There are several strands of research which have recently converged in the analysis of exam results. A long-standing line of enquiry has been to assess the impact of deprivation on educational performance. At the school level, recent work has concentrated on trying to assess the 'added value' of the education provided by a school, taking into account the abilities of the children on entry. The need to assess influences operating at different levels (notably, the individual and the school) has led the analysis of exam results to be a core area in which multilevel models (MLMs) are being developed. A key question which remains is whether conditions in the neighbourhood in which the child lives exert a separate influence over and above the individual's characteristics and those of the school. This paper is an examination of the value of different data sources for explanatory MLMs of 16 year olds' exam results in Newcastle upon Tyne. Several different types of data source are assessed in order to profile individuals, schools, and neighbourhoods. The process of linking the data sets is described, highlighting some problems which are inherent to the innovation here of drawing upon administrative data from several separate information systems. As a result of these and other limitations, the MLMs which are then developed are essentially exploratory. The aim is to indicate which of the data sources provide the variables which appear to have the most predictive power in these analyses. The results are interesting and intuitively reasonable and enable judgments to be made as to, for example, which data sources provide the more indicative measures of the effects of deprivation (which the MLMs show to be operating independently at two different levels, with schools and wards cross-classified at the higher level).