Empirical Properties of Diversion Ratios

Conlon and Mortimer (2021) A copy of the paper is here: https://chrisconlon.github.io/site/diversion.pdf

Before running code

To download the repo simply type:

git clone https://github.com/chrisconlon/DiversionReplication

Dataset Size and Memory

Total runtime on a 2015 i7 iMac with 64GB of RAM is around 40 minutes. Runtime on a 2017 MacBook Air with 8GB of RAM is around three hours.
All of the datasets saved should take up less than 120 MB of drive space.
The file run_all_cases.py need only be run once to populate data/dict. This is 90% of the computational time.

How to run code

Change to the directory containing this file and run "bash run_all.sh" on the terminal. The code should take approximately 1-2 hours to run in its entirety. Tables and figures will be produced as described below.

Windows Users: instead use "run_all.bat" from the command prompt.

File of origin for tables and figures

Table / Figure Number	Generating File	File Name
Table 1	(by hand)	N/A
Table 2	(by hand)	N/A
Table 3	tab34_params.py	tab3_params.tex
Table 4	tab34_params.py	tab4_stats.tex
Table 5	tab56_diversion.py	tab5_div.tex
Table 6	tab56_diversion.py	tab6_rel.tex
Table 7	tab7_wtp.py	tab7_wtp.tex
Table B1, Panel A	tab34_params.py	tableB1_blp.tex
Table B1, Panel B	tab34_params.py	tableB1_nevo.tex

Within-File Dependencies:

run_all_cases.py

aux_nevo_cases
aux_blp_cases

tab34_params.py

aux_table_functions / load_pyblp_dict, get_params_nevo, get_params_blp, outreg

tab56_diversion.py:

aux_table_functions / load_pyblp_dict

tab7_wtp.py:

aux_table_functions / load_blp_base, do_single_market, do_single_market_indiv, reshape_wtp
aux_plot_config
aux_nevo_cases / get_nevo_base

fig12_decomp.py:

aux_plot_config
aux_double_plot / plot_double_mpc
aux_table_functions / draw_blp_agents, load_blp_base

fig34_late.py:

aux_plot_config
aux_late / plot_late_single
aux_table_functions / draw_blp_agents, load_blp_base

Python dependencies

Python (version 3.4 or above) - install dependencies with

pip3 install -r requirements.txt

numpy, pandas, matplotlib, scipy, tabulate, pyarrow, brotli

Note: Windows 10 has trouble with older versions of Python 3 and multiprocessing (unrelated to this project). Please use 3.7.7 or above.

Files Provided

data/raw:

blp_product_data_opt.parquet: BLP auto data corrected using pre-computed optimal instruments. See Conlon and Gortmaker (2020) for a detailed discussion.
nevo_product_data_opt.parquet: Nevo (fake) cereal data corrected using pre-computed optimal instruments. See Conlon and Gortmaker (2020) for a detailed discussion.