OECD Regions and Cities at a Glance 2024

Factors like real productivity growth, industrial specialisation and price changes influence regional economic performance over time. When the baskets of goods produced across regions are similar and/or inflation is low, the price effect will typically be negligible in explaining differences in regional performance, including over time. However, in nominal prices, the basket of goods produced can vary significantly across regions, especially in resource-rich regions and recent years have seen inflation reach its highest levels in decades. As such, the underlying assumption that regional economic performance can be assessed using national price indices has become increasingly strained. However, newly available data on regional price indices are now available, providing the scope to better understand relative performance across regions and how this translates into long-term economic growth.

This edition of Regions and Cities at a Glance introduces two significant methodological improvements: the use of regional deflators instead of national deflators in real gross domestic product (GDP) and gross value added (GVA) series at the large (TL2) and small (TL3) region levels and the use of regional consumer price indices (CPIs) to estimate real household incomes at the TL2 region level. Accounting for regional price changes provides a more accurate picture of economic growth drivers and changes in regions’ living conditions. These methodological improvements build on the recent release of regional deflators in European Union (EU) countries, complemented by data collection for other OECD countries.

Until recently, most countries did not produce volume estimates of regional GDP in basic prices that reflected the actual price changes of GDP in regions. Volume estimates, where available, were derived using national GDP deflators. While most countries do not currently produce regional estimates of GDP in volume terms using actual price change data observed by economic actors resident in the region, many are now generating volume estimates of regional GDP in basic prices by deflating regional value-added in a given industry using the national deflator for that activity. Whilst not perfect, this significantly improves a whole economy’s national GDP deflator. Moreover, whilst there may be significant differences in the price of a given product across regions in a country, there is less likely to be a significant difference in the price change of that product across the country, which means that volume measures of regional GDP produced in this way can be considered robust.

The primary advantage of employing regional deflators is the enhanced accuracy in measuring real (or volume) estimates of economic activity across distinct regions. National deflators, which average price changes across the entire country, can hide significant regional disparities in price changes. Using regional deflators better accounts for actual price changes producers and consumers face in each region. This methodological refinement facilitates more informed policy making, enabling the development and implementation of strategies better tailored to the specific economic circumstances of individual regions.

However, the transition to regional deflators presents certain challenges. Collecting and maintaining accurate regional price data require substantial resources and sophisticated statistical methodologies. Variations in data quality and availability across regions can affect the consistency and comparability of regional deflators. Additionally, while regional deflators provide a more accurate local perspective, chain-linked measures can result in a loss of additivity between deflated subnational measures and the equivalent national totals. Despite these challenges, the benefits of adopting regional deflators – such as enhanced accuracy of regional economic analysis – outweigh the drawbacks, ultimately leading to more effective economic policies.

This section explains the concept of regional deflators and, to help illustrate their importance, compares economic indicators with previous series using national deflators. The first section focuses on output prices, the second on consumer prices.

To arrive at the price deflators for regional GDP (hereafter called Regional Output Price Index, ROPI), current regional GDP estimates are divided by the regional GDP chain-linked volume series (with the series rebased on the reference year 2015). Regional GDP chain-linked volumes are available for most OECD and EU countries at the TL2 level. In addition, a few countries, such as Czechia, Denmark, Estonia, Finland, Japan, Korea, the Netherlands, Portugal, the Slovak Republic, Sweden, Türkiye, the United Kingdom and the United States, have data available at the more granular TL3 level. New Zealand, Norway and Switzerland do not have GDP series in volume at the regional level (Table ‎C.1).

Some national statistical offices (NSOs) provide public information about the methodology used to compile regional volume growth (Table ‎C.1). In essence, countries use two distinct approaches, in part reflecting the scarcity of regional‑level price data in many countries, as outlined in Chapter 6 of the EU Manual on Regional Accounts Methods (Eurostat, 2013[1]). As such, many countries use a top-down approach, applying national-level price changes to the region’s industrial structure (meaning that at the product level, all price changes are the same across the country) and differences across regional aggregates reflect only composition and not price effects. This compares with those countries able to adopt bottom-up methods, which capture differences in price changes across regions. In some cases, both approaches are combined, depending on data availability and regional characteristics.

