSGVP was conducted as a supplement to the Labour
Force Survey. With the increasing importance of donations to research, funding was granted to
establish a permanent survey within Statistics Canada in 2001, which was renamed the CSGVP.
All three survey thousands of Canadians aged 15 and over, in the ten provinces from September
to December. All donations reported were made within the last 12 months before the survey,
and donations made to charities for purposes of international benefit are listed separately.
There is not much that is different between the two surveys, aside from a few questions asked
13
The small number of clusters (i.e. 30 clusters for a three-year quasi panel, 20 clusters for a two-year quasi-panel)
places restrictions on the number of standard errors (parameters) that can be estimated in each regression.
14
The tobit command allows for clustering in Stata 10, but not Stata 9.
12. P a g e | 12
in the NSGVP that have been dropped from the CSGVP. Two questions of interest are whether
the respondent voted in the provincial election, and whether the respondent watches
television at least a few times a week. A few questions are also differently worded; to control
for any effects this might have on the responses, survey-year dummies are included.
As the federal tax treatment of charitable donations in the US is different from that in
Canada, so, too, are the implications for data choice. Charitable donations in the US can be
deducted from taxable income, while in Canada they are used as non-refundable tax credit and
subtracted from the total tax paid. However, in Canada, the first $200 dollars of donations
claimed per tax return is given a credit of 17% in 1997 and 200015 (16% in 200416), while
anything in excess is given 29% credit. There is no similar credit scheme in the US. There is
thus an incentive for Canadians to hold onto charity donations receipts and claim all of them at
once, anytime within five years of the year the donation was made, to receive a higher credit
rate. It is highly probable that the use of tax returns in Canada as data would be misleading,
depending on how detailed the data is, as the year in which donations are claimed may not be
the year in which the donor felt compelled to donate. Furthermore, as mentioned above, not
all donations made are claimed. A taxpayer who has made a single small donation, for
example, might feel it unnecessary to bother to keep or track down the documentation to gain
15 st
“TI-97 Donations and Gifts” Canada Revenue Agency, 1997, Aug. 1 2008
< http://www.cra-arc.gc.ca/formspubs/prioryear/t1/1997/5006-s9/5006-s9-10-97e.pdf >
st
“TI-00 Donations and Gifts” Canada Revenue Agency, 2000, Aug. 1 2008
<http://www.cra-arc.gc.ca/formspubs/prioryear/t1/2000/5000-s9/5000-s9-00e.pdf>
16 st
“TI-2004 Donations and Gifts”, Canada Revenue Agency, 2004, Aug. 1 2008
<http://www.cra-arc.gc.ca/formspubs/prioryear/t1/2004/5000-s8-9/5000-s8-9-04e.pdf >
13. P a g e | 13
a small tax credit, making it likely that smaller donations will be systematically omitted from tax
returns. Because donation documentation is not required on the survey, a donor is more likely
to report it on a survey than list it on a tax return17.
What could be a major shortcoming of using survey data, however, is that it does not
allow for direct tax-incentive controls. Whereas most contributions to the literature control for
the price of giving, p=1-t, where t is the marginal tax rate on the first dollar donated, income is
noted in survey data in broad categories, thereby making it difficult to determine the applicable
tax rates for each respondent. The effect of tax credit on donations is partially captured by an
income variable. The higher the income, the higher the ability to donate and the higher the tax
benefits to donating. Nevertheless, p should have a negative effect on charitable giving, and its
absence suggests that the coefficient on government transfers may be negatively biased.
Government expenditure is represented by provincial transfers, the data on which are
obtained from CANSIM. The federal government does exert some spending power over the
four subsectors through conditional and unconditional grants to the provincial governments18,
for example, but not enough data exists to be able to create statistics for total per-province
(provincial and federal), per-subsector expenditure. Nonetheless, education, healthcare, and
17
On the other hand, donors are more likely to exaggerate donations or even make them up.
