the effect of social influences on gggiving evidence from ... · n 243,974 10,044 243,974...

36
Centre for Market and Public Organisation Public Organisation The effect of social influences on giving evidence from the (running) field Sarah Smith and Edmund Wright University of Bristol

Upload: others

Post on 19-Jun-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Centre for Market and Public OrganisationPublic Organisation

The effect of social influences on giving –g g

evidence from the (running) field

Sarah Smith and Edmund Wright

University of Bristol 

Page 2: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Overview

• How much do people donate when they can see how much other l b f th h d t d?people before them have donated? 

• Unique dataset of donations to charity made on behalf of peopleUnique dataset of donations to charity made on behalf of people running in the 2010 London marathon

• Donations are made online to individual fundraising pages on two b it Wh th t th d ll i d tiwebsites. When they go to the page, donors see all previous donations

• Two main questions:o a quest o s:

• What is the effect of previous donations on how much people give?

• Do the identities of the fundraiser and other donors matter? 

– Do men give more to male fundraisers?

– Do men give more if other male donors have given more?  

Page 3: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Our contribution

• A number of lab and field experiments have looked at the effect of i l i fl / i l i f ti F lk t l (2003) F dsocial influences/ social information – Falk et al (2003), Frey and 

Meier (2004), Shang and Croson (2009) 

• Why analyse naturally‐occurring field data?  

• Obvious drawback that social information cannot be manipulated

• But potential advantages

– Scale of fundraising activity: >300,000 donations to >12,000 fundraisers running on behalf of 1,000+ charities  u d a se s u g o be a o ,000 c a t es

– Richness of data: Donors see full history of previous donations. In earlier field experiments, donors have been given a single reference level (“a previous donor” “the typical donor”)reference level (“a previous donor”, “the typical donor”)

Page 4: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Main findings

• Mean of all past donations has a small, negative effect on how much l i i t t ith “ d t”people give, consistent with “crowd out”

• Threshold effects are important – reaching the target has a negativeThreshold effects are important  reaching the target has a negative effect on how much people give

• Recent donations – and even more so the minimum of past donations  – seem to have some “anchoring” effect

• No evidence that shared identity (gender) has any effect

Page 5: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 6: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 7: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Possible effects of social influences on donations

• Crowd‐out

• Other donations reduce individuals’ contributions• Other donations reduce individuals  contributions

– Perfect crowd out – Warr (1982), Roberts (1984)

– Imperfect crowd out – Andreoni (1989, 1990)

– Threshold effects (Andreoni, 1998)

d i• Crowd‐in

• Other donations increase individuals’ contributions

– Social norms – reciprocity (Sugden 1984) conformity (Bernheim 1994)Social norms  reciprocity (Sugden, 1984), conformity (Bernheim, 1994)

– Status (+ assumption that relative generosity matters) – Glazer and Konrad (1996), Harbaugh (1998)

– Signalling charity quality – Vesterlund (2003)

– Cognitive effects – della Vigna (2009), Scharf and Smith (2010) 

Page 8: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Evidence on the effects of social influences

• Crowd‐out

• Most papers focus on the effect of government grants (not individual donations). Evidence on extent of crowd out is mixed

• Andreoni and Payne (2009) – 73% crowd out, but most is indirect, i.e.Andreoni and Payne (2009)  73% crowd out, but most is indirect, i.e. through charity’s reduction in fundraising expenditures 

• Crowd in

• Survey data – Andreoni and Scholz (1998) 

• Lab experiments – Falk et al (2003) Bardsley and Sausgruber (2005)• Lab experiments – Falk et al (2003), Bardsley and Sausgruber (2005) 

• Field experiments – Frey and Meier (2004), Alpizar et al (2008), Shang and Croson (2009)

Page 9: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Evidence on the effects of social influences

• Boeg et al (2008)

• Analyse small sample of 365 fundraisers on justgiving website and associated donationsassociated donations

• Focus on effect on early donations (first two days)

• Early mode acts as an upper bound on later donations

Page 10: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Evidence on the effects of social influences

• Identity

• Meer (2008) – people give more if they are asked by someone they know and by someone with shared characteristics (race)

• Shang and Croson (2008) – people give more if they are told thatShang and Croson (2008)  people give more if they are told that someone like them (gender) has given a large donation