Whilst these differences may impact international comparability, and indeed within-country comparability, the impact is mitigated somewhat by the fact that the methods only require estimates of price change of a given product rather than price levels, which is generally a weaker assumption (albeit with some exceptions, for example when price changes ripple from one part of the country to another, as is sometimes seen in housing markets).

Table ‎C.1 summarise different methods countries use to produce GDP and GVA volume series at the regional level, with some using bottom-up and others applying mixed or top-down methods. The information is provided for the seven countries where the methods are publicly available.

Effect of ROPIs on GDP measures for large regions

ROPIs vary substantially compared to the national (implicit GDP) deflator previously used to deflate the regional GDP current series. Over the period covered, most regions with the largest gaps between regional and national deflators are either specialised in the primary sector or are capital-city regions (Figure ‎C.1). Using the ROPI deflator has two main effects in countries with regions specialised in the primary sector (e.g. Chile, Colombia) over the period covered: i) the GDP per capita of regions specialised in natural resources decreases, as over the period covered price changes on the prices of natural resources increased at a significantly higher rate than for other products; and ii) in turn the GDP per capita of other regions, in particular capital-city regions, increases because the regional deflator is smaller than the national deflator.

To illustrate regional deflators’ impact on measures of real GDP growth, Figure ‎C.2 below compares regional shares of GDP based on nominal values (equivalent to shares derived via deflating with a national deflator) and equivalent shares using regional deflators. As expected for the period in question, shares for many regions specialised in the primary sector substantially decreased when using regional deflators. Additionally, the share of some capital regions, especially countries with regions specialising in the primary sector, increases.

The effects of the deflator are also significant enough to impact some inequality measures, such as the Theil index; which uses regional deflators and generally exhibits less variability than national deflators, especially when measuring within-country inequality (Figure ‎C.3). However, there is a consistent downward pattern when using regional deflators.

The effect of regional price measures can also impact estimates of other measures of polarisation (top 20% over bottom 20%, Figure ‎C.4). However, some care is needed in interpretation. Nominal comparisons of regional inequalities should remain the primary vehicle for comparing GDP per capita estimates. Deflated estimates, however, provide a means to understand to what extent the drivers of change in GDP per capita over time (convergence or divergence) reflect price changes (which may only be temporary and short-lived) or other structural changes (and, in particular, productivity growth).

Using regional deflators for regional GDP affects additivity, meaning that the sum of regional GDPs no longer equals the national GDP. This discrepancy arises because industry-specific price changes influence inflation rates in different regions. When aggregated, these variations in industry deflators can result in mismatches with national GDP, especially if a region experiences significant price changes in a dominant sector that do not reflect national trends. However, the additivity gap is generally small, with 37 countries showing a gap below +/-0.5% on average over the 2020-22 period. Notable exceptions include Latvia (0.5%), Romania (1.7%), and the Slovak Republic (0.9%). The additivity gap tends to be larger for years beyond the 2015 reference year. For example, in 2002, the gap was -2.3% for Denmark, -1.6% for Latvia, 9% for Romania and -3% for the United Kingdom (Table ‎C.2).

Effect of ROPIs on GDP measures for small regions

Currently, 13 OECD countries have ROPI deflators available at TL3 levels listed in Annex B. To conduct analysis at the TL3 level using regional deflators for all countries, the available information has been optimised by primarily using TL3 deflators where available. For countries where TL3 deflators are unavailable, TL2 deflators of the parent regions have been used to estimate GDP TL3 series in constant prices. National deflators have been applied for three countries lacking regional deflators (New Zealand, Norway and Switzerland).

The changes in regional inequality in GDP per capita within countries display less variance when using ROPI deflators (Panel A, Figure ‎C.5) compared to national GDP deflators (Panel B, Figure ‎C.5). In addition, the pattern of raising inequalities within countries for small regions differs from the observed downward trend in inequality between large regions (Figure ‎C.3).

Inflation reduces households’ disposable incomes and, other things being equal, more so in regions experiencing higher inflation. Regional differences in price changes are due to two factors: regional differences in price changes of individual products included in consumer baskets and different compositions (weights) of consumption baskets across regions. For example, housing costs and their weight in consumption baskets are typically higher in cities. This means that relative price falls in housing costs translate, in theory, into smaller price changes in urban versus non-urban areas.

