Official statistics measuring the cost of living are known to suffer from several biases because they often do not accurately capture substitution patterns, product entry/exit, and preference shifts. In particular, the latter two biases have been shown to be large on average. This paper shows that the size of the biases can also vary with economic conditions. Using AC Nielsen homescan data for Switzerland around the 2015 appreciation of the CHF against the EUR, which resulted in a substantial shift in relative prices, it is first shown that official price indexes can be tracked using homescan data when applying the same methodology as that of the statistical office. Then, the often-acknowledged traditional substitution bias arising from lagged expenditure weights is shown to be relatively small. Based on an approach that allows for preference shocks and product entry/exit (unied price index), the two further biases, that is, the consumer valuation bias and the variety adjustment bias, are evaluated. Both are large on average and of similar size, together resulting in a 3.7 percentage point bias in the annual in ation rate. Furthermore, the bias is particularly large in the aftermath of the 2015 appreciation, increasing to 5.3 percentage points. In particular consumer valuation is shown to contribute signicantly to the increase in the bias after the exchange-rate shock. The unified price index declines by 2.4 times more than the traditional Fisher index, suggesting that taking into account the time-variation in the bias is important.
Homescan data relative price shifts inflation measurement bias inflation indexes