Rudi Seljak, Lea Bregar, Sanda Colić and Maja Dozet
In spite of theoretical inferiority, a large majority of statistical institutes use non-probabilistic sampling techniques in price surveys. The main disadvantage of non-probabilistic sample design is that the risk of biased results is increased. Attempting to handle this risk in the domain of service producer price indices (SPPI) of professional services, the Croatian Bureau of Statistics (CBS) developed an innovative methodology and implemented it relying on the probability proportional to size (PPS) sample design. The purpose of the paper is to evaluate the impact of the probabilistic sampling strategy on the quality of price indices as shown in the case of SPPI for professional services at the CBS. The paper outlines respective methodological upgrading of SPPI compilation at the CBS, including also the method for variance estimation. The effect of the probabilistic sampling on SPPI is analysed by comparison with traditional purposive sampling surveys. The quality of SPPI based on the probabilistic sample approach is examined by coefficients of variation and confidence intervals.