Population change and migration in Britain since 1981: evidence for continuing deconcentration

Transcription

1 Environment and Planning A 994, volume 26, pages Population change and migration in Britain since 98: evidence for continuing deconcentration A G Champion Department of Geography, University of Newcastle upon Tyne, Newcastle upon Tyne NE 7RU, England Received 0 March 993; in revised form 0 December 993 Abstract. The counterurbanisation decade of the 970s appears to have been followed by a period of more mixed trends in migration between metropolitan and nonmetropolitan areas. This author examines the experience of Great Britain against the background of developments reported for other countries. The British Census small-area statistics are used to calculate 98-9 rates of population change for a typology of local labour-market areas in order to test for the existence of population deconcentration, and the results are compared with the rates for the three previous intercensal decades. Annual population estimates are then used to examine the migration component of 98-9 population change and to investigate the extent and timing of fluctuations in growth rates since the early 960s. The results indicate that the differentials in the population growth rate between metropolitan and nonmetropolitan Britain narrowed somewhat between the 970s and the 980s, but the negative relationship between urban status and population change remained very clear. Moreover, contrary to the experience of the USA and a number of European countries, in the mid-980s Britain saw a resurgence of nonmetropolitan growth which had widespread impact across the country. These results raise questions which can in part be addressed by in-depth research on the 99 Census and related data sets. Introduction Population redistribution between metropolitan and nonmetropolitan areas has been generating much academic and policy interest for decades in Britain and elsewhere (Pacione, 984; Robinson, 990). At one time, the focus was largely on metropolitan growth and rural depopulation (Bracey, 970; Lawton, 973; Saville, 957; Zelinsky, 962). Since the early 970s, however, the centre of attention has switched to metro decline and the rural population turnaround (Brown and Wardwell, 980; Bryant et al, 982; Champion and Watkins, 99; Lewis and Maund, 976). The nature and significance of these developments have been discussed widely in the context of the so-called 'counterurbanisation debate' (see Champion, 989a, for a review). The developments of the past decade are not likely to lead to any diminution in this interest. Early observers of counterurbanisation and metropolitan migration reversal (Berry, 976; Vining and Strauss, 977) believed that these developments constituted a major turning point in the organisation of settlement systems, and Fielding (982) envisaged the 980s as seeing the completion of the switch from urbanisation to counterurbanisation in Western Europe. Since then, however, evidence of a revival of metropolitan growth, most notably for the USA (Frey, 993), raises many questions over whether this means a return to the processes of the 960s or even 950s, as well as prompting a reexamination of the migration patterns of the 970s. Also, not all countries appear to have experienced this latest reversal (for instance, see Champion, 992a). The purpose in this paper is to investigate the recent experience of Britain in the light of trends revealed by recent research on other countries, drawing on the results of the latest Population Census and other sources. The paper begins with a

2 502 A G Champion review of population redistribution trends in the USA, Australia, and Europe, highlighting the main differences between the 980s and the previous decade. Against this background, I then go on to examine evidence relating to trends in population change in Britain, using two geographical frameworks which permit a distinction between the more metropolitan and more rural parts of the country. Primary attention is given to the changes which have taken place over the ten-year period 98-9, but these recent changes are also set in the context of the evolving patterns provided by Census data from 95 onwards, and annual estimates are used to identify trends within as well as across intercensal periods. 2 The international context of metro/nonmetro population change The evidence for a reversal of the 970s rural population turnaround is particularly impressive for the United States. According to Frey (993) the overall population growth for nonmetro America for was only 3.9%, well down from the 4.3% for and only barely above the 2.2% figure for Even more impressively, it was well below the rate for metro America and, within metro America, it was the largest metro areas that were recording the strongest growth: 2.% for these as opposed to 0.8% for other metros. Moreover, the reversal appears to have intensified during the course of the 980s. Frey's (993) figures indicate an overall nonmetro population growth of only 0.3% for compared with 3.6% for the first half of the decade. Part of the explanation for the downturn in the demographic resilience of America's nonmetro areas lies in the continuing convergence of fertility rates, eroding the nonmetros' traditional advantage of high birth rates (Fuguitt et al, 99). Migration, however, provides the main explanation behind the switch to lower growth rates. According to residual net migration analyses produced by Forstall (99) and shown in table, the intercensal period was marked by a substantial reduction since the previous decade in the gains made both by metro areas with under 0.5 million people and by all types of nonmetro counties except those that Table. Average annual population and migratory change for metropolitan and nonmetropolitan areas in the USA, (source: extracts from Forstall, 99, tables B and C). United States metropolitan nonmetropolitan Metropolitan (by population > 5 million -5 million 0.5- million million <0.25 million Nonmetropolitan (983 definition) became metro by 990 >5% commuting 0%-4% commuting 5%-9% commuting under 5% commuting Populatior i change (%) ) Residual net migration (%) Note: 990 definitions unless stated otherwise. Net migration is calculated by the residual method.

