We start with all observations in Execucomp database for the period between 2000 and 2010 with a defined annual chief executive officer
Due to the unique regulatory and reporting environment of financial services and insurance companies,
we remove observations with an SIC between 6000 and 7000. We also eliminate observations
that are missing data to calculate executive overconfidence, missing data in Compustat ,
missing auditor information , missing audit committee data, missing data
for Delta and Vega, city-industry-year combinations with less than two observations,
and companies using a non-Big 5/4 auditor .12The final sample contains 7661 company-years.
We winsorize all continuous variables at the 1st and 99th percentiles