This project assembles an eleven-part panel dataset of Form 477 broadband data from December 2014 through December 2019, sourced from the Federal Communications Commission. The Form 477 dataset reports fixed broadband service at the Census block level in the United States. The panel also includes demographic data for the years 2014 to 2019 from the U.S. Census Bureau, including:
- Population Density for U.S. block groups,
- Housing Density for U.S. block groups, and
- Median Income for U.S. tracts.
This repository can be used to replicate the dataset and econometric results of articles by Kotrous and Bailey (see Replicating Research below).
This project uses the MIT License, so you are free to copy and modify the code as it suits your research needs. The panel retains geographic identifiers for Census blocks, block groups, tracts, counties, and states, so joining additional demographic data should be fairly straightforward. I'm currently developing a project that joins High-Cost Support data from USAC by Census block id, for example.
The code was most recently executed in Stata 16.1. Earlier versions of this code could be executed in Stata 14, so I expect this code will be compatible with most flavors of Stata.
The do file for econometric tests, broadbandcompetition-panel.do, uses the estout package to export regression tables in RTF/Word and LaTeX formats. I strongly recommend installing it for this and other Stata projects!
In the repo working directory, create two new directories: crosssection (note three s's) and source.
crosssection: This directory is used to store each of the nine cross-sections of Form 477 data that has been joined with Census data. You are okay to leave this directory empty. It will be populated with files as the Stata scripts execute.
source: This directory is used to store all source files (in csv format) from FCC, the U.S. Census Bureau, and U.S. Bureau of Labor Statistics. This directory must be populated with source files in order to work. Contact me on Twitter or my personal website to request these files. These files require just under 50 GB of storage. Note that when you extract the source files, the directory structure in your local environment (relative to the repo working directory) should look like source/2014, source/2015, and so on.
If you wish to see the final dataset or use it to run econometric tests, you can request the final dataset by messaging me on Twitter or contacting me on my personal website. This dataset includes observations for all U.S. Census tracts in the 48 contiguous states and Washington, D.C.
Executing the scripts is quite simple. Simply open master.do in Stata, define the working directory on line 7, and execute the do file. There's no need to edit or modify the do files in the scripts directory. The master do file will execute the supporting scripts in the appropriate order.
Assembling the full national panel for all U.S. Census tracts in the 48 contiguous U.S. states and Washington, D.C. requires at least 64 GB of RAM. The full national panel is about 40 GB in size.
If you have insufficient RAM to work with the full national panel, you can select a sample by randomly drawing a given percentage of U.S. census tracts. Modify master.do to execute append-random.do rather than append-full.do. You will also need to edit lines 10, 29, 47, 60, and 73 of scripts/broadbandcompetition-panel.do to use the dataset US-Fixed-Panel-Random-Merged.dta.
By default, scripts/append-random.do will sample 25 percent of U.S. census tracts. Edit line 36 of scripts/append-random.do to change the sample size.
You can define the seed to allow for replication of a sample on line 35 of scripts/append-random.do. The seed is set to 5663451. You can request a dataset (~10 GB) that uses this seed to draw a sample of 25 percent of U.S. census tracts from Dec. 2014 to Dec. 2019. Please contact me on Twitter or my personal website for that file.
When the full master.do script is executed, four files will be placed in your working directory.
- US-Fixed-Panel-Merged.dta (default, or US-Fixed-Panel-Random-Merged.dta if
append-random.doscript is executed in master file instead ofappend-full.do) - broadbandcompetition-panel.log (Stata log with full econometric results)
- regression-tables.rtf (Word-friendly format for regression tables)
- regression-tables.tex (LaTeX file for regression tables)
You will also notice that the crosssection directory is populated with Stata dta files for each merged cross-section of Form 477/ACS data, as well as various Stata log files with merger summaries and the like.
The 1.1.1 release of this repository can be used to replicate the dataset and econometric results of Kotrous and Bailey (2021) in the Journal of Information Policy.
The 1.0.1 release can be used to replicate Kotrous and Bailey (2021), a working paper released by the Center for Growth and Opportunity in Jan. 2021. That working paper analyzes a panel for 2014 - 2018 that samples 25 percent of U.S. Census tracts at random. The seed for replicating that sample is 5663451. To draw a different sample of tracts, edit the seed on line 35 of scripts/append-random.do.
Collecting the data took considerable effort, and storing the source files and allowing you to retrieve them is not free. If you use or enjoy this repository, I would appreciate you buying me a beer! 🍺
I would love to hear from you if you publish any research that uses or was inspired by this repository! If you do so, please cite this repository and:
Kotrous, Michael, and James Bailey. "Broadband Speeds in Fibered US Markets: An Empirical Analysis." 2021. Journal of Information Policy 11(2021): 478-522. https://doi.org/10.5325/jinfopoli.11.2021.0478.
A previous release of this repository can be used to replicate the dataset and results in the working paper version of this article, which can be cited as follows:
Kotrous, Michael, and James Bailey. "Broadband Speeds in Fibered Markets: An Empirical Analysis." Working Paper, The Center for Growth and Opportunity at Utah State University, January 2021. https://www.thecgo.org/research/broadband-speeds-in-fibered-markets-an-empirical-analysis/.