18 st
“The Spending Power – Scope and Limitations”. Library of Parliament. 2008. Aug. 1 2008.
<http://www.parl.gc.ca/information/library/PRBpubs/bp272-e.htm>
14. P a g e | 14
welfare are all within provincial jurisdiction,19 and provincial spending statistics provide an
adequate picture of where taxpayers in each province see their tax dollars going20. The overlap
between what the measures of government spending encompasses and what is measured for
private donations in the survey is imperfect. It should matter little to our analysis, however.
More detailed descriptions of what each transfer and donation variable encompasses is given in
Table 1.
In order to account for differently sized provinces, all government transfer variables are
converted into per capita terms using provincial population data from CANSIM. Although it
would have been informative to include them, the data are incomplete for the Yukon and
Northwest Territories and they are excluded from this analysis. Government transfer variables
are constructed by taking an average of spending in the previous two years. For example, if the
expenditure measure is to correspond to private donations in the year 2000, an average was
taken of provincial spending in the year 1998 and 1999. The lagged values correct for
simultaneous correlation. For example, an increase in government transfers to healthcare in
1997 might be brought about by a demand for increased funding to hospitals in the same year.
The demand for increased funding to hospitals will, in the same year, encourage private
donations to healthcare as well.
19
Environment is a recent subsector and the federal-provincial lines are unclear. For more, see Gibson, Dale,
“Constitutional Jurisdiction over Environmental Management in Canada”, University of Toronto Law Journal, 23(1):
1973, p. 54.
20
Brooks (2000) finds state spending, not federal spending, has a significant effect on donations, which suggests
that donors are more sensitive to local government actions.
15. P a g e | 15
Table 1
Transfers and donations
Health care Organizations engaged in out-patient health-related activities
Private and support services, education and research for specific
donations conditions (i.e. Heart and Stroke Foundation), hospitals.
Education Organizations and activities administering, providing,
supporting education and research. Includes primary,
secondary, post secondary, adult, and vocational education
as well as research organizations.
Social Organizations providing social services to the community.
welfare Includes emergency relief and income support and
maintenance.
Environment Organizations promoting and providing environmental
conservation, pollution control, environmental education and
health, animal protection.
Health care Medical care, preventative care, other health service
Government Education Elementary and secondary education, post secondary, special
transfers retraining, other
Social Income maintenance, social security, family allowances, etc
welfare
Environment Water purification and supply, pollution control, other
environmental services
Note: Information on private donations found in data codebook. Information on government transfers
found on CANSIM accompanying data series.
Table 2
Summary statistics
Median Mean MinStd Max
Private donations Deviation
variables used Health care 20 62.8773 0 17140 260.9197
Education 0 16.1991 0 23162 212.5742
Soc. welfare 3 37.4763 0 49759 369.8202
Environment 0 7.2211 0 7400 95.2769
Government Health care 1701.907 1844.337 1310.34 2488.774 328.5941
transfers variables Education 1295.276 1284.244 857.3799 2060.979 287.6167
used Soc. welfare 1047.109 1050.18 562.2576 1425.514 194.9936
Environment 52.7986 63.0587 30.1699 215.058 32.8093
16. P a g e | 16
RESULTS AND DISCUSSION
Overall, the regression results do not support the theory that government expenditures
will crowd out (or in) donations. This result is fairly consistent across all four subsectors;
although some specifications indicate crowding out in some sectors and crowding in for others,
those results suffer from biased standard errors and are of questionable significance. The
control variable coefficients have the expected signs and are consistent with the similar
estimates in the literature.
Table 3 displays results from the first (baseline) specification. Private donations are not
significantly affected by government transfers. Donation amounts increase with age, although
the areas of education and the environment are, understandably, less of a priority for retirees
relative to their middle-aged counterparts. As we might suspect, respondents with higher
household income donate more, and as mentioned before, it is likely that the coefficients show
the effects of both an income (ability to donate) and a tax credit (incentive to donate). Having
only a high school diploma or less lowers the amount donated by 17-33 dollars, depending on
the subsector, relative to those with a higher education. Where a respondent was born has no
significant effect on donation amount, and neither does religiosity, with the exception of the
environment, to which being non-religious increases donations. Membership in a political or
neighbourhood organization has a significant positive effect on donations to education, social
welfare, and the environment.
