Inflation inequality in the European Union and its drivers

Publishing date
19 January 2024

First published: 25 October 2022

Latest update: 19 January 2024

Headline inflation tracks the change in the price of the average household’s consumption basket. It thus masks how different individuals across society are impacted differently by rising prices. Households can face different inflation rates because their spending patterns are different. For instance, while a sharp rise in the cost of fuel increases overall inflation, a household with no car will be less affected than one with a car.

The sharp rise in inflation in the European Union since mid-2021 makes it particularly important to investigate the extent to which inflation is impacting different groups of the population differently. The main objective of this database is to document how low and high-income households face different inflation rates because of the differences in their consumption patterns, and to identify the categories of expenditure driving these differences.

Standard inflation data cannot do this because headline inflation based on the Harmonized Index of Consumer Prices (HICP) index – constructed by national statistical institutes and Eurostat – uses spending patterns of the average consumer. To solve this, we use Household Budget Surveys (HBSs), national-level surveys on consumer expenditure that contain information on household characteristics (income level, education level, age of person of reference, number of children, etc.) to estimate the consumption baskets of households in different income brackets.

This allows us to compute alternative spending shares for each consumption category for different income quantiles. For instance, high-income households spend more of their income on luxury goods, and so any price change in this category of goods has a larger impact on the inflation rate they face than on that faced by poorer households. Using this approach, we can compute the inflation rates faced by households in different income brackets for each EU country.

The figure above provides an overview of recent inflation rates and our measure of inflation inequality in the EU. Countries that have experienced the highest rates of inflation are often also those in which low-income groups face a much higher inflation rate than high-income groups. However, this is not always the case as discussed in a separate blog post. The dataset features detailed figures for each country, which are described below.

Figure 1: Inflation rate for top and bottom quantile

This graph displays multiple measures of year-on-year (yoy) inflation since January 2019. The green line represents inflation faced by the top income quantile, the red line by the bottom quantile and the dashed grey line by the total population. If the red line is above the green, low-income households are more affected by rising prices than high-income ones. The light grey dotted line is the HICP inflation, which we include to see whether our inflation measure for the total population based on the HBS corresponds broadly to the one reported by Eurostat every month. Given that the weights used for the HICP are calculated differently (see technical appendix below), some slight differences can be observed between the two measures.

Figure 2: Expenditure categories driving inflation inequality

The black line represents ‘inflation inequality’, defined here as the difference in inflation rates faced by low- and high-income households displayed in Figure 1 undefined Strictly speaking inflation inequality exists whenever this measure diverges from 0, but in the rest of the text we abusively use the term ‘inflation inequality’ to qualify regressive inflation inequality, ie when the difference is positive. . If the black line is in positive territory, it means that low-income households face higher rates of yoy inflation than high-income ones. To understand the factors driving inflation inequality, it is useful to look at the role played by each category of goods and services.

The contribution of each consumption category to inflation inequality over time is included in Figure 2, where they are aggregated into broader categories. The stacked bars represent the different categories of goods and services that drive these differences. For instance, if a category is positive in a given month, this means that price changes in the goods and services falling under that classification contributed to the difference in inflation rates faced by low- and high-income households. This could be either because this category represents a higher share of the low-income consumption basket compared to the high-income basket and its price is rising (eg food in Portugal), or conversely because its share is relatively smaller for the lower-income consumption and the price is decreasing (eg insurance in Ireland).

Figure 3: Price growth and difference in importance of consumption categories

For a consumption category to affect inflation inequality, two things must hold: the price of that category of goods must change and it must make up a different share of the consumption bundle of low- and high-income households. If all households place the same weight in their overall consumption on a certain good, then even very large increases in its price will not affect inflation inequality. Conversely, even a small increase in prices can drive inflation inequality up if that good makes up a substantially larger share of low-income household consumption.

Figure 3 allows us to distinguish between these two drivers of inflation inequality, by mapping both the average rate of inflation for 2022 (y-axis) and the difference in the share of consumption between the top and bottom quantiles (x-axis) for each consumption category. If a category is in the top right corner, it means that its price has increased and that it makes up a greater share of low-income households’ consumption than of high-income ones.


This dataset will be updated monthly when Eurostat releases final HICP numbers. Data may change for earlier months due to revisions in the underlying data.

The data behind each graph is available for download from the link at the top of the page. For example, the sheet labelled 'AT' in the file ‘data_figure1.xlsx’ provides the data used to create Figure 1 for Austria. Eurostat country codes used for each country can be found here. Figures can be reproduced without permission as long as Bruegel is used as a source.

For any questions, comments or suggestions of additional data sources, please contact [email protected]

Technical Appendix:

Data sources:

The Household Budget Surveys (HBSs) were launched in several EU countries in the early 1960s. They report household consumption expenditure across different categories of goods and services, which are divided into ‘Classification of individual consumption by purpose’ (COICOP) codes. These are available at various degrees of detail. For instance, CP01 (two-digit COICOP) refers to food and non-alcoholic beverages, CP011 refers just to food and CP0111 refers to bread and cereals, etc.

The granularity with which countries report this data is at their discretion, as are the household characteristics they include in their releases. For instance, some countries report the consumption analysis broken down by occupation, family structure and income at a detailed level (4-digits COICOP), while others are much less precise.

Since 1988, Eurostat has collected, harmonised and published this data at the national, EU and euro area-levels. In theory, this should happen at least every 5 years. However, one important limitation of the HBS is that “since the survey is conducted based on a gentlemen’s agreement, each Member State decides the objectives, methodology and frequency of the survey” as stated by Eurostat itself. As a result, some countries conduct the survey annually (eg Italy), others every two years (eg Belgium), and others every five years (eg France).

A second limitation of the HBS is that some countries do not report their data to Eurostat despite having completed the survey. For example, while Italy reports its data annually on the national statistical office’s website up until as recently as 2021, the most recent Eurostat entry for the country is 2005. This means that the harmonised data (Eurostat report the data at the 3-digit COICOP level for each income quintile) is not available for all countries from the same year. Clearly, more recent data would be beneficial, as it would allow us to gather a more accurate picture of households’ consumption patterns in recent years.

In contrast, the expenditure weights used in the construction of the HICP are adjusted yearly and using various sources of information (see here for details). This means that the HICP figure will sometimes not precisely match the measure of inflation we calculate for the overall population using the HBS. We do however include the HICP line in Figure 1 to see how far off the HICP indicator our metric is.

For these reasons, for some countries (France, Italy, Belgium, Czechia, Portugal, Cyprus), instead of using data published by Eurostat, we use the HBS data published directly by their national statistical institutes because it is either more granular, more recent or more frequent than the data provided by Eurostat. Within this group of countries, there are some countries for whom we have multiple waves of the HBS within the period of our analysis (2019-2022), which allows for a more precise measure of inflation inequality. For all other countries, we use the HBS data published on Eurostat every 5 years. The exact details of the data used for each country is available in Table 1.


  • For calculating inflation rates for different income categories:

Depending on the level of data available, for each country we use the HBS to calculate the percentage of consumption that each 2, 3 or 4-digit COICOP category makes up for the top and bottom income quantiles, as well as for the overall population. In keeping with the Eurostat HICP methodology, we do not include imputed rent (CP042) in our analysis. We use these percentages as weights for our analysis, and multiply the inflation rate for each category with its weight to calculate the overall rate of inflation faced by each group, as shown in Figure 1.

For the countries in which we have annual data (ie, Italy and Czechia), the frequency and level of granularity provided allow us to mimic the original HICP method: weights computed for year t are used to compute inflation rate in year t+1. For Belgium, we use 2018 weights for 2019 and 2020, and 2020 weights for 2021 and 2022. For all other countries, we unfortunately have to use the same weights over the entire 2019-2022 period.

  • For calculating the contributions of the main drivers of inflation inequality:

By weighting the inflation rate of a good/service by the difference in the good’s relative importance in low- versus high-income households’ total consumption, we can compute the good’s contribution to inflation inequality. For example, if a given good represents 20% of total expenditure of low-income households and only 10% of total expenditure of high-income households, and its inflation rate is 10%, the good will increase inflation inequality by 1 percentage point [i.e.:(0.2-0.1)*0.1 = 0.01]. If a given good has the same weight among low- and high-incomers, its inflation (however high) will have no impact whatsoever on inflation inequality. This data is displayed in Figure 2, while Figure 3 allows to identify easily which categories for which there are either strong differences in consumption patterns or  high rates of inflation.

Potential limitations of our measure

The lower frequency of the surveys in some countries is obviously problematic. Ideally, we would like to be able to take into account recent behavioural changes, in particular the potentially significant ones that took place during the pandemic and since the beginning of the energy crisis. This would give us a precise idea of recent inflation rates really faced by different income quantiles.

This is because inflation inequality can evolve due to two different dynamics. First, changes in inflation rates of particular goods can drive inflation inequality when these goods have different weights in the consumption baskets of different groups. This effect is well captured by our measure.

Second, behavioural changes in reaction to extreme changes in prices might differ from one income group to another. For instance, if low-income households have a harder time smoothing their consumption given the rise of prices in the energy sector, the share of energy in their expenditures will rise more for low-income households, given their need to sacrifice other types of consumption to continue heating their houses and driving their cars. This consumption shift would come on top of the increase in energy inflation, thereby exacerbating inflation inequality.

This second channel is harder to capture with the data at hand, given that for all countries apart from three, we must use the same HBS weights from 2019 to 2022. Such behavioural responses are quite likely and would mean that our measure potentially underestimates this effect and thus overall inflation inequality. This is also true for the HICP, as it uses also weights from previous years and not contemporaneous weights, which could lead to minimising inflation really faced by consumers (a problem discussed in a previous blog post). However, this issue could be exacerbated in our case, in particular for countries for which we do not have frequent HBS waves.

However, using Italy as an example, Table 2 suggests that, while behavioural changes over the course of a couple of years were significant, the percentage point difference in the weights of the different consumption categories between low- and high-income households has remained relatively stable, which is what most interests us. This means that, despite changes in overall consumption patterns, the impact of recent behavioural changes on inflation inequality may not be too high, given that it appears that changes in consumption often go in the same direction for different income groups. If this holds for other countries too, this suggests that using the same weights for the entire period is not too significant an issue.

About the authors

  • Grégory Claeys

    Grégory Claeys, a French and Spanish citizen, joined Bruegel as a research fellow in February 2014, before being appointed senior fellow in April 2020.

    Grégory Claeys is currently on leave for public service, serving as Director of the Economics Department of France Stratégie, the think tank and policy planning institution of the French government, since November 2023.

    Grégory’s research interests include international macroeconomics and finance, central banking and European governance. From 2006 to 2009 Grégory worked as a macroeconomist in the Economic Research Department of the French bank Crédit Agricole. Prior to joining Bruegel he also conducted research in several capacities, including as a visiting researcher in the Financial Research Department of the Central Bank of Chile in Santiago, and in the Economic Department of the French Embassy in Chicago. Grégory is also an Associate Professor at the Conservatoire National des Arts et Métiers in Paris where he is teaching macroeconomics in the Master of Finance. He previously taught undergraduate macroeconomics at Sciences Po in Paris.

    He holds a PhD in Economics from the European University Institute (Florence), an MSc in economics from Paris X University and an MSc in management from HEC (Paris).

    Grégory is fluent in English, French and Spanish.


  • Lionel Guetta-Jeanrenaud

    Lionel worked at Bruegel as a Research Assistant until August 2022. He studied economics at the Ecole normale supérieure de Lyon, in France. Before joining Bruegel, Lionel worked as a research assistant at the Department of Economics of Harvard University.

    His Master’s thesis investigated the impact of newspaper closures on anti-government sentiment in the United States. In addition to media economics and political economy, his research interests include fiscal policy and the digital economy.

    Lionel is a dual French and American citizen.

  • Conor McCaffrey

    Conor works at Bruegel as a Research assistant. He studied Philosophy, Political Science, Economics and Sociology in Trinity College Dublin for his undergraduate degree, where he specialised in Economics and studied in Tilburg University for a semester. He also holds an MA in Economics from the Vancouver School of Economics in the University of British Columbia, Canada, and his thesis considered the impact of welfare reforms on educational outcomes in the UK.

    Prior to completing his Master’s degree, Conor completed a traineeship in the European Parliament, where much of his work was focused on the Special Committee on Foreign Interference in European Democracies. He also worked as an intern in the Institute for International and European Affairs in Dublin, and held roles as both a Research Assistant and a Teaching Assistant over the course of his Master’s degree. He is particularly interested in labour and public economics.

    Conor is a native English and Irish speaker. 

  • Lennard Welslau

    Lennard is a Research analyst at Bruegel. His research interests lie in the fields of macroeconomics, international economics, and data science. He studied Philosophy, Politics and Economics in Freiburg and Buenos Aires and holds an MSc in Economics from the University of Copenhagen. In his thesis, he used a small open economy heterogeneous agent New Keynesian (HANK) model to investigate the role of wage flexibility in countries facing external demand and interest rate shocks.

    Prior to joining Bruegel, he worked as a trainee with the European Central Bank, held research assistant positions at the Walter Eucken Institute and the Copenhagen Business School, and contributed to the work of the UN Economic Commission for Latin America and the Caribbean (ECLAC) as a research consultant, working on empirical trade modelling.

    Lennard is a native German speaker and is fluent in English and Spanish.

Related content