We provide evidence on the effect of elementary index choice on inflation measurement. Using scanner data for 15844 individual items from 42 product categories and 10 euro area countries, we compute product category level elementary price indexes using nine different elementary index formulas. Measured inflation outcomes of the different index formulas are compared with the Fisher Ideal index to quantify elementary index bias. Across product categories, mean levels of annual elementary index bias vary between -0.53 percentage points and 0.55 percentage points depending on the index, while the standard deviation is larger than 1 percentage point. National indexes based on aggregation of the elementary indexes remain biased. The average effect of elementary index bias on national inflation ranges from -0.45 to 0.45 percentage points depending on the index. The results show that elementary index bias is quantitatively more important than upper level substitution bias.
elementary index HICP inflation measurement bias lower level substitution bias