What is the carbon inequality in Luxembourg? How do consumption emissions compare with the climate targets? Who are the highest and lowest emitters, and what do they consume? These are questions I discuss here, the answers of which will be key in the design of socially inclusive carbon policies in Luxembourg.
Carbon footprints
The household carbon footprint corresponds to the total amount of greenhouse gases such as carbon dioxide that are associated with household consumption. For example, this includes tailpipe emissions from driving a car or burning fuel for heat that households directly witness (often referred to as direct emissions). But also the emissions required to produce and distribute the goods and services we consume, such as the car or the gas stove (often referred to as indirect emissions). One advantage of the carbon footprint approach is that it captures economic activity and emissions that are outsourced; for example, even if production processes are relocated to another country and thus don’t show up in Luxembourg’s territorial accounts, they still make up a part of the Luxembourg’s carbon footprint.
Figure 1 shows the total and per capita average carbon footprints in Luxembourg between 2010 and 2020. The total footprint has increased 11% between 2010 and 2020, from 8.0 to 8.9 MtCO2eq, while the per capita footprint has dropped 11%, from 15.8 tCO2eq/cap to 14.1 tCO2eq/cap (or tons CO2- equivalents per person). For comparison, this is about 40% higher than the EU per capita average[1]. Note that these reflect the household carbon footprint, or the carbon footprint associated with household purchases.
Figure 1: Total and per capita household carbon footprint in Luxembourg in 2010, 2015 and 2020.
Own calculations based on EXIOBASE 3.8.
Multiregional input-output analysis, the method used for household carbon footprints quantifications, allows for the bridging of the geographical distance between consumption and its global supply chains. As discussed above, this is a chance to account for emissions (see Technical Note (1)) that are embodied in the products and services consumed by households in Luxembourg; for example, equipment and clothes that were imported and bought in Luxembourg but were produced elsewhere (indirect emissions). They are particularly important to identify in a small, open economy like Luxembourg’s, where few of the goods consumed are produced locally.
Carbon footprint distribution
Figure 2 depicts the average footprint of top 1%, top 10%, middle 40% and bottom 50% households, where top 1% refer to the 1% of households with the highest carbon emissions embodied in their consumption. The top 1% emitters have an average carbon footprint between 60 and 80 tCO2eq/cap across the reviewed years (Figure 2). This is consistent with an earlier analysis, estimating the carbon footprint of a typical super-rich household of two to be about 130 tCO2eq[2]. Results should still be interpreted with some caution. While carbon footprints reduce consistently across time and carbon groups (top and bottom emitters), the results of the top 1% may also be heavily influenced by the presence of extreme outliers.
Figure 2: Average household carbon footprints in Luxembourg in 2010, 2015 and 2020.
Top 1% households are included in the top 10% category. Calculations based on EXIOBASE 3.8 and Eurostat Household Budget Survey.
Figure 3: Distribution of carbon footprints per capita.
The lines in the boxes describe 25th percentiles, median and 75th percentiles respectively, and the whiskers describe the minimum and maximum values in the absence of outside values.
The average carbon footprint of the top 10% carbon emitters also reduced from 43 to 36 tCO2eq/cap. The reductions in the carbon footprint of top emitters can be (at least partly) explained by the drop in carbon intensity of consumption in that period, from 0.6 to 0.4 kgCO2eq/EUR. The carbon intensity here refers to the amount of carbon dioxide equivalents per euro spent by households.
The average carbon footprint of the top 10% emitters is still five times that of the bottom 50%, and substantially above the climate targets. The share of resident respondents with carbon footprints within 2.5 tCO2eq/cap – the climate target for 2030 (see Technical Note (2)) – is below 1% across all years. The average carbon footprint of the bottom 50% is three times the target. That is, the carbon intensity of the consumption of the lowest 50% of emitters needs to reduced three times to fit within climate limits.
Figure 3 shows the distribution of the carbon footprint across the three years, highlighting a more concentrated distribution in 2020. This is likely influenced by the drop in demand associated with the COVID pandemic. I describe the method for footprint calculations across the different household types in Technical Note (3).
