Almost assuredly the 4% number is biased lower than the true population statistic (unless censoring was accounted for --- very likely not). If you don't include

Almost assuredly the 4% number is biased lower than the true population statistic (unless censoring was accounted for --- very likely not). If you don't include drivers having lived longer than a year the figure will be lower (same if you truncate their lifetime to be one year). One solution would be to wait a year after the start date of all drivers as the censoring problem would go away (but of course your information is a year out of date!). Which is it, biased lower than the true value or a year out of date?