• Observability

• People give more if their donations are observed – Rege and Telle (2004), Andreoni and Petrie (2004), Soetevent (2005), Alpizar et al,(2004), Andreoni and Petrie (2004), Soetevent (2005), Alpizar et al, (2008)

Page 11: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 12: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

hCharities 

O li f d iOnline fundraisers15,000 “charity” runners, others who fundraise after winning a 

place in the ballot

Donors – typically friends, family and colleagues

Page 13: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Online fundraising

• Justgiving (JG)

• Set up in 2001• Set up in 2001

• Profit‐making company, charging charities £15 monthly fee, and also taking 5% gross donations (i.e. including the value of tax relief) 

• 15,000 fundraising pages for the 2010 London marathon

Vi i M Gi i (VMG)• Virgin Money Giving (VMG)

• Set up in 2009

• Not‐for‐profit, charging charities a one‐off, set‐up fee of £100 andNot for profit, charging charities a one off, set up fee of £100 and taking 2% nominal donations. 

• Estimated 9,000 fundraising pages for the 2010 London marathon

• Other, much smaller websites – bmycharity, mycharitypage, mygift

Page 14: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

30 donations per page 50 donations per page

Page 15: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 16: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Our sample Web pages13 369 from

18,126 pages622,968 donationsM d ti £32 16

13,369 from JG; 4,757 from VMG

Mean donation = £32.16

Page 17: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Our sample Web pages13 369 from

18,126 pages622,968 donationsM d ti £32 16

13,369 from JG; 4,757 from VMG

Mean donation = £32.16

Race results:Gender, age, nationalitynationality, race time for fundraisers 12,750 pages

440 130 donations440,130 donationsMean donation = £31.66

Page 18: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Characteristics of fundraisers compared to other runners 

Matched 

fundraisers

JG VMG Other       

runners

Mean hours 4.68 4.67 4.72** 4.46

Male .623 .620 .631 .693

British .987 .988 .985 .888

Age 18 – 39  .687 .693 .675* .508

Age 40 – 49 .239 .233 .251* .321

Age 50 +   .074 .074 .075 .171

N 12,750 8,716 4,034 23,774

Differences between the online fundraisers and the other runners are all statistically significant at the 1% level; 

*, ** differences between JG and VMG fundraisers are statistically significant at the 5%, 1% levels.

Fundraisers are slower, less male, more likely to be British and youngerFundraisers are slower, less male, more likely to be British and youngerthan other runners

Page 19: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Our sample Web pages13 369 from

18,126 pages622,968 donationsM d ti £32 16

13,369 from JG; 4,757 from VMG

Mean donation = £32.16

Race results:Gender, age, nationalitynationality, race time for fundraisers 12,750 pages

440 130 donations440,130 donationsMean donation = £31.66

Database of baby names:yGender for donors

12,634 pages336,298 donationsMean donation = £30.31

Page 20: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Sample summary statisticsSample summary statistics

Mean St. dev. Median

All fundraisers

Number of donations per page 34.5 25.4 29

Total raised online £1,093 £1,401 £778

Total raised offline £335 £1,115 £0

Average online donation – all   £30.31 £66.02 £20

Average online donation – made by men  £35.38 £78.36 £20

Average online donation – made by women  £24.96 £49.22 £20

Proportion of donors who are male 513Proportion of donors who are male .513

Number of fundraisers 12,750Note: All amounts exclude value of Gift Aid

Page 21: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Distribution of donationsDistribution of donations

Page 22: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Justgiving Virgin Money Giving

Page 23: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Distribution of donations, JG and VMG

Amount Fraction of donations in JG Fraction of donations in VMG

£100 .065 .076

£50 .105 .123

£20 .234 .282

£10 .263 .270

TOTAL .667 .751

Page 24: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Sample summary statistics, by fundraiser gender

Mean St. dev. Median

Male fundraisers

Number of donations per page 34.5 25.9 29

Total raised online £1,121 £1,523 £775

Total raised offline £338 £1,290 £0

Average online donation – all   £31.19 £68.08 £20

Average online donation – made by men  £35.88 £80.20 £20

Average online donation – made by women  £24.73 £45.68 £20

Proportion of donors who are male .580

Number of fundraisers 7,957

Female fundraisers

Number of donations per page 34.6 24.6 30

Total raised online £1,047 £1,170 £782

Total raised offline £330 £739 £20

Average online donation – all £28.87 £62.50 £20

Average online donation – made by men  £34.21 £73.91 £20

Average online donation – made by women  £25.23 £53.01 £20

Proportion of donors who are male .406

Number of fundraisers 4,803

Page 25: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 26: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Estimating the effect of previous donations

