Problems and methods in evaluating the reliability of national accounts

Abstract

The reliability of national accounts depends on a great variety of data sources and estimating techniques. The article focuses on some major methodological problems in assessing national accounts reliability. It is argued that the combined use of data of varying quality, and the additional operations of aggregation, estimation and adjustment, make the assessment of overall reliability almost impossible. The inquiry provides a framework for reliability analysis based on partial criteria suitable for measuring the main aspects of reliability. The following criteria are presented and critically discussed (a) subjective reliability ratings, assigned by experts; (b) consistency criteria, that measure the discrepancy between two or more values of the same economic aggregate (typical cases are the sratistical discrepancy between the expenditure and production approaches to GDP, or GNP, and the periodic revisions of preliminary estimates); (c) accuracy criteria, that involve an evaluation of the observed values with reference to the 'true' values. The emphasis is on the specification of appropriate measurement models, crucial to investigate the characteristics of errors.