The regional consumer price indices (RCPIs) are available at different subnational levels for 13 OECD countries (see Annex B for data sources and methods for countries with available data). The RCPI reflects changes in consumer prices experienced by households at their place of consumption. Whilst the basket of goods and services consumers purchase over time is typically the same across regions, the weighting schemes usually differ, as shown below.

In Australia, capital city CPIs are based on the 2011 Australian Statistical Geography Standard (ASGS) Greater Capital City Statistical Areas. These indices measure price movements over time in each city individually but they do not measure differences in retail price levels between cities.

In Canada, price surveys are based on a sample of representative goods and services for which prices are observed in selected areas of a city or province. The geographical distribution of the sample varies by product. The most geographically dispersed price samples are for goods and services where prices are likely to be heavily influenced by local market conditions (e.g. locally determined prices such as rents, water charges, local transit fares and property taxes). In contrast, prices for car registration or postage fees are collected from provincial or national agencies.

In Czechia, CPIs are calculated for all households in the country and specifically for households living in the capital city of Prague. The weighting scheme for households in Prague is based on their expenditure structure, with average prices calculated as the simple arithmetic mean of prices surveyed in the city. For certain expenses (such as therapeutic stays at a spa, ski lift tickets, university canteen catering services and university accommodation services), the average consumer prices for the whole of Czechia are used. Differences in weighting schemes between household groups reflect their distinct economic and living conditions, resulting in varied expenditure structures. The weighting schemes for pensioner households do not include imputed rentals.

In Italy, CPIs are calculated by aggregating product indices compiled at the provincial level. The all-items indices by province are derived as the weighted average of these provincial product indices. The weighting coefficients are based on the household consumption share of each product aggregate, defined at the regional level.

In Japan, the relative price index for each item is calculated for each municipality and then averaged using the respective weight for each municipality to obtain the average price index by item nationwide. This index is then averaged with the weight for each item to produce upper-level groups and the total index for Japan. The same procedure is also applied to city classes and districts.

In New Zealand, regional indices cover five broad regions (based on regional council areas): Auckland, Canterbury, the Rest of North Island, the Rest of South Island and Wellington. The regional indices currently being published1 are not considered fit for purpose as they rely on national movements for about 30% of the basket (based on expenditure weight), even when regional variation is possible. This includes important goods and services such as constructing new dwellings, dwelling rentals and used cars. The indices for the five broad regions will now incorporate price movements specific to their respective regions for these categories.

Regional vs. national CPIs to adjust regional GDP

Perhaps not surprisingly as baskets and weights of consumers are broadly much more comparable than comparisons of regional specialisations, the range of regional variation in CPIs compared to the national CPI is much smaller. For the period in question, in four out of eight countries with available data, CPIs in capital-city regions were lower than the national average across 2022 (Figure ‎C.6).

Regional vs. national CPIs to adjust household income

Overall, regional variations in inflation rates are not too high. However, in recent years, in line with higher inflation overall (2019-23), they have widened by 70% compared to the period from 2014‑18 (Figure ‎C.7).

The specialisation in the primary sector is based on GVA in several industries: agriculture, forestry and fishing (A); mining and quarrying (B); electricity, gas, steam and air conditioning supply (D); water supply, sewerage, waste management and remediation activities (E).

The specialisation is assessed with the Normalized Revealed Comparative Advantage (NRCA). This index measures the share of GVA in a particular industry $i$ of a region $r$ relative to the national GVA share in the same industry. The normalisation helps interpret the index; for example, a positive NRCA ($NRCAir>0$) indicates that region $r$ has a specialisation or comparative advantage in industry $i$ (Yu, Cai and Leung, 2008[2]). The NRCA was calculated for each year in 2000-22. Regions are identified as specialised if their NRCA is greater than 0.25 for at least half of the available data period or if they have been specialised in the last years of the period.

For Canada, since GVA by industries is only available in chained volumes, which affects the weight of natural resources industries, the regions are identified as specialised if their NRCA is greater than 0.

