We tested our predictions using data from GitHub, the largest OSS hosting community. The data allows us to observe different types of activities among community participants: founding repositories, contributing code, advice exchange (=opening issues), or project maintenance (=managing incoming contributions) and related communication – with the last activity being reserved to project founders only. Data had all the features we needed to operationalize two behavioral approaches founders can adopt and link them to the growth of founders' projects. As mentioned earlier, we found evidence for our hypotheses.

However, as every empirical study is imperfect, we needed to perform a few robustness checks in order to be able to claim the validity and generalizability of our results. Here are a few points/limitations we need to mention:

1) There might be an issue with the construct validity for our key independent variable that explains founder’s project maintenance work. Generally speaking, project management and maintenance work consists of two crucial parts - managing incoming contributions and providing feedback to contributors. Our measure, however, only includes the former since we had no data records available on the latter commenting part. Theoretically, this omission could distort our results. However, practically, based on this data observation, we can only assume that founders do not consider this activity as being particularly important for the project. In fact, in our related interview studies, a lot of project founders mentioned that this activity takes away their valuable time and they'd rather delegate it, if it was possible.

2) For our final data sample we had to exclude inactive GitHub accounts and empty repositories. For robustness, we reran all analyses with such zero records - results are largely in line with our main findings.

3) Our sample might still be considered as little old. Perhaps, this might in part explain why we still do not observe a lot of pull request activity. For robustness, we tested our hypotheses on a subsample of newer data - results remain unchanged.

Overall, we are confident that our findings are robust. Currently, we are working on the "study contributions" section.