Journal of Economic and Social Measurement

National accounts data provide the most comprehensive overview available of developments in national economies. They are of great interest to a wide range of users of economic information who have needs which may differ in many respects. However, since their interest generally stems from the likelihood that they will make better decisions if they are well informed about economic developments, they all have a strong interest in the accuracy of economic data. This paper examines the reliability of preliminary quarterly national accounts statistics. In particular, it considers the longer-term behaviour of the provisional estimates to GDP growth and its main expenditure components for the seven largest OECD countries, through an examination of the revisions to those estimates. The results suggest that preliminary estimates for output growth have not been statistically biased. The average size of revisions has been large although smaller than those exhibited by the demand components of the quarterly national accounts.

Counts of the US Hispanic population are available every ten years from the decennial census, but for the years following or between censuses, estimates have to be created using data and techniques that are expected to track changes in that population over time. Such estimates are a recent development and there is currently no standard methodology that has been widely used, carefully documented, and rigorously tested. In this article, we describe an experimental methodology for estimating the Hispanic population of states and counties. Postcensal data on births, deaths, and school enrollment are used for estimates of the total Hispanic population and data from the two most recent decennial censuses are used for estimates of the age, sex, and race distribution of that population. We discuss the strengths and weaknesses of this methodology and illustrate its application by making estimates of the Hispanic population for counties in Florida.

This paper, originally presented at the 1997 annual meeting of the Society for Computation in Economics, offers a perspective on the creation of econometric software for general use. To date there have been a number of attempts on the part of economist-developers to describe their work, but often these have been restricted to a presentation of specific algorithms, rather than taking the form of presenting an overall view of the process. While it stops short of proposing a formal evaluative methodology the paper attempts to establish criteria for the evaluation of econometric software that include the man-machine interface, database management capabilities, specific econometric functionality, and the general environment in which the software is used. The point of reference for the paper is the author's experience as it relates to the development of the MODLER BLUE econometric software package and related systems.