3 Population change and migration in Britain since were incorporated into metro America between 984 and 990 in all these cases, this was a bigger fall than the national shift towards a lower net migration balance between the two decades. The net migration rates for these settlement categories have not fallen back to the level of the 960s, but the picture is markedly different from the 970s. In particular, note the return to net migration loss of the nonmetro category that is most remote from metro influence, namely counties with less than 5% of employed residents commuting to metro areas in 980. Annual migration estimates confirm the existence of overall net migration losses from nonmetropolitan America during the latter half of the 980s (DeAre, 992). As shown in table 2, the flows from nonmetro to metro areas were higher than flows in the opposite direction for each of the five years shown. Estimated levels of net loss to metropolitan areas were running at a particularly high level at the middecade, with continued immigration into nonmetro areas from outside the USA only partially offsetting the effect on population levels in nonmetro America. Table 2. Migration (in thousands) in nonmetropolitan America, (source: extracted from DeAre, 992, table C). Year Metro to nonmetro Nonmetro to metro Net flow to nonmetro Net immigration Total net migration The US evidence therefore suggests a return to seemingly more traditional urbanisation patterns after the 'counterurbanisation decade' of the 970s. Indeed, it is largely consistent with explanations which saw the 970s turnaround as part of a broader industrial restructuring of the US economy. This 'regional restructuring' perspective (see Frey, 987) forecast a return to population clustering, but not so much a general process as one focused on metropolitan areas containing specific economic niches such as high-level services and new industries a pattern encapsulated by the term 'back to the future' (Frey and Speare, 99). A similar conclusion has been drawn from analysis of the 980s experience in Australia, though in this case the original turnaround took on a spatial form rather different from that in the USA, and also the subsequent shift back from population dispersal has been less pronounced. As noted by Hugo (989), one of the most distinctive characteristics of the turnaround in Australia in the 970s was its spatial concentration in selected zones, notably the well-watered and environmentally attractive nonmetropohtan areas on the southeast and east coasts. These areas were found to be largely responsible for the major switch in population growth from metropolitan to nonmetropolitan Australia between and Since then, as a result of the five-yearly incidence of censuses in Australia, the most detailed research relates to the period According to Hugo and Smailes (992), the turnaround in Australia slowed down during the early 980s, but the nonmetropolitan population was still growing faster than that of the large cities an increase of 7.8% between 98 and 986 for the former and 6.7% for the latter, compared with rates of 9.5% and 6.3% for 976-8, respectively. Also noted was an intensification of the spatial concentration of the nonmetropolitan areas experiencing net migration gains, with the most favoured places being characterised by their proximity to large metropolitan areas, tourist potential, and/or scenic environment.

4 504 A G Champion Preliminary analysis of patterns, as described by Hugo (994), appears to produce somewhat more mixed results. On the one hand, the differential between metro and nonmetro growth rates widened again to around the level of the later 970s, with nonmetropolitan Australia's population growing by 0.4% over the five-year period, compared with the 6.6% increase for metropolitan areas. On the other hand, the degree of spatial concentration seems to have become even more pronounced, increasing further the degree of diversity to be seen across nonmetropolitan Australia. On balance, this second case leads Hugo to follow Frey and Speare (99) and opt for the 'back to the future' scenario in anticipating trends to the end of the century. The situation in Europe appears to be even more complicated, according to the evidence provided by two cross-national analyses. Using a core-periphery regional framework, Cochrane and Vining (988) found fairly consistent signs of renewed metropolitan attractiveness in the late 970s and early 980s in peripheral countries of Europe, notably Scandinavia, but much less uniformity in northwest Europe. Whereas a return to faster core growth was observed for the Netherlands and Denmark, the core region in West Germany continued to lose migrants and in France the rate of core population loss was accelerating. A more direct comparison with the evidence presented above on the USA and Australia is provided by an update of Fielding's (982) innovative work on Western Europe. This approach uses smaller regional units than those of Cochrane and Vining, and the situation in the late 980s is distinguished from that earlier in the decade. The specific method of analysis is very different, however, partly because of the virtual absence of officially defined metropolitan areas in Europe. Regions are distinguished according to their level of population density and this is correlated with net migration rate, in order to identify tendencies towards population concentration (a positive correlation between the two indicates 'urbanisation') or deconcentration (a negative relationship referred to as 'counterurbanisation'). The results of this work (table 3) suggest a very mixed picture. Comparison of the correlation coefficients for the 980s with those of the previous decade suggests that the countries are fairly evenly split between those which moved towards clearer counterurbanisation or less-clear urbanisation and those which experienced some degree of reversal towards the more traditional urbanisation pattern. Within the 980s, however, the trend was much more uniformly away from counterurbanisation, paralleling the US experience through the decade. On the other hand, even in the period since 984, almost half the countries were still displaying a counterurbanisation tendency. The question I will address in the remainder of this paper is the extent to which Britain's experience of population change in the past decade matches any of those just described for other countries. Previous research on concentration and deconcentration tendencies in Britain (see Champion, 989b) has shown that Britain was in the vanguard of the population turnaround, with the first stages of the rural population recovery being evident during the 950s and the metropolitan migration reversal occurring at the beginning of the 960s. At the same time, however, it was found that the pace of population deconcentration peaked in the first half of the 970s and then waned steadily through the remainder of that decade, suggesting the possibility of a cyclic nature in the process. What, then, of the 980s? The recent appearance of data from the 99 Census provides a timely opportunity to examine the latest trends and assess their significance in terms of longer-term population developments in Britain. In the main part of the analysis I draw on the small-area Census population counts assembled to