Access to hospital indicators was estimated using point of interest (POI) data, a one-kilometre resolution population (GHS-POP 2020) and degree of urbanisation (GHS-SMOD 2020) grids (Schiavina et al., 2023[3]) and the Mapbox Isochrone application programme interface (API).

Details on the hospital POI data sources for each country are available in Annex. For countries where data were available only in the form of postal addresses, these were converted into geographic co‑ordinates using the Python package Geocoder (https://pypi.org/project/geocoder/; ArcGIS provider https://developers.arcgis.com/rest/geocode/api-reference/overview-world-geocoding-service.htm).

The Mapbox Isochrone API2 provides isochrone polygons for a facility using its geographical co‑ordinates as inputs for time ranges up to 60 minutes for 3 modes of transport: driving, cycling and walking.

For each hospital, the 45-minute driving isochrone (i.e. the area that is reachable in 45 minutes by car around the considered hospital) was downloaded using the Mapbox Isochrone API. Isochrone polygons for all hospitals within the same country are dissolved into a single polygon to obtain the area within reach of at least one hospital and avoid overlaps. Isochrones are dissolved regardless of the regional borders, i.e. the area within reach of a hospital located in a given region is not limited to that region. These dissolved isochrones are then combined with the population grid, regional administrative borders and the degree of urbanisation grid to obtain the population within each region and the degree of urbanisation that can access a hospital in a 45-minute drive.

The climate zone for each TL3 region was defined using the Köppen-Geiger classification for 1991-2010 (Beck et al., 2023[4]). Figures C.10 and C.11 show the macro climate zones obtained in OECD countries. The climate zones for each TL3 region are obtained by first computing the population share in each climate zone using GHS-POP 2020 (Schiavina et al., 2023[3]). The climate zone of a TL3 region corresponds to the zone with the highest population share. Figures ‎C.8 and C.9 show respectively the distribution of regions and population by climate zone and country.

Climate projections by SSP scenario are derived from the NASA NEX-GDDP-CMIP6 dataset (Thrasher et al., 2022[5]). This dataset consists of global daily downscaled climate scenarios derived from the General Circulation Model (GCM) runs conducted under the Coupled Model Intercomparison Project Phase 6 (CMIP6). The CMIP6 GCM runs were developed to support the Intergovernmental Panel on Climate Change Sixth Assessment Report (IPCC AR6). The following CMIP6 models were used: ACCESS-CM2, ACCESS-ESM1-5, BCC-CSM2-MR, CMCC-ESM2, CanESM5, FGOALS-g3, GISS-E2-1-G, MIROC-ES2L, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, MRI-ESM2-0, NorESM2-LM, NorESM2-MM, TaiESM1 and UKESM1-0-LL.

Three climate indicators were derived from this dataset: air temperature, icing days and hot days. Hot days are days when the maximum air temperature exceeds 35 degrees Celsius (°C). Icing days are days during which the maximum air temperature does not exceed 0°C. These variables are then compared to the 1981‑2010 baseline climatology.

To estimate population exposure to hot days, the 2030 GHS-POP layer (Schiavina et al., 2023[3]) was used to compute population-weighted averages of hot days by region.

GHG emissions at the subnational level were estimated using the Emissions Database for Global Atmospheric Research (EDGAR) version 8.0 developed by the European Commission Joint Research Centre and the Internation Energy Agency (EC JRC/IEA, 2023[6]). EDGAR provides annual sector-specific grid maps for the four GHGs (carbon dioxide or CO₂, methane or CH₄, nitrous oxide or N₂O and fluorinated gases or F-gases) at a 0.1° spatial resolution (approximately 11 km). The different sectors and subsectors covered are:

• Energy: Power generation.

• Industry: Combustion in the manufacturing industry, oil refineries and transformation industry, chemical processes, fuel exploitation, iron and steel production, non-energy use of fuels, non‑ferrous metals and non-metallic minerals production, solvents and products use.

• Transport: Ground transport: road, trains and off-road transport. Shipping and aviation are excluded in the subnational GHG estimates for the transport sector.

• Building: Energy for buildings.

• Agriculture: Agricultural soils, agricultural waste burning, enteric fermentation, manure management, indirect N₂O emissions from agriculture.

