This chapter analyses the geographic distribution of population and service provision in Korea and proposes a methodology to identify regional hubs. It begins by describing the methodology to define settlements and their reachability in Korea in an internationally comparable way. It then presents key facts on the spatial distribution of the population and services in Korea compared to the rest of the OECD. The last part combines data on population sizes and driving times with rich data on service locations in Korea to identify 37 regional hubs.
Enhancing Compactness, Connectivity and Accessibility in Korea
Abstract
This chapter analyses the geographic distribution of population and service provision in Korea and proposes a methodology to identify regional hubs. The chapter aims to answer questions such as: “did the population in villages near an urban centre decrease less than in more remote settlements?” or “do towns that are larger than their surroundings have more services than other towns?”. To arrive at answers that consider size and relative position of settlements, the chapter applies the same typology as OECD (2024[1]) to all countries with available data on service locations. The combination of internationally consistent settlement delineations, driving time data, and cutoff values allows for a comparison between Korea and other OECD countries that is driven as little as possible by country-specific definitions.
The next section describes the methodology to define settlements and their reachability in Korea in an internationally comparable way. The third section presents five key facts on the spatial distribution of the population and services in Korea compared to the rest of the OECD. The last section combines data on population sizes and driving times with rich data on service locations in Korea to identify 37 regional hubs.
The analysis relies on internationally standardised definitions, in particular settlements from level 2 of the degree of urbanisation framework (OECD et al., 2021[2]). This framework specifies thresholds regarding population density and population size to aggregate contiguous high density population grid cells into urban centres, towns, and villages (see Box 2.1 for more details).
Urban centres are smaller than functional urban areas that also include commuting zones. In the degree of urbanisation framework, “cities” correspond to administrative definitions, i.e. the administrative boundaries of the municipalities that contain an urban centre. In this report, reference is made to “urban centres” instead of “cities” as the unit of analysis are grid-based areas instead of administratively defined areas.
Box 2.2 illustrates different urban definitions using the example of Seoul. Defining urban centres as settlements within a unified framework that includes towns and villages enables a comprehensive assessment of their spatial relationships. Moreover, using an internationally standardised definition of settlements helps overcome challenges associated with comparing local areas of vastly different sizes. For instance, the average population of a municipality exceeds 200 000 in Korea but falls below 2 000 in France. As a result, municipalities alone offer limited value for meaningful comparisons in terms of population concentration and service provision.
Position in the settlement network
Besides a settlements’ own size, its relative position in the settlement network also matters. For example, the geographic location relative to other settlements can affect service provision. A town for which the next bigger settlement is a long drive away might heavily depend on having a hospital within its boundaries. In contrast, people in a similar-sized town close to a large urban centre might have relatively good access to a hospital even when there is none located in the town itself.
Following the typology used in a previous OECD report (OECD, 2024[1]), this report classifies settlements into three reachability types:
1. Access to an urban centre denotes towns and villages within 30 minutes driving time from an urban centre. The threshold approximately corresponds to the average time for one-way commutes in OECD countries. Driving times consider the distance from the centre of the town or village to the edge of the nearest urban centre. They rely on simulated driving times using the Mapbox Isochrone Application Programming Interface and TomTom road network data.
2. Regional centres are the biggest settlements within 30 minutes driving time. This definition can apply to urban centres, towns, or villages. Towns and villages with access to an urban centre cannot be regional centres, as it is possible to reach the boundaries of a more populated settlement within 30 minutes from them.
3. No access to an urban centre describes towns and villages that have a more populated town or village within 30 minutes driving time, but no urban centre. The three groups are mutually exclusive; towns and villages with no access to an urban centre are all those towns and villages that are neither regional centres nor towns or villages with access to an urban centre in 30 minutes. It is possible to set different time thresholds and adjust all three definitions accordingly.
Korea has 71 regional centres: 27 urban centres, 26 towns and 18 villages. Figure 2.2 shows the geographic distribution of Korean settlements and their reachability type. Urban centres can be a regional centre (depicted in dark red) or not a regional centre when they are close to a larger urban centre (light red). Towns and villages with access to an urban centre (light blue) surround the urban centres. Regional centers (medium blue) and towns and villages without access to an urban centre (dark blue) can be found at greater distance from urban centres.
There are two major clusters of urban centres and smaller settlements with access to them, the first in the northwest around Seoul, and the second in the southeast around Busan and Ulsan, including large parts of the South Gyeongsang province. On the mainland, regional centres are typically towns, while several islands have villages that are regional centres. About three fourths of Korean smaller settlements with no access to an urban centre are villages. Moreover, the different types of towns and villages vary systematically in size, both in Korea and for the average across OECD countries (Table 2.1).
Source: Based on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
The degree of urbanisation framework provides an internationally standardised definition of settlements. It was developed by the OECD and EC, together with four other international organisations. The United Nations Statistical Commission endorsed it in 2020 as the recommended method for international statistical comparisons between cities and other settlements along the rural-urban continuum.
It identifies clusters of adjacent 1km x 1km cells from a regular population grid. Rules regarding population density in each cell and total population in each cell cluster classify the grid cells into different groups, listed in Table 2.2.
Level 1 partitions countries into urban centres, towns & semi-dense areas and rural areas. Level 2 includes urban centres, towns and villages. Table 2.3 presents the corresponding population threshold values. Settlements are the units of observation for most of the analysis in this chapter. Annex Table 2.A.1 presents the population grids used for the settlement delineation by country.
Source: OECD (2021[2]), “Applying the Degree of Urbanisation: A Methodological Manual to Define Cities, Towns and Rural Areas for International Comparisons”, https://doi.org/10.1787/4bc1c502-en.
Figure 2.1 shows the administrative boundaries of Seoul (in red) and the SMA that comprises the TL3 regions Seoul, Incheon, and Gyeonggi-do (turquois). Administrative boundaries typically follow historical circumstances rather than functional criteria and are difficult to compare internationally. Functional urban areas (FUAs) are internationally standardised and combine a dense urban centre with surrounding local areas that have strong commuting connections with the urban centre (Dijkstra, Poelman and Veneri, 2019[4]). Depending on the definition, Seoul has about 9.5 million (administrative boundaries), 24.2 million (FUA), 25.7 million (Seoul metropolitan area), or 18.8 million (settlement) inhabitants, amounting to 18%, 47%, 50%, or 37% of the Korean population.
Figure 2.1 depicts the Seoul FUA in blue and other FUAs in orange. It also distinguishes between the Seoul settlement in green and other settlements in yellow. As FUAs do not cover many towns and villages, this report uses settlements as the unit of analysis. If a commuting zone contains several settlements, the analysis considers them individually.
Note: The map does not show settlements beyond Korea’s national borders.