17. P a g e | 17
Table 3
Model 1 results:
1997, 2000, & 2004 Health Education Soc. welfare Environment
Government Health care 0.0286
transfers (0.0342)
Education 0.0237
(0.03432)
Social welfare -0.1187
(0.0799)
Environment -0.2291
(0.1592)
Age group 25-34 12.2374 54.0603 80.1311 98.0889
(6.9692)* (16.3541)*** (36.5742)** (18.9231)***
(15-24 35-44 26.2319 90.8502 113.6355 97.4129
dropped) (9.7440)** (20.9377)*** (39.9830)*** (25.7663)***
45-54 36.6666 75.1268 156.2897 129.4887
(10.6890)*** (21.1438)*** (58.7002)*** (25.8502)***
55-64 40.3498 105.4233 168.4237 168.1662
(13.3195)*** (28.8368)** (57.9801)*** (34.5422)***
65 + 49.1279 47.2298 175.698 155.7813
(17.5665)*** (17.1464)*** (55.4448)*** (29.3557)***
Household [$20,000, 26.0152 39.4371 42.2160 40.2973
income group $40000) (4.5445)*** (10.9623)*** (12.0534)*** (10.6862)***
before taxes [40,000, 30.9435 77.2505 73.1700 101.7197
60,000) (6.7434)*** (18.6936)*** (15.3577)*** (17.0583)***
(< 20,000 [60,000, 36.3122 101.1627 80.6681 126.223
dropped) 100,000) (8.1535)*** (20.1301)*** (15.5790)*** (17.0864)***
100,000 + 46.8610 179.6319 132.0525 166.3118
(19.6171)** (38.8470)*** (20.1697)*** (27.4914)***
Completed ≤ High school -32.9966 -17.3121 -21.5186 -32.6341
education (3.8398)*** (5.2183)*** (6.8992)*** (10.5048)***
(≥ some post
secondary
dropped)
Socio- Canadian born 5.5589 12.6921 11.0734 7.8413
demographic (7.2916) (12.0446) (7.7887) (10.7013)
(Foreign born Non religious 2.1660 -8.9684 6.3486 84.6836
and Religious (4.0580) (8.0010) (14.8293) (15.2844)***
dropped)
Gender Female 9.9728 11.96205 24.3468 73.9473
(3.1593)*** (6.0981)** (10.6126)** (12.7412)***
Membership in Political 5.6181 69.43304 94.8513 109.7404
organization (6.7667) (15.7013)*** (43.3661)** (14.2284)***
Neighbourhood 7.9391 70.9151 40.8863 27.3570
18. P a g e | 18
(4.5380)* (13.4335)*** (9.7843)*** (9.3449)***
Constant -140.4908 -560.9161 -250.4825 -1044.543
(70.6658)** (135.6356)*** (93.3967)*** (147.8973)***
Wald chi-square 4689.89 5852.01 671.6900 1330.76
(p-value) (0.0000) (0.0000) (0.0000) (0.0000)
Observations 46726 46726 46726 46726
Note: *, **, and *** indicate significance at the 90, 95, and 99 percent level, respectively.
Standard errors are reported in brackets.
Provincial and Year dummies were used in all models.
Standard errors are clustered by province-year.
Although results from the baseline model suggest a victory for the ‘warm glow’
hypothesis, it is likelier that imperfect information is the culprit. Whether or not government
transfers affect private donations depends on public awareness of government actions. The
second specification uses two measures of respondent awareness: whether or not they voted in
the last provincial election, and how often they watch the news. Unfortunately, the two
questions corresponding to these measures were dropped from the 2004 CSGVP survey, so a
quasi-panel of two years, 1997 and 2000, is used instead. This reduces the number of province-
year clusters from 30 down to 20, further limiting the number of parameters that can be
estimated. Rather than omit a number of key variables from the model to fit in new variables,
the clustering of standard errors is dropped. Thus, the significance of coefficients in Tables 4
and 5 is to be interpreted with caution.
