Replication Read Me

Jonathan Elliott, Georges V. Houngbonon, Marc Ivaldi, and Paul Scott

This document describes the code used to reproduce Market Structure, Investment, and Technical Efficiencies in Mobile Telecommunications (Elliott, Houngbonon, Ivaldi, and Scott, Journal of Political Economy, 2025). This document is a subset of the full replication package read me. The full replication package is available here.

This document is organized as follows:

• Description of code to run model
• Description of code to clean data and produce paper results
• Example of how to run equilibrium simulation code (can skip to here), which can be adapted for other settings

Readers interested in computing an equilibrium given a specified market structure should consult section 3 and the notebook example_counterfactual_simulation.ipynb, which demonstrates how to compute an equilibrium.

---

1. Code to Run Model

The following Python files implement the model of supply and demand presented in the paper. There are three packages: demand, supply, and welfare, each containing several modules.

demand/

• demandfunctions.py: Supplies functions that summarize demand, including market shares, elasticities, and diversion ratios
• coefficients.py: Supplies functions that return demand parameters for each type of consumer
• dataexpressions.py: Supplies functions that return expected data consumption and expected utility derived from data consumption
• demandsystem.py: Defines a class DemandSystem, containing market data and related functions
• iteration.py: Supplies functions used to solve for ξ’s quickly, adapted from PyBLP (Conlon and Gortmaker, 2020) to use autograd

supply/

• costs.py: Supplies functions to back out marginal costs based on observed investment and pricing decisions
• infrastructureequilibrium.py: Computes the market equilibrium given firms’ costs, demand, bandwidth allocations, and spectral efficiencies
• infrastructurefunctions.py: Describes signal quality based on investment levels
• priceequilibrium.py: Describes pricing first-order conditions and elasticities
• transmissionequilibrium.py: Calculates average download speeds

welfare/

• welfare.py: Summarizes welfare based on prices and investment levels

---

2. Code to Produce Paper

Running the following files in order reproduces the paper:

Process market sizes

• msize_alt.do: Computes market sizes
• agg_shares.do: Constructs aggregate market shares

Process choice set

• chset.do: Constructs the choice set

Process download speeds

• import_ookla.do: Combines speed test datasets together
• get_insee_code.py: Determines municipality from latitude-longitude
• ookla_quality.do: Determines average download speeds by operator-market

Process data consumption

• dbar.do: Constructs average data usage from Orange customer records

Process incomes

• income_alt.do: Constructs income distributions

Process infrastructure

• pop_dist.R: Computes population densities
• base_stations.do: Imports coordinates of base stations
• base_stations.R: Combines population density and base station data
• infrastructure_calibration.do: Constructs infrastructure data by operator-market

Process data for demand estimation

• demand_data.do: Constructs dataset used for estimating demand

Create descriptions of data

• contracts_table.do: Summarizes phone contract data
• data_consumption_patterns.do: Summarizes data consumption over the day
• contracts_table.py: Creates phone plans table
• descriptive_graphs.py: Creates descriptive graphs
• summary_stats.py: Creates data summary tables
• data_consumption_patterns.py: Data consumption statistics

Estimate demand

• demand.py: Estimates demand parameters

Process demand estimations

• process_demand.py: Creates tables and graphs of demand estimates

Run counterfactual simulations

• preprocess_counterfactuals.py: Constructs counterfactual simulation inputs
• counterfactuals.py: Runs counterfactual simulations
• process_counterfactual_arrays.py: Constructs standard errors on counterfactual outcomes

Process counterfactual simulations

• process_counterfactuals.py: Summarizes counterfactual results

Auxiliary files

• paths.py: Provides file locations
• moments.py: Lists moments used for demand estimation
• variancematrix.py: Variance matrix functions
• weightingmatrix.py: GMM weighting matrix functions
• estimation.py: Functions to estimate demand via GMM
• gmm.py: GMM objective function

---

3. Equilibrium Simulation Example

This package includes a notebook example_counterfactual_simulation.ipynb that shows how to:

• Compute an equilibrium in prices and infrastructure given demand, products, bandwidth, costs, and market characteristics
• Modify market structure or product offerings

Requirements to Run Notebook

• Python version: 3.9.19
• NumPy: 1.22.4
• SciPy: 1.13.1
• Pandas: 1.1.5
• Autograd: 1.3
• Matplotlib: 3.8.4

Ways to Set Up Environment

1. Manually install the specified versions of the packages.
2. Use environment.yml to create a conda environment:

   conda env create -n telecom_test -f environment.yml

3. Use the Singularity overlay image provided in the full replication package (Linux required, details provided in the full replication package here).

---

References

Conlon, Christopher, and Jeff Gortmaker. 2020. "Best practices for differentiated products demand estimation with pyblp." The RAND Journal of Economics, 51(4): 1108–1161.