Source: The settlement delineation and population counts rely on the 2021 Korean population grid provided by the National Geographic Information Institute of Korea.
Source: Based on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Service locations
Comparing service locations between countries requires high-quality point-of-interest data for an internationally consistent set of countries. Unfortunately, such data is not readily available. Following OECD (2024[1]), the comparative part of the analysis focuses on five key services with available data from 32 OECD member and accession countries (Annex Table 2.A.2). The five key services are schools, banks, pharmacies, hospitals, and higher education institutions (HEIs).
This section describes five key facts about the spatial distribution of population and service provision in Korea:1
1. The population in Korea is very concentrated in and around urban centres.
2. While the distribution of urban centre sizes in Korea is consistent with the size and urbanisation rate of the country, Seoul is extraordinarily large in absolute terms.
3. Recent population growth has been highly concentrated in the northwest and – to a smaller extent – the southeast of the country.
4. While the recent increase in concentration by settlement size is comparable to other periods or countries, it still places a significant burden on Seoul.
5. Unlike other OECD countries, Korean towns and villages further away from urban centres do not have more schools, pharmacies, or banks than towns and villages close to an urban centre.
The next sections subsequently discuss each of the five key facts and contextualise the situation in Korea by comparing it with other OECD member and accession countries.
The population of Korea is highly concentrated in and around urban centres
Korea is a highly urbanised country. A large share of its population lives in urban centres and even those living in towns and villages can often reach an urban centre within 30 minutes. The relatively small population share outside urban centres is not more clustered in settlements than elsewhere.
The share of people living in urban centres in Korea is considerably higher than in most other OECD countries. Figure 2.3 shows the population share by level 1 of the degree of urbanisation (Box 2.1) for all OECD countries. More than three out of four people in Korea live in urban centres. The country with the second highest share living in urban centres is Japan with 70%, while the (population-weighted) OECD average falls just short of 50%. In 21 out of 38 countries, less than 40% of people live in urban centres.2
Source: Based on population data from GHS: Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
The comparatively high level of population concentration in Korea persists when considering the clustering of towns and villages around urban centres, even if Korea is less of an outlier in that regard. About 80% of people in Korean towns and villages can reach an urban centre within 30 minutes driving time, while the average for 32 countries with available data is 50% (Figure 2.4). Korea has similar levels of accessibility to Switzerland, and lower levels compared to Belgium, Luxembourg, and the Netherlands, where 90% to 93% of people living in towns and villages have access to an urban centre within 30 minutes driving time.
Source: Based on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Korea is also the most concentrated country when considering the share of people living in urban centres together with the share of people living in smaller settlements with access to an urban centre. Figure 2.5 shows the corresponding cross-country comparison. On average, about half of the population lives in settlements more than 30 minutes away from any urban centre or outside settlements, i.e., in an area with less than 300 inhabitants per square kilometre. In Korea, that share is only 13%, with the remaining 87% living either in urban centres or in towns and villages with access to an urban centre.
Source: Based on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
However, the extraordinarily high levels of population concentration in Korea are limited to urban centres and the towns and villages surrounding them. Annex Figure 2.B.1 compares the number of people living in towns and villages to the total number of people living outside of urban centres. With a share of 48%, Korea is right in the middle of the cross-country distribution. This means that about as many people in Korea live outside of settlements (defined according to Box 2.1) as in towns and villages.
While the distribution of urban centre sizes in Korea is in line with its urbanisation rate, Seoul is extraordinarily large in absolute terms
The Korean urban centre size distribution is consistent with a country of its size and high level of urbanisation. It closely follows Zipf’s law - an empirical regularity which states that city sizes in a country follow a specific distribution - and is similar to the respective distributions of city sizes in Canada and the United States (see Box 2.3). Seoul is the largest urban centre in the country, both in absolute terms and relative to the total country population. However, when compared to the total population living in urban centres, its size is exactly in the middle of the first-ranked urban centres in the sample. This underlines the high level of urbanisation in Korea and suggests that the size of Seoul might be a consequence of it. Nevertheless, the large absolute size of Seoul might imply large agglomeration costs. Similarly, while the growth rate of the Seoul urban centre was not extraordinary, compared to other first-ranked urban centres, the large population increase in absolute terms implies substantial challenges concerning housing and infrastructure.
In its most basic form, Zipf’s law predicts the largest city to have twice as many inhabitants as the second largest city, three times as many inhabitants as the third largest city, and so on. It can be tested by regressing the logarithm of the rank minus one-half on the logarithm of the settlement population (Gabaix and Ibragimov (2011[6]), Chauvin et al. (2017[7])). If Zipf’s law holds true, the regression should result in a very high fit with an R-squared close to one, which studies have confirmed for many countries. The slope parameter indicates whether population rises faster or slower as rank falls. The strict version of Zipf’s law suggests a slope parameter of minus one, but values somewhat above or below are common. If large urban centres are considerably above (below) the fitted line, they are bigger (smaller) than expected.
The urban centre size distribution of Korea closely follows Zipf’s law. The corresponding R-squared is 0.98. The slope parameter is -0.90, suggesting that urban centres rise somewhat faster in population size as rank falls, compared to what the strict version of Zipf’s law predicts. However, this is fully in line with the estimates for other countries. As Figure 2.6 shows, Seoul is somewhat larger and the second, third, and fourth ranked urban centres (Busan, Daegu, and Gwangju) are somewhat smaller than predicted.
Figure 2.6 shows Zipf’s law for six countries with many urban centres.3 The Korean slope parameter (-0.90) is very close to the ones from Canada (-0.90) and the United States (-0.89).4 European countries tend to have steeper slopes (-1.10 for Germany and Italy, -1.20 for Spain). This implies that city population in European countries increases slower as rank falls, with their largest urban centres typically being smaller than in Canada, Korea, or the United States.
Note: The points and regressions lines represent the relation between log10(rank-1/2) and log10(settlement population). The axis labels show values without the log-transformation for better readability.
Source: Based on population data specified in Annex Table 2.A.1.
Despite the regularities in the rank of urban centres by their size, Seoul is extraordinarily large in absolute and relative terms. Annex Table 2.B.1 lists the largest settlements across OECD countries. The settlement of Seoul comes on top of the list with 18.8 million inhabitants, followed by New York City (15.4 million), Los Angeles (14.2 million), and Paris (9.6 million).5 Figure 2.7 provides a measure of primacy by dividing the size of the largest urban centre of a country by that country’s total population. In that relative way of measuring size, Seoul is the largest settlement in the sample as well, hosting 37% of the Korean population.
Note: Based on population data specified in Annex Table 2.A.1.