5 Population change and migration in Britain since form the nearest equivalent of the 'metropolitan area' concept currently available in Britain, looking initially at the patterns of change in 98-9 and then comparing these with the three previous decades. Subsequently, use is made of the annual estimates of population change for local authority districts, which though providing a less satisfactory depiction of Britain's urban and regional framework does allow the separate examination of the natural change and migration components of population change and provides a basis on which to examine trends in population change within intercensal periods. Table 3. National trends in urbanisation and counterurbanisation in the 970s and 980s, for selected European countries (source: compiled from Fielding, 982; 986; 990; and calculations from data supplied to the author by national statistical agencies. Reprinted from Champion, 992a, page 47). Country (number of regions) 970s 980s Shift Shift Austria (6, 8) Belgium (9) Denmark () Finland (2) France (22) FRG (30, 2) Ireland (9) Italy (3, 20) Netherlands () Norway (8) Portugal (7) Sweden (2, 24) ? -? Switzerland () a? Notes: NC no significant change. In shift columns, shift towards 'urbanisation', - shift towards 'counterurbanisation'. Also note that the correlation coefficients should be interpreted with care because their significance level depends on the number of regions. Data are correlation coefficients of relationships between the net migration rate and population density. a Data for population change (not migration). - NC - 3 Population redistribution for a typology of Britain's local labour market areas The framework of local labour-market areas (LLMAs) was devised by the Centre for Urban and Regional Development Studies (CURDS) at Newcastle University specifically for the analysis of urban and regional change in Britain (Coombes et al, 982). It involves a functional regionalisation of Britain, leading to the identification of urban-centred regions which are relatively self-contained in journey-to-work terms. The 28 LLMAs that make up Great Britain (defined as England, Scotland, and Wales, but excluding Northern Ireland, the Isle of Man, and the Channel Islands) can be classified according to their degree of dependence on the nation's principal urban centres. These centres are known as the 'metropolitan dominants', and the LLMAs which are linked to them by significant commuting ties are termed 'metropolitan subdominants'. These two, along with a small number of 'metropolitan rural areas' (LLMAs also linked by commuting flows to either the dominants or subdominants, but with a main settlement which falls below a critical threshold size), make up metropolitan Britain. The remainder of Britain consists of LLMAs which are relatively independent of the nation's major metropolitan centres and thus are labelled 'freestanding', these being subdivided into 'urban regions' and 'rural areas' according to the size of their main settlement. The patterning of Britain on this basis is shown in figure.

6 506 A G Champion Metropolitan Dominant Subdominant Rural Freestanding H] Urban FTT Rural Figure. The local labour-market area framework. In the following analysis I draw on a data set which has been assembled from data from the last five decennial Censuses in Britain, spanning the period 95-9 (and excluding the 0% sample Census of 966). All the small-area statistics from the 97, 98, and 99 Censuses have been aggregated to the 98 Census wards which form the building blocks for the LLMAs, whereas comparisons of 95, 96, and 97 are made by reference to the aggregation of parish and ward

7 Population change and migration in Britain since population counts extracted from the published Census county reports. For the most recent intercensal period, 98-9, the population is defined in terms of residents, but for the earlier decades comparisons over time are possible only on the basis of the enumerated population; that is, persons present on Census night. The results of calculations of population changes for the metropolitan/urban typology of Britain's LLMAs for the most recent decade are shown in table 4. The pattern is very clear. The population of metropolitan Britain grew somewhat between 98 and 99, but the freestanding LLMAs in aggregate increased their number of residents by a much larger margin in both relative and absolute terms. Within metropolitan Britain, the dominants lost population over the ten-year period and their surrounding subdominants fell short of the national rate of growth, so both saw a shrinkage in their share of Britain's total population. Here, only the small group of metropolitan rural LLMAs grew faster than the national rate. Within freestanding Britain, both the urban regions and the rural areas exceeded the national growth rate, with the rural areas outperforming the urban regions. For the 980s, therefore, there exists a very strong metropolitan/urban dimension to the rate of population change in Britain, with strong contrasts between metropolitan and freestanding Britain and also between the urban regions and rural areas in both these parts of the country. Table 4. Population distribution and change, by metropolitan or urban type, 98-9 (source: calculated from Population Census small areas statistics). LLMA type 98 population 99 population 98-9 change thousands %GB thousands %GB thousands % Metropolitan dominant subdominant rural!i Great Britain Note: % GB is the percentage of the Great Britain total. Table 5 sets the experience of the 980s in the context of longer-term trends from 95. First, from the broad distinction between metropolitan and freestanding Britain, it can be seen that the latest decade does not represent a continuation of earlier trends which involved the progressive widening of the gap in the rate of population change from the 950s situation of almost identical growth. Although freestanding Britain between the 970s and 980s appears to have maintained its growth rate at 6%, metropolitan Britain recovered from substantial population loss in the 970s. As a result, the differential in the growth rate narrowed from 8.3% points in 97-8 to 5.6% points in This point is demonstrated most clearly in the four right-hand columns of table 5, where the change rates are presented as percentage-point deviations from the national rate, and in figure 2, in which these data are plotted. The lower panel of table 5 shows the results of the same form of analysis applied to the four main metropolitan/urban types of LLMAs (the very small category of metropolitan rural is ignored for this purpose). The key features of the trends over time are again most clearly identifiable after allowance is made for fluctuations in the national growth rate, as presented in the right-hand columns of

