To mitigate the influence of outliers and potential data errors we truncate accounting items needed in the calculation of our proxies at the first and 99th percentile and delete firm-year observations where accounting items are exactly equal to zero, most likely indicating missing data.
For robustness, we check that our results do not hinge on either of these two design choices. The final sample consists of 287,354 firm-year observations from private and publicly traded, non-financial companies over the fiscal years 1997 to
2001 across 13 European countries