The Geography of Choice

6. Conclusions

Abstract

This concluding chapter synthesises the report’s main findings and considers their implications for the use of functional geographies. It highlights the complementary strengths of the approaches examined and identifies key directions for further methodological development, including improving consistency between urban and rural delineations, testing the approaches in different national contexts and exploring their application to cross-border daily systems.

This report examined how rural territories can be organised into internationally comparable functional areas, addressing a gap between the OECD's two existing internationally comparable territorial typologies namely the TL3 regional typology, too broad to isolate rural conditions with precision, and the Degree of Urbanisation's grid-cell classification, too fine to correspond to any functionally coherent area.

Chapter 2 introduced the concept of "daily system," a self-contained territorial system within which people can access homes, jobs and services on a daily basis, often spanning several local administrative areas with their own distinct boundaries and policies. Local areas such as municipalities are a familiar statistical unit, but they are not always the most suitable one for capturing how people actually live. FUAs address this for cities, but disregard people living far from any urban centers, the gap this report sets out to fill. The chapter reviewed existing delineations against a set of criteria developed for daily systems with nationwide coverage, spanning rural as well as urban contexts:

• LMAs, built on actual, observed commuting trips between where people live and work.

• Bassins de Vie, built on theoretical trips, simulated on the road network, connecting homes to service locations.

• FRAs, built on theoretical trips, simulated on the road network, connecting homes to population centres.

Chapter 3 described work commuting flows in FRAs+FUAs using administrative data. Moreover, it compared the three delineation methods against a common set of criteria, using the same number of units to distinguish functional structure from artifacts of simple geographic clustering. It also benchmarked them against random spatial clustering to show whether they pick up interconnections beyond what would be expected from neighbouring local areas. The chapter found that:

• FRAs contain not only rural areas but also many towns and suburbs, while some rural areas are situated in FUAs.

• Most working residents work within the FRA where they live, although FRAs are generally less self-contained than FUAs. Some FRAs have strong commuting links to another functional area, typically to a FUA.

• LMAs and FRAs+FUAs substantially outperform their random benchmarks on most measures, including self-containment, balance and cohesion, indicating that both capture meaningful functional structure rather than arbitrary spatial patterns.

• Minimum self-containment is more nuanced and varied across countries. In France, a small number of individual FRA+FUA areas show self-containment no stronger than random clustering, though global self-containment remains considerably higher than the benchmark, showing these cases do not dominate the overall pattern.

• Homogeneity is the one dimension where neither delineation clearly outperforms random benchmarks, an expected result given that real settlement patterns and agglomeration economies naturally produce functional areas of uneven size.

• Bassins de Vie, operating at a considerably finer scale with many more units compared to LMAs and FRAs+FUAs, show lower self-containment and higher cohesion than LMAs and FRAs+FUAs in direct comparison, reflecting this difference in scale rather than weaker performance. Benchmarked instead against random groupings of the same size, Bassins de Vie also capture meaningful spatial structure.

• No method was found to be uniformly superior. The most useful choice depends on the policy question being addressed and the geographic scale at which the analysis needs to operate.

In Chapter 4, LMAs and FRAs+FUAs were tested using mobile phone data, drawn from either mobile network operator (cell-tower) signals or GPS-enabled applications, each with distinct trade-offs between spatial accuracy, population coverage and representativeness bias across three countries (Korea, Japan and Australia). The main findings of the chapter include:

• In Korea, work-related trips cross local-area borders within the same FRA considerably more than other trip types, indicating that mobility behaviour differs by trip purpose.

• In Japan, where the FRA+FUA configuration was calibrated to closely match the number of official LMAs, the two delineations perform comparably overall. FRAs+FUAs show somewhat higher self-containment, while LMAs show a somewhat higher global interaction index.

• In Australia, where unit counts were not perfectly matched, results again show broadly similar performance.

• This practical testing with mobile phone data reinforces the findings from Chapter 3, which are based on administrative commuting data, and shows that they also apply to trips beyond work commuting. FRA+FUA and LMA capture comparably meaningful functional structures where they can be fairly compared, with neither showing a decisive overall advantage.

Chapter 5 applied the novel FRA methodology, selected for its international comparability, to examine territorial trends across the OECD with the following results:

• Old-age dependency rises steadily with rurality, and foreign-born population share, a potential source of demographic renewal, declines along the same gradient. The areas facing the greatest demographic pressure are also the least able to draw on migration to offset it.

• Access to a general hospital within 45 minutes falls sharply with increasing rurality, and internet speeds show a comparable decline, limiting digital access as a substitute for physical services.

• Housing markets show a consistent gradient with rurality. Dwelling vacancies rise and rents fall as areas become more rural, with the two moving together.

• Per-capita greenhouse gas emissions are far higher in the most rural areas, largely reflecting where production activities are physically located. At the same time, more rural areas have seen far greater decline since 2010, concentrated in the power sector.

• Built-up areas have continued to expand in FRAs even where population has declined, indicating that the physical footprint requiring services and infrastructure is not shrinking in step with the population it serves.

These findings are broadly consistent with existing OECD research conducted at the TL3 level while offering considerably more granularity than that level of analysis allows.

Taken together, the findings show that each of the three analysed approaches captures meaningful functional relationships, while offering different advantages depending on the purpose and scale of the analysis. This conclusion is supported by two complementary strands of evidence. First, Chapter 3 shows that LMAs, Bassins de Vie and FRAs combined with FUAs generally outperform random spatial benchmarks with comparable numbers of units. All three therefore identify functional structures that extend beyond the simple grouping of neighbouring areas, although their relative performance varies across criteria. The distinction between observed commuting trips (LMA) and simulated, theoretical trips (Bassins de Vie and FRA) did not determine which method performed better in practice. Second, Chapter 4 uses mobile phone data to assess mobility beyond commuting. The Korean analysis shows that trip purpose influences how movements are organised within functional areas, while the comparisons in Japan and Australia find broadly similar performance for LMAs and FRAs combined with FUAs. These results reinforce the conclusion that different delineations can provide useful perspectives on daily systems, provided that they could be fairly compared on matched unit counts (i.e. that each approach has the same number of areas).

The FRA delineation applied in Chapter 5 was selected for practical reasons specific to that exercise, not as an implicit ranking of the three approaches. Its application across OECD countries reveals a consistent set of patterns highlighted above.

This publication is intended to complement, not replace, the wide range of national definitions of rural areas already in use, which remain the most appropriate tool for national policy design. Where internationally comparable functional areas are used to inform policy, their relevance and application should be assessed in light of national territorial contexts and available data. Further work, through the ongoing WPRUR and WPTI expert group process, will continue to test and refine functional approaches in collaboration with member countries, including their potential application to cross-border daily systems. Future methodological development could also examine how to improve consistency between functional delineations across urban and rural territories. This process can strengthen the internationally comparable evidence base for understanding rural diversity and support policies that better reflect how people live, work and access services across territories.