8 508 A G Champion table 5 and plotted in figure 3. Particularly impressive is the transformation in the relative growth of the four types between the 950s and the 970s. In the first two decades, the metropolitan subdominants were in the ascendency, with below-average rates for both ends of the settlement spectrum, but by the 970s a regular gradient had emerged, with the subdominants being outpaced by the freestanding urban regions and with the freestanding rural areas displaying the highest percentage growth rate. The change in the performance of the rural areas over these three decades is very marked, shifting from 5.5% below the national rate in the 950s (and an actual decline of 0.5%) to 8.8% above it in As noted above, the same regularity across the four types is found for the 980s, but the steepness of the gradient is less, with the differential between the two extremes narrowing to 8.8% points from its level of 5.7% points in the 970s. Table 5. Decennial rates of population change, 95-9, by metropolitan or urban type [(source: calculated from Population Census small area statistics and published reports (see text)]. LLMA type Population change rate (%) Rate standardised to GB rate Great Britain Metropolitan Freestanding Dominant (UR) Subdominant (UR) Freestanding (UR) Freestanding (RA) Note: UR urban region; RA rural area. Four right-hand coliftnns show the percentage-point deviation of rates from the national level for the relevant decade $s Freestanding Britain Metropolitan Britain ~ 4 ' L Figure 2. Population change, 95-9, for metropolitan and freestanding Britain. Data relate to the percentage-point deviation from the national rate for each decade (see table 5). From this evidence it would appear that Britain's experience in the 980s puts it into an intermediate position in the international context outlined earlier in the paper. On the one hand, Britain did not undergo the complete reversal of the original turnaround that characterised the situation in the USA and a number of more peripheral European countries at this time. On the other hand, however, the

9 Population change and migration in Britain since pace of population deconcentration in Britain was running more slowly between 98 and 99 than in the previous decade, and there was certainly no further intensification of the dispersal process as had occurred between the 960s and 970s. Freestanding rural Freestanding urban Metropolitan sub dominant Metropolitan dominant Figure 3. Population change, 95-9, for metropolitan and urban LLMA types. Data relate to the percentage-point deviation from the national rate for each decade (see table 5). 4 Population change and migration for types of local authority districts The Census-derived data for LLMAs cannot address all the points of comparison with the studies reviewed earlier. One aspect is that by definition these data do not, by themselves, allow the examination of population trends within intercensal periods and thus, with only a ten-yearly Census in Britain, they cannot reveal whether significant changes took place between the first and second halves of the 980s, as was found for Australia, the USA, and some European countries. Second, it is not possible to focus in on the migration component of population change by using the Census for the LLMA framework. This is partly because in Britain, unlike Australia and France for instance, the Census collects no information on people's addresses at the previous Census, and partly because registration data on births and deaths, which could be used to calculate net migration as a residual, are not available for the small areas (enumeration districts) which form the building blocks for the LLMAs. In any case, given the much greater problem of underenumeration encountered by the 99 Census compared with earlier ones (see OPCS, 993), it would be difficult to know how much credence to give to migration residuals calculated in this way at least until the results of fuller checks on the 99 Census data have been released. The obvious alternative data are the annual estimates produced by the official population agencies, namely the Office of Population Censuses and Surveys (OPCS) for England and Wales and the Registrar General for Scotland. Population estimates, also known as the mid-year estimates because they refer to the position on 30 June each year, are now available for 99. These are superior to the raw Census data because not only do they take into account the results of the 99 Census but they also make allowance for the known deficiencies in the Census, notably the underenumeration problem, insofar as this is possible with the intelligence available

10 50 A G Champion to the agencies. They also provide a time series of annual figures which can be taken back as far as 96. For analyses of urban-system change, the chief drawback is that the data at their most disaggregated relate to the local government districts which are not only relatively large (459 areas for a population of over 56 million, i.e. averaging over people) but also bear rather slim resemblance to Britain's settlement structure. Nevertheless, a typology of these districts is now widely used to look at metropolitan and rural tendencies in population redistribution (for example, Champion, 987; OPCS, 993), and data on births and deaths are readily available for these areas, allowing the estimation of the net migration balance. The typology of districts is shown in table 6. A distinction is drawn between metropolitan and nonmetropolitan parts of Britain essentially in terms of what are commonly called the 'metropolitan counties' and the 'shire counties'. In England the metropolitan part comprises Greater London and the six Metropolitan Counties (Greater Manchester, Merseyside, South Yorkshire, Tyne and Wear, West Midlands, and West Yorkshire), and in Scotland the term is used to cover the districts approximating to the extent of the former Central Clydeside Conurbation. Within metropolitan Britain, London is treated separately from the others owing to its much larger size, its special status, and its location in the south, and is separated into its inner and outer boroughs. Amongst the metropolitan districts elsewhere in Britain a distinction is drawn between their principal cities (Manchester, Liverpool, Sheffield, Newcastle upon Tyne, Birmingham, Leeds, and Glasgow, respectively) and the rest. For the rest of England and Scotland and for the whole of Wales (termed 'nonmetropolitan Britain' for the purposes of this study), districts are classified according to a mixture of size, planning status, and sociodemographic characteristics, based primarily on a cluster analysis produced by Webber and Craig (976). The allocation of districts to the eleven basic categories recognised in this typology follows fairly closely that used by OPCS for England and Wales (see OPCS, 99, appendix 5) and has been extended to cover Scotland on the same principles (see Champion and Townsend, 990a, pages 5-54). This typology provides two basic levels at which population concentration and dispersal tendencies can be examined. The first is the broad distinction between the metropolitan and nonmetropolitan parts of Britain. At this level the data on changes in 98-9 in table 6 confirm the strong shift of population away from the metropolitan areas. The only significant difference from the results of the LLMA analysis above is that table 6 indicates a sizeable population loss for metropolitan Britain, with its overall number in 99 being or.8% lower than in 98, as opposed to the slight gain shown in table 4. This difference is due largely to the tighter definition of metropolitan Britain in the present analysis. Second, it can be seen that there are major differences in growth rates between the individual district types. Particularly notable is the urban-rural distinction within nonmetropolitan Britain. Here the four district types which are defined primarily on the basis of urban status follow a clearcut gradient: a.2% decline in population over the decade for the large cities, through gains of.8% for small cities and 5.7% for urban and mixed urban/rural districts, to a 7.6% increase for remoter, largely rural districts. For the other three nonmetropolitan types, a clear (and predictable) distinction can be seen between the below-average growth of the manufacturing towns and the strong performance of the resort and retirement towns and the districts containing New Towns. Set against these two lines of evidence for relatively strong deconcentration tendencies over the 98-9 period, however, is the case of London. The rest of metropolitan Britain, with its 2.7% decline, fits in well with the urban-rural gradient

