A review and evaluation of the use of longitudinal approaches in business surveys
Jul 27, 11:00
Business surveys are not generally considered as components of a longitudinal framework, and when longitudinal methods are used, they are often ad hoc approaches to particular issues. However, the largest businesses are almost always included in repeating surveys because of the properties of business populations (Rivière 2002), and short-period business surveys frequently make use of rotating panel designs to induce sample overlaps between different periods.
In this paper we review the ways in which business surveys may be considered to be longitudinal, and the methods and approaches which can be used to improve the accuracy, relevance and interpretation of business survey outputs, giving examples of their use. Specifically we consider the way samples and rotation patterns are designed to take advantage of longitudinal approaches to improving estimates of change, and how coordinated sampling schemes can be designed and implemented to put them into operation. Then we examine how best to impute for missing data in a rotating panel. There is a variety of estimation approaches which take account of the common units between two time periods, and we examine how this information can be used to improve estimates of period to period change. Finally we examine the interaction between the outputs from these surveys and the way that seasonal adjustment and similar time series models work.
We also look the other way, considering the aspects of longitudinal survey design to see what additional methods have potential to improve the operation of business surveys, and suggest ways in which these could be evaluated and implemented.