Is Pennine England becoming more polycentric An analysis

Is Pennine England becoming more polycentric? An analysis of commuting flows in a transforming industrial region, 1981 -2001 Tony Champion and Mike Coombes Acknowledgments: Colin Wymer (CURDS) Oliver Duke-Williams (Leeds University & CIDER) Data drawn from Special Workplace Statistics of 1981/1991/2001 Censuses (Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland) Paper for IGU Urban Commission meeting, Canterbury, 14 -20 August 2010

Structure of presentation • • Polycentricity and Pennine England Testing for urban system polycentricity: review Testing for growing polycentricity: our approach Test 1: Simple gravity model Test 2: Expanded gravity model Summary of results Next?

Polycentricity and Pennine England 1 • The world’s first major urban-industrial region, still the economic heartland of Northern England home to 10 million people (and over 4 million workers) • Developed as a many-centred region based on the geographical division of labour based on local specialisms within sectors like textiles • A true ‘conurbation’ because it resulted from the growing together of smaller urban areas into a larger one (unlike the London region, with its essentially monocentric evolution) • Recognised by map in the 1851 General Census Report and observed by Mackinder (1902), Geddes (1915) and Fawcett (1922, 1932), e. g. “the second great urban region of Britain, the area within a 50 -mile radius from Manchester Town Hall” (Fawcett, 1932, p. 109) • But whose overall population had hardly grown 1921 -1931, again unlike London (ibid. )

Polycentricity and Pennine England 2 • By the 1970 s it was seen to include [a] four separable conurbations (which became Metropolitan Counties of Gtr. Manchester, Merseyside, South Yorkshire, West Yorkshire), as well as [b] a more polycentric Central/NE Lancashire area around Preston (see map 2) • In the 2000 s Pennine England became (temporarily!) part of a huge ‘Northern Way’ planning region (see map 1), seeking regional balance versus the London region • Close spacing of urban areas ‘overtaken’ by long-term increases in average commuting trip lengths, partly due to the growing share of the labour force who can afford longer -distance commuting • More recent growth in jobs (in knowledge economy etc) seen to be mainly city-centred: has this reversed any trend towards polycentricity by reviving centripetal flows?

Map 1: The Northern Way planning region and the 5 Pennine England City Regions within it

Map 2: The 5 Pennine England Cities and their City Regions Cities are defined as Core City local authority areas. City Regions are identified by meta-analysis of other definitions.

Testing for urban system polycentricity: review • Dearth of empirical studies (Burger, 2011) and these mostly based on pattern (morphology, degree of local specialisation) rather than process (spatial integration) • Some process-oriented studies examine non-people flows (trade, interfirm links), most have been based on people flows, notably commuting (due to data availability) • Most commuting-based tests for polycentricity have been cross-sectional, but some compare 2 time points (e. g. Clark & Kuijpers-Linde, 1994; Burger et al, 2011) • De Goei et al (2010) test for growing polycentricity at both intra-urban and inter-urban levels, but with two separate sets of models • Present study uses commuting data from 3 censuses to test for trends in polycentricity for Pennine England at intra-urban and inter-urban levels simultaneously

Testing for growing polycentricity: our approach • Commuting data for 1981, 1991 and 2001 from the local authority level SWS (adjusted to 2001 boundaries and allowing for the 10% sample basis of 1981 and 1991) • For all people both living and working in Pennine England (i. e. excluding commuters across the PE boundary but covering all ages, genders, occupations, transport modes) • Gravity model of the standard form Tij=a*Oi*Dj*dij, where Tij is the predicted number of trips between zones i and j Oi is the total number of trips originating in zone i Dj is the total number of trips terminating in zone j dij is a measure of distance between zones i and j. • Plus 2 dummy variables for testing Within-CR (dum 1) and Between-CR (dum 2) polycentricity [see later] • Logged for estimating parameters from each 49 X 49 matrix (N=2, 401), after adding 1 to all cells to eliminate zero counts: log(Tij)=a*log(Oi)+b*log(Dj)+c*log(dij)+e 1*log(dum 1)+e 2*log(dum 2) +log(f) [the intercept]

Test 1: Simple gravity model Main aims of this test: (1) Is the region behaving increasingly like a single urban system? Based on closeness of R 2 to 1. 0. (2) Is the region becoming more spatially integrated? Based on trend in the distance deterrence (DD) parameter. Results of the 3 models: Year 1981 1991 2001 adjusted R 2 0. 780 0. 795 0. 857 -3. 538 -3. 476 -3. 304 origin zone size 0. 499 0. 583 0. 674 destination zone size 0. 865 0. 939 0. 961 intercept 11. 19 10. 35 9. 22 distance deterrence Answers: (1) Yes, R 2 rising ; (2) Yes, DD reducing – but …

