Urban Stud 2005 Visual Methods Johnston 1221 7

where, xi is the number of members of theethnic group in census ED i; ti is the totalpopulation of census ED i; X is the totalpopulation of the ethnic group across allEDs; Ii is the index of isolation forethnic group i; and summation is over all ncensus EDs.

For Simpsons hypothetical city, the indexof isolation for his Group 1 (the ethnicminority group) is 0.36 at the first date and0.45 at the second. This suggests an increasein segregation between the two datesof theorder of 25 per cent. Other authors (forexample, Cutler et al., 1999; and Noden,2000) have argued that the index as originallyformulated should be corrected to takeaccount of the relative size of the ethnic

group (see also Sin, 2003).

MIi Ii Xi

T

(2)

where, Ii is the index of isolation for ethnicgroup i as computed by formula (1); Xi isthe total population of ethnic group i; Tis the total population of the city; and MIi isthe modified index of isolation for ethnicgroup i.

In Simpsons example, the citys popu-lation has grown from 15 000 to 15 800,entirely as a result of growth in the size ofthe ethnic minority group. Taking this intoaccount, the index of isolation at the twodates is 0.253 and 0.298 respectively.

Figure 1. Simpsons Figure 1 (Hypothetical regions of 15 areas with 2 groups. Lower graph showsGroup 1 increased by 50 per cent). Source: Simpson (2004, p. 667).

1222 RON JOHNSTON ET AL.

at UNIVERSITAT AUTNOMA DE BARCELONA on August 1, 2013usj.sagepub.comDownloaded from

Holding constant the minority groupsincreased proportion of the total population,its isolation increases.

Which index you use depends on whichaspect of the minority groups distributionacross the set of EDs you wish to focus on.The index of segregation used by Simpsonshows that the distribution of the two groupsdid not change between the two dateseachED has the same percentage of that groupspopulation living there at the second date asit did at the first. The unevenness of theirdistributionsrelative to each otherisunchanged. The index of isolation indicatesthe probability that a member of the groupunder considerationin this case, the ethnicminorityis likely to meet another memberof the same group by chance, given theethnic composition of the ED that he/shelives in. This probability increases over time,because the relative size of that group in thefive EDs increases: at time 1, group 1 com-prises 0.2, 0.2, 0.3, 0.4 and 0.5 of the popu-lation of areas 1115 respectively; at time 2,it comprises 0.27, 0.27, 0.39, 0.5 and 0.6respectively.

Neither index is right and neither is wrong:it depends on what you want to measure. Ourpreference between the two is for the index ofisolation, since it encapsulatesto a greaterextent than the index of dissimilarity/segregationwhat we understand to be thecentral feature of a segregated population,following Philpott (1978) and Peach (1996,1999). This involves the degree to which itdominates the population of areas rather thanshares those areas with members of othergroups, and hence the degree to which itis encapsulated from the remainder ofsociety. On that measure, Simpsons group 1has become more segregated over time: it ismore dominant in the five areas and sharesresidential space less with group 2.

Residential Segregation in Bradford

If we apply the index of isolation toBradfords South Asian population at the EDscale over the decade 19912001, usingSimpsons data (which he kindly supplied to

us), we get indices of isolation for the SouthAsian population of 0.53, 0.57 and 0.62 in1991, 1996 and 2001 respectively. Over thedecade, the South Asian population (whichincreased from 65 469 to 94 251, or some 44per cent) became somewhat more isolated.The South Asian group also became a largerproportion of Bradfords total, as the non-South Asian population declined: in 1991,the South Asians formed 14 per cent of thetotal; in 1996, 16 per cent; and, in 2001,nearly 20 per cent. Correcting the index ofisolation to take this relative growth intoaccount, and thus provide strictly comparablefigures, gives modified indices of 0.39, 0.41and 0.42 respectively. As it is growing inabsolute and relative size in Bradford, theSouth Asian population is becoming slightlymore isolated residentially.

An alternative way of looking at the degreeof segregationbased on the same definitionas the index of isolationuses visualmethods (Poulsen et al., 2002). These deployconcentration profiles to indicate what pro-portion of the members of an ethnic grouplives in areasEDs in this casewheretheir co-ethnics form a given percentage ofthe total population. Figure 2 shows such pro-files for Bradfords South Asians at the threedates in Simpsons data-set.

To construct the profiles in Figure 2, eachED percentage of South Asian and otherswas computed. A series of thresholds0,10, 20, 30 . . . 100was then identified andeach ED placed in its relevant between-threshold band (i.e. an ED with 35 per centSouth Asians was placed in the 3040threshold band). The cumulative distributionof South Asians across the threshold bandswas then calculated, to give the concentrationprofile. In Figure 2, the solid line shows theprofile for South Asians in 1991. At the 50per cent threshold, for example, this showsthat 60 per cent of South Asians lived inEDs with populations that were 50 per centor more South Asian. At the 70 per centthreshold, 30 per cent of Bradfords SouthAsians lived in EDs where their co-ethnicscomprised at least that share of the totalpopulation.

ON RESIDENTIAL SEGREGATION 1223

http://usj.sagepub.com/

The three graphs in Figure 2 show theincreased segregation of Bradfords SouthAsians over the decadeusing the PhilpottPeachPoulsen et al. definition. Each profileis closer to the upper-right-hand corner thanthat for the previous date, indicating greaterconcentration. That increased concentrationis particularly noticeable in the areas withthe relatively largest South Asian com-ponents. In 1991, for example, some 12 percent of South Asians lived in EDs where 80per cent or more of the population consistedof their co-ethnics; 5 years later, the figurewas 22 per cent; and, in 2001, it was 35 percent. At the end of the decade, over one-third of Bradfords South Asians lived inEDs where they comprised more than four-fifths of the population, almost three timesthe ratio 10 years earlier.

This graph of concentration profilesprovides very strong visual evidence of theincreasing segregationi.e. isolation fromthe remainder of the populationofBradfords South Asian population over that10-year period. This conclusion is confirmedby a second graph of concentration profiles

(Figure 3), which shows the distribution ofSouth Asians relative to the others (i.e. theremainder of the population, most of whomare White). In this case, the thresholds aredefined by the percentage of each EDs popu-lation comprised of others, so the profilesindicate the percentage of South Asiansliving in EDs with pre-defined other popu-lations. Thus in 1991, for example, justunder 40 per cent of South Asians lived inEDs where the other population was atleast 50 per cent of the total. Ten years later,the figure was 30 per cent. This increased iso-lation of South Asians from others inBradford is particularly noticeable at thelower thresholds: in 1991, for example, 87per cent of South Asians lived in EDs whereat least 20 per cent of the population wasfrom other ethnic groups; 10 years later,the figure had fallen to 65 per cent.

On the Meaning of Segregation: Patