/Safe_options_risk_attitudes_gender_data_and_analysis

Data and Analysis for the Crosetto & Filippin paper "Safe options and gender differences in risk attitudes"

Primary LanguageR

Data and analysis scripts for the paper "Safe Options and Gender Differences in Risk Attitudes"

This repository contains all the data and analysis scripts to reproduce all tables and figures in the paper Safe Options and Gender Differences in Risk Attitudes by Paolo Crosetto and Antonio Filippin.

Organisation of the repository

  • Data for the whole project can be found in the Data folder. Data is split into 9 different (.csv and .dta) files for convenience; all the data stems from the .csv files.
  • You generate figures by running the Plots.R script in R. The generated plots are saved in the Figures folder. This script works based on the three .csv data files.
  • You generate all tables and tests -- with the exclusion of Maximum Likelihood estimations presented in Table 10 and 11 in the paper by running the Tables.R script in R. Tables were mostly written down in latex by hand from the results that are to be found in the .csv output from each table. This script works based on the three .csv data files.
  • the Power.R script performs ex-ante and ex-post power analyses for the paper.
  • You generate the Maximum Likelihood estimations by running the Table_10.do and Table_11.do dofiles in Stata (13+). The raw latex tables outputted by Stata are to be found in the Tables folder. These estimations require data transformations that follow the procedure of Apesteguia, J., Ballester, M.A., 2018. Monotone stochastic choice models: The case of risk and time preferences. Journal of Political Economy 126, 74–106. The likelihood function is provided in define_likelihood.do. The scripts work based on the data contained in the six .dta data files. Those datasets are just transformations of the original .csv data following the procedure outlined by Apesteguia and Ballester. These transformations are not provided here.

Dependencies

The analysis was carried out partly in R(plots), partly in Stata (tests, maximum likelihood estimations).

For R

The R scripts depends only on the tidyverse library. This is easily installed from CRAN, using install.packages("tidyverse") (go grab a coffee, it takes a while)

For Stata

The scripts run with Stata 13 or higher. They also depend on dedicated .do files that apply the data transformation and computation required to run the Apesteguia et al. TODO ADD REF Random Parameter model; but those files are provided within this repository.