Towards Test 2 Test 1 merely indicates a tendency towards longer commuting distances, but need not mean greater polycentricity – could be due to more longer trips into one or more large cities So, Test 2 includes two measures designed to capture two spatial dimensions specifically related to polycentricity (beyond any effect of change in distance deterrence): 1) Intra-urban: Within CRs, have the cities lost dominance as attractors of commuters from their own hinterlands? Dum 1=1 for flows to city from other zones in same CR. Expect parameters to be positive (main city attracts more than other zones) but reducing over time. 2) Inter-urban: Between CRs, has there been an increase in commuting across the boundaries between the 5 CRs? Dum 2=1 for flows that cross a CR boundary. Expect parameters to be negative (CR boundary between 2 zones reduces flow), but reducing over time.

Test 2: Expanded gravity model Results of the 3 expanded models: Year 1981 1991 2001 adjusted R 2 0. 797 0. 807 0. 863 -3. 051 -3. 080 -3. 019 0. 325 0. 251 0. 152 -0. 465 -0. 379 -0. 275 origin zone size 0. 549 0. 624 0. 702 destination zone size 0. 875 0. 948 0. 971 8. 98 8. 55 7. 92 distance deterrence dum 1: to city (of same city region) dum 2: between city region intercept Answers: (1) Dum 1: Yes, ceteris paribus, the city exerts an attractive pull, but this has fallen progressively over time (2) Dum 2: Yes, the CR boundary acts as a deterrent, but its effect has fallen progressively over time too NB: the DD parameter no longer falls progressively

Summary of results • Commuting becoming more uniform across Pennine England in its response to Mass and Distance (rising R 2) • General increase in commuting distance for people living and working in region (falling distance deterrence) • Key elements in this are those relating to polycentricity: - reducing attractive power of the 5 cities within their CRs - reducing deterrence effect of the CR boundaries • Allowing for these 2 elements, the distance deterrence parameter fell much less than in the basic gravity model • Further analysis of the SWS datasets indicates: - little deconcentration of jobs across the region (cf. pop), nor tendency towards one centre dominating the region - increase in % flows between non-city zones as well as in between-CR flows and fall in share of intra-zone flows

Next? • So far, this has been an essentially exploratory study aimed at validating our approach and specific polycentricity metrics • Is it worth experimenting with alternative forms of gravity modelling (constrained, Poisson, velocity rather than Tij)? • Possibility of modelling labour force sectors separately, e. g. male/female, high/low-skill – using each’s Mass counts? • Scope for updating analyses using the Annual Population Survey (but small sample size), then 2011 Census • Use a finer-grained zone geography than local authority area that better delineated employment centres? • Or is employment growth not restricted to a small number of centres but more dispersed in a form of ‘a-centric’ growth? • Apply our approach to another case study area, e. g. the more dynamic (and traditionally monocentric) London region

Is Pennine England becoming more polycentric? An analysis of commuting flows in a transforming industrial region, 1981 -2001 Tony Champion and Mike Coombes tony. champion@ncl. ac. uk mike. coombes@ncl. ac. uk Acknowledgments: Colin Wymer (CURDS) Oliver Duke-Williams (Leeds University & CIDER) Data drawn from Special Workplace Statistics of 1981/1991/2001 Censuses (Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland)

Annex: The 5 City Regions and their constituent zones

Hoover Index of the distribution of workforce and jobs across the 49 zones of Pennine England date 1981 1991 2001 Workforce : area 37. 84 36. 47 36. 41 Jobs : area 39. 87 38. 81 38. 83 7. 19 7. 80 7. 95 Workforce : jobs Workforce (i. e. at home) became more dispersed, but mainly in 1981 -1991 Jobs (i. e. at workplace) became more dispersed in 1981 -1991 only The distribution of workforce and jobs became more different, but mainly in 1981 -1991

Background data on change in commuting flows 1981 Between-CR commuters 1991 2001 -1981 121, 150 158, 610 222, 141 +100, 991 3. 0 4. 1 5. 4 +2. 4 12. 5 15. 0 17. 6 +5. 1 4, 058, 490 3, 881, 940 4, 084, 747 +26, 257 970, 960 1, 056, 760 1, 263, 937 +292, 977 Between-CR as % of: All workers Inter-zonal commuters Inter-zonal % all workers 23. 9 27. 2 30. 9 +7. 0 People commuting between the 5 City Regions are becoming: (1) a larger share of all those who live and work in Pennine England, and also (2) a larger share of those who commute between any of the 49 zone pairs

Change in the share of all Pennine commuting flows by type of flow (% point)

Changing composition of Pennine commuting flows (A) Intrazonal’s % of all; (B) Types’ % of interzonals

Inflow to City from Rest of City Region as % all betweenzone flows within City Region

City’s share of City Region’s total workers (counted at workplace), 1981 -2001