The problem of disparities between different estimates of GDP is well known and widely discussed. Here, the authors describe a method for examining the discrepancies using a technique allocating them with reference to data reliability. The method enhances the reliability of the underlying data and leads to maximum-likelihood estimates. It is illustrated by application to the UK national accounts for the period 1920-1990. The book includes a full set of estimates for this period, including runs of industrial data for the period 1948-1990 which are longer than those available from any other source. The statistical technique allows estimates of standard errors of the data to be calculated and verified; these are presented both for data in levels and for changes in variables over 1-, 2- and 5-year periods. A disk with the dataset in machine readable form is available separately.