• Waste: Solid waste incineration, landfills, wastewater handling.

Emissions from land use and land cover change are not included. National GHG emissions are disaggregated by using subsector-specific geospatial proxies. GHG emissions are expressed in CO₂ equivalents using 100-year global warming potential from the IPCC 5^th Assessment Report (AR5), i.e. 28 for CH₄ and 265 for N₂O.

To estimate an emissions reduction target for 2030 for advanced economies as a whole, and that can serve as a reference to compare OECD regions, this publication derives a per capita GHG emissions 2030 target from the IEA Net Zero Emissions Scenario (IEA, 2021[7]). In this scenario, CO₂ emissions from energy-related and industrial processes in advanced economies, i.e. in OECD countries and the 27 European Union member states, drop from 12.3 gigatonnes of CO₂ equivalent (Gt CO₂-eq) to around 5.5 Gt CO2-eq in 2030, which corresponds to a drop from 8.8 t CO2-eq per person in 2019 to 3.8 t CO₂-eq per person in 2030. The share of CO₂ emitted by global energy-related and industrial processes from advanced economies falls from 34% to 26%. Given that CO₂ from fossil fuel and industry, CH₄ and N₂O account for respectively 12.8 Gt CO2-eq, 2.2 Gt CO2-eq and 0.8 Gt CO2-eq of net anthropogenic GHG emissions in advanced economies (EC JRC/IEA, 2023[6]), we assume the same distribution of GHG in 2030. GHG emissions per capita, would have to drop to 4.7 t CO2-eq per person in 2030.

Forest cover loss at the subnational level was estimated using the Hansen Global Forest Change (2000‑22) data (Hansen et al., 2013[8]). This dataset provides forest extent and changes at 30 metre (m) resolution. It is derived from Landsat satellite images. Trees are defined as vegetation taller than five m in height and are expressed as a percentage per output grid cell as 2000 Percent Tree Cover. Forest Cover Loss is defined as a stand-replacement disturbance, or a change from a forest to a non-forest state, from 2000 to 2022. Due to data availability, only forest cover loss is included in this publication and not net forest change, which would also include forest gain from 2000 to 2022.

Protected areas are delineated using the World Database on Protected Areas (UNEP-WCMC/IUCN, 2024[9]), the most comprehensive global database on terrestrial and marine protected areas. It is a joint project between the United Nations Environment Programme (UNEP) and the International Union for Conservation of Nature (IUCN) and is managed by the UNEP World Conservation Monitoring Centre (UNEP-WCMC), in collaboration with governments, non-governmental organisations, academia and industry.

Protected areas are classified into 100% terrestrial, coastal, marine and terrestrial, and 100% marine. In this publication, marine protected areas are excluded. The IUCN also classifies protected areas by management categories:

• Ia: Strict nature reserve: Areas strictly protected to preserve biodiversity and possibly geological features, where human visits are strictly controlled.

• Ib: Wilderness area: Large, unmodified or slightly modified areas retaining their natural character without permanent or significant human habitat.

• II: National park: Large natural or near natural areas set aside to protect large-scale ecological processes.

• III: Natural monument or feature: To protect a specific natural monument, which can be a landform, mountain or geological feature.

• IV: Habitat/species management area: To protect particular species or habitats where management reflects this priority.

• V: Protected landscape/seascape: An area where the interaction of people and nature over time has produced an area of distinct character with significant ecological, biological, cultural and scenic value.

• VI: Protected areas with sustainable use of natural resources: Areas that conserve ecosystems and habitats together with associated cultural values and traditional natural resource management systems.

This publication defines protected areas with strict management based on Categories Ia, Ib, II and IV.

The boundaries of protected areas are intersected with OECD regions to get the share of protected areas.

The Kunming-Montreal Global Biodiversity Framework aims to protect at least 30% of land by 2030 (Convention on Biological Diversity, 2023[10]).

Burnt areas by land cover were obtained using the JRC global wildfire dataset to analyse fire regimes and fire behaviours (Artés et al., 2019[11]), based on MODIS burnt area product Collection 6. This dataset provides yearly individual fire perimeters from 2000 to 2022. Burnt areas are combined with Copernicus annual 300 m land cover (CCI-LC) data (C3S CDS, 2019[12]) to get burnt forest areas.

