Article

Jaspersen, J. G., Ragin, M. A., and Sydnor, J. R. (2022). Predicting insurance demand from risk attitudes. Journal of Risk and Insurance, 89(1), 63– 96.

• Published version
• Manuscript
• Online appendix

Funding from the Alfred P. Sloan foundation, Grant No. G-2016-7312

Code and data

The only initial data files needed are MainAll_long.dta and MainAll_wide.dta. Everything else is created by the Stata and Matlab code.

_MasterDoFile.do

• Sets global filepaths for all Stata code.
• Runs all other Stata do-files (cleaner, analysis, etc.).

MainStudy_cleaner.do

• This must be run before any of the other Stata do-files (other than _MasterDoFile.do, which calls the cleaner first). Once this is run, the other do-files can be run in any order.
• Cleans and labels raw experiment data.
• Includes a shell script to run Matlab code for estimation of parameters and coinsurance predictions. Sometimes this freezes or takes a long time, so it may be best to comment it out and run the Matlab code separately.
• Merges raw experiment data with parameters and predictions estimated in Matlab.
• Outputs clean data in both "wide" format (mostly used for summary stats and graphics) and "long" format (mostly used for regressions).

getparams_final_01.m

• Takes lottery choices in Excel format (created by MainStudy_cleaner.do) and estimates preference parameters.
• Calls functions from u.m, u_loss.m, and weightp2.m.
• Uses relative filepaths for input/output that may need to be changed.

makepredictions_final_01.m

• Takes estimated preference parameters (created by getparams_final_01.m) and estimates coinsurance demand.
• Uses relative filepaths for input/output that may need to be changed.

MainStudy_FigsTabs.do

• Creates...
  • Histograms of preference measures
  • Summary stats tables
  • Correlation tables
  • Demand curves

MainStudy_Reg_Coins_Prefs.do

• Regresses actual coinsurance demand on preference measures, individually and jointly (Tables 5-7).

MainStudy_HorseRace.do

• Creates "horse race" table comparing predictions of each model to actual insurance demand (Table 8).