Table 4 shows the results with two new controls entering the model linearly. The
inclusion of controls for awareness does not fundamentally change the insignificance of the
government transfers variables. The only significant government transfer coefficient is on
19. P a g e | 19
transfers to healthcare, suggesting a very small crowd-out effect; however, a 10% level
significance is not enough to ensure that the coefficient would retain its significance if standard
errors were to be clustered, assuming there is no negative correlation within clusters. Even if
the coefficient remains or increases significance with correct standard errors, the crowd-out
effect would be very small, with a 4-cent drop for every dollar the government spends on
healthcare. The rest of the controls are similar in direction and magnitude to the above
baseline specification, with religiosity and country of birth gaining in significance. Interestingly,
being non-religious decreases the amount donated to healthcare, education, and social welfare,
but increases the amount donated to environmental causes, relative to religious persons.
Table 4
Model 2 results:
1997 & 2000 only Health Education Soc. welfare Environment
Government Health care -0.0417*
transfers (0.0247)
Education 0.0003
(0.0156)
Social welfare -0.0253
(0.0211)
Environment (0.4980)
(0.4454)
Age group 25-34 32.9580 19.7096 17.4371 86.8062
(6.0978) *** (5.9437)*** (5.2215)*** (18.8418)***
(15-24 35-44 54.7322 36.0786 37.2763 69.6226
dropped) (5.9192)*** (5.7599)*** (5.0718)*** (18.6077)***
45-54 67.9803*** 22.5844 52.55821 99.7956
(6.2340) (6.1270)*** (5.3443)*** (19.2926)***
55-64 87.8471 35.8431 57.5336 127.8144
(6.6084)*** (6.4922)*** (5.6694)*** (20.4315)
65 + 110.1534 17.6885 62.1672 102.3794
(6.4370)*** (6.4530)*** (5.5324)*** (20.4223)***
Household [$20,000, 26.48938 19.0640 19.6679 28.76814
income group $40000) (4.4081)*** (4.4644)*** (3.7924)*** (14.2342)**
20. P a g e | 20
before taxes [40,000, 60,000) 48.3113 34.3361 33.9047 96.5704
(4.7125)*** (4.6950)*** (4.0317)*** (14.4429)***
(< 20,000 [60,000, 64.3909 47.6133 42.0658 116.7078
dropped) 100,000) (4.9399)*** (4.8700)*** (4.2209)*** (14.9005)***
100,000 + 136.8106 84.9076 83.2176 143.7738
(6.3383)*** (6.0972)*** (5.4132)*** (17.9086)***
Completed ≤ High school -18.8762 -13.4092 -19.8968 -31.33838
education (3.0754)*** (2.9120)*** (2.5402)*** (8.7087)***
(≥ some post
secondary
dropped)
Socio- Canadian born 24.3308 10.6151 1.9778 -9.9964
demographic (4.5585) *** (4.5514)** (3.8984) (12.5109)
Non religious -6.6945 -10.9073 -7.3586 75.3667
(Foreign born (3.7679)* (3.7287)*** (3.2635)** (9.9085)***
and Religious
dropped)
Gender Female 14.0588 4.9351 5.7179 63.03706
(2.9293) *** (2.8754)* (2.5179)** (8.6524)***
Organization Political 24.78059 36.7529 25.9944 78.6520
membership (6.1248)*** (5.7652)*** (5.1791)*** (15.7713)***
Neighbourhood 13.7636 28.8334 21.4380 27.4496
(3.9793)*** (3.7742)*** (3.4066)*** (10.7670)**
Measures of Voted last Prov. 32.9339 7.1035 18.2323 13.2138
awareness Election (3.9889)*** (3.9160)* (3.4366)*** (11.5186)
Follow news ≥ 17.85913 -1.6747 13.854 27.7707
(Follow news ≤ several times a (8.0723)** (7.7296) (6.8522)** (24.7680)
few times a week
month
dropped)
Constant -167.4568 -220.7575 -114.8057 -1033.534
(38.4414)*** (23.8496)*** (29.3000)*** (58.6470)***
Wald chi-square 1838.41 736.41 954.70 748.36
(p-value) (0.0000) (0.0000) (0.0000) (0.0000)
Observations 28221 28221 28221 28221
Note: *, **, and *** indicate significance at the 90, 95, and 99 percent level, respectively.
