Incentivizing Innovation in Open Source: Evidence from the GitHub Sponsors Program

NBER Working Paper 31668

This paper seeks to understand the effects of the introduction of GitHub Sponsorships on digital innovation.

I work on the project as a co-author with Annamaria Conti (IE Business School), Maria Roche (Harvard Business School), and Jorge Guzman (Columbia Business School) - contributing to the data pipeline, analysis, and the experimental setup.

My responsibilities include:

1. Designing a data pipeline from scratch to create an unbalanced panel dataset of ~15000 GitHub users and their repositories.
2. Executing the data pipeline in Python on a Unix environment, and processing the data into a panel using STATA.
3. Merging datasets collected using GraphQL and GitHub Archive, both of which had differing formats and structures.
4. Conducting data analyses in STATA, and contributing to experimental design as well.
5. Improving the empirical strategy for the paper to fix the challenges we encounter.
6. Creating a survey and a discrete choice experiment to understand users’ motivations behind their contributions to open source.

Abstract:

Open source is key to innovation, but we know little about how to incentivize it. In this paper, we examine the impact of a program providing monetary incentives to motivate innovators to contribute to open source. The Sponsors program was introduced by GitHub in May 2019 and enabled organizations and individuals alike to reward developers for their open source work on the platform. To study this program, we collect fine-grained data on about 100,000 GitHub users, their activities, and sponsorship events. Using a difference-in-differences approach, we document two main effects. The first is that developers who opted into the program, which does not entail receiving a financial reward, increased their output after the program’s launch. The second is that the actual receipt of sponsorship has a long-lasting negative effect on innovation, as measured by new repository creation, regardless of the amount of money received. We estimate a similar decline in other community-oriented tasks, but not in coding effort. While the program’s net effect on users’ innovative output appears to be positive, our study shows that receiving an extrinsic reward may crowd out developers’ intrinsic motivation, diverting their effort away from community and service-oriented activities on open source.