11 Population change and migration in Britain since 98 5 found in nonmetropolitan Britain, but according to this analysis London managed to maintain its population level over the decade and thus proved more resilient than the other metropolitan areas and the larger nonmetropolitan cities (table 6). This is clearly at odds with the basic tenet of the 'counterurbanisation' or 'population deconcentration' perspectives on population growth patterns, which as noted earlier in the paper require a negative relationship between the size (or urban status) of a place and its rate of population growth. London's 980s experience is also rather distinctive in terms of the lack of a major difference in growth rates between its inner boroughs and its suburbs, with if anything the inner boroughs appearing to perform the more strongly. At the same time, it can be suggested that the metropolitan districts will have proved less dynamic than would be predicted by their size and urban status, because of their high degree of specialisation in older declining industries. Table 6. Population change in Great Britain, by type of district, District type Great Britain All metropolitan Britain London inner outer Metropolitan districts principal cities other districts Nonmetropolitan Britain large cities small cities industrial areas New Towns resort, port, and retirement urban and mixed urban/rural remoter, largel y rural Population 98 (thousands) Overall change 98-9 thousands % Natural change (%) Net migration (%) Note: Rates are calculated from official mid-year estimates (provisional estimates for 99), adjusted for boundary changes. The district classification is a modified version of the OPCS district typology; 'metropolitan' includes the Central Clydeside Conurbation area. 4. Net migration rates, 98-9 So far, this analysis of the population estimates has been concentrated on the rates of overall population change, but as mentioned earlier this data set has the advantage of a readily available breakdown into the natural increase and net migration components. This is important because the counterurbanisation debate has always been concerned principally with patterns of migration between metropolitan and less urbanised areas. The question thus arises as to whether the deconcentration dimension in the 980s appears so important if only the migration component is examined, and whether the positions of London and the three functionally defined nonmetropolitan types appear as distinctive on the basis of migration alone as they do in terms of their rates of overall population change. The answer can be gauged from the rates of natural increase and net migratory change shown in the two right-hand columns of table 6. In the first place, the basic distinctions between metropolitan and nonmetropolitan Britain and

12 52 A G Champion across the settlement-size gradient appear even stronger when migration alone is considered. Metropolitan Britain recorded a net migratory loss of 4.2% between 98 and 99 in contrast to the 3.3% gain for the rest of the country. Second, the four nonmetropolitan district types that are essentially defined on the basis of urban status appear in the same rank order as for population change and exhibit a wider range of performance, stretching from a 3.0% migratory loss for the larger cities to the 8.% gain of the remoter, largely rural districts. Moreover, the metropolitan districts (excluding London) extend this progression, whether we look at their overall 4.5% migratory loss or differentiate the smaller metropolitan districts (3.4% loss) from the principal cities (6.6% loss). Furthermore, the position of London appears rather less distinctive when we look at migration on its own. It can be seen from table 6 that London is characterised by a rate of natural increase which is very high by national standards, its 3.6% level for the decade being twice the national rate and second only to the New Towns amongst the district types. This clearly provides a strong boost to London's overall population change rate. As a corollary, its migratory loss of 3.6% between 98 and 99 is not far short of the 4.5% rate for the metropolitan districts. At the same time, it should be borne in mind that the migratory balances shown in table 6 are an amalgam of both internal and international migration exchanges. Estimates based on the International Passenger Survey reveal that over the period London was a net gainer through international migration to the tune of around people, and lost through net internal migration (Champion and Congdon, 992). If this relationship held for the full decade 98-9, London's internal migration losses will have amounted to over 6%, making London's position conform much more with the deconcentration relationship found across the rest of Britain which was much less affected by net international migration. As regards the three functionally defined nonmetropolitan district types, the breakdown of their population change into the two components renders them even more distinctive from each other than when expressed in terms of the rates of overall change (table 6). The industrial areas, characterised by a level of natural increase significantly above the national rate, appear in an even less attractive light when their migration balance is considered than when the overall population growth is used, being the only net migration loser among the nonmetropolitan types besides the large cities. Also, the strong overall growth of the districts with New Towns can now be seen to be almost entirely because of a natural increase, with their net migratory gain of 2.3% falling roughly in line with the 'urban and mixed urban/ rural' category which is very different from the rate for remoter, larger rural districts. By contrast, the resort and retirement districts emerge as an even more distinctive category, characterised by a high rate of natural decrease and an extremely strong level of net inward migration which dwarfs even that of the rural districts. According to this analysis, therefore, population redistribution in 980s Britain appears to have been characterised by a clear deconcentration dimension, even though the effects of economic restructuring are also evident, notably in the contrast between London and the metropolitan districts and in some of the growth-rate differentials in nonmetropolitan Britain. So far, however, the analyses have been undertaken at the national level, with only the contrast between London and the metropolitan districts giving an indication of regional differences in rates of population change. The question now arises as to how pervasive across Britain is the general tendency towards population deconcentration from the major urban centres and net migration into less urbanised parts of the country. Given the marked reopening of the North-South divide in Britain in the 980s (for instance, see

