January 2019: fifth update of the consumer price index (2013=100)

Consumer prices
January 2019: fifth update of the consumer price index (2013=100)
  • Update of the consumer price index in 2019 based on a unanimous opinion of the Index Commission, approved by the Minister of Economy.
  • As of 2019, indices will be published for 4 additional product groups. The index basket has been extended with 9 representative items : student room rental, hourly rate of carpenters, door lock replacement, thermal compresses, second-hand cars, cookbooks, consultation of a dietician, tattoo and baby carriage.
  • A specific methodology has been developed for 2 new representative items (student room rental and purchase of second-hand cars) and the calculation method for 3 existing representative items (airplane tickets, holiday villages and banking services) has been adapted.
  • he share of data collection via scanner data and webscraping for the consumer price index increases in 2019 from 23.0 % to 26.6 %.
  • All the weights in the index basket have been updated to 2018 in order to reflect the current pattern of expenditure of consumers.

Why update the consumer price index every year?

The consumer price index (CPI) with reference year 2013 = 100, which was introduced in January 2014, is updated every year in January. It is therefore the fifth annual update in a row. The purpose of the annual updates is to keep the index representative throughout the years and to avoid misrepresenting the measured inflation as the index ages. This can be achieved by keeping the product basket up-to-date, adjusting calculation methods, integrating new price sources and by keeping a representative shop sample.

Will the base year or reference year also change?

The most recent index reform of January 2014 implied a switch from an index with fixed base year to a chain index. The reference year is still 2013 = 100, which will also be the case in the following years. Therefore, the annual updates do not require any adjustment of the selected reference year.

With the chain index method, the prices of the 12 months of the current year are each time compared with prices from December of the previous year. By multiplying these short-term indices, and so creating a chain, we get a long-term series with a fixed reference period. The sub-indices are aggregated with the weights that reflect the pattern of expenditure of the previous year.

What changes have been made in the index basket?

4 new groups and 9 new representative items are added to the basket. In principle, no representative item is removed, but in function of the transition to webscraping - where, just like for scanner data, the traditional concept of "representative item" no longer exists - a total of 13 representative items disappear from the basket for footwear (10), package holidays (2) and hotel rooms (1) in Belgium.

However, the removal of these 13 representative items is more than compensated by webscraping. In fact, it means that the content/definition of 5 COICOP groups on the fifth level (most detailed level published) is modified or extended, and that the 13 representative items removed are included in the higher COICOP groups. So, in principle, they remain part of the index basket.

Furthermore, the definitions for 3 representative items; namely airplane tickets, banking services and holiday villages, are modified. The table below gives an overview of the changes to the index basket.

New representative items Student room rental Hourly rate of carpenters Door lock replacement Thermal compresses (cold/hot packs) Second-hand cars Cookbooks Consultation of a dietician Tattoo Baby carriage (buggy)
Definition change / expansion of groups as a result of webscraping Footwear for men Footwear for women Footwear for infants and children Seaside and Ardennes weekends Hotel rooms
Other definition changes Airplane tickets Holiday villages Banking services
Representative items included in group indices as a result of webscraping Leather lace-up shoes (integrated in Footwear: sneakers (integrated in Jogging shoes (integrated in Hiking shoes (integrated in Tennis shoes (integrated in Leather low-fronted shoes (integrated in Leather boots (integrated in Ballerinas (integrated in Sports shoe (integrated in Ankle boots (integrated in Ardennes weekend (integrated in Seaside weekend (integrated in Hotel rooms (integrated in

No index is published at the level of the representative items, but the introduction of the following 4 new representative items

  • Student room rental
  • Hourly rate of carpenters
  • Door lock replacement
  • Second-hand cars

will result in the publication of indices for 4 additional COICOP groups as of January 2019:

  • group Actual rentals paid by tenants for secondary residences
  • group Services of carpenters
  • group Other services for maintenance and repair of the dwelling
  • group Second-hand motor cars.

On what basis are representative items added?

Adding representative items is of course not done arbitrarily. There are several reasons:

  • expansion based on the relative importance of the product group to which the product belongs: if there are relatively few representative items in the basket for an important weight, representative items can be added: this is for example the case for the group "Other medical products n.e.c." (a heterogeneous group), which previously only contained two representative items (plasters and thermometers). The representative item "thermal compresses" is also added to this group;
  • addition based on consumer purchase behaviour: this is for example the case for the representative items "consultation of a dietician" and "tattoo", which were added to the group "Personal grooming treatments", which until now contained three representative items (tanning shop membership, sauna and pedicure);
  • some groups with a significant weight which had not been followed up to now: it was for example the case for the group "Actual rentals paid by tenants for secondary residences" with the representative item "student room" and the group " Second-hand motor cars" with the representative item "second-hand motor cars".

Which methodological changes are brought to the index?

Methodological changes have been applied to airplane tickets, holiday villages and banking services.

Airplane tickets

For airplane tickets, 30 destinations are followed. Until the end of 2018, a virtual booking was made for each of these destinations 4 months before the date of departure. As of 2019, the number of booking periods will increase form one to three: prices of airplane tickets for reservations made 2 months and 2 weeks before departure will also be followed.

Holiday villages

The sample of holiday villages will be substantially extended and each of the holiday villages in the index calculation will receive a weight based on tourism statistics. Furthermore, the weekly prices per holiday village are now aggregated with a geometric mean instead of an arithmetic mean. Given that prices within holiday villages vary significantly between the types of housing, this gives a more correct price evolution.

Banking services

Up to now, the calculation methodology for banking services in the consumer price index was based on the costs related to a current account for the five largest banks. A new methodology has been developed in order to better grasp the price evolution of banking services. The new calculation is based on current accounts and/or account packages of the eight largest banks.

The calculation is made with four user profiles:

  • electronic profile with credit card;
  • electronic profile without credit card;
  • traditional profile with credit card;
  • traditional profile without credit card.

Every profile is characterised by a number of operations:

  • money withdrawals at an ATM of other banks;
  • money withdrawals at the counter;
  • paper transfers.

For the calculation, each current account or account package is assigned to one or several user profiles. The number of clients per current account (and option credit card) is supplied every year by the bank sector. In combination with price lists, an average price per profile is calculated. The four profiles are then aggregated via the number of accounts per profile. Finally, this gives the aggregated index for banking services.

What does the new weighting scheme look like?

The chart below displays the weightings of the twelve main groups (on a thousand) in 2017, 2018 and 2019.


The general CPI is the first level. Under the second COICOP group level, composed of 12 main groups (illustrated above), there are three lower group classifications. Finally, the sixth level is the level of representative items. The weighting scheme is always published down to the lowest COICOP group level.

How is the weighting scheme built in practice and why is the weighting scheme adjusted every year?

As in the previous years, the weighting scheme is based on the new European ECOICOP nomenclature. This European 'Classification of Individual Consumption by Purpose' (COICOP) is a classification consisting of 12 main groups. To create the 2019 weighting scheme, new representative items and groups were added to the 2018 index basket.

The idea is that the weighting scheme of the consumer price index matches the pattern of expenditure of an average household as closely as possible. It is based on the household budget survey, a biannual survey that analyses the expenditure patterns of the households. The weights are based, just like last year, on the household budget survey 2016. This is the most recent survey, the results of which have been published in September 2017. These weights, that refer to 2016, have been updated to 2018.

Since a chain index is used, in which the prices of the twelve months of the current year are compared with December of the previous year, the weighting scheme needs to be adjusted annually based on the price evolution between the year in which the values are expressed to the same year as the reference month. This procedure is known in professional literature as the 'price update' of weights. Since the weighting scheme is based on the household budget survey from 2016, the weights are updated to December 2018 in order to correctly measure the price evolution in 2019.

In practice, the price update of weights means that the values of the weighting scheme, based on household expenses in 2016, are updated to December 2018 based on the index evolution of each group, since December 2018 is the new reference month. This is done by multiplying the weights at the most detailed level published with the price evolution measured between December 2018 and the year 2016 (that is the ratio between the index in December 2018 and the average index of 2016). For product groups with significant seasonal variations, the annual average of the 2016 and 2018 indices is used for the price update.

After these calculations, the results are rescaled to convert the total weight again to 1000‰. And finally, the upper levels are recalculated based on the rescaled underlying weights obtained.

In concrete terms, the price update of the weights means that the weighting scheme is updated on the basis of the evolution of each group's index, which ensures that the current expenditure pattern of households is systematically used for the index calculation.

To what extent are big data used in the index calculation?

Since 2015, Statbel uses big data as data source for the consumer price index. In this context, this is the price information obtained via scanner data and webscraping. See also our publications about scanner data and webscraping for more background information. In total, the price evolution of 27 % of the weight of the index basket in 2019 is measured using big data. In 2018, this was 23 %. The use of scanner data and webscraping improves the accuracy of the CPI. Indeed, the price index of a product group should no longer be based on a relatively limited sample of products, but we can process the prices of multiple items sold. The new data sources lead to an index that more closely matches actual consumption habits.

Scanner data have been progressively introduced in the CPI since 2015. The weight of the basket followed using scanner data amounts to 22.5 % in 2019. These are the cash register data from the largest supermarkets. These scanner data are supplemented with traditional price recordings in shops (example: baker, butcher, etc.).

In addition to scanner data, tariff prices, catalogue prices and price recordings in shops, prices are also collected via webscraping. This is a technique for automatically scraping data from web pages. The data from web pages are collected and processed in a structured way, so that they can be used for statistical purposes. Given the growing importance of webshops and the online sale of "classic shops", it is appropriate to include these data in the calculation of price indices.

The use of big data also allows to improve the efficiency of the data collection. In addition, the representativeness of the price indices increases, as the prices of a multitude of products are followed, compared to the traditional price recordings.

In 2018, webscraping results were already incorporated in the index for DVD's, Blu-ray discs, video games and international train tickets. These covered a total weight of 0.2 % of the basket. As of January 2019, the number of price recordings via webscraping will be considerably increased by the introduction of this technique for the following segments:

  • footwear (already existing group with initially 10 representative items)
  • weekends by the sea and in the Ardennes (already existing representative item)
  • hotel rooms (cities) (already existing representative item).

In addition, as of January 2019, 2 of the new segments will also be followed via webscraping, namely:

  • student room rental
  • second-hand cars.

As a result of this expansion, the weight share of product groups followed via webscraping will amount to 4.1 % in 2019.

The table below gives an overview of the product groups that are followed via scanner data or webscraping, and their corresponding weight in the index basket.

CIOCOP Big data in the CPI 2018 (‰) 2019 (‰)
01 Food and non-alcoholic beverages 169,39 167,33
02 Alcoholic beverages and tobacco 23,10 23,42 Miscellaneous small tool accessories 3,04 2,79
05.6.1 Cleaning and maintenance products 8,26 8,10 Products for pets 7,21 6,97 Paper products 1,03 0,98 Other stationery and drawing materials 1,70 1,70
12.1.3 Other appliances, articles and products for personal care 14,31 13,70
Total scanner data 228,03 224,99 International train travel 0,49 0,51 Blu-ray disc 0,52 0,49 DVD (music or film) 0,52 0,49 Video game for console 0,39 0,35 Weekend in the Ardennes   1,67 Weekend at the seaside Hotel room   6,72 Second-hand motor cars   17,24 Renting a student room   3,25 Footwear for men   2,78 Footwear for women   5,05 Footwear for infants and children   2,69
Total webscraping 1,93 41,24
Total big data 229,96 266,23

How is the calculation of the indices based on webscraping done in practice?


For footwear, the websites of the largest shoe shops in Belgium are scraped. These are companies with both physical stores and an online store. Several times a week, there is a bulk webscraping. This means that all products on the website are scraped and that no product selection (or limitation) is done in advance. Data cleaning and selection is done during the analysis phase.

For each retailer, a stratification is first made for footwear for men, women and children, after which each segment is further subdivided into shoe type. This is done using the classification provided for on the retailers' websites, as the offer may vary from site to site. A simple Jevons index is calculated at this lowest level over all items. In a Jevons index, the relative prices (representing the price ratio for the same items between two months) within a consumption segment are aggregated via a geometric mean.

For each retailer, an index for footwear for men, women and children is thus obtained, from which an overall index for each of these three categories is calculated based on the turnover of the respective retail chains. For the weights, which are a measure of turnover, the annual accounts or VAT declarations are used for the various chains.


The traditional method of manual price searches, which was used until the end of 2018, resulted in prices being noted once a month for a weekend stay of two nights in a double room. These were the prices obtained for virtual bookings four weeks before arrival. The selection was a sample for the coast, the Ardennes and other cities. A price per hotel is calculated.

On the other hand, the webscraping method, which will be used as of January 2019, will daily record the prices for bookings made 4 and 8 weeks before arrival. These bookings always concern a weekend stay of 2 nights in a double room. There is a stratification by region, destination, booking time and comfort level (number of stars). A price per stratum is calculated. The resulting indices are then aggregated into an index "Weekend by the sea or in the Ardennes" and an index "Hotel rooms" which will replace the corresponding representative items as of January 2019.

Student rooms

The rental of a student room was not yet included in the index. This is a separate COICOP group, the weight of which, in the absence of a representative item, was added to the group Actual rentals paid by tenants. As of January 2019, this new representative item in the index basket will be included in the COICOP group Actual rentals paid by tenants for secondary residences.

The survey of the rentals of student rooms is very difficult to carry out in practice using a traditional survey and is cost-intensive, given the short rental periods and the expected low response rate. This is why, for this new representative item, rentals are collected via webscraping. The scraped prices of various websites concerned the university cities of Antwerp, Ghent, Leuven, Hasselt, Brussels, Louvain-la-Neuve, Mons, Namur and Liège.

The data collected are cleaned, geocoded and stratified according to homogeneous segments per city, using characteristics such as surface area and type of student room (room or studio). The calculation is done separately for each of the 9 cities: for each stratum a Jevons index is calculated (geometric mean of the prices of a stratum in year J compared to the geometric mean of the prices of the same stratum in year J-1). The resulting indices are aggregated into an index per city, where the weight of each stratum is the sum of the prices of the previous year within that stratum. Finally, the indices of the 9 cities are aggregated into one index for the student rooms based on the number of students in the 9 university cities. The index calculation takes place once a year and is incorporated in the consumer price index of October.

Second-hand cars

The purchase of second-hand cars was not included in the index until now. The weight of this expenditure was therefore added to the group New motor cars. As of 2019 this new representative item with its corresponding weight will be included in COICOP group Second-hand motor cars. The data collection is done via webscraping. The most popular websites with a range of second-hand cars are scraped. The sample is based on a database of the FPS Mobility and Transport (DIV) and contains the most popular brands and types of second-hand cars.

As the supply of second-hand cars varies from model to model from period to period, not only the prices should be scraped for the brands / models, but also the characteristics, so that the depreciation can be correctly incorporated in the index. For this segment, the index calculation must correct the systematic difference in the supply of second-hand cars between the two periods considered. The age, the number of kilometres, the type of fuel, ... is different for each type of car. In addition to brand, model and price, all these characteristics of the offered vehicles for a determined sample of second-hand cars are scraped.

With these characteristics, a so-called hedonic regression can be applied to calculate the index. A hedonic regression assumes that the price of a product can be expressed as a function of its characteristics. Since there is no market for characteristics, let alone that a characteristic is sold separately, they cannot be observed separately. Implicitly, a characteristic naturally contributes to the price of a good through the law of supply and demand. These implicit contributions can be estimated using regression techniques. This is an estimate of what the price of a new model would be in the base period or of a disappeared model in the current period.

What role do the Index Commission and the Minister of Economy play in this?

The Index Commission is a joint commission composed of a president and 21 members representing the academic world and the workers’ and employers’ organisations. The commission is supported by statisticians of Statbel. It advises the Minister of Economy on consumer price index matters and also issues a monthly opinion on the index calculated by Statbel.

The Index Commission issued a unanimous opinion on this update to the Minister of Economy on 21st December 2018. The minister has decided to approve the working method proposed by the Index Commission for the composition and calculation of the index in 2019. The administration will then calculate the index from January 2019 onwards according to the new modalities.