The Effect of Social Information on Online Giving Behaviors

Post on 25-May-2015

147 views 2 download

Tags:

description

from a paper presentation at ARNOVA 42nd conference at Hartford, CT, USA, 2013

Transcript of The Effect of Social Information on Online Giving Behaviors

Shusaku SASAKI Osaka School of International Public Policy, Osaka University Daigo SATO JustGiving Japan Foundation

ARNOVA 42nd Annual Conference, 21 November, 2013

1

The Effect of Social Information on Online Giving Behaviors

Conformity in Charitable Contributions

2

Conformity in Charitable Contributions

3

We often compare our own choice with others’ choices, because…

����������"����%%��"��!�"�����!"�����"�������#�"

��������� ������ ���� ��"��������!����!

������#!"�%��"�"�����"���!����"�����!�"����"�� !

����������� �!��

�� ����$��������������� ��"�� ���"����"�� ! ������ ���"�������� �!

����� ���

Conformity in Charitable Contributions

4

Do we “Always” care about the others?

Conformity in Charitable Contributions

5

Do we “Always” care about the others?

When and How conformity is likely to be an important factor in charitable contributions?

Previous studies

6

Previous studies

7

Several experiments have already demonstrated

the existence of conformity.

Previous studies

8

������ ������

��������������

��������������

������ ���������

���

Previous studies

9

������ ������

��������������

��������������

������ ���������

���

l  Average Participation Rate l  Average Donation Amount

<

Limitations

10

Limitations

11

������ ������

��������������

��������������

������ ���������

���

<

How much Representative this amount is among others’ contributions?

Our study

12

Our study

13

We investigate the impact of the degree of variation of others’ contributions.

Our study

14

We investigate the impact of the degree of variation of others’ contributions.

Our study

15

6

Figure1 Sample of an online fundraising campaign page

3.2 Data

� We use data on consisting of all the fundraising campaigns launched on JustGiving.jp between March 2010

and December 2012. Our unit of observation is each fundraising campaign, and our basic data include 4,046

observations. With an average of 22 donors per campaign, the data reflect about 88,187 donations. The data

for each campaign include the following: target price of the campaign, total sum donated, total number of

donors, duration from the first donation to the last donation, and whether the campaign has set the ending date

(some do not have the ending date). In addition, the categories of the campaign purpose are included in the

data5: a campaign for reconstruction from the Great East Japan Earthquake and Tsunami, for international

cooperation, education, sports promotion, child-care, family-care, medical issues, reconstruction from the

other disasters, a culture of donation, environmental conservation, animal protection, local development, etc.

The information could be closely related to the trend of personal attributes of donors (e.g., the larger number

of female donors could donate to the campaigns for child-care).

� Crucial for our analysis, the data include information about the distribution of the donations: not only mean,

median, and mode price of the donations, but also number of donors who donate in the mode and standard

deviation of the donations. What is more important is that the data can include the similar information within

a certain period: mean, median, and mode price of the donations in the first 3 days, number of donors who

donate in mode price, and standard deviation of the donations in the first 3 days.

� In order to strengthen the assumption that potential donors would read the list of the previous donations, we

drop 474 outlier campaigns with much larger number of total donors. It is because that there are more

possibilities that potential donors cannot easily read and know the distribution of all the previous donations in

5 We classify a category of each campaign’s purpose by checking both the title and message of the page and the activity field of a charity that is to benefit from the campaign.

The$Title$of$$the$Fund.raising$Campaign�

Fund.raiser’s$name�

Message$××××××××××××××××××××�××××××××××××××××××××�××××××××××××××××××××���

GOAL$$$��○○$yen$RAISED��○○$yen�

○○��

Donate$Now!!!!�

DonaGon$by$○○$on$hh:mm,$MM/dd/yyyy$� ○○○○$yen�

DonaGon$by$●●$on$hh:mm,$MM/dd/yyyy$� ●●●●$yen�

DonaGon$by$○●$on$hh:mm,$MM/dd/yyyy$� ○●○●$yen�

 Donors read and know the distribution of the previous donations.

Our limitations

16

Our limitations

17

Online dataset & Privacy Policy

Our study

18

Individual unit dataset Campaign unit dataset

Our study

19

��� ��� ��� �� ��

��� � ���� ��� ��

��� ��� ��� �� ��

��� � ��� ��� ���

����������� ����������������������

����������

����������

��

In each fundraising campaign, we make Two kinds of information.

Our study

20

��� ��� ��� �� ��

��� � ���� ��� ��

��� ��� ��� �� ��

��� � ��� ��� ���

����������

����������

������������ ��� ������ �����������

����������������������������������!�

Our study

21

� � ��� ����

� ���

����������

������������ ����"�� ��� �� ���� ���

����������

��!���������������� ����� ��������������

����������"� ����������� �����#�

Results

22

Results

23

Table 2.A Whether the mode of the rest of the period is equal to the mode of the first 3 days or not?

(dsame=1, if the later mode is equal to the early mode)

Regression model: Logit (1) (2) (3) (4)

rate_number_early_mode 2.161*** 2.483***(0.449) (0.488)

early_coeffecient_of_variation -0.564*** -0.590***(0.184) (0.194)

early_mode 0.0305 0.0344(0.156) (0.162)

early_mean 0.114 0.118(0.0819) (0.0845)

total number of donors 0.0462*** 0.0548*** 0.0318*** 0.0357***(0.0117) (0.0125) (0.0109) (0.0114)

rate of unique donors 0.365 0.560 0.118 0.269(0.463) (0.490) (0.453) (0.477)

target price 0.000522 0.000328 0.000578 0.000405(0.000681) (0.000696) (0.000705) (0.000718)

campaign with the deadline -0.161 0.169 -0.106 0.209(0.203) (0.243) (0.200) (0.241)

duration to the last donation -0.000323** -0.000318** -0.000290** -0.000280**(0.000134) (0.000141) (0.000132) (0.000140)

the other controls No Yes No Yes

Constant -2.593*** -5.262*** -0.563 -2.733***(0.616) (1.049) (0.469) (0.914)

Log likelihood -448.79659 -429.08553 -455.86097 -437.81809 Pseudo R^2 0.0383 0.0806 0.0232 0.0619

Observations 698 698 698 698Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

Dependent variable: dsame Rate of number ofearly mode Coefficient of variation

Strength of the trend of the others' contributions

Control for the Level of the trend of the others' contributions

Controls for Basic Attributes of each campaign

When the distribution of the early donations is smaller, the later donors are more likely conform to the trend among the early donations.

Discussion and Conclusions

24

25

If the stronger trend of conformity among the early donations exists,

the later donors are more likely to conform to the trend.

Discussion : The strong magnitude of the trend among the early donations might encourage the attitude of conformity of the later donors or gather the later donors who have already had high attitude of conformity. This study contributes to a further understanding toward when and how the conformity is likely to happen in the charitable contributions.

Discussion and Conclusions

Message

26

27

Our research is largely supported by JustGiving Japan, one of Japanese NPOs. They provided us valuable dataset very kindly and understood the importance of our research. I hope that there will be lots of collaborative works between practitioners and researchers. Thank you very much for your attention.

Message