Consumption across consumption categories
Figure 4 shows how average carbon footprints are distributed across consumption category. Interpretations should focus on the differences across household types within the same year, to account for variations in carbon intensities across years.
Certain categories such as transport are particularly dominant in the carbon footprint of the highest emitters. For example, the transport share increases from 24% for the bottom 50% emitters to 61% for the top 10% emitters in 2010. This includes both private vehicle and air travel. At the same time, the carbon intensity of transport is remarkably high, 0.5 – 1.7 kgCO2eq/EUR spent across the years. On the other hand, basic categories such as food and shelter vary much less across the emitting groups in absolute terms.
Figure 4: Average household carbon footprints across consumption categories.
The EXIOBASE carbon multipliers vary substantially across the years, which explain the wide variation in the consumption category distribution from one year to another.
In the absence of careful consideration, blanket approaches to carbon mitigation can worsen existing inequalities. For example, understanding the elasticity of consumption of goods and services to the level of consumer income is key for reducing carbon emissions. High-income elasticity goods and services are those that are very sensitive to changes in income, with consumption that rises sharply in response to income increases. Inelastic products are those with relatively stable quantity demanded, even at varying household income. A comprehensive analysis of consumption will highlight products and consumption sectors that are both high-carbon AND highly elastic to income changes. These are generally good candidates for carbon policies, as they are less likely to worsen social inequalities.
Figure 5 highlights changes in product demand associated with increases in total expenditure and income. Income and expenditure are strongly correlated, as spending -and carbon footprints- increase proportionate to the income. Rises in total expenditure tend to translate into increases in the consumption of categories such as C05 Furnishings, household equipment and maintenance, C07 Transport, and C09 Recreation, sport and culture (Figure 5); in other words, these are the most elastic consumption categories. For example, doubling total expenditure (100% increase) results in an increase of transport spending by 150-160% (shown by expenditure elasticities of 1.5 and 1.6 across the years). However, the income elasticity of transport demand is as low as 0.8 for 2020. According to economic theory, an income elasticity below 1 indicates a necessary/normal good or service, as opposed to the consumption of luxury or ‘superfluous’ products. This suggests that certain transport activities are necessary and basic, and transport-related carbon policy needs to consider the wider context to ensure effectiveness and justice. This includes considerations such as whether there are available low-carbon alternatives, and how to make such alternatives affordable, accessible and convenient for all.
A substantial number of categories had income elasticities below one in 2015 and 2020, suggesting that they are basic, and so blanket policies that do not distinguish necessary from luxury consumption will likely be regressive and affect low-income consumers disproportionately.
Figure 5: Expenditure and income elasticities across survey categories.
Household income information is missing for 2010. Calculations based on the Eurostat Household Budget Survey.
Socio-demographic characteristics
Table 1 in Technical note (4) presents a regression analysis that assesses the role of socio-demographic characteristics for emissions, with level and source of income, household type and degree of urbanisation explaining around 56% of the variation in carbon footprints. The regression analysis also shows socio-demographic features that characterise the top 10% and bottom 50% emitters.
The top 10% emitters cluster is associated with high-income people, who tend to live in sparsely populated areas. It is more likely for single adults to be within the high-emitters group, compared to multi-person households with or without children. As more people live together, the carbon footprint of the household rises, but the average per capita footprint falls with household members sharing space and resources.
The bottom 50% carbon emitters are associated with lower incomes and higher risk of poverty. Income is a key determinant of carbon footprints, where a doubling of income results in a 70% increase in the carbon footprint. Having more household members is associated with a higher likelihood of being within the lowest emitting group; the likelihood is particularly high in the context of more than two adults with dependent children. People living on benefits (secondary income) are not more likely to be part of either top or bottom cluster.
The degree of urbanisation is significant in explaining the variation in the carbon footprint, but is insignificant among the bottom 50% emitters. That is, while the highest emitters tend to live in more sparsely populated areas, there are also low emitters living at similar population density. This again highlights the importance of accounting for social difference and intersectionality.