1 2 1 , 2ˆ

ifct i f C t fc t ifct ifctD X Z D W uα β β γ δ λ λ−= + + + + + + +

• donation of donor i giving to fundraiser f for charity c at time t

• Xi – genderXi gender

• Zf – gender, age, nationality, race time, target amount

• Indicators for charities (400 with 5+ pages), date relative to when page set up and date of marathon

• – measure of previous donations. Mean of all previous donations; mean of last ten donations. All regressions exclude first 10 donations. 

fc tD −

ea o ast te do at o s eg ess o s e c ude st 0 do at o s

• Wifct – proximity to target, order on page

• “Reflection problem”: relationship between amount given and past donations will be affected by correlated characteristics

Page 27: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Estimating the effect of previous donationsg p

• Include donor fixed effects

8 000 d i JG l i t th f d i• 8,000+ donors in JG sample give to more than one fundraiser

• Similar to Falk et al (2003) – can potentially identify effect of previous donations through variation in peer groups

• But may also pick up unobserved fundraiser characteristics

• Include fundraiser fixed effects

• Exploit sequential nature of donations

• Identification from within‐page variation in history of past donations• Identification from within‐page variation in history of past donations

• Controls for date of donation, place order in page (and day)

• Argue that exogenous variation comes from exactly when donors come to the site which will be subject to random factors – when they turn on computer, check e‐mails, get round to donating etc 

Page 28: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

OutlineOutline

• Effects of social influences• Effects of social influences

• Online fundraising

• Our sample

• Identification

• Results

Page 29: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Female donor ‐6.696** (0.102) ‐6.352** (0.103)

Female fundraiser 0 327** (0 113) 1 098 (0 563)

OLS regressions: Dependent variable = individual donations (£) 

Female fundraiser 0.327 (0.113) 1.098 (0.563)

Donor same sex as fundraiser ‐0.296** (0.102) 0.589 (0.500) ‐0.570** (0.103)

Non‐British fundraiser 5.593** (0.538) 1.795 (3.394)

Fundraiser 18‐39Fundraiser 18 39

Fundraiser 40‐44 1.362** (0.152) ‐0.091 (0.806)

Fundraiser 44‐49 2.399** (0.189) ‐1.733 (1.013)

Fundraiser 50‐54 2.090** (0.274) ‐1.076 (1.555)

Fundraiser 55‐59 2.368** (0.446) ‐4.969 (2.667)

Fundraiser 60‐64 1.121 (0.651) ‐5.113 (3.972)

Fundraiser 65‐69 3.769* (1.520) 9.816 (14.360)

Fundraiser 70+ ‐1.876 (2.089) ‐16.740 (11.070)

Target (0/1) 1.453** (0.204) ‐0.080 (1.055)

Target amount ‐0.000 (0.000) 0.000 (0.000)

Fraction of target achieved ‐3.079** (0.263) ‐2.629* (1.305) ‐9.529** (0.549)

Within day, donation number ‐0.177** (0.021) ‐0.101 (0.108) ‐0.161** (0.022)

Mean of all past donations (£) 0.294** (0.002) 0.125** (0.012) ‐0.036** (0.007)

Mean of last 10 donations (£) 0.185** (0.002) 0.094** (0.009) 0.012** (0.002)

Donor fixed effects No Yes No

Fundraiser fixed effects No No Yes

N 243,974 10,044 243,974

Additional controls for marathon time, place on page, days until marathon, days since page creation and charities with 5+ pagesStandard errors in parentheses, * p < 0.05, ** p < 0.01All regressions exclude first ten donations to a page and top 1% donations

Page 30: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

What about other moments?

In relation to all previous donations

the donation is.... 

In relation to the last ten donations

the donation is....

Less than Equal to More than Less than Equal to More than

Minimum 017 130 853 041 195 764Minimum .017 .130 .853 .041 .195 .764

Mode .259 .308 .433 .244 .296 .460

M i

Note: Sample excludes the first ten donations on each fundraising page

Maximum .945 .030 .025 .871 .066 .063

Page 31: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Estimated effects associated with different reference levels

What about other moments?

