Furthermore, some variables include outliers in only the upper values, so we truncate the sample at the 99 percentile to exclude these outliers.
When variables include outliers in both the upperand lower values, we truncate at both the 0.5 and 99.5 percentiles of the sample.
The data collected on the 80,625 firms from 2006 to 2009 include 91 items from their balance sheets and profit and loss statements.
The total number of firm-year observations is 278,611.
The CRD collects small business data to construct scoring mod-els that predict default by small businesses.
The CRD uses data on the small business clients of 124 financial institutions and 52 pub-lic guarantee corporations with regular member status (as at July 2013)