However, Korea’s size and high urbanisation level can explain the large size of Seoul. Figure 2.8 compares the size of the largest urban centre of a country by that country’s total urban centre population. This is another measure of primacy, but unlike the one in Figure 2.7, it does not show Seoul among the (relatively) largest urban centres. With Seoul’s share of hosting 46% of all people living in urban centres, it ranks at position 17 of 32. The fact that so many Koreans live in urban centres (Figure 2.3) explains this striking difference between the two primacy measures.
Note: Based on population data specified in Annex Table 2.A.1.
Recent population growth concentrated in the northwest and – to a smaller extent – the southeast of the country
The geographic concentration in the two main population clusters in the northwest and, to some extent, in the southeast of the country increased between 2005 and 2020.6 Figure 2.9 shows for each Korean settlement whether its population increased (depicted in yellow) or decreased (in blue) between 2005 and 2020. While the average population of Korean settlements slightly increased over that period, 59% of settlements lost inhabitants. Urban centres in the cluster around Seoul grew, as did most towns and villages with access to them. The picture in the cluster around Busan and Ulsan is more mixed. Most other settlements had less inhabitants in 2020 than in 2005.
Only 49 of 80 (61%) urban centres exhibited population growth. Urban centres that are regional centres (part of Panel A, see Annex Figure 2.B.2 for a breakdown) did not grow more than other urban centres (Panel B). Most of the urban centres that did grow are in the area around Seoul or, to a lesser degree, in the area around Busan and Ulsan in the southeast of the country. Notably, population growth in the southeast did not happen in urban centres that are regional centres like Busan and Ulsan themselves but in smaller urban centres around them. Beyond the two main clusters, a clear majority of urban centres lost population.
Source: The settlement delineations and typology rely on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Towns and villages with access to an urban centre (panel C) show a similar picture of increased geographic concentration. Around Seoul, an overwhelming majority of smaller settlements with access to an urban centre gained population. Near Busan and Ulsan, the situation is very mixed, with some settlements losing and others gaining inhabitants. In the rest of the country, having access to an urban centre did not prevent most settlements from population losses. When considering all regions, 52% of towns and villages with access to an urban centre experienced a population decline between 2005 and 2020.
The overwhelming majority of settlements further away from urban centres lost population. This is true regardless of whether these towns and villages are regional centres or not. Only 14% of regional centre towns and villages (part of Panel A, see Annex Figure 2.B.2) and 17% of towns and villages with no access to an urban centre (Panel D) had more inhabitants in 2020 than in 2005. Those with population growth tend to be in the extended vicinity of Seoul or on islands.
While the recent increase in concentration by settlement size is comparable to other periods or countries, it still places a significant burden on Seoul
Recent population concentration trends in Korea are part of a longer trend that was stronger in the past. Compared to the more recent period, the country urbanised more rapidly between 1990 and 2005. The population in urban centres other than Seoul grew 0.7% over that period, compared to only 0.2% between 2005 and 2020 (Figure 2.10). Villages lost 1.1% of inhabitants between 1990 and 2005, and towns also experienced a (minimal) population decline. Between 2005 and 2020, villages still depopulated, but at a smaller rate of 0.4%. At the same time, towns gained population at a rate comparable to that of urban centres other than Seoul. The growth rate of Seoul was similar over the two time periods.
Source: The settlement delineations rely on population data specified in Annex Table 2.A.1. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Korean urban centres experienced faster population growth than urban centres in other OECD countries between 1990 and 2005, but they grew at a similar rate between 2005 and 2020. Seoul’s population growth rate was about twice as high as the OECD average of first-ranked urban centres between 1990 and 2005 and slightly below average between 2005 and 2020 (see Annex Figure 2.B.3).7
Future projections predict a population decrease across all settlement sizes. In both Korea and the OECD average, villages will lose population at the highest rate, followed by towns and then urban centres that are not the largest of their country. However, the predicted growth rate between 2020 and 2030 is below average (more negative) for Korean settlements of all sizes. The prediction of a population decline within the settlement of Seoul might be counterbalanced by additional growth at its fringe, with the aggregate of the two forces possibly resulting in modest population growth.
Population concentration trends among Korean urban centres of different sizes have not been stronger than in other OECD countries in recent years. The annual population growth rates by urban centre size in Korea are not particularly large or different among each other compared to other OECD countries (Figure 2.11).8 This is not just an effect of relatively smaller first-ranked urban centres catching up: there is almost no correlation between the 2005 (log) population size of each countries’ largest urban centre and their population growth rate between 2005 and 2020.9 In 21 of 30 countries, the largest urban centre grew faster than the respective aggregates of large urban centres, mid-size urban centres, and small urban centres.
Note: The order of countries follows the difference in the compound population growth rates between the largest and all small urban centres.
Source: The settlement delineations rely on population data specified in Annex Table 2.A.1. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Seoul’s population growth rate between 2005 and 2020 was not higher than those of other first-ranked urban centres. Table 2.4 list the largest urban centre for each country with available data and displays its population growth between 2005 and 2020. While Seoul’s annual growth rate during that period was markedly above the Korean average with 0.6% (see Figure 2.10), half of the first-ranked urban centres experienced higher growth rates. Their average annual growth rate was 0.7% and individual urban centres like Auckland, Luxembourg, and Stockholm had values as high as 2.1%, 3%, and 1.9%, respectively.
However, a high absolute level and increase of population might become increasingly costly for very large urban centres. Table 2.4 also shows that Seoul’s absolute population increase between 2005 and 2020 was higher than in all other first-ranked urban centres. Adding additional people requires housing and infrastructure and exacerbates congestion. This challenge is complicated by the fact that Seoul is already one of the most densely populated urban centres in the sample, limiting the potential for infill. Individuals do typically not take their own marginal contribution to congestion (or other positive and negative) externalities into account for their moving decisions. When weighting their own benefit of moving to (or staying in) Seoul against the associated costs, Seoul still seems to be a very attractive location.
Source: The settlement delineations rely on population data specified in Annex Table 2.A.1. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Korean towns and villages far from urban centres do not have more services than those closer to urban centres
The service completeness index describes the availability of schools, banks, hospitals, pharmacies, and HEIs in a settlement. The index provides information about the extensive margin (for example, whether a settlement has at least one school). An index value of one means a settlement provides precisely one of these five service types to its inhabitants. In contrast, an index value of five means a settlement hosts all five service types. The index does not include information about the number of distinct service locations of the same type; having one or nine schools adds one point.
Larger settlements provide a higher number of different services than smaller settlements (Figure 2.12). This trend is visible across all countries, including Korea.10 The average village offers between one and three of the five different services, while the average town has three to four service types. Urban centres typically have at least four service types, and almost all larger urban centres above 250 000 inhabitants have at least one pharmacy, school, bank, hospital, and HEI. For large parts of the distribution, the average values for Korea are close to the international sample averages.
