Are Donation Badges Appealing?:
A Case Study of Developer Responses
to Eclipse Bug Reports
Keitaro Nakasai, Hideaki Hata, and Kenichi Matsumoto
IEEE Software ( Volume: 36 , Issue: 3 , May-June 2019 ), ESEC/FSE 2019 Journal First
Now you can
donate to Eclipse
Dec 2007
special mirror site
‘Friend of Eclipse’ logo
Bug 282088 -
Expand Friends of
Eclipse Program
Jul 2009
Linux Foundation offered
various benefits
New Friend of Eclipse
Program Launched
Oct 2014
T-shirt, discount on
EclipseCons, O'Reilly
discount, etc.
Bug 514954 - Make
"Friends of Eclipse"
more appealing
Apr 2017
Bug 434249 - Add
decorator for
Friends of Eclipse
Nov 2014
Badges implemented
Bug 496899 - Remove
Friends mirror
Jun 2016
The friends mirror is no longer
needed
Bug 514954 - Make
"Friends of Eclipse"
more appealing
Apr 2017
I want to give back since Eclipse technology is free 84.44%
I want to support open source 71.85%
I want a t-shirt/other benefits 5.93%
I want to help improve Eclipse technology 72.59%
I feel a sense of closeness to a community or group 9.63%
Giving people more value as a Friend has proved not to make a
difference over the past 3 years and our various campaigns. That is
why we removed it in September 2016.
We ran a survey in June 2016 and over 84% of our respondents we
non committers and most of them (80%) have been using Eclipse for
over 5 years. What they said was:
We no longer promote the Friend status, but kept it at 35$ in the
background and FAQs.
A friend of Eclipse badge
(Donation badge)
Research Question
Are Donation Badges
Appealing?
Do donation badges affect developer
response time?
Difference in Differences
One of the core methods in today’s
econometrics toolkit to address causal
questions.
Regression analysis estimating the effect of
sharp changes in the economic environment
or changes in government policy.
DID Model
Trend in the control group
Trend in the treatment group
Counterfactual trend
in the treatment group
Treatment effect
Quantile DID
Bug 434249 - Add
decorator for
Friends of Eclipse
Nov 2014
Badges implemented
Nov 13, 2014
Before (2 years) After (2 years)
Nov 13, 2014
Before (2 years) After (2 years)
At least one bug report in both periods
957
31
Nov 13, 2014
Before (2 years) After (2 years)
# of months worked in Bugzilla
# of bug report submissions in the periods of before and after
# of commits in Git repositories
957
31
31
Propensity Score Matching
Nov 13, 2014
Before (2 years) After (2 years)
31
31
617
453 245
507
Metrics for bug reports: Severity, OS, Component, Community, Social relationship
Metric Description Coeffs P-value
Donor Reporter is a donor or not 2.259 0.007
Period Submitted time is before or after -1.598 0.193
Badge It has a donation badge or not -2.219 0.037
Enhancement Severity is enhancement or not 0.668 0.412
Windows Issue is related to Windows or not 1.119 0.106
Linux Issue is related to Linux or not 0.675 0.477
MacOS Issue is related to MacOS or not 0.793 0.382
Component Response days in median for the belonging components 0.317 0.109
Community # of contributors in the belonging components 0.000 0.917
Time A numerical order of time in months 0.152 0.193
Relationship # of reports in which the reporter and the first responder have worked together 0.005 0.371
Metric Description Coeffs P-value
Donor Reporter is a donor or not 2.259 0.007
Period Submitted time is before or after -1.598 0.193
Badge It has a donation badge or not -2.219 0.037
Enhancement Severity is enhancement or not 0.668 0.412
Windows Issue is related to Windows or not 1.119 0.106
Linux Issue is related to Linux or not 0.675 0.477
MacOS Issue is related to MacOS or not 0.793 0.382
Component Response days in median for the belonging components 0.317 0.109
Community # of contributors in the belonging components 0.000 0.917
Time A numerical order of time in months 0.152 0.193
Relationship # of reports in which the reporter and the first responder have worked together 0.005 0.371
Donation badges decrease developer response time 2 hours (median).
The median of developer response time in 1,822 studied bug reports is 3.5 hours.
Implications
Donation badges can work as
signals, similar to badges in GitHub
[Trockman et al., ICSE 2018].
Monetary factors can affect software
development processes.
Statistical causal inference can be a
useful toolkit for supporting decision
making.
Limitations
Threats to internal validity: there can
be unknown factors.
The badge effect is not eternal.
Designing and maintaining the
system is needed.
DID cannot work in other conditions.
Other methods: [Zhao et al., ASE
2017], [Murphy-Hill et al., ICSE 2019].
Are Donation Badges Appealing?: A Case Study of Developer Responses to Eclipse Bug Reports

Are Donation Badges Appealing?: A Case Study of Developer Responses to Eclipse Bug Reports

  • 1.
