|
Abstract:
|
With the proliferation of user-generated content online, many content platforms have been exploring ways to monetize content. A recent approach adopted by many platforms is to provide content providers with the freedom to launch paid content. This thesis proposes a framework that is suitable for demand estimation in content markets. We generalize Orthogonal Random Forest, a flexible nonparametric method for estimating heterogeneous treatment effects in the presence of high-dimensional confounders, by enabling Deep Neural Network in estimating nuisance functions. We apply this approach to Zhihu, a leading knowledge-sharing platform in China, which started as a free Q&A platform and later allowed users to deliver exclusive live-streaming talks. We estimate how price elasticity varies over the temporal distance to the live session and across seller and product characteristics. We also evaluate the externalities of monetizing content on the platform “ecosystem.” Leveraging the context of Zhihu, we find that content providers holding priced talks tend to contribute more free content than those using the free Q&A platform only. Potential mechanisms for such behavior are explored.
|