Note: Each marker averages settlements of population bins.
Source: Based on population data specified in Annex Table 2.A.1 and service data specified in Annex Table 2.A.2 and Annex Table 2.A.3.
Small villages in Korea offer more different service types than their counterparts elsewhere. Even villages in the smallest bracket, with not much more than 500 inhabitants, have 1.8 different services on average in Korea. The corresponding international sample average is only 1.2. The gap narrows for larger villages and towns but does not completely close for these smaller settlements. In contrast, there is no difference between Korean urban centres and urban centres elsewhere.
Regression analysis (detailed in Box 2.4) shows that in most OECD countries, smaller settlements that have access to an urban centre have fewer service locations than equally sized settlements that do not. A plausible explanation is that residents can more easily obtain services in the nearby urban centre. Many might also commute there for work, allowing them to access services along the way. In this sense, services in urban centres act as substitutes for those in towns or villages.
By contrast, in OECD countries people living in regional centres or settlements without access to an urban centre face greater challenges in reaching alternative service locations. These differences influence the demand for market-based services, affecting their profitability and firms’ willingness to provide them. For state-provided services, planners may prioritize locations where alternative access is more limited. Regional centres, as hubs for surrounding smaller settlements, often serve as service providers in less densely populated areas where offering a service location in every settlement is impractical. As a result, the largest settlement within a 30-minute drive often becomes the primary service hub. Consistent with this pattern, regional centres in most OECD countries host more services than other equally sized settlements without access to an urban centre.
In OECD countries, regional centres have the highest average number of services for all three services. In the case of towns, they have between two and three more schools (Figure 2.13), about one more pharmacy (Figure 2.14), and almost two more banks (Figure 2.15) than settlements with access to an urban centre. For towns, settlements without access to an urban centre fall clearly between the other two categories. Villages also have the most service locations when they are regional centres. The absolute difference is smaller than for towns, in line with population size being a good predictor of service locations (Figure 2.12). Villages with no access to an urban centre also have more service locations than villages with access to an urban centre, but the difference is small.
Korea shows a positive regional centre effect for towns but not for villages. Regional centre towns have, on average, 0.3 more schools (Figure 2.13), 1.9 more pharmacies (Figure 2.14), and 0.8 more banks (Figure 2.15) than towns with access to an urban centre. In contrast, villages show a large negative regional centre effect for schools, with regional centre villages having almost two fewer schools than villages with access to an urban centre. The regional centre effect is also slightly negative for pharmacies and essentially zero for banks.
Towns and villages in Korea with no access to an urban centre have equally many or fewer service locations than their counterparts with access to an urban centre. In contrast to the OECD average, the corresponding marginal effects are negative for schools (Figure 2.13). For settlements that are not regional centres, having access to an urban centre does not affect the provision of pharmacies in Korean towns or villages (Figure 2.14). For banks, there is again a negative effect on towns and no effect on villages (Figure 2.15).
Korean settlements without access to an urban centre are less remote than their counterparts in other countries, which can explain why they do not have more services than settlements closer to urban centres: 82% of Korean towns and villages without access to an urban centre in 30 minutes driving time can reach an urban centre within 60 minutes. The corresponding cross-country average is only 60%. The difficulty to consume a service in the next urban centre increases in the time it takes to reach it and Korea does not have many places that are very far away from the next urban centre.
Merely comparing the average number of service locations by settlement type is prone to omitted variable bias, given how relevant population size is for service provision (see Figure 2.12 and Table 2.1). A regression framework that controls for population size can help mitigate that problem. The analysis follows OECD (2024[1]) in using a negative binomial regression to assess the number of service locations for different types of settlements. It separately considers schools, pharmacies, and banks. In the case of schools, the counts are the number of schools in a settlement, and the zeros correspond to settlements without a school.
The estimation includes population and population squared as covariates and an interaction term between population size and being a regional centre. The quadratic term controls for a nonlinear effect of settlement size. The interaction term allows the effect of being a regional centre to vary by settlement size. The full regression equation is:
where is the number of schools, pharmacies, or banks, stands for the natural logarithm and having access to an urban centre and being a regional centre are two indicator variables. Every settlement that is a regional centre does not have access to an urban centre by construction.
The analysis predicts the number of service locations in each of the following six settlements for every country and service:
1. A town with 10 000 inhabitants that has access to an urban centre within 30 minutes driving time and is therefore not a regional centre.
2. A town with 10 000 inhabitants that is a regional centre and has therefore no access to an urban centre within 30 minutes driving time.
3. A town with 10 000 inhabitants that has no access to an urban centre within 30 minutes driving time and is not a regional centre.
4. A village with 1 000 inhabitants that has access to an urban centre within 30 minutes driving time and is therefore not a regional centre.
5. A village with 1 000 inhabitants that is a regional centre and has therefore no access to an urban centre within 30 minutes driving time.
6. A village with 1 000 inhabitants that has no access to an urban centre within 30 minutes driving time and is not a regional centre.
The analysis first considers each country separately.11 Some countries might have systematically fewer but bigger service locations, while others might offer the same service in many small locations. If these differences coincide with the distribution of settlement types, including all countries in the same regression can yield results that are hard to interpret.
Service count predictions of standardised settlements allow the computation of internationally comparable marginal effects, answering questions like: “How many more (or fewer) schools does a regional centre town with 10 000 inhabitants have, compared to a town of equal size that has access to an urban centre”? OECD (2024[1]) explains the methodology in more detail.
Source: Based on service data specified in Annex Table 2.A.2 and Annex Table 2.A.3. The settlement delineations and typology rely on population data specified in Annex Table 2.A.1 and Mapbox (2024[3]), Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Source: Based on service data specified in Annex Table 2.A.2 and Annex Table 2.A.3. The settlement delineations and typology rely on population data specified in Annex Table 2.A.1 and Mapbox (2024[3]), Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Source: Based on service data specified in Annex Table 2.A.2 and Annex Table 2.A.3. The settlement delineations and typology rely on population data specified in Annex Table 2.A.1 and Mapbox (2024[3]), Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Regional hubs can act as a service centre for their inhabitants and the surrounding area's population. Their identification can guide the design of policies that follow a compact-and-connected development. For instance, they can be considered in policies that aim at concentrating services in areas facing depopulation while maintaining an adequate level of access for all citizens, including children and seniors, as part of the Comprehensive National Territorial Plan discussed in the previous chapter.
For the purpose of identification, regional hubs must fulfil two criteria: 1) they host many services, particularly services of many different types; and 2) they act as a centre for their surroundings and are not themselves close to an even larger settlement.12 This section develops a methodology to identify regional hubs from population data, driving times, and point of interest (POI) data on service locations in Korea and showcases the resulting distribution of regional hubs.