Standard errors are reported in brackets.
Provincial and Year dummies were used in all models.
Standard errors are not clustered; see text for more information.
Intuition suggests that the crowding-out (or in) effect may be more evident in those
people who are more aware of the provincial government’s actions. Table 5 shows results from
21. P a g e | 21
the third estimation. In this specification, government transfer variables are interacted with
dummies indicating whether the respondent voted in the last provincial election, and whether
he or she follows the news at least several times a week. The results are mixed across
subsectors.
For healthcare, estimates show that a dollar of government transfers crowds in 6 cents
of donations in general, and environmental spending crowds in a little over a dollar’s worth of
donations in general. A small crowd-out effect for healthcare is seen in respondents who voted
in the previous provincial election, while a small crowd-in effect is seen in respondents who
follow the news. This is perhaps because those who vote in the provincial election are
informed about promises to spend on healthcare, while those who watch the news may pick up
on criticising reports about the need for healthcare funding. Similarly, education donations are
crowded-in among respondents who follow current events. On the other hand, social spending
crowds out donations among those who follow the news. Figure 1 shows that health and
education are the two top spending areas in Canada, so it should come as no surprise that they
are under the most scrutiny in the media. It is harder to understand why social spending, which
comes in at a close third in spending priority, would yield a different result. That said, as the
standard errors may be biased, it would not be much use to overanalyze the results in Model 3.
22. P a g e | 22
Table 5
Model 3 results:
1997 & 2000 with interactions Health Education Soc. welfare Environ.
Government Health care 0.0668
transfers (0.0293)**
Education 0.0003
(0.0179)
Social welfare -0.0162
(0.0277)
Environment 1.04763
(0.6313)*
Government Health*vprov -0.0402
transfers interacted (0.0213)*
with awareness Education*vprov -0.0039
measures (0.0122)
Social*vprov 0.0164
(vprov = voted in last (0.0162)
provincial election; Environ*vprov -0.3048
news = follow news ≥ (0.4096)
several times a week) Health* news 0.0519981
(0.0222)**
Education*news 0.0249
(0.0112)**
Social *news -0.0221
(0.0135)*
Environ*news -0.3526
(0.3805)
Age group 25-34 12.3973 19.8035 17.4565 86.8180
(5.8549)** (5.9438)*** (5.2212)*** (18.8461)***
(15-24 dropped) 35-44 24.8726 36.1853 37.2839 69.8343
(5.7043)*** (5.7601)*** (5.0714)*** (18.6152)***
45-54 37.0906 22.7229 52.5738 99.8983
(6.0470)*** (6.1273)*** (5.3438)*** (19.2976)***
55-64 43.7689 36.0285 57.4469 127.7742
(6.4437)*** (6.4930)*** (5.6691)*** (20.4371)***
65 + 59.4326 17.9342 62.0567 102.1525
(6.2400)*** (6.4540)*** (5.5325)*** (20.4260)***
Household [$20,000, 27.5766 19.0360 19.5624 28.8345
income group before $40000) (4.3454)*** (4.4643)*** (3.7927)*** (14.2382)**
taxes
[40,000, 33.6685 34.3128 33.8364 96.6992
60,000) (4.6577)*** (4.6951)*** (4.0317)*** (14.4456)***
(< 20,000 dropped)
[60,000, 40.1472 47.6658 41.9983 116.7097
100,000) (4.8948)*** (4.8701)*** (4.2210)*** (14.9026)***
100,000 + 79.6616 84.6627 83.1560 143.86
(6.3955)*** (6.0986)*** (5.4136)*** (17.9133)***
Completed education ≤ High school -24.8126 -13.4608 -19.8385 -31.2525
(2.9959)*** (2.9123)*** (2.5402)*** (8.7099)***
(≥ some post
secondary dropped)