13 Population change and migration in Britain since Lewis and Townsend, 989), there is a distinct possibility that the trends shown in these national tables could have resulted from large-scale changes occurring only in the southern half of the country. In practice, this does not seem to be the case. The evidence in table 7 is that population growth and net in-migration were relatively widespread across rural Britain over this decade. In aggregate, the category of remoter, largely rural districts recorded gains in all the standard regions where these districts are represented. Not surprisingly, there are variations between regions, varying from the South West and East Anglia (with a 0%-% increase in the population in this district type) to Scotland (with a rate of not much more than a third of this). This range, however, broadly reflects the differences in the overall regional growth rate. Indeed, it can be seen from table 6 that the rural 'premium' (measured in terms of the percentagepoint difference between the rural category and the regional average) is highest in the more northerly regions that are experiencing overall population decline. Furthermore, it is clear that migration is the driving force behind this rural growth in all regions (table 6). Table 7. Population change, 98-9, for remoter, largely rural districts, by region. Region Overall population change RLRDs only RLRDs only thousands % all types \' ) natural (%) migration (%) Great Britain East Anglia South West Yorkshire and Humberside West Midlands East Midlands South East Wales Northern Scotland Notes: RLRDs Remoter, largely rural districts; there are no districts of this type in the North West region. Figures may not sum exactly because of rounding. Rates are calculated from official mid-year estimates (provisional estimates for 99), adjusted for boundary changes. If this impression of the widespread nature of rural growth needs any further confirmation, it is amply provided by data at the level of individual local authority districts. As shown in table 8, there is a much greater degree of variation at this level, but according to this source only seven of the 02 districts making up this 'remoter, largely rural' category are estimated to have experienced population decline between 98 and 99. Moreover, the most extreme of these, the Shetlands, can be treated as a special case owing to the large number of temporary construction workers present in 98 during the main period of North Sea oil development. This is not to say that the majority of rural settlements in Britain recorded population growth, because it is recognised that each local government district covers a large geographical area comprising a multitude of separate parishes with small villages, hamlets, and isolated dwellings. Previous research on 97-8 population change has shown that most of these rural districts, even the ones which

14 54 A G Champion had grown most rapidly, contained a significant proportion of declining parishes (Champion and Townsend, 990b; Flowerdew and Boyle, 992; Weekley, 988), and the same is likely to be the case between 98 and 99. The central point, however, is that over the majority of rural Britain the overall trend in the 980s was one of population growth produced by net in-migration. Table 8. Population change in Britain's remoter, largely rural districts, 98-9, extreme rates. Districts of decline Shetlands Islands Western Isles Caithness (Highland) Teesdale (Durham) Sutherland (Highland) Wigtown (Dumfries and Galloway) West Lindsey (Lincolnshire) Argyll (Strathclyde) Derby Dales Roxburgh (Borders) % Districts of growth Kincardine & Deeside (Grampian) Gordon (Grampian) Suffolk Coastal South Hams (Devon) North Cornwall Caradon (Cornwall) Fenland (Cambridgeshire) East Cambridgeshire East Yorkshire (Humberside) South Shropshire % Notes: Scilly Isles excluded. Rates are calculated from official mid-year estimates (provisional estimates for 99), adjusted for boundary changes. 4.2 Fluctuations in the rates of population change since 96 The other central question which the population estimates can be used to address is whether the reduction in the pace of deconcentration noted in the LLMA analysis between 97-8 and 98-9 represents a sustained reversal of the trends which developed between the 950s and 970s or results from the timing of short-term factors what Frey (987) terms 'period effects' in relation to intercensal decades. In particular, the yearly run of data provided by the population estimates from 96 provides the opportunity to examine whether the rates of rural population growth have been falling steadily since their peak in the early 970s. Because of data deficiencies for the years up to 97, the following analysis is restricted to the overall population change rather than net migration and excludes Scotland. It should also be noted that the annual data for the individual years have not yet been revised by OPCS in the light of the 99 Census results, so they are not comparable with the estimates for 99 and should therefore be treated with caution. The main features of the fluctuations in the rate of population change over time since 96 can be seen from table 9. The same classification of local authority districts is used here as in the previous analyses, except that the large and small nonmetropolitan cities are treated as a single group and as just mentioned the data refer only to England and Wales. The first data column provides the benchmark, showing that population dispersal was already well developed by the 960s. Both London (particularly its inner boroughs) and the principal metropolitan cities were already losing population heavily, against the national trend of relatively strong growth, and the chief beneficiaries of this process were the four least urbanised district types. At this time, however, it is notable that the strongest gains (averaging over twenty per thousand a year) were recorded by the New Towns and the 'urban and mixed urban/rural' districts, with the remoter, largely rural districts growing at less than half this rate. The pattern changed significantly in the first few years of the 970s, according to the evidence of the estimates in table 9. While the rate of population loss from

