4. Two questions
1. Why do previous studies find different results?
2. How do individuals differ in their response to
changes in government contributions?
5. Our first question
1. Why do previous studies find different results?
2. How do individuals differ in their response to
changes in government contributions?
6. Meta-analysis
Systematic literature review
We collect effect sizes published in previous
research
We seek to explain differences in effect sizes
between studies by characteristics of samples
and publications
7. Meta-analysis: collecting studies
Y = Amount of private donations
X = Government contribution
Retrieval in Web of Science through EndNote
Our search now extends back to 2007
We include only original empirical quantitative
results
N = 218 estimates from 34 articles
20. Findings
Analyses of tax records and lab experiments
produce more crowding out than surveys and
field experiments.
Analyses of organizational level data produce
more crowding out than individual level data.
Studies from Europe find the weaker estimates
of crowding out than US studies.
21. Units of analysis
Multilevel random-effects regression on COE estimates (excl. outliers)
Units of analysis
Individuals (ref.)
Organizations
-0,18 (0,20)
(Constant)
-0,27 (0,10)
-0,20 (0,13)
Between-study SD
Rho
Studies
Observations
0,42
0,72
21
85
0,43
0,73
21
85
22. Type of government contribution
Multilevel random-effects regression on COE estimates (excl. outliers)
Type of govt contribution
Subsidies to orgs (ref.)
Expenditures
Rebate
Match
Taxing respondents
0,34 (0.17) *
0,87 (0,21) **
0,47 (0,16) **
- 0,12 (0,17)
(Constant)
-0,47 (0,11) **
Between-study SD
Rho
Studies
Observations
0,25
0,49
21
85
23. Awareness
Multilevel random-effects regression on COE estimates (excl. outliers)
Rs aware of govt contributions
No (ref.)
Yes
0,18 (0,20)
Rs aware of need donated to
No (ref.)
Yes
0,14 (0,20)
(Constant)
-0,37 (0,16) *
-0.33 (0,14) *
Between-study SD
Rho
Studies
Observations
0,43
0,73
21
85
0,43
0,73
21
85
25. Our second question
1. Why do previous studies find different results?
2. How do individuals differ in their response to
changes in government contributions?
27. The scenario experiment
• In the Giving in the Netherlands Panel Survey
2012 we included a scenario experiment.
• 1,448 participants evaluated 3 scenarios,
constructed randomly by combining
information on budget cut levels and sectors.
• Participants were reminded of their
households’ contribution in the past year.
28. Example of scenario
“With your household you donated €100 to
health in the past year. If the government cuts
5% in this area, how would you react?”
Response categories:
• I will give the same as last year
• I am willing to give more
• I will also give less
[if more/less] What will be the new amount?
29. How the Dutch respond to cutbacks
Average response across all
4,344 scenarios
32. Support for the civic voluntarism model
Odds ratios from logistic regression of willingness to contribute more after government
cutback in at least one scenario (GINPS12, n=1,478; including controls for gender, age,
income from wealth, home ownership, number of donation areas)
33. Values, reputation and efficacy
Odds ratios from logistic regression of willingness to contribute more after government
cutback in at least one scenario (GINPS12, n=1,478)
34. Conclusions of meta-analysis
• On average, a $1 reduction in government support is
associated with a $0.28 increase in private
contributions.
• However, crowding-out estimates vary considerably
from study to study.
• Differences in the methodology used to measure the
influence of government contributions on private
giving are driving these differences.
35. Conclusions of scenario experiment
• Individuals also vary systematically in their
responses to changes in government contributions.
• Those with more resources, receiving more
solicitations and more generous donors are more
likely to contribute more after government
cutbacks.
• The principle of care, reputation and charitable
confidence are key mechanisms in crowding-out.
• The principle of care is the only characteristic
predicting the level of crowding-out.
36. Contact details
• René Bekkers, r.bekkers@vu.nl and Arjen de
Wit, a.de.wit@vu.nl
• ‘Giving in the Netherlands’, Center for
Philanthropic Studies, Faculty of Social
Sciences, VU University Amsterdam,
www.giving.nl