Historically, the development of probabilistic methods in survey sampling has been closely linked to official statistics. One of the first large-scale attempts to infer about a target population from partial observations came in the 1890s when Anders Kiaer, the director of Statistics Norway, challenged the discussion of “representative” samples—not yet based on probabilistic theory—with partial observations of the Norwegian population (Schweder, 1999). The well-known session of the International Statistical Institute in 1895 condemned such a practice—that is, inferring from partial observations to the total population. Kiaer argued in vain for his “representative” method at the next three consecutive sessions of the International Statistical Institute. Finally, he also had to admit defeat at home as some important estimates from such surveys were proved to be wrong in 1906 (Lie, 2002). On the other hand, the theoretical development of the probabilistic approach was slowly expanding to its first peaks in the 17th ...