Thread by @elliott_jt

1/ Check out NBER SI IO @ 11 am EDT today to listen to @ptscott present our paper "Market Structure, Investment and Technical Efficiencies in Mobile Telecommunications" (joint with Georges V. Houngbonon and @marcivaldi). Link: youtube.com/channel/UC79EL…

2/ A quick overview: we study how scale efficiencies and market power in the mobile telecommunications industry impact how we should think about antitrust policy, as well as the allocation of spectrum.

3/ Economists have made substantial progress in evaluating market power but have fewer tools for evaluating scale efficiencies.

4/ To quantify returns to scale in this industry, we use data from France and build a model based on physical principles to map the level of investment into experienced download speeds.

5/ Essentially, we build the production function from engineering relationships, which we then use to consider regulation in this industry.

6/ Main result: Scale efficiencies are sizable in this industry (larger carriers increase the download speeds), but consumers ultimately prefer a relatively high number of firms even at the cost of lower download speeds.

7/ The engineering model exhibits two types of scale efficiencies: economies of density and economies of pooling. Economies of density result from path loss: it is more efficient to transmit data across smaller distances.

8/ The more base stations a firm has, the shorter the distance between consumers and stations, and the higher the download speeds. Merging base stations networks of two firms results in shorter average distances = higher download speeds (holding investment fixed).

9/ Economies of pooling result from queuing (when multiple consumers request data at the same time).

10/ Merging two firms' networks allows the merged firm to more efficiently allocate network capacity among consumers, which means less congestion and therefore higher download speeds.

11/ What do we find? As we increase the number of firms, the number of base stations built per firm exhibits an inverse U-shaped relationship, increasing from 1 to 2 firms, then declining after 2. Download speeds, however, are monotonically declining in the number of firms.

12/ We find that consumer surplus is maximized at 7 firms, while total surplus is maximized at 4. This headline result masks heterogeneity across consumers. Low-income consumers prefer a high number of firms, which yields low prices but low download speeds.

13/ High-income consumers, on the other hand, prefer a low number of firms, which yields high prices and high download speeds.

We next use our model to consider the value of spectrum.

14/ Typically, spectrum has been allocated through auctions, which capture the value of the spectrum to the firm. The social value of spectrum, however, may be different from the firm's willingness to pay.

15/ Indeed, we find that the marginal consumer surplus from additional spectrum is roughly 5x a firm's willingness to pay.

Lastly, we ask: how should new spectrum be allocated within the industry?

16/ We consider two possibilities: allocating bandwidth across incumbent firms or adding a new firm endowed with the new bandwidth. While in terms of total surplus, the former is better, in terms of consumer surplus, the latter is.

17/ Finally, a bonus for IO economists: due to data limitations we only have product-level market shares for 1 firm and national market shares for others. We show in paper that we can rationalize agg. market shares for the other firms, and the BLP contraction mapping still works.

18/ We are very happy to hear comments and suggestions!
Paper: jonathantelliott.com/files/telecom_…
Slides: jonathantelliott.com/files/telecom_…
Code: github.com/jonathantellio…