As of April 2018, Statistics Netherlands (CBS) has made upward adjustments of a few percent to the Consumer Price Index (CPI and HICP) for men’s, women’s and children’s clothing. CBS research had shown that, as a result of a new price observation method for clothing, the expected average year-on-year changes in 2018 relative to 2017 would be too low otherwise. The adjustment of the indices for clothing articles also affects the overall CPI and HICP results. The year-on-year change in the CPI for April 2018 has been revised upwards from 0.9 to 1.1 percent. The results for the first three months of 2018 are not revised.
Explanation of the adjustments
CBS is constantly developing new and better methods for the compilation of statistics. In recent years, a new method for observation of clothing prices has been introduced in a number of successive steps. In the past, prices of predefined articles were observed by price collectors in the shops. The new method uses so-called webscrapers. This makes it possible for prices of all products sold on the website of a clothing shop chain to be collected automatically, potentially on a daily basis. This method provides Statistics Netherlands with many more price observations, which means improved measurement of price developments at a far lower cost. This method is in line with the buying behavior of consumers, who increasingly purchase clothes from webshops.
The new method was introduced for one retail chain in 2015. In 2016 and 2017, several chains were added and price collection in shops was reduced. As of 2018, price collection in shops is fully replaced by webscrapers.
For the introduction of new methods in the HICP, there are European regulations which are usually also applied to the national CPI. These regulations prescribe that new methods or new sample of products or retailers must be introduced with the change from December to January. The index development according to the new method is linked to the index of December that was still calculated using the old method. In most cases, changes in method or samples hardly have any impact on the year-on-year changes of the index. These regulations were also applied to the first few months of 2018.
However, a change in method may sometimes affect the year-on-year change of the index of certain article groups, particularly in three cases:
- when the seasonal pattern of the index according to the new method differs from the old pattern,
- when the last results using the old method were incidentally very high or very low, or
- when the first results using the new method are incidentally very low or very high.
During the transition in December 2017 from the last price observations in shops to price observations only by webscrapers, all three cases happened to coincide. As a result, the price index for clothing articles had a very low outcome in the first months of 2018. The outcomes are significantly lower than the outcomes that would result if the old method had still been used, but also significantly lower than if the new method had already been used in 2017. In addition, clothing price indices have a significant weight, which means that the impact of the new method is notable in the headline results of CPI and HICP.
This was reason for CBS to investigate this case further and make an estimate of the impact for the remainder of the year 2018. The outcome is that if the rules are to be strictly followed, the average year-on-year change of clothing prices in 2018 compared to 2017 will be about 5 percentage points lower than without the method change. The impact on the overall CPI and HICP will on average be around -0.2 percentage. Therefore, CBS has decided to apply a level correction to the indices for clothing prices.
Application of the level correction
The preliminary April index results for clothing articles have been raised and finalised. For men’s clothing, the adjustment is 4.7%, for women’s clothing it is 6.5% and for children’s and baby clothing it is 3.6%. CPI results for January, February and March were already final. They will not be adjusted.
Without this level correction, the annual average indices for 2018 would be about 0.2 percentage too low for the total CPI and 0.3 percentage too low for the HICP. After the correction, these biases is virtually eliminated. What will remain is that until March 2019, the differences between the seasonal patterns of the old and new methods may cause the year-on-year change in some months to be rather high or rather low. This is unavoidable.