Research in Public #11: A first draft manuscript is now complete -- summary of results^[1]

I won't link to the manuscript, since it instantly doxxes my identity and most people here wouldn't want to read the full paper anyway. It's also not really public in the sense that I haven't posted it to any public outlets, but it is available to read on my github for anyone who wants to dig for it. When I do post it to a public outlet is arxiv or SSRN, I will let Stackers know.

But I did want to summarize the main results here and get some feedback. I decided to pare down some of the paper's ambitions. For example, I am no longer attempting to investigate adoption of self custody in this paper. It just felt like adding that to the paper would be making the paper "try to do too much." I thought it was much more cohesive to simply focus on the incentive structures on Stacker News and how that affects the quality of discourse. The paper is of decent length already (35 pages), and I think it'll get longer as more exercises and robustness checks are suggested. I plan to return to the question of self-custody in a later project.

So anyway, here are the main results and some brief commentary on each. Please keep in mind that the results and conclusions are preliminary, and may change as more work is done.

1. Pay-to-post increases content quality:
   
   We demonstrate this by showing that when a territory increases its posting fee, the number of posts goes down but the quality of posts go up (as measured by either amount of zaps received or number of comments received).
   
   The thought experiment goes like this: imagine two similar territories A and B. A increased its posting fee but B did not. The data shows that territory A will subsequently experience a decline in the number of posts but an increase in post quality, relative to territory B.
   
   The results suggest that pay-to-post is an effective tool for screening out low quality posts and encouraging high quality ones.
   
    
2. Value-for-value increases content quality:
   
   We demonstrate this by showing that users learn from their previously received zaps. Users who received more of their zaps from high quality posts are more likely to subsequently make high quality posts.^[2]
   
   The thought experiment goes like this: imagine two users who have made the same number of posts, including the same number of high quality and low quality posts. But by chance user A received more zaps on their high quality posts than user B. The data shows that user A will then subsequently be more likely to make a high quality post than user B.
   
   Perhaps because of this effect, the share of high quality posts on Stacker News has been increasing.
   
   The results suggest that value-for-value incentives can shape the type of content that gets posted and increase the quality of content over time.
   
    
3. Profitiability affects user retention:
   
   We demonstrate this by showing that users who have been unprofitable for the last 8 weeks are more likely to become inactive than users who have been profitable. This effect is stronger when the Bitcoin-USD price has recently appreciated.
   
   The thought experiment goes like this: imagine two users who made the same number of posts over the last 8 weeks. However, user A was profitable and user B was unprofitable. The data shows that user B is more likely to become inactive than user A, and that if the Bitcoin price recently appreciated, this difference would be even greater.
   
   The result suggests that Stacker News tends to retain users who make posts that other people like, but churns out users who make posts that others don't like.
   
    

The abstract summarizes our results nicely:

We study the role of financial micro-incentives in driving user behavior on a Bitcoin-based discussion platform. On the platform, users must pay Bitcoin to post but can also earn tips from other users for posting good content. We find three main results. First, when the cost to post in a subforum increases, the number of posts in that subforum decreases but the quality of the posts increases. Second, the incentive to earn tips leads to higher post quality over time as users learn what content others value. Third, unprofitable users are more likely to leave the platform while profitable users are more likely to stay. The findings suggest that financial micro-incentives may be effective tools to improve the quality of discourse in online discussion platforms. Moreover, the incentives present on the platform are only possible because of Lightning Network's ability to instantaneously process small, global micro-payments, thus highlighting an important use case of Bitcoin.

Feedback is welcome!

1. This is a series of public research journals for an academic research project using Stacker News data. Please see #1243188 and #1269944 for more details. ↩

2. High quality, in this case, is measured not by zaps but based on the number of words in the post and whether the post contains images and links contained in the post body. Essentially, high quality tends to mean long-form multimedia content. ↩