Socio-demographic Canadian born 7.8322 10.6992 1.967524 -9.9273
23. P a g e | 23
(4.5266)* (4.5527)** (3.8982) (12.512)
(Foreign born and Non religious -2.4094 -10.8968 -7.4310 75.3260
Religious dropped) (3.7711) (3.7293)*** (3.2635)** (9.9089)***
Gender Female 9.2512 4.9745 5.6903 63.1425
(2.9545)*** (2.8755)* (2.5178)** (8.6545)***
Organization Political 19.1798 36.6943 26.0324 78.6631
membership (6.3924)*** (5.7662)*** (5.1791)*** (15.7772)***
Neighbourhood 2.5653 28.9523 21.5016 27.3134
(4.1426) (3.7745)*** (3.4067)*** (10.7695)**
Measures of Voted last Prov. 78.3671 11.5018 1.6637 30.9524
awareness Election (34.9407)** (14.9483) (16.7025) (26.4340)
Follow news ≥ -44.9961 -30.49852 36.9615 51.7527
(Follow news ≤ few
several times a (35.7800) (15.0973)** (15.7963)** (35.3560)
times a month
week
dropped)
Constant -242.5931 -224.2166 -127.7774
(46.3132)*** (25.8524)*** (35.3868)***
Wald chi-square 689.78 741.36 958.14
(p-value) (0.0000) (0.0000) (0.0000) (0.0000)
Observations 28221 28221 28221 28221
Note: *, **, and *** indicate significance at the 90, 95, and 99 percent level, respectively.
Standard errors are reported in brackets.
Provincial and Year dummies were used in all models.
Standard errors are not clustered; see text for more information.
CONCLUSION
Although government spending in Canada is perpetually under scrutiny in the media and
academia alike, its effects on donations has been little studied. Using recent survey data from
the NSGVP and CSGVP, I estimate the effect of government transfers on private donations for
four public goods: healthcare, education, social welfare, and the environment. The baseline
estimations suggest no crowd-out (or in) effect. The second and third specifications control for
the respondent’s awareness of government actions. The results from including these new
controls show that awareness might be an important factor in whether determining whether or
not donations are affected by government spending. Health and education experience a crowd-
24. P a g e | 24
in among those who regularly watch the news, while health experienced a weakly significant
crowd-out among those who voted in the last provincial election. Social spending causes a
weakly significant crowd-out among followers of the news. Healthcare spending, in general,
crowds in donations, while environmental spending, in general, crowds out donations.
The main conclusion to take away from this exercise is that, as far as correct standard
errors are concerned, there does not appear to be any support for either a crowding-out or a
crowding-in effect. Private donation variables across healthcare, education, social welfare, and
the environment are unresponsive to respective changes in government transfers. Results from
the third specification may suggest that the effect of government transfers on donations varies
across individuals with different levels of awareness about government actions. However
interesting these results may be, their economic as well as policy significance is questionable.
Where the estimated crowding effect is large, the significance is weak and where the
significance is higher, the magnitude of the effect is negligible and not worth tailoring spending
policy around. There is thus ample room in the literature for more investigation into how
awareness interacts with changes in government expenditures to affect donations.
25. P a g e | 25
REFERENCES
Abrams, Burton and Mark Schmitz. “The ‘Crowding-Out’ Effect of Governmental Transfers on Private
Charitable Contributions”. Public Choice. 33(1): 1978. 29-39.