Minimum of  Mode of  Maximum of Maximum

Dependent variable = Amount given (£)

previous 

donations (£)

previous 

donations (£)

previous 

donations (£)

MaximumMaximum

x VMG

All dAll donors

All donations 0.082**

(0.024)

‐0.016**

(0.005)

0.001*

(0.000)

‐0.001

(0.001)

0.003**

(0.001)(0.024) (0.005) (0.000) (0.001) (0.001)

Last 10 

donations

0.030*

(0.013)

0.007*

(0.003)

0.002**

(0.000)

0.002**

(0.000)

0.001

(0.001)

Additional controls for gender of donor, place on page, days until marathon, days since page creation, proximity to target

Page 32: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Estimated effects associated with different reference levelsEstimated effects associated with different reference levelsDependent variable = Amount given (£)

Minimum of previous  Mode of previous  Maximum of previous donations (£) donations (£) donations (£)

All donors

All donations 0.082** ‐0.016** 0.001*

Last 10  0.030* 0.007* 0.002**

All donors

All – same sex 0.012 0.005 0.002***

Last 10 0.010** 0.017*** 0.004***

Female donors

All 0.001 ‐0.009 0.002**

Last 10 0.005 0.006 0.002**

Male donors

All 0 144** ‐0 012 0 001All 0.144 ‐0.012 0.001

Last 10 0.042* 0.010 0.002**

Additional controls for gender of donor, place on page, days until marathon, days since page creation, proximity to target

Page 33: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Discussion

• Consistent with experimental evidence, we find that recent donations have a positive effect on how much people give Strongest effecthave a positive effect on how much people give. Strongest effect comes through the minimum. Suggests social/ cognitive anchoring. 

• But the overall mean has a negative effect (as does hitting target level of donations), consistent with crowd out

• Points to importance of looking at all previous donations, not just a single reference level of donations as earlier studies have donesingle reference level of donations as earlier studies have done 

• No effects of observable shared  characteristics – but donors are likely to know each other, and fundraiser

• External validity? If anything would expect crowd out effects to be smaller than in general charity fundraising campaigns because of personal relationships

Page 34: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Future work… 

• This is the first look at a sub‐sample of a potentially much larger d t tdataset 

• > 4 million donations have been made through Justgiving

– Larger donor panel and networks of donorsLarger donor panel and networks of donors

– Bilateral relationships – You sponsor me, I sponsor you 

– Rolling out an online survey of all previous donations to collect demographic and socio‐economic information 

Page 35: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page
Page 36: The effect of social influences on gggiving evidence from ... · N 243,974 10,044 243,974 Additional controls for marathon time, place on page, days until marathon, days since page

Charity

Coefficient 

estimate t‐statistic

No. of 

pages

MACMILLAN CANCER SUPPORT  [medical] ‐0.2503 ‐0.4903 356

WHIZZ‐KIDZ  [children/disability] 1.7265 3.1865 266

CANCER RESEARCH UK  [medical] ‐1.2321 ‐2.2903 241[ ]

CHILDREN WITH LEUKAEMIA  [children/medical] ‐0.5160 ‐0.7756 232

CLIC SARGENT  [children/medical] ‐1.4433 ‐2.4907 232

NSPCC [ hild ] 1 2271 2 2284 232NSPCC  [children] ‐1.2271 ‐2.2284 232

GET KIDS GOING!  [children/disability] 0.5407 0.9370 204

HELP FOR HEROES  [veterans] ‐2.9563 ‐4.6137 179

ALZHEIMERS SOCIETY  [medical/elderly] ‐2.0164 ‐3.4035 163

SENSE  [disability] ‐1.5002 ‐2.1452 152

NATIONAL DEAFBLIND AND RUBELLA ASSOC’N  [disability] ‐1.6543 ‐2.7074 146[ y]

BRITISH HEART FOUNDATION  [medical] ‐0.1670 ‐0.2762 139

THE CHILDREN'S TRUST  [children/disability] 0.3402 0.5660 138

ASTHMA UK [ di l] 2 5325 4 0139 129ASTHMA UK  [medical] ‐2.5325 ‐4.0139 129

MS SOCIETY  [disability] ‐0.9050 ‐1.4123 115

SHELTER  [homeless] ‐0.2845 ‐0.4346 109

PHAB  [community] ‐1.5972 ‐2.4102 105

ST JOHN AMBULANCE  [medical/ambulance] ‐2.0714 ‐3.1594 104

HELP THE HOSPICES  [elderly/medical] 0.1251 0.1969 102