Repeated surveys and the Kalman filter

Abstract

The time series nature of repeated surveys is seldom taken into account. The few studies that take this into account usually smooth the period-wise estimates without using the cross sectional information. This leads to inefficient estimation. I present a statistical model of repeated surveys and construct a computationally simple estimator based on the Kalman filter which efficiently uses the whole underlying data set, but which is computationally very simple as we only need the first and second empirical moments of the data.