Identifying regional hubs from the presence of services and connectivity to other settlements
The analysis uses rich POI data for Korea that allows for detailed differentiation of service types (though a twelve-digit identification code). The methodology in this section is flexible enough to allow for adjusted mappings, which can result in a slightly different set of regional hubs. All POIs are geocoded, which allows assessing whether a service type exists in a settlement and counting how often it exists.
The methodology involves four steps:
1. Defining education, healthcare, commercial, leisure, and general hubs. For example, education hubs include all settlements that host at least six of the following seven service types: childcare, an elementary school, a high school, a kindergarten, a language school, a middle school, and a university. Equivalent definitions apply to the other four kinds of hubs.
2. Defining service hubs as settlements that are education, healthcare, commercial, leisure, and general hubs simultaneously. In other words, a settlement needs to host a variety of services to classify as a service hub, both within kinds of services and across them.
3. Identifying regional centres that are the largest settlements within 30 minutes of car driving time. This step relies on criteria of size and proximity rather than the presence of services in a settlement. Incorporating such a definition can provide insights into a settlement’s role within the broader settlement network. For instance, consider two large towns of similar size. If one is the largest settlement in its area and hosts numerous services, it is likely to attract people from surrounding areas seeking services. It may also serve as an employment hub, particularly for service-related jobs. In contrast, if the other town is located near Seoul, a different dynamic is probable. While it may still offer services to its residents, people in the surrounding area are more likely to travel directly to Seoul for their needs.
4. Classifying settlements as regional hubs when they are both service hubs and regional centres.
Regional hubs are spread across the country and specialise in different services
Table 2.5 shows the service types included in each kind of hub, the rule to determine whether a settlement is a hub, and the number of hubs of each kind in Korea. Between 20% and 28% of settlements qualify for each kind of hub, except healthcare hubs, where this number is 38%.
Note: In this context, “/” implies that a settlement must offer either the former, or the latter, or both to classify as having the service type.
Source: Based on Point of Interest Data provided by the National Geographic Information Institute of Korea.
The frequency of the different service types varies considerably, both in the extensive margin (whether a settlement has at least one service of a given type) and in the intensive margin (how many services of a given type a settlement has). For example, cafes / teahouses, restaurants, and hairdressers are relatively ubiquitous, with many settlements having more than 20 corresponding service locations. On the other hand, universities, old-age residencies, and banks are relatively rare. There are also service types of which many settlements have one or two, like post offices and kindergartens. Annex Table 2.B.2 shows the median number of service locations by service type for urban centres, towns and villages, respectively.
The different hubs spread across the country, mainly following the population distribution. Many settlements are hubs for different kinds of services or not a hub in any of the five kinds. However, some settlements are hubs in one or two service kinds, while their inhabitants will often travel to another urban centre or town to consume services of another kind.
Figure 2.16 shows the distribution of service hubs. Almost all (77 out of 80) Korean urban centres with more than 50 000 inhabitants are service hubs. This includes many urban centres near Seoul or in the population cluster in the southeast. 47 out of 309 towns (5 000 to 49 999 inhabitants) are also service hubs, while no village (500 to 4 999) classifies as a service hub. However, some villages meet the criteria to be a hub for one or several kinds of services. Figure 2.17 displays the geographic distribution of the 71 regional centres.
Figure 2.18 shows the 37 settlements that meet the regional hub criterion. Regional hubs include the largest urban centres in Korea. However, they also include some towns that are much smaller in absolute size but that likely play an essential role as a centre for their less densely populated surroundings. Annex Table 2.B.3 lists all regional hubs and the TL3 regions where their centre is located. Note that large urban centres can span multiple TL3 regions. This is particularly true for the urban centre of Seoul, which has substantial parts in the TL3 regions of Seoul, Incheon, and Gyeonggi.
Note: Service hubs are settlements that are all the following: an education hub, a healthcare hub, a commercial hub, a leisure hub, and a general hub.
Source: Based on Point of Interest Data provided by the National Geographic Information Institute of Korea. The settlement delineations rely on population data specified in Annex Table 2.A.1.
Note: Regional centres are the biggest settlements within 30 minutes driving time.
Source: Based on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
Note: Regional hubs are all those settlements that are both a regional centre and a service hub.
Source: Based on Point of Interest Data provided by the National Geographic Information Institute of Korea, population data specified in Annex Table 2.A.1, and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
This section assesses the presence of inter-regional transport infrastructure and the accessibility to essential services and railway stations within regional hubs. The analysis focuses on children and older individuals, in line with recent work by the OECD on making cities work for all ages. For transport infrastructure, it considers the existing and planned railway stations. For essential services, it considers three essential services that may be frequently accessed by those groups: childcare, doctors and pharmacies (see Box 2.5 for an explanation of how childcare and health services are organised in Korea).
Early childhood education / childcare
Early childhood education is not compulsory in Korea, but it has become increasingly universal. The enrolment rate for early childhood education, children aged 3 to 5, was 94.0% in 2020, above the OECD average of 87.4%. Early childhood education is operated through a dual system of nurseries serving children aged 0-5 years old and kindergartens serving children aged 3-5 years old. These entities operate separately and are subject to different administration and financial laws and regulations. The government revised related laws of childcare centres and kindergartens and unified the ministries in charge of kindergartens and childcare centres as the Ministry of Education from 2024. The Ministry of Education will draw up an action plan and implement the unified early childhood education and childcare system from 2026.
Kindergartens can be government/public, corporate or individual. The Housing construction standards regulations require the presence of a kindergarten in any housing complex for more than 2 000 households. The head of the regional education office regulates the opening of facilities, and families with children are free to choose any kindergarten that works best for their situation.
The location of the nurseries, in turn, is subject to two rules. First, the Infant care act requires the location of public nurseries in vulnerable regions such as urban low-income residential areas and rural farming and fishing villages. Second, the housing construction standards regulations require the presence of a nursery when developing an apartment complex for more than 300 households. The mayor of the municipality controls facility openings, and parents can choose any nursery based on their own needs.
Healthcare
In Korea, separation of prescribing and dispensing has been in effect since 2000. Dispensing separation is a system in which doctors are responsible for treatment and prescription, and pharmacists are responsible for drug preparation.
There is no regulation where to locate a pharmacy or doctor’s office. However, pharmacists need to obtain permission from the mayor of the municipality. Any doctor who intends to open a doctor’s office should report to the mayor of the municipality.
