How to use this repo to reproduce the paper "Yellow Vests, Pessimistic Beliefs, and Carbon Tax Aversion" (AEJ:EP)?
Repository Project ID: openicpsr-128143
- Thomas Douenne, University of Amsterdam, Roetersstraat 11, 1018 WB Amsterdam, Netherlands (email: t.r.g.r.douenne@uva.nl)
- Adrien Fabre, ETH Zürich (D-MTEC), Zürichberstrasse 18, 8032 Zürich, Switzerland (email: fabre.adri1@gmail.com)
- The code is in R and Python 2.
- The main file is code/papier.R: it enables researchers to reproduce all the empirical findings of the paper section by section.
- In order to run this script, a few preliminary steps are needed. They are described below.
- First, go to code/preparation.R and run this script. It will prepare the dataset (called "s") used all along the paper, and load the packages needed.
- Second, go to code/papier.R and run the main code.
- In this repository you will find some .csv files, some of which are produced by the .py files. You can reproduce these .csv yourself.
- These .csv files use official statistics, in particular to compute the objective distribution of gains and losses from the carbon tax and dividend.
- Step 1: go to model_reforms_data/prepare_dataset.py and run the script to build a dataset from data_menages.csv. This .csv provides data from the survey "Budget de Famille 2011".
- Step 2: then the script model_reforms_data/gains_losses_data.py can be used to compute the incidence of the tax and dividend policy on all households in the dataset.
From there several things can be done:
- consistency_bdf_ptc.py compares the distribution of the tax incidence computed from official statistics to the ones of our survey (hence, need to run code/preparation.R before). This script can also be used to reproduce df_subjective_gains.csv, df_objective_gains.csv, or df_objective_gains_inelastic.csv (by changing parameters' values in the computation of the tax incidence).
- model_reforms_data/regression_feedback.py enables to regress the policy's incidence for housing energies. This will be used to estimate a respondent-specific tax incidence. It can be done either from the consumer ('bdf') or housing ('enl') survey. Both datasets for housing energies can be generated using model_reforms_data/prepare_dataset_housing.py
- test_predictions_ols_regression_with_transports.py, test_predictions_binary_models.py, and code/matching_predict_winner.R assess the precision of our respondent-specific estimation of the tax incidence. This is used to provide a customized feedback on respondents' expected net gain (see Section 4 of the paper).
- Then, you can run code/preparation.R, and the main code code/papier.R
code/agglos.csv: data used to match town code to information on town size, provided by the survey company Bilendi code/correspondance-code-insee-code-postal.csv: data from Insee, matches postal code with characteristics of the town, unused code/df_menages_domicile_teg.csv: data used in preparation.R but not in the paper: subsample from transport survey used in companion paper. Transport survey: https://www.insee.fr/en/metadonnees/source/serie/s1277 code/df_objective_gains.csv: distribution of respondents objective gains plotted in section 3 and procuded by consistency_beliefs_losses.py code/df_objective_gains_inelastic.csv: distribution of respondents objective gains assuming inelastic demand, plotted in section 3 and procuded by consistency_beliefs_losses.py code/df_subjective_gains.csv: distribution of respondents subjective gains plotted in section 3 and procuded by consistency_beliefs_losses.py code/matching_predict_winner.R: performs statistical matching as an alternative way to predict winners and losers code/packages_functions.R: loads all the necessary packages code/papier.R: main code, enables to reproduce all the findings of the paper code/preparation.R: necessary to run this script before executing papier.R, prepares the dataset to be used code/questionnaire.qsf: enables to import the questionnaire in Qualtrics code/quotas.xlsx: provides objectives quotas from Insee, information obtained from Insee's website code/rev_i_erfs2014.csv: data from survey ERFS used to get information on actual income distribution of the French population (selected variables) code/survey.csv: raw data from Qualtrics survey code/survey_prepared.csv: data from Qualtrics survey after preparation conso-eff-function.xls: data from National accounts used to inflate our figures.
images/: different images generated from papier.R, also includes images from companion paper
model_reforms/define_tax_incidence.py: defines formulas to study the tax incidence model_reforms/diesel_standard_example.py: specific example applied to diesel model_reforms/domestic_fuel_standard_example.py: specific example applied to domestic_fuel model_reforms/gas_standard_example.py: specific example applied to natural gas model_reforms/gasoline_standard_example: specific example applied to gasoline
model_reforms_data/computation_co2_emissions.py: compute the reduction in CO2 emissions from the policy model_reforms_data/define_tax_incidence_data.py: defines formulas to study the tax incidence on households in surveys model_reforms_data/gains_losses_data.py: computes tax incidence on households in surveys model_reforms_data/prepare_dataset.py: prepares the dataset (selects variables, translates names to English, inflates sectoral expenditures) model_reforms_data/prepare_dataset_housing.py: prepares the dataset for housing energies only (selects variables, translates names to English, inflates sectoral expenditures) model_reforms_data/regression_feedback.py: regress households' expenditures in housing energies on househlds' characteristics model_reforms_data/standardize_data_bdf_ptc.py: defines functions to compare new survey with official statistics (in particular subjective vs objective gains), functions are used by consistency_beliefs_losses.py model_reforms_data/data_menages.csv: data to be used in prepare_dataset.py : comes from the matching of the consumer and transport surveys done in Douenne (2020, The Energy Journal). (selected variables). Consumer survey: https://www.insee.fr/en/metadonnees/source/serie/s1194 ; Transport survey: https://www.insee.fr/en/metadonnees/source/serie/s1277 model_reforms_data/data_matching_bdf.csv: data to be used in prepare_dataset_housing.py : comes from the French consumer survey (selected variables). Consumer survey: https://www.insee.fr/en/metadonnees/source/serie/s1194 model_reforms_data/data_matching_enl.csv: data to be used in prepare_dataset_housing.py : comes from the French housing survey (selected variables). Housing survey: https://www.insee.fr/en/metadonnees/source/serie/s1004 model_reforms_data/prediction expenditures.csv: data on objective gains and losses produced in test_predictions_ols_regression_with_transports.py. and used to assess precision of our prediction model_reforms_data/prediction expenditures (2).csv: alternative specification for the prediction, called in code/preparation.R but unused in the paper model_reforms_data/prediction housing expenditures.csv: same thing but specific to housing energies, called in code/preparation.R but unused in the paper model_reforms_data/prediction housing expenditures (2).csv: same thing but specific to housing energies, called in code/preparation.R but unused in the paper
Questionnaire/: files used for questionnaire on Qualtrics
consistency_beliefs_losses.py: compares new survey with official statistics (in particular subjective vs objective gains) and produces figures C.3 and df_objective_gains.csv, df_objective_gains_inelastic.csv, and df_subjective_gains.csv df_donor_enl.csv: data produced by test_predictions_ols_regression_with_transports.py from housing survey and used for the statistical matching (alternative method to predict winners and losers). Housing survey: see https://www.insee.fr/en/metadonnees/source/serie/s1004 df_objective_gains.csv: distribution of respondents objective gains plotted in section 3 and procuded by consistency_beliefs_losses.py df_objective_gains_inelastic.csv: distribution of respondents objective gains assuming inelastic demand, plotted in section 3 and procuded by consistency_beliefs_losses.py df_subjective_gains.csv: distribution of respondents subjective gains plotted in section 3 and procuded by consistency_beliefs_losses.py df_receiver_bdf.csv: data produced by test_predictions_ols_regression_with_transports.py from consumer survey and used for the statistical matching (alternative method to predict winners and losers). Consumer survey: https://www.insee.fr/en/metadonnees/source/serie/s1194 LICENSE: License to use the present repository test_predictions_binary_models.py: tests accuracy of our respondent-specific win/lose prediction for alternative specifications and methods test_predictions_ols_regression_with_transports.py: tests accuracy of our respondent-specific estimation of the tax incidence utils.py: defines useful functions to plot graphs
Fully available. URL: https://github.com/thomasdouenne/yellow_vests_aej_ep/tree/main/code/survey_prepared.csv DOI: http://doi.org/10.3886/E128143V2.
Fully available. URL: https://github.com/thomasdouenne/yellow_vests_aej_ep/tree/main/model_reforms_data/data_menages.csv
Fully available to researchers upon completion of a form. Presentation of the survey: https://www.insee.fr/en/metadonnees/source/serie/s1194 URL: http://www.progedo-adisp.fr/enquetes/XML/lil.php?lil=lil-0831
Fully available to researchers upon completion of a form. Presentation of the survey: https://www.insee.fr/en/metadonnees/source/serie/s1277 URL: http://www.progedo-adisp.fr/enquetes/XML/lil.php?lil=lil-0634
Fully available to researchers upon completion of a form. Presentation of the survey: https://www.insee.fr/en/metadonnees/source/serie/s1004 URL: http://www.progedo-adisp.fr/enquetes/XML/lil.php?lil=lil-1022
Fully available to researchers upon completion of a form. URL: http://www.progedo-adisp.fr/enquetes/XML/lil.php?lil=lil-1177
Fully available. File: conso-eff-function.xls URL: https://www.insee.fr/fr/statistiques/3547386?sommaire=3547646#titre-bloc-20
Fully available. It consists of the following quote: “3,4 millions de résidences principales sont encore chauffées au fioul en France. Cela représente 12 % des foyers”. URL: https://www.lesechos.fr/industrie-services/energie-environnement/le-chauffage-au-fioul-devient-de-plus-en-plus-cher-147372
Software requirement: R, Python 2. The following software and language versions were used: RStudio 1.3.1073; R 4.0.3; Python 2.7. Tables are generated without their Notes by default. To generate them with the Notes, uncomment the "write_clip(...)" lines after "stargazer(...)" in papier.R; noting this requires an interactive (GUI) environment like RStudio. A particular version (0.99.22) of the R package "memisc" is needed. If another version of "memisc" is installed, packages_functions.R will automatically uninstall it and install the appropriate version; using the package "installr" that runs only on Windows. Length necessary to install all packages (on a powerful cloud machine): 15 min. Length of necessary computation (on a powerful laptop): 2 min.