How does Google price the items on Google Play?

Question

I am a fledgling computer science researcher working in intersection of economics and computation. Please excuse me if this question seems out of place to the administrator.

I am currently studying different pricing models and studying the "optimal" way or price to sell items. I have noticed that on Google Play there are movies which one could rent or buy. Also, for movies for similar popularity I noticed that there is often a huge difference in their price. (about 10 euros). Conventionally, one looks at the valuation distribution for an item. However, that won't explain this difference in cost. There may be issues of purchasing copyrights too. However, more generally, how do companies like Google price these digital items? Do they update these prices regularly based on sale information? Is there any prior work about how can one go about doing this?

I am just looking for directions so any help is greatly appreciated.

Answer 1

The simple answer is they estimate the demand curves for each product and, using their cost structure and market characteristics (competition structure, etc.) set price to maximize profits. This is standard for any firm, though.

How Google in particular and these big firms in general (Amazon, Microsoft, etc.) estimate demand curves is somewhat different than the usual economist might do it. For usual demand estimation, a researcher would have to make use of market idiosyncrasies to identify demand. For example, using supply shifters with 2SLS for basic demand estimation, BLP for discrete choice with heterogeneous products, etc. Identification is such a big issue for demand estimation because a researcher generally just observes equilibrium (p, q) combinations, not the actual demand curve. We are also often constrained purely by the amount of data available.

For a big firm like Google, however, they 1) have the ability to enact exogenous perturbation in price to see how sales change and 2) have access to tons and tons of data. Using 1) they are constantly running little experiments to see how consumer behavior changes. They can then use the results to actually trace out the demand curve. In these experiments, the firm could easily take into account things like movie popularity, genre, etc. With respect to 2), Pat Bajari, chief economist at Amazon and one of the biggest names in modern empirical IO, has a (at this point) working paper with Nekipelov, Ryan, and Yang on how to use machine learning to estimate demand curves across products with lots of sample points bunches of characteristics (think thousands of product characteristics). As a "fledgling computer science researcher," you'd probably be in to this. This approach is especially relevant for people/firms with access to tons of data (like Google, Amazon, etc.)