In general, there’s no specific restriction for the location of hospitals either. However, any doctor who intends to open a hospital should get permission from the governor or mayor of the regional government, who shall not grant permission if the planned hospital does not comply with the basic policy for bed supply and the regional bed supply plan.
Source: Answers to an OECD questionnaire from the Ministry of Health and Welfare and the Ministry of Education of Korea.
Accessibility measures
Accessibility measures consider both travel costs and opportunities. Travel costs depend on the spatial distribution of people and service points, as well as the transport network connecting them.
This analysis focuses on two key indicators:
1. Travel time to the nearest facility – measuring how long it takes to reach the closest facility offering a given type of essential service.
2. Population share with access to at least one facility – assessing the proportion of residents with access to a service point within a given travel time.
The accessibility indicator estimates walking time from each 1 km × 1 km grid cell to the nearest service point (i.e. a railway station or a service facility) using the r5r package.13 The unit of analysis is the regional hub, defined above and shown in Figure 2.18. To minimize border effects and account for the immediate surroundings of the settlement, the study areas extend 5 km beyond the boundaries of the urban centres or towns classified as regional hubs.
The analysis includes service points both within and beyond the settlement borders, ensuring a comprehensive understanding of access patterns. The comparison across regional hubs focuses on walking times to highlight inequalities faced by individuals without car access and those with limited mobility.
The transport network data is sourced from OpenStreetMap (OSM). While OSM assumes streets are fully walkable, real-world conditions may present accessibility challenges—especially for individuals with reduced mobility, such as older adults and caregivers pushing strollers. Although future research could incorporate adjustments for these constraints, this analysis represents travel times under optimal street network conditions.
Figure 2.19 illustrates an example of the travel time indicator for Suncheon. It depicts the underlying street network, pharmacy locations, travel time values for each grid cell, and the boundaries of the settlement, including the 5 km buffer.
Source: Based on data from the National Geographic Information Institute of Korea.
To evaluate the potential demand for essential services, this analysis considers three population groups within each grid cell:
Total population
Older adults (65+ years)
Young children (0–5 years old)
For the same regional hub, Figure 2.20 presents the population density at the grid cell level for both the total population and the 65+ age group. A comparison between these maps highlights differences in the spatial distribution of older adults compared to the total population. Since this analysis assumes constant travel times across groups, any disparities in accessibility between older adults and the general population stem from variations in their geographic distribution within the study area rather than differences in transport infrastructure or network efficiency.
Note: Grid cells correspond to an area of 1 km by 1km.
Source: Based on data from the National Geographic Information Institute of Korea.
To obtain the population share with access to at least one service point, the analysis sums all grid cells within the study area that have walking access from the cells’ centroid to the nearest facility in 15 minutes or less. The threshold of 15 minutes is chosen considering a realistic travel time to a daily service by walking.
Railway connectivity is not fully developed across regional hubs
The distribution of railway stations in Korea reveals significant variation in accessibility across different regional hub sizes. Larger regional hubs, particularly those with populations over 1.5 million, are well-served by express trains, ensuring high-speed and long-distance connectivity (Figure 2.21). In contrast, many smaller regional hubs, especially those with fewer than 50 000 people, lack railway stations entirely, making them more reliant on motorised transport. The western and southern regions have a higher density of train stations, including express services, while some eastern coastal areas still await expansion. Overall, most settlements benefit from railway connectivity, enhancing regional mobility and economic integration. However, gaps remain, as some regional hubs still lack railway access, potentially limiting their economic development. The presence of new stations under construction suggests ongoing efforts to improve the transportation infrastructure and address these accessibility disparities.
Note: This map shows how each of the 37 regional hubs is connected to the train network. Whenever a settlement has at least one express station within its borders or five-kilometre buffer, it is classified as “express station”. An exception was made for Ulsan, where the express station is more than five kilometres away from the urban centre. When there is no express station but at least one normal station, a settlement is classified as “normal station”.
Source: Based on data from the National Geographic Information Institute of Korea.
Larger regional hubs generally have a higher share of their population with access to railway services within a 15-minute walk. In contrast, smaller regional hubs, either lack railway stations altogether or have a small share of the population that can access those stations by foot. Many midsized regional hubs show moderate accessibility levels, with 10–30% of their population benefiting from railway connections. These disparities highlight the need to evaluate multi-modal transport in settlements of all sizes, to ensure people have easy access to inter-regional transport connections and do not over-rely on private motorised transport modes.
Note: This map is based on train network connections for the 37 regional hubs. Whenever a settlement has at least one express station within its borders or five-kilometre buffer, it is classified as “express station”. An exception was made for Ulsan, where the express station is more than five kilometres away from the urban centre. When there is no express station but at least one normal station, a settlement is classified as “normal station.
Source: Based on Open Street Maps and Point of Interest Data provided by the National Geographic Information Institute of Korea.
While most people can easily get to services in regional hubs, smaller regional hubs unsurprisingly offer lower accessibility levels than larger regional hubs
In an ideal scenario, everybody would be able to get to at least one essential service point within a 15- minute walk in regional hubs. In Korea, the share of people that have this level of access is relatively high, with variations across regional hub sizes. By the 15-minute mark, 75-100% of residents in all regional hub size bins tend to have access to childcare, pharmacies and doctors.
Unsurprisingly, larger regional hubs have better accessibility for all three services. The largest regional hubs (> 1.5 million population, dark blue) have the highest cumulative share at any given travel time for doctors and pharmacies, meaning a greater percentage of the population reaches these services faster. They also offer above average access to childcare at any given travel time. Conversely, the smallest regional hubs (< 50,000, very light blue) consistently show lower cumulative shares, indicating more limited access. A reason for this difference may be a larger share of rural areas within the 5 km buffer around smaller regional hubs compared to larger ones. Access to a doctor shows the greatest gap between small and large regional hubs, with residents of smaller regional hubs taking significantly longer to reach a doctor.
Note: Travel time by urban cluster size based on walking time from each grid cell within a settlement and a 5 km buffer around it and the nearest service point (inside or outside the 5 km buffer).
Source: Based on Open Street Maps and Point of Interest Data provided by the National Geographic Information Institute of Korea.
While children’s accessibility to early childhood services is better than for the overall population, some regional hubs lag behind
Childcare services appear to be more prevalent in areas with young children, as seen in the lower access gap for this group compared to the general population (Figure 2.24). This is to be expected, given that people choose their home location accordingly when they have or expect small children, and at the same time, facilities (public and private) open in locations close to where families with children concentrate. Substantially higher accessibility levels for the entire population compared to children, like in the cases of Hongseong and Wonju suggest that some regional hubs face a spatial mismatch between the supply and demand for childcare services. These cases deserve closer examination to evaluate if the geographical distribution of early childhood facilities corresponds with the geographical distribution of demand for these services.