15 Population change and migration in Britain since London and the metropolitan districts accelerated, the growth rates of both the New Towns and 'urban and mixed urban/rural' districts fell back, in both cases by more than the reduction in national growth rate. By contrast, the annual growth rate of the rural category moved up to fifteen per thousand, against the national trend, making it one of the country's two fastest-growing categories alongside the New Towns. After 974, however, the growth rate for the remoter, largely rural districts moved down quite sharply and for the rest of the decade settled at a level broadly comparable with the 'urban and mixed urban/rural' category and the resort and retirement districts. By the early 980s all four of the less urbanised district types, including the New Towns, were sharing similar annual average rates of growth of between five and eight per thousand. This change from the situation ten years earlier in is paralleled by the marked reduction in population losses from London and the principal metropolitan cities (table 9). Table 9. Population change, 96-90, by type of district for England and Wales (annual average rate per thousand people) (source: OPCS Monitor PP 88/; calculations from OPCS mid-year estimates). District type Annual average rate England and Wales London inner outer Metropolitan districts principal cities other districts Nonmetropolitan Britain cities industrial areas New Towns resort, ports, and retirement urban and mixed urban/rural remoter, largely rural Thus far the patterns appear to accord strongly with the experience of the USA and other countries where the rural population turnaround went into reverse after the 970s, except perhaps that the peak of rural growth would seem to have been reached somewhat earlier in Britain. After 984, however, the remoter, mainly rural districts recorded a further upturn in the growth rate. This was matched by an acceleration in the growth rate of the resort and retirement category and was accompanied by the cessation of overall population losses by both Inner and Outer London (table 9). This apparent paradox of upward shifts in growth rates at both ends of the metropolitan/nonmetropolitan scale can be explained in terms of the resurgence in the national growth rate and by reference to the findings of the components-of-change analysis noted above from table 6. During the 980s, London, and notably its inner part, was the recipient of a large net influx of international migrants as well as being characterised by above-average levels of natural increase, both of which in statistical terms offset the effects of substantial levels of net outmigration to other parts of the United Kingdom, notably the less urbanised nonmetropolitan districts and not least the remoter rural areas.

16 56 A G Champion The annual trends for the two district categories at the extremes of the settlement hierarchy provide a good summary of the British experience of fluctuating rates of population deconcentration. In figure 4 the annual population growth rates for Inner London and the remoter, largely rural districts are represented in terms of percentage-point deviations from the national rate, so as to permit a direct comparison of their relative performance. This clearly shows how the growth rate of the rural category began to run ahead of the national rate in the early 960s and surged further ahead at the beginning of the 970s, but also how after 974 the growthrate differential narrowed steadily until the start of the 980s, after which the gap widened again to reach in 987/88 a level not far short of its 97/72 peak. Inner London forms something of a mirror image in terms of the phasing of its performance relative to the national rate, though it reached its nadir in 970/7 one year before the largest rural surge and subsequently made a recovery which appears much more dramatic than the degree of cutback in rural growth. The annual estimates, however, raise more questions than they can directly answer, particularly in view of the developments which took place in the closing stages of the 980s. As shown in figure 4, the resurgence of rural growth seems to have come to an abrupt halt after mid-988, whereas Inner London's relative position improved further after the setback of 987/88. Indeed, the remarkable position was reached in 989/90 when Inner London's rate of population growth was higher than that for the remoter, largely rural areas. What significance should be attached to this observation and what can the annual trends since the 960s tell us about the likely future importance of population deconcentration? The immediate response is to counsel against reading too much in the data for just one or two years. This is particularly the case here where the annual data from 98 onwards have not yet been revised in the light of the Census-based estimates for 99. It may also be significant that changes were taking place in 990 in the method of processing the records from which internal migration flows are calculated, and also that this was the period when the introduction of a register for the r / S~~^ >X,,,, k ^ ^ * - L \ r \ \ \ ' ' ' l_ / \ Remoter, largely rural / ^ ^ S y -/ ~x ^- / \ ' N \ i " X i i i i i i i i i i i i i i i i I * I \ N / / -'\/'' x ^ ^J - - ^ ' ' \ i /' \ i ' / i V v / Inner London v 96/62 966/67 97/72 976/77 98/82 986/87 Figure 4. Annual population change for remoter, largely rural districts (England and Wales only) and Inner London, 96-90, expressed as percentage-point deviations from the rate for Great Britain (see table 9).