Conclusions
This article examines carbon inequalities around consumption, as well as some of the social and infrastructural differences that underpin them. I recommend that future research furthers this work by adding more detailed perspectives on the main carbon-intensive and highly elastic consumption sectors, which may be particularly suitable for demand-side mitigation approaches such as carbon tax on luxury items. Furthermore, it will be beneficial to have a more detailed account of who the lowest and highest emitters are in Luxembourg. It will be particularly useful to incorporate well-being perspectives to be able to connect carbon footprints with estimates on essential consumption for a good life.
Finally, we cannot design carbon policies with fair and sustainable outcomes without considering the origins and effects of that carbon. Luxembourg is strongly connected to the rest of the world through consumption, and any changes in it will produce ripple effects through the global supply chains. This has important implications for ensuring that consumption aligns with net-zero and well-being targets, both in Luxembourg and elsewhere.
Technical notes
This analysis follows the structure of Ivanova and Wood (2020)[1]. I performed the footprint analysis in Python based on EXIOBASE version 3.8.2, and the statistical analysis in Stata.
1. Carbon footprints were calculated using the Global Warming Potential 100 (GWP100) metric communicating the amount of CO2, CH4, N2O (from combustion and non-combustion) and SF6 in kgCO2-equivalents per year.
2. For comparison, a climate target of 2.5 tCO2eq/cap for 2030 is commonly presented as one that is consistent with emission pathways limiting global warming to 1.5°C[1].
3. I developed a consistent classification of consumption categories across the years in consideration of EXIOBASE’s product detail and the average consumption expenditure across each category. Across these categories, I calculated carbon intensities across all years. I derived the total household expenditure in basic prices in Luxembourg from the supply-use tables across all years and products, accounting for both domestic production and imports. Based on the transport and trade margins and product taxes, I estimated household expenditure in purchasers’ prices.
Using multiregional input-output analysis, I quantified indirect household carbon footprints across years and consumption categories. Based on the footprints and expenditure in purchasers’ prices, I could quantify indirect carbon intensities in kgCO2eq per EUR spent in purchasers’ prices. I distributed the direct household carbon footprint to the categories of home energy (gas, liquid, and solid fuels) and operation of personal transport equipment and quantified direct and total carbon intensities based on the expenditure in these categories. I applied the total carbon intensities to the expenditure in the survey to quantify household-level carbon footprints. Finally, I compared average footprints from the survey (adjusted by household weights) and EXIOBASE, aligning the survey distribution so it matched EXIOBASE’s per capita average.
The approach reflects the spending of residents filling out the household budget surveys; this is particularly key where Luxembourg is strongly affected by notable differences in territorial and residential accounts[3].
4. The regression table includes an OLS model on the continuous carbon footprint, and two logistic regressions on the binary variables describing the top 10% and bottom 50% emitters. More details about the independent variables: Densely populated in the Degree of urbanisation variable reflects at least 500 inhabitants/km2; intermediate – between 100 and 499 inhabitants/km2; and sparsely populated – less than 100 inhabitants/km2. Primary in the main source of income variable reflects income from wages or salary, income from self-employment, and property income; secondary reflects pensions, retirement benefits, unemployment benefits, and other current benefits and income.
Table 1: Multivariate regressions.
The regressions feature data from 2015 and 2020, as the 2010 dataset does not include income information. The intercepts have been removed.
References
1. Ivanova, D. & Wood, R. The unequal distribution of household carbon footprints in Europe and its link to sustainability. Global Sustainability 3, 1–12 (2020).
2. Otto, I. M., Kim, K. M., Dubrovsky, N. & Lucht, W. Shift the focus from the super-poor to the super-rich. Nat Clim Chang 9, 82–84 (2019).
3. Usubiaga, A. & Acosta-Fernández, J. Carbon emission accounting in MRIO models: The territory vs. the residence principle. Economic Systems Research 27, 458–477 (2015).