Comparing the performance of regional hubs across size ranges reveals that the worse performing regional hubs in terms of walking accessibility to childcare are all below 50 000 inhabitants. The smallest regional hubs also show the largest variation between each other. In 4 out of 10 regional hubs with less than 50 000 inhabitants, at least 20% of children are unable to reach a childcare facility within a 15-minute walk. In Hongseong, the regional hub with the lowest accessibility, more than half of the children cannot access a childcare facility within a 15-minute walk. At the other end of the spectrum, Donghae and Haenam rank amongst the best regional hubs in Korea in terms of walking access to childcare. This is remarkable given their size and their much lower accessibility levels for the entire population.
Children in most regional hubs with 50 000 – 250 000 people have better access to childcare than in larger regional hubs, including Seoul, Busan, and Daegu. They also show significant differences in the level of access for children compared to the entire population in all cases. The case of Boryeong illustrate the importance of considering differences in the residence patterns of families with young children compared to the total population. The indicator shows that 24% of people (regardless of age) do not have access to a childcare facility, but the figure for children is only 5%. The differences between the entire population and children largely disappear for midsize and large regional hubs, suggesting that accounting for the location of families with young children is particularly important in smaller regional hubs.
Finally, while large regional hubs have relatively good access to childcare, parents may face more competition for existing facilities rather than the lack physical access to a facility. This could be captured in future work with an indicator that accounts for the number of places per capita.
Note: Travel time by urban cluster size based on walking time from each grid cell within a settlement and a 5 km buffer around it and the nearest service point (inside or outside the 5 km buffer).
Source: Based on Open Street Maps and Point of Interest Data provided by the National Geographic Information Institute of Korea.
Older individuals have worse access to pharmacies and doctors compared to the total population, especially in smaller regional hubs
Walking access to healthcare services for older individuals can improve medication adherence and reduce complications from delayed treatment. It can also help reduce reliance on caregivers, lower transportation costs, ensure quicker emergency response and facilitate regular preventive care. However, unlike childcare services that are typically located near young families, pharmacies and doctors serve people of all ages, meaning their locations are not necessarily aligned with where older adults live. This can create accessibility challenges, particularly for seniors who rely on walking access due to mobility restrictions—such as the inability to drive or difficulties using public transport that is not adapted for their needs.
A comparison of 15-minute walking access to pharmacies (Figure 2.25) and doctors (Figure 2.26) across regional hubs reveals that older adults in regional hubs with fewer than 1.5 million residents generally have lower access than the overall population. As outlined in the methodology, this disparity is largely driven by differences in residential patterns between older adults and other age groups. The mismatch between the location of healthcare services and the areas where older adults live is particularly pronounced in regional hubs with fewer than 250 000 residents. In Haenam, Jindo, Boryeong, Geochang, Seosan, Jangheung, and Geongeup, the share of older adults without walking access to a pharmacy is at least 10 percentage points higher than for the general population. In Haenam, a regional hub with fewer than 50 000 residents, 33% of the total population lacks walking access to a pharmacy, but this figure rises to 48% for older adults.
Regional hubs with the worst pharmacy access (all under 50 000 inhabitants), including Taebaek, Jindo, and Jangheung, also have the poorest access to doctors, suggesting a pattern of co-location between these services that further exacerbates accessibility issues for older adults.
In contrast, large regional hubs show relatively low shares of older adults without walking access to pharmacies or doctors, with minimal differences between age groups. This high level of local accessibility is especially critical for older individuals in large urban areas, where traffic congestion and high transportation costs can further limit mobility. However, this analysis does not account for the quality of pedestrian infrastructure, which can significantly impact the actual walkability of healthcare services for older adults. Addressing these accessibility gaps, particularly in smaller regional hubs, is crucial to ensuring equitable healthcare access for aging populations.
Note: Travel time by urban cluster size based on walking time from each grid cell within a settlement and a 5 km buffer around it and the nearest service point (inside or outside the 5 km buffer).
Source: Based on Open Street Maps and Point of Interest Data provided by the National Geographic Information Institute of Korea.
Note: Travel time by urban cluster size based on walking time from each grid cell within a settlement and a 5 km buffer around it and the nearest service point (inside or outside the 5 km buffer).
Source: Based on Open Street Maps and Point of Interest Data provided by the National Geographic Information Institute of Korea.
This chapter analysed population concentration and service provision in Korea against other OECD countries. While the country has experienced faster urban population growth than the OECD average in the past, recent trends indicate that Korea’s urban centres have followed similar growth patterns to other OECD countries. The distribution of urban centres aligns with Korea’s size and level of urbanisation, even if Seoul stands out due to its extraordinarily large population. At such large sizes, further increases of population might become increasingly costly and difficult to manage.
Korea can strive for rebalancing urban development along its rich network of urban centres by boosting connectivity and accessibility to services for people of all ages. Nevertheless, population growth has been more prominent in smaller urban centres near major cities like Busan and Ulsan, while many urban settlements beyond the main clusters have experienced population decline. Projections indicate that this concentration will persist, with villages and smaller towns expected to see the highest population losses in the future. Addressing inter-regional connectivity gaps is key to develop new axes of population growth in the country.
Accessibility to services also varies across urban and rural areas, with regional hubs playing a crucial role in providing essential services. While small settlements in Korea offer a broader range of services than their counterparts in other OECD countries, disparities remain, particularly in healthcare access, where smaller regional hubs show significantly longer walking times to reach doctors. Childcare services, however, appear to be more efficiently distributed based on local demand. Despite urban concentration trends, smaller towns and villages continue to serve as critical hubs, especially in less densely populated regions. Maintaining balanced regional development and ensuring equitable access to essential services will be key challenges as Korea navigates future demographic shifts.
References
[7] Chauvin, J. et al. (2017), “What is different about urbanization in rich and poor countries? Cities in Brazil, China, India and the United States”, Journal of Urban Economics, Vol. 98, pp. 17-49, https://doi.org/10.1016/j.jue.2016.05.003.
[4] Dijkstra, L., H. Poelman and P. Veneri (2019), “The EU-OECD definition of a functional urban area”, OECD Regional Development Working Papers, No. 2019/11, OECD Publishing, Paris, https://doi.org/10.1787/d58cb34d-en.
[8] Eurostat (2011; 2021), GEOSTAT Population Grid, https://ec.europa.eu/eurostat/web/gisco/geodata/population-distribution/geostat.
[6] Gabaix, X. and R. Ibragimov (2011), “Rank — 1/2: A Simple Way to Improve the OLS Estimation of Tail Exponents”, Journal of Business & Economic Statistics, Vol. 29/1, pp. 24–39, https://www.jstor.org/stable/25800776.
