Application of selected ideas from statistical overlapping samples theory to tendency surveys: Designed panel vs resulting overlapping samples

Most tendency surveys are organized to be based on a fixed sample of units across time. This fixed panel constitutes a designed sample. But in practice the resulting sample always differs from the designed one, sometimes quite considerably. In tendency surveys, like in all real surveys, some sampled units refuse to participate, some agree to cooperate but forgo several periods later, some respond irregularly. Consequently, the resulting samples across time never constitute a perfect panel, they form an overlapping sample pattern. In the paper we propose a formula for adjusted balance statistics that takes into account distortion of a sample. The main idea of adjusted balance statistics is analogous to estimators known from statistical overlapping samples theory. Theoretical part of the paper is extended by empirical analysis of monthly business tendency survey data. In particular, the response pattern is studied and comparison of original and adjusted balance statistics is conducted