    Are Donation BadgesAppealing?: A Case Study of Developer Responses to Eclipse Bug Reports Keitaro Nakasai, Hideaki Hata, and Kenichi Matsumoto IEEE Software ( Volume: 36 , Issue: 3 , May-June 2019 ), ESEC/FSE 2019 Journal First
  • 3.
    Now you can donateto Eclipse Dec 2007 special mirror site ‘Friend of Eclipse’ logo Bug 282088 - Expand Friends of Eclipse Program Jul 2009 Linux Foundation offered various benefits New Friend of Eclipse Program Launched Oct 2014 T-shirt, discount on EclipseCons, O'Reilly discount, etc. Bug 514954 - Make "Friends of Eclipse" more appealing Apr 2017 Bug 434249 - Add decorator for Friends of Eclipse Nov 2014 Badges implemented Bug 496899 - Remove Friends mirror Jun 2016 The friends mirror is no longer needed
  • 4.
    Bug 514954 -Make "Friends of Eclipse" more appealing Apr 2017 I want to give back since Eclipse technology is free 84.44% I want to support open source 71.85% I want a t-shirt/other benefits 5.93% I want to help improve Eclipse technology 72.59% I feel a sense of closeness to a community or group 9.63% Giving people more value as a Friend has proved not to make a difference over the past 3 years and our various campaigns. That is why we removed it in September 2016. We ran a survey in June 2016 and over 84% of our respondents we non committers and most of them (80%) have been using Eclipse for over 5 years. What they said was: We no longer promote the Friend status, but kept it at 35$ in the background and FAQs.
  • 6.
    A friend ofEclipse badge (Donation badge)
  • 7.
    Research Question Are DonationBadges Appealing? Do donation badges affect developer response time?
  • 8.
    Difference in Differences Oneof the core methods in today’s econometrics toolkit to address causal questions. Regression analysis estimating the effect of sharp changes in the economic environment or changes in government policy.
  • 9.
    DID Model Trend inthe control group Trend in the treatment group Counterfactual trend in the treatment group Treatment effect Quantile DID
  • 10.
    Bug 434249 -Add decorator for Friends of Eclipse Nov 2014 Badges implemented Nov 13, 2014 Before (2 years) After (2 years)
  • 11.
    Nov 13, 2014 Before(2 years) After (2 years) At least one bug report in both periods 957 31
  • 12.
    Nov 13, 2014 Before(2 years) After (2 years) # of months worked in Bugzilla # of bug report submissions in the periods of before and after # of commits in Git repositories 957 31 31 Propensity Score Matching
  • 13.
    Nov 13, 2014 Before(2 years) After (2 years) 31 31 617 453 245 507 Metrics for bug reports: Severity, OS, Component, Community, Social relationship
  • 14.
    Metric Description CoeffsP-value Donor Reporter is a donor or not 2.259 0.007 Period Submitted time is before or after -1.598 0.193 Badge It has a donation badge or not -2.219 0.037 Enhancement Severity is enhancement or not 0.668 0.412 Windows Issue is related to Windows or not 1.119 0.106 Linux Issue is related to Linux or not 0.675 0.477 MacOS Issue is related to MacOS or not 0.793 0.382 Component Response days in median for the belonging components 0.317 0.109 Community # of contributors in the belonging components 0.000 0.917 Time A numerical order of time in months 0.152 0.193 Relationship # of reports in which the reporter and the first responder have worked together 0.005 0.371
  • 15.
    Metric Description CoeffsP-value Donor Reporter is a donor or not 2.259 0.007 Period Submitted time is before or after -1.598 0.193 Badge It has a donation badge or not -2.219 0.037 Enhancement Severity is enhancement or not 0.668 0.412 Windows Issue is related to Windows or not 1.119 0.106 Linux Issue is related to Linux or not 0.675 0.477 MacOS Issue is related to MacOS or not 0.793 0.382 Component Response days in median for the belonging components 0.317 0.109 Community # of contributors in the belonging components 0.000 0.917 Time A numerical order of time in months 0.152 0.193 Relationship # of reports in which the reporter and the first responder have worked together 0.005 0.371 Donation badges decrease developer response time 2 hours (median). The median of developer response time in 1,822 studied bug reports is 3.5 hours.
  • 16.
    Implications Donation badges canwork as signals, similar to badges in GitHub [Trockman et al., ICSE 2018]. Monetary factors can affect software development processes. Statistical causal inference can be a useful toolkit for supporting decision making.
  • 17.
    Limitations Threats to internalvalidity: there can be unknown factors. The badge effect is not eternal. Designing and maintaining the system is needed. DID cannot work in other conditions. Other methods: [Zhao et al., ASE 2017], [Murphy-Hill et al., ICSE 2019].