Sampling techniques in the CPI measurement

The procedure used by a National Statistical Office (NSO) for collecting prices to produce the Consumer Price Index (CPI) is based on sample surveys. The universe (or population) of items has three dimensions: product, geographical, and time, all of which are described in the paper. This paper presents and discusses general concepts and techniques of survey sampling that are crucial for the construction of price indices. In particular, both probability and non-probability sampling techniques are discussed and illustrated with the real-world examples. A separate section discusses sampling scanned products. One of the approaches used for such data is the dynamic approach, which involves monthly sampling by applying appropriate data filters. This technique can be seen as a special form of cut-off s ampling. The empirical study investigates the effect of data filtering on the level of price i ndices. The main pragmatic conclusion is that the low-sales filter has the most significant impact on reducing the size of the scanner dataset. The second important conclusion is that changing the order of data filtering has minimal impact on the value of the price index.