Transactions bias arises because properties that trade are not a random sample of the total housing stock. Price indices are potentially susceptible to this bias because they are typically based on transactions data. Existing approaches to this problem have relied on Heckman-type correction methods where a binary probit regression is used to capture the differences between properties that sell and those that do not sell in a given period. However, this approach can only be applied where there is reliable data on the whole housing stock. In many countries ñ such as the UK ñ no such data exists and so there is little prospect of correcting for transactions bias in any of the regularly updated mainstream house price indices. This paper offers an alternative approach based on information at postcode sector level. The probability of a property transacting is modelled by applying fractional probit regression (FPR) to the proportion of properties that sell in each postcode sector. Transactions data on 1.4 million house sales distributed across 1,200 post code sectors in the South East of England over the period 1996 to 2003 are used to create a correction term for in a simple monthly hedonic house price regression. Corrected and uncorrected price indices are compared.