Enrolment in HEIs refers to enrolment in International Standard Classification of Education (ISCED) 5 to 7 programmes (including ISCED 7 long degrees). The presented indicators include institutions with at least 200 students enrolled at ISCED 5 or above and where at least 1 person graduated at ISCED 5 or above in the latest year for which data are available.

The methodology follows the OECD-International Transport Forum/European Commission Urban Access Framework (UAF) (ITF, 2019[13]). The UAF addresses comparability issues arising from city size and administrative boundaries. The framework proposes the “transport performance ratio”, a measure that standardises the accessibility measure by a proximity indicator. The rationale for the indicator is to compare what can be reached using the transport network to what could have been reached if one had no barriers to move in space.

$actual-to-maximum=\frac{absolute accessibility}{absolute proximity}$

Accessible destinations are measured by the absolute cumulative opportunity index (the number of destinations reachable within a given time by a given mode). In contrast, nearby destinations are measured with the proximity indicator (total number of destinations within a certain Euclidian distance radius buffer). The time and distance ratios are described in the following table. Implicitly, the ratios assume a travel speed of 4 km per hour for walking and 16 km per hour for cycling, driving and public transportation.

Transport performance ratios were computed using General Transit Feed Specification (GTFS) data from the Mobility Database (https://database.mobilitydata.org/) and additional GTFS data sources as described in Annex B, OpenStreetMap road network data, a 1-kilometre resolution population grid (GHS-POP 2020) (Schiavina et al., 2023[3]), elevation data (Terrain Tiles on Amazon Web Services)3 and the r5r R package.4

Access to services in functional urban area (FUA) indicators was estimated using point of interest (POI) data, OpenStreetMap (OSM) road network, a 100-metre resolution population grid (GHS-POP 2020) (Schiavina et al., 2023[3]) and the r5py Python library (Fink, 2022[14]).

Details on the POI data sources for each country are available in Annex B. For countries where data were available only in the form of postal addresses, these were converted into geographic co‑ordinates using the Python package Geocoder (https://pypi.org/project/geocoder/; ArcGIS provider https://developers.arcgis.com/rest/geocode/api-reference/overview-world-geocoding-service.htm).

The r5py library allows computing origin-destination (OD) matrices from two sets of co‑ordinates (origin and destination points). OD matrices provide the travel time (in minutes) between each origin and each generation. Origin points correspond to the facilities’ co‑ordinates for the service of interest (e.g. pharmacies). Destination points correspond to population and are computed as the centroids of the population grid cells. Due to computational limitations, the 100-metre GHS-POP 2020 grid was re-scaled to a 500-metre population grid. The OD matrices were computed using a walking speed of 4 km/h.

The share of the population that lives within a certain time (e.g. 15 minutes) from the nearest facility (e.g. a pharmacy) is derived from the OD matrices.

Access to green area indicators in FUAs’ urban centres was estimated using Open Street Map data. Selected amenities include the following tags: grave_yard; land use: cemetery, forest, recreation_ground, village_green; leisure: nature reserve, park, playground, recreation_ground, garden; tourism: zoo; natural: wood, scrub, heath, grassland, wetland. The population within 400 metres of the nearest green area was calculated using a 400-metre buffer around the green areas’ polygons and the 100-metre resolution GHS‑POP 2020 population grid (Schiavina et al., 2023[3]).

Regional disparities are also measured by a Theil entropy index, which is defined as:

$Theil=\sum _{i=1}^N\frac{y_i}{\stackrel{-}{y}}ln\left(\frac{y_i}{\stackrel{-}{y}}\right)$

where $N$ is the number of regions in the OECD, $y_i$ is the variable of interest in the $i$-^th region (i.e. household income, life expectancy, homicide rate, etc.) and $\stackrel{-}{y}$ is the mean of the variable of interest across all regions.

The Theil index can be easily decomposed into two components: i) the disparities within subgroups of regions, where, for example, a subgroup is identified by a set of regions belonging to a country; ii) the disparities between subgroups of regions (i.e. between countries). The sum of these two components is equal to the Theil index.