[3] Mapbox (2024), Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
[1] OECD (2024), Getting to Services in Towns and Villages: Preparing Regions for Demographic Change, OECD Rural Studies, OECD Publishing, Paris, https://doi.org/10.1787/df1e9b88-en.
[2] OECD et al. (2021), Applying the Degree of Urbanisation: A Methodological Manual to Define Cities, Towns and Rural Areas for International Comparisons, OECD Regional Development Studies, OECD Publishing, Paris/European Union, Brussels, https://doi.org/10.1787/4bc1c502-en.
[5] Schiavina, M., M. Melchiorri and M. Pesaresi (2023), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
The analysis relies on grid-level population counts and data on service provision. For both cases, the report attempts to use data that is as accurate and internationally comparable as possible. The data follows mostly OECD (2024[1]).
Population grids can follow top-down or bottom-up approaches. Top-down approaches start from population counts of small administrative units and disaggregate them to individual grid cells. An example is the GHS population grid that uses the share of built-up space in a grid cell for disaggregation. Bottom-up approaches use geocoded data points, which they aggregate into grid cells. Eurostat produces such a grid for Europe. Korea has bottom-up population grids for 2018 and 2023. Annex Table 2.A.1 shows the population grids used for
The delineation of the settlements according to the degree of urbanisation.
All comparisons of the current level of population concentration.
The typology of towns and villages by reachability.
The analysis of service provision.
Note: The built-up area 1 km2 grids are based on GHS-BUILT (2014), derived from satellite data.
Sources: GEOSTAT (Eurostat, 2011; 2021[8]) data and GHS population data from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Annex Table 2.A.2 provides brief definitions of the service types for the international comparison, along with their respective country coverage.
Note: No service data is available for Cyprus and Malta.
Annex Table 2.A.3 lists the data sources by country and service type. Australia and New Zealand have a variety of public data sources.14 For the European Union, two types of GIS databases (GISCO and ESPON) cover private and public sector establishments. In Canada, Statistics Canada maintains databases of healthcare, commercial and educational facilities. Data from Korea come from an official database on points of interest, provided in a confidential manner to the OECD from Korea’s Ministry of Land, Infrastructure and Transport. For the United States, most data come from the Homeland Infrastructure Foundation Level.
References
[8] Eurostat (2011; 2021), GEOSTAT Population Grid, https://ec.europa.eu/eurostat/web/gisco/geodata/population-distribution/geostat.
[5] Schiavina, M., M. Melchiorri and M. Pesaresi (2023), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Source: Based on population data specified in Annex Table 2.A.1.
Note: The population counts refer to the number of inhabitants in the urban centre as defined by the degree of urbanisation.
Source: Based on population data specified in Annex Table 2.A.1.
Note: The settlement delineations and typology rely on population data specified in Annex Table 2.A.1 and on Mapbox (2024[3]) Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Source: The settlement delineations rely on population data specified in Annex Table 2.A.1. Population grids come from Schiavina, Melchiorri and Pesaresi (2023[5]), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
Source: Based on Point of Interest Data provided by the National Geographic Information Institute of Korea. The settlement delineations rely on population data specified in Annex Table 2.A.1.
References
[3] Mapbox (2024), Mapbox Isochrone API, https://docs.mapbox.com/api/navigation/isochrone/.
[5] Schiavina, M., M. Melchiorri and M. Pesaresi (2023), GHS-SMOD R2023A - GHS settlement layers, application of the Degree of Urbanisation methodology (stage I) to GHS-POP R2023A and GHS-BUILT-S R2023A, multitemporal (1975-2030), https://doi.org/10.2905/A0DF7A6F-49DE-46EA-9BDE-563437A6E2BA.
← 1. Only those OECD member and accession countries with detailed data on the location of services (Annex Table 2.A.2) are included in the analysis.
← 2. Korea remains an outlier when considering only people in midsize or large urban centres (of at least 250 000 inhabitants) or the population in urban centres, towns, and villages combined. The gap is wide for midsize or large urban centres (Korea with 65% is followed by Canada with 45%) and narrower when considering all settlements, where Korea (89%), Spain (85%), and the Netherlands (81%) have the highest shares.
← 3. It includes the three countries within and the three countries outside Europe with the largest number of urban centres. France and Poland complete the list of countries with the largest number of urban centres. Their slope parameters are very similar to those of Germany and Italy (-1.08 for France, -1.10 for Poland).
← 4. Using a different city definition, Chauvin et al. (2017[7]) find a coefficient of -0.91 for the United States.
← 5. These population counts correspond to the population of the urban centres according to the degree of urbanisation. The population of the corresponding functional urban areas is typically even larger.
← 6. The population development analysis in this and the next section uses historic grid-level population data from the Global Human Settlement (GHS) project. GHS provides such data in five-year intervals, as well as projections for 2025 and 2030. While the delineation and classification of the Korean settlements still builds on Korean population grids, GHS offers a valuable addition, as it allows tracking time trends internationally. The analysis in this part considers settlement boundaries fixed to their current extent.
← 7. Other urban centres in Korea experienced much higher population growth than their counterparts in other OECD countries between 1990 and 2005 and grew at a similar rate between 2005 and 2020. This convergence in population concentration trends between Korea and the OECD average persists for smaller settlements. While a substantial share of villages is losing population in many countries, the average OECD country still experienced an increase in the population of villages. In contrast, Korea saw its population in villages decrease substantially. However, both developments slowed in 2005-2020 compared to 1990-2005. The differences are minor for towns.
← 8. Some settlements that have above 50 000 inhabitants with the national population grids have less than 50 000 inhabitants with the GHS data that this historical comparison builds on. This cross-country comparison excludes these settlements. Including them pushes Korea even further to the left, indicating a small difference between the population growth rate of Seoul and that of small urban centres.
← 9. The Pearson correlation coefficient between the population in 2005 and the compound population growth rate between 2005 and 2020 is -0.07.
← 10. The 26 countries with data on all five service types are Austria, Belgium, Bulgaria, Croatia, Czechia, Finland, France, Denmark, Germany, Greece, Hungary, Ireland, Italy, Korea, Lithuania, Luxembourg, Netherlands, Norway, Poland, Portugal, Romania, Slovak Republic, Slovenia, Spain, Sweden, and the United States.
← 11. The intensive-margin analysis excludes Luxembourg, as it only has one regional centre, making it impossible to estimate all parameters.
← 12. This analysis still relies on the same settlement definition described above but it focuses on Korea.
← 13. https://cloud.r-project.org/web/packages/r5r/vignettes/r5r.html.
← 14. In Australia, the Royal Flying Doctor Service and School of the Air provide important access to health and educational services in remote areas but are not counted as belonging to particular settlements.