ohtanilson / shipmerger

Estimating Endogenous Coalitional Mergers: Merger Costs and Assortativeness of Size and Specialization.

written by Suguru Otani, 2021.08.05

Note: The code format is adjusted to Atom editor, so I recommend to open the code on Atom to improve readability.

PATH must be on shipmerger/. The following folders contain all R implementations and raw data.

• shipmerger/main: This contains execution files.
• shipmerger/input: This contains a raw dataset.
• shipmerger/output: This contains cleaned dataset for Julia.
• shipmerger/figuretable: This contains figures and tables shown in my paper.
• shipmerger/figure_table_ppt: This contains power point file creating figures of sankey diagrams.

Since computing all processes of Julia on a single laptop may take a few weeks, I locate the following folders restoring current Julia result files on ../shipmerger/.

• shipmerger/julia_merger_result : This contains all intermediate files.
• shipmerger/julia_merger_table: This contains all tables converting intermediate files into latex style. XXX_generate_latex_YYY named functions are only used for the conversion.
• shipmerger/julia_merger_figure: This contains all figures.

The following files restore (and update current files if you run a code) the results on these folders. You can replicate all results by following processes sequentially. However, as randomness of DE algorithm and sampling, some results may vary. To keep a record of my final results, I restore all intermediate files in shipmerger/julia_merger_result.

Data Cleaning and Reduced Form Analysis (Rproject)

• shipmerger/main/01_make_figuretable.R This cleans raw data file (210627firmdata_processed.csv) in shipmerger/input. Then, this produces cleaned data file (data_for_maximum_rank_estimation.csv) for Julia in shipmerger/output. This also generates related Figures and Tables in my paper on shipmerger/figuretable. It takes just a minute.
• shipmerger/main/02_make_sankey_diagram This generates sankey diagram in Figure 6. Input data is manually assigned based on results on ship_merger_counterfactual.jl.

Matching Maximum Rank Estimation (Julia)

• ship_merger_score_estimation_HHI.jl (1). the simplest matching maximum rank estimation in Appendix (4 hours for 50 point-estimates, 6 hours for CI bootstrapping 200)
• shipmerger/ship_merger_score_estimation.jl This computes 8 specifications with bootstrap 95% CI for (1). multivariate matching maximum rank estimation in Appendix (22 hours for 200 point-estimates, 21 hours for CI bootstrapping 200) (2). matching maximum rank estimation with two variables in Table 5 (39 hours for 200 point-estimates, 38 hours for CI bootstrapping 200) (3). matching maximum rank estimation with two variables for main 12 firms in Table 7 (1 hours for 200 point-estimates) All results are restored in julia_merger_result folder. If you want to run the each estimation block, please switch want_to_run_multivariate_estimation = "not_run" to "run" want_to_run_two_variable_estimation = "not_run" to "run" want_to_run_only_main_firms = "not_run" to "run" If you do not switch (= default), ship_merger_score_estimation.jl just reads current results from julia_merger_results folder and generates (and updates) latex tables in ship_merger_table folder. This generates Table 4, 5, 6 in the main text and Table 11, 12 in Appendix as Latex format in julia_merger_table folder. If you switch temp_subsidy_type = "shared" into temp_subsidy_type = "to_buyer" on line 40, you can run all processes for robustness checks with another subsidy specification shown in Monte Carlo Section.

Counterfactual simulation (Julia)

• shipmerger/ship_merger_counterfactual.jl It takes 100 hours (360402.808865 seconds) to compute 20 simulated matching outcome for 60 scenarios (6 subsidy amounts * 10 subsidy thresholds).

Auxiliary files and datasets (Julia)

The following scripts are loaded only as module and data in ship_merger_score_estimation.jl.

• shipmerger/functions.jl
• shipmerger/output/data_for_maximum_rank_estimation.csv

The following intermediate scripts are used only in ship_merger_score_estimation.jl.

• shipmerger/ship_merger_score_estimation_plotting.jl
• shipmerger/ship_merger_score_estimation_point_estimates_and_confidence_intervals.jl
• shipmerger/ship_merger_score_estimation_generate_latex.jl
• shipmerger/ship_merger_score_estimation_generate_latex_all_firms.jl
• shipmerger/ship_merger_score_estimation_generate_latex_only_main_firms.jl

The following intermediate scripts are used only in shipmerger/ship_merger_counterfactual.jl.

• shipmerger/ship_merger_counterfactual_plotting.jl
• shipmerger/ship_merger_counterfactual_for_sankey_diagrams.jl
• shipmerger/ship_merger_counterfactual_functions.jl

For Appendices (Julia)

You can run the following files for generating results in Appendices.

• shipmerger/ship_merger_score_estimation_HHI.jl

• shipmerger/ship_merger_monte_carlo.jl

• shipmerger/ship_merger_comparative_statics.jl

• shipmerger/ship_merger_monte_carlo_large_firms

• shipmerger/ship_merger_comparative_statics_large_firms.jl

The following intermediate scripts are used only in the above files.

• shipmerger/ship_merger_score_estimation_point_estimates_and_confidence_intervals_HHI
• shipmerger/functions_for_monte_carlo.jl (differs only in cost specification)
• shipmerger/ship_merger_counterfactual_functions.jl