In order to decompose the Theil index, let us start by assuming $m$ groups of regions (countries). The decomposition will assume the following form:

$Theil=\sum _{j=1}^M\sum _{i=1}^N{s}_j\frac{y_{ij}}{\stackrel{-}{y_j}}ln\left(\frac{y_{ij}}{{\stackrel{-}{y}}_j}\right)+\sum _{j=1}^M{s}_{j}ln\left(\frac{y_j}{\stackrel{-}{y}}\right)$

where the first term of the formula is the within part of the decomposition equal to the weighted average of the Theil inequality indices of each country. Weights, $s_i$, are computed as the ratio between the country average of the variable of interest and the OECD average of the same variable. The second term is the between a component of the Theil index and represents the share of regional disparities that depends on the disparities across countries.

The Theil index ranges between zero and $\infty$, zero representing an equal distribution and higher values representing a higher level of inequality.

The index assigns equal weight to each region regardless of its size; therefore, differences in the values of the index among countries may be partially due to differences in the average size of regions in each country.

[11] Artés, T. et al. (2019), “A global wildfire dataset for the analysis of fire regimes and fire behaviour”, Scientific Data, Vol. 6/1, p. 296, https://doi.org/10.1038/s41597-019-0312-2.

[4] Beck, H. et al. (2023), “High-resolution (1 km) Köppen-Geiger maps for 1901–2099 based on constrained CMIP6 projections”, Scientific Data, Vol. 10/1, p. 724, https://doi.org/10.1038/s41597-023-02549-6.

[12] C3S CDS (2019), Land Cover Classification Gridded Maps from 1992 to Present Derived from Satellite Observation, Copernicus Climate Change Service Climate Data Store, https://doi.org/10.24381/cds.006f2c9a.

[10] Convention on Biological Diversity (2023), 2030 Targets (with Guidance Notes), https://www.cbd.int/gbf/targets.

[6] EC JRC/IEA (2023), EDGAR (Emissions Database for Global Atmospheric Research) Community GHG Database: IEA-EDGAR CO2, EDGAR CH4, EDGAR N2O, EDGAR F-GASES version 8.0, JRC (Datasets), European Commission.

[1] Eurostat (2013), “The compilation of regional GVA in prices of the previous year; volume development of GVA”, in Manual on Regional Accounts Methods, Eurostat, https://ec.europa.eu/eurostat/documents/3859598/5937641/KS-GQ-13-001-EN.PDF/7114fba9-1a3f-43df-b028-e97232b6bac5.

[14] Fink, C. (2022), r5py: Rapid Realistic Routing with R5 in Python (0.0.4), https://doi.org/10.5281/zenodo.7060438.

[8] Hansen, M. et al. (2013), “High-resolution global maps of 21st-century forest cover change”, Science, Vol. 342/6160, pp. 850-853, https://doi.org/10.1126/science.1244693.

[7] IEA (2021), Net Zero by 2050, IEA, Paris, https://www.iea.org/reports/net-zero-by-2050.

[13] ITF (2019), “Benchmarking Accessibility in Cities: Measuring the Impact of Proximity and Transport Performance”, International Transport Forum Policy Papers, No. 68, OECD Publishing, Paris, https://doi.org/10.1787/4b1f722b-en.

[3] Schiavina, M. et al. (2023), GHS-POP R2023A - GHS Population Grid Multitemporal (1975-2030), https://doi.org/10.2905/2FF68A52-5B5B-4A22-8F40-C41DA8332CFE.

[5] Thrasher, B. et al. (2022), “NASA Global Daily Downscaled Projections, CMIP6”, Scientific Data, Vol. 9/1, p. 262, https://doi.org/10.1038/s41597-022-01393-4.

[9] UNEP-WCMC/IUCN (2024), Protected Planet: The World Database on Protected Areas (WDPA) and World Database on Other Effective Area-based Conservation Measures (WD-OECM), UNEP-WCMC and IUCN, Cambridge, UK.

[2] Yu, R., J. Cai and P. Leung (2008), “The normalized revealed comparative advantage index”, The Annals of Regional Science, Vol. 43/1, pp. 267-282, https://doi.org/10.1007/s00168-008-0213-3.