Abrams, Burton and Mark Schmitz. “The Crowding-Out Effect of Governmental Transfers on Private
Charitable Contributions: Cross-Section Evidence”. National Tax Journal. 37(4): 1984. 563-67.
Andreoni, James. “Impure Altruism and Donations to Public Goods: A Theory of Warm-Glow Giving”.
The Economic Journal, 100(401): 1990. pp. 464-477
Brooks, A.C. “Is there a Dark Side to Government Support for Non-Profits?” Public Administration
Review. 6(3): 2000. 211-17.
“Tax Advantages of Donating to Charity” Canada Revenue Agency. 2008. Aug. 1st 2008. <http://www.cra-
arc.gc.ca/E/pub/tg/rc4142/rc4142-05e.pdf>
Darley, J. M. and B. Latané. “Bystander Intervention in Emergencies: the Diffusion of Responsibility”.
Journal of Personality and Social Psychology. 8(4): 1968. pp. 377-83.
Estimates of population, by age group and sex . [Data file]. Retrieved from http://cansim2.statcan.ca
(Table 051-0001). Statistics Canada.
Federal, provincial and territorial general government revenue and expenditures [Data file]. Retrieved
from http://cansim2.statcan.ca (Table 385-0002 ). Statistics Canada.
Friedman, M., and Rose D. Friedman. Free to Choose: A Personal Statement. New York: Harcourt Brace
Jovanovich, 1980.
Gibson, Dale, “Constitutional Jurisdiction over Environmental Management in Canada”, University of
Toronto Law Journal, 23(1): 1973, pp. 54-87.
Hood, R.D. et al, “Economic Determinants of Individual Charitable Donations in Canada”, Canadian
Journal of Economics, 10( 4): 1977), pp. 653-669
Internal Revenue Service Publication 526 (2007), “Charitable Contributions”, 2007. Aug. 1st 2008.
<http://www.irs.gov/publications/p526/index.html >
Khanna, Jyoti et al. “Charity Donations in the UK: New Evidence Based on Panel Data.” Journal of Public
Economics. 56(2): 1995. pp. 257-272
Khanna, J. and T. Sandler. “Partners in Giving: The Crowding-In Effects of UK Government Grants”.
European Economic Review. 44(8): 2000. pp. 1543-56.
“Publication 501”, IRS - US Department of Treasury, Aug. 1st 2008, Aug. 1st 2008
<http://www.irs.gov/publications/p501/ar02.html#d0e5569>
Roberts, Russell D. “A Positive Model of Private Charity and Public Transfers”. The Journal of Political
Economy, Vol. 92, No. 1 (Feb., 1984), pp. 136-1.
26. P a g e | 26
Schiff, Jerald. “Does Government Spending Crowd Out Charitable Contributions?” National Tax Journal.
38(4): 1985. pp. 535-546.
“The Spending Power – Scope and Limitations”. Library of Parliament. 2008. Aug. 1st 2008.
<http://www.parl.gc.ca/information/library/PRBpubs/bp272-e.htm>
“TI-97 Donations and Gifts” Canada Revenue Agency, 1997, Aug. 1st 2008 < http://www.cra-
arc.gc.ca/formspubs/prioryear/t1/1997/5006-s9/5006-s9-10-97e.pdf >
“TI-00 Donations and Gifts” Canada Revenue Agency, 2000, Aug. 1st 2008 <http://www.cra-
arc.gc.ca/formspubs/prioryear/t1/2000/5000-s9/5000-s9-00e.pdf>
“TI-04 Donations and Gifts”, Canada Revenue Agency, 2004, Aug. 1st 2008 <http://www.cra-
arc.gc.ca/formspubs/prioryear/t1/2004/5000-s8-9/5000-s8-9-04e.pdf >
Veall, M.R. and Klaus Zimmermann. “Goodness of Fit Measures in the Tobit Model”. Oxford Bulletin of
Economics and Statistics. 56(4): 1994. 485-501.