17 Population change and migration in Britain since Community Charge ('poll tax') provided people with a clear incentive to disappear from all administrative lists. On the other hand, the rise and fall of rural growth during the middle and closing stages of the 980s, respectively, correspond closely with other evidence on migration drawn from housing and labour-market sources and can be related to the 'Lawson boom' and subsequent collapse of Britain's economic recovery from the recession. Perhaps the most important extra insight provided by the annual data is to demonstrate that the population deconcentration process has followed a cyclic pattern and that this pattern is not very adequately represented by the intercensal decades. In relation to the experience of rural Britain, one such cycle would seem to have run from the early 960s to the early 980s, with the acceleration of growth over the first ten years of upswing being interrupted only temporarily by the economic recession ushered in by the 967 Arab-Israeli war. Rather than the 970s being the 'decade of counterurbanisation' as revealed by the Census-based calculations, the annual estimates suggest that it was the period up to 97 which saw the acceleration of rural in-migration, with the subsequent ten years being a period of deceleration. Similarly, although the 980s saw rural areas gaining population more slowly on average than in the 970s, they would appear to have formed a period of accelerating rural growth until, that is, the abrupt halt associated with the latest recession. 5 Concluding comments In this paper I have presented the results of an empirical investigation into recent population redistribution trends in Britain, with the primary aim of discovering how far this has paralleled developments in other countries with respect to shifts between metropolitan and more rural areas. The international review at the beginning of the paper provided a somewhat confusing variety of patterns for the 980s. The most common feature, however, was a slow down or even reversal of the deconcentration or 'counterurbanisation' tendency which had been operating widely in the previous decade. Perhaps not surprisingly in these circumstances, the examination of redistribution trends in Britain identified both similarities and differences compared with this wider experience. The decade-based analysis has shown that the growth-rate differential between metropolitan and nonmetropolitan Britain narrowed somewhat between the 970s and the 980s. On the other hand, there was certainly no return to population concentration along the lines observed for the USA. Indeed, the negative relationship between metropolitan or urban status and the rate of population change, which was very clear in the period 97-8, held firm during the 980s and, moreover, appears even stronger for rates of net migratory change than for overall population growth. It has also been found that, contrary to the experience of the USA and a number of European countries, Britain did not witness an acceleration in the swing away from deconcentration over the decade but, instead, saw a resurgence of rural population growth in the mid-980s. The most distinctive features of Britain's 980s experience the acceleration of migration from metro to nonmetro areas during the decade and its widespread impact across rural Britain have already stimulated a number of studies designed to check these observations and gather more information about them, most notably those carried out under the aegis of the Institute of British Geographers (IBG) Working Party on Migration in Britain (Stillwell et al, 992a). The fluctuations in the pace of metro to nonmetro movement can be linked partly to trends in overall migration propensities (Stillwell et al, 992b). These trends closely follow the economic

18 58 A G Champion cycles found in the labour and housing markets, which during the 980s were dominated by the post-982 recovery of London and the South East but were also subject to a 'ripple effect' which produced progressively greater impact on more remote areas as the decade progressed (Champion and Congdon, 992). Reinforcing this pattern is the finding that, for most nonmetropolitan areas, the rates of gross in-migration fluctuate more widely than do out-migration rates (Fielding, 992). As regards the spatial impact of in-migration on nonmetropolitan areas, case studies have confirmed the existence of both extended suburbanisation and more remote rural growth. For instance, Flowerdew and Boyle (992) have shown that the most substantial levels of absolute growth through in-migration to Hereford and Worcester occurred close to that county's border with the metropolitan districts of the West Midlands conurbation. This experience is paralleled in other regions, including the overspill from Merseyside into Clwyd in north Wales (Green, 992), losses from the principal cities of northern England in favour of their surrounding metro districts and the nonmetro areas beyond (Rees et al, 992), and 'peripheral counterurbanisation' from Glasgow into the more rural parts of Strathclyde and from Edinburgh into the adjacent regions of Borders, Fife, and Tayside (Jones, 992). Nevertheless, as noted in table 8, the strongest gainers of the 980s include relatively remote areas of Devon and Cornwall, and Green (992) has observed the highest rates of migratory growth in Wales to be for the most rural counties of Powys, Gwynedd, and Dyfed. Longer-distance moves have also been observed in northern England (Rees et al, 992) and in the 'white settler' areas of northern Scotland such as Mull, Skye, Wester Ross, and the Orkneys (Jones et al, 986; Jones, 992). Yet, despite the recent growth of research on internal migration in Britain, there remain many unanswered questions and important opportunities for further work (Champion, 992b). Indeed, the principal aims of the IBG Migration Working Party (see Champion and Stillwell, 99) were to produce an agenda for future research and to provide a platform of empirical knowledge based on existing data sets which could be updated and extended as new data become available from the 99 Census and other sources. In particular, 994 should see the appearance of the 99 Census Special Migration Statistics and Special Workplace Statistics, as well as the release of the Samples of Anonymised Records and the 99 Census link-up to the OPCS Longitudinal Study of England and Wales. These should provide much greater opportunities than ever previously existed to look at the individual characteristics of migrants between metropolitan and nonmetropolitan areas and to relate such migration to local circumstances in both origin and destination areas. Furthermore, they will allow the examination of how localities have changed in nature over the past decade and, in particular, they will cast more light on the crucial question of which places can be considered metropolitan and nonmetropolitan for the purposes of studying population concentration and deconcentration tendencies. Acknowledgements. This is a revised and updated version of a paper presented at the UK/US Seminar on Migration in Post-industrial Society, held at the University of Southern California, Los Angeles, in August 992. The author gratefuly acknowledges the helpful comments of Professor WAV Clark and two anonymous referees, as well as comments received during the Seminar itself. Participation in the Seminar was made possible by funding from the Royal Society and the Economic and Social Research Council (award no. A ). The 99 Census data were obtained from the ESRC/JISC-funded holding at the Manchester Computing Centre. The author is particularly grateful to the support of colleagues at Newcastle University in assembling the Population Census small area statistics on a Local Labour Market Area basis, namely David Atkins, Martin Charlton, David Choi, Mike Coombes, Daniel Dorling and Colin Wymer. Ann Rooke drew figures 2 and 3, and Eric Quenet drew figure 4. The original data from the Census and population estimates are Crown Copyright.