This is a public repository by the data team of the Belgian newspaper De Tijd. Using the featured scripts and data we try to take a deep dive in the Belgian inflation figures. The current version is a work in progress. The main purpose of this repo is to make our research available for review.

If you are a researcher willing to provide feedback please contact us directly.

Step 1

We were able to reproduce the officially published inflation figure (see the script from line 20 to 108). In this code we unchain and aggregate the indices.

Step 2

In line 113 till 177 we lwe load and clean the historical hbs data from 2012 until 2020.

Step 3

From line 182 until 207 we include and exclude certain product groups as provided in the methodology of NBB

In order to calculate an inflation figure by income group we understand the need to filter out the HBS data for products that aren’t used to calculate the CPI. We are aware of the fact that some corrections made to the official inflation calculations of StatBel can’t be reproduced as they would need to make use of microdata. Yet NBB defines a selection of products and categories that should be either included or excluded in order to approach the officially published inflation figures as closely as possible.

The list of INCLUDED coicop-levels (including all child elements) is listed below.

COICOP	desc
CP01	VOEDING EN NIET- ALCOHOLISCHE DRANKEN
CP02	ALCOHOLISCHE DRANKEN, TABAK
CP03	KLEDING EN SCHOENEN
CP041	Reële huur
CP043	Onderhoud en herstelling woning
CP044	Leidingwater, afvalophaling, andere kosten verbonoden aan de huisvesting
CP0451	Elektriciteit
CP0452	Gas
CP0453	Huisbrandolie
CP0454	Vaste brandstoffen
CP05	MEUBELEN, HUISHOUDTOESTELLEN EN ONDERHOUDSPRODUCTEN
CP06	GEZONDHEID
CP071	Aankoop van voertuigen
CP0721	Banden, vervangstukken en onderdelen voor persoonlijk vervoer
CP07221	diesel
CP07222	benzine
CP07223	andere brandstoffen
CP07224	smeermiddelen
CP0723	Herstelling en onderhoud van voertuigen
CP0724	Overige uitgaven voor persoonlijk vervoer
CP073	Diensten vervoer
CP08	COMMUNICATIE
CP09	CULTUUR EN VRIJE TIJD
CP10	OPLEIDING
CP11	RESTAURANT EN HORECA
CP12	PERSOONLIJKE VERZORGING EN DIENSTEN

Additionally, we exclude the following products (and their child elements)

COICOP	desc
23	drugs
122	prostitution
1251	life insurance
12530B	hospital insurance
12530C	health insurance

Further information about additional corrections that can be made using publicly available data would be very welcome.

• Do we need to rescale the filtered HBS data (based on categories above) per mille before additional price updates?

• How do we calculate the price updates correctly?

We currently load all publicly available HBS data from 2012 on. We update the weights using the following formula (applied on coicop level 3)

I.E: Updated weight in 2021 = HBS weight in 2018 * (index in December 2020 / mean index in 2018).

• A further breakdown to coicop4 causes more category names and codes that change over time. Can an accurate price update be done by sticking to coicop level 3?

• Can we calculate the inflation on coicop level 1 as the weighted average of the deeper coicop levels?

Step 4:

From line 218 to 245 we rebuild the hbs-weights after we filtered specific product codes

Step 5: Price Update

Starting from line 260 we pdate the weights according the the methology provided by statbel.

---

Credits

HBS and CPI data is property of StatBel, the Belgian Office of Statistics. The methodology used the reproduce the inflation is based on earlier reports by NBB (National Bank of Belgium)

Inflation calculations De Tijd © 2021 by De Tijd is licensed under Attribution-NonCommercial-ShareAlike 4.0 International

This license requires that reusers give credit to the creator. It allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, for noncommercial purposes only. If others modify or adapt the material, they must license the modified material under identical terms.