Do Binary Collateral Outcome Variables Proxy Collateral Levels? The Case of Collateral from Start-ups and Existing SMEs

This paper is designed to test whether the factors which affect the decision to collateralise business loans affect the level of collateralisation in the same manner. If the level of collateralisation does matter, the provision of collateral becomes more than a goodwill gesture to placate banks and more a device to ameliorate the risk of lending. We use a thitherto unseen dataset from a U.K. retail bank comprising 4,618 transfers and start-ups (TS group) who applied for business loans and overdrafts between January 1998 and January 2000. The control sample comprised 9,596 existing businesses from the same period. Our unique dataset permits an analysis of this kind for the first time because it contains a continuous variable for collateral unlike previous studies. Existing businesses exhibit a higher frequency (binary outcome) and level (tobit outcome) of collateral than the businesses who are start-ups or have transferred from another bank only when distortions within the data are not controlled for. These distortions negate the value of binary collateral variables. Factors such as business type and loan purpose are useful at explaining the likelihood of a borrower having his loan collateralised and the level of collateralisation for borrowers who provided collateral or not.