pull down to refresh
42 sats \ 3 replies \ @SimpleStacker OP 11h \ parent \ on: Research in Public #03: TWFE regression of # posts on territory fees econ
Any linear effect of the number of available territories should be captured in the week fixed effects. Of course, there could be nonlinear relationships that aren't modeled here.
More to the point, at some point it may be worth writing down what we think the structural model of user behavior is.
Good point. I was imagining including the log of the count, but a lot of that variation would still be absorbed by the fixed effects, especially if there are always about the same number of territories.
The more relevant number of substitutes per territory isn't captured by the fixed effect, but we also don't know it.
That model of user behavior will definitely have expected returns in it, or we'll both lose our licenses to practice economics.
reply
It might be fun to try and write down a structural model that captures every incentive.
Posters are motivated by i) expected earnings on zaps, ii) post cost, iii) inherent desire to post the content. They generate post ideas at random of varying quality, then decide whether or not to post, and which territory to post in.
Territory owners are motivated by i) expected earnings on fees, ii) billing costs, iii) inherent desire to see certain content. Each month they decide whether or not to pay the billing (we'll abstract from the options for yearly/perpetual payment)
Zappers are motivated by i) cost of zaps, ii) inherent desire to see certain content.
For now, we'll make it all separable, so a person's decision to post, run a territory, and zap are all independent of each other. I think it'd be way too hard to model a single entity with a three-fold role of poster, territory owner, and zapper.
Oh, and I guess rewards would have to factor in somehow too, but that seems tricky without knowing the formula for rewards.
reply
We know the formula for rewards, or at least it's public and @k00b would presumably explain it if we asked. The bigger problem is that the rewards are very complicated (intentionally so) and stackers likely factor it into their choices in a vague way, rather than precisely. I've actually been meaning to make a ~meta post about the saliency of Stacker News rewards.
Within "inherent desire to post" are a couple of separate things that we can split out and use proxxies for.
- Engagement: How many comments does a post generate? How many zaps do those comments receive?
- Reach: How many zaps were received? How many different stackers commented?
I was thinking about the choice of territory to post in and I think there are three driving forces:
- Desire to post in the most relevant territory
- Posting costs in the relevant territories
- Relationship with territory owners
I doubt we could quantify this claim, but perhaps we can handwave something about profits being maximized in a relevant territory to the subject matter of the post because of visibility, owner zapping, downzapping, and community engagement.
I really like the idea of random idea formation. I have no idea what distribution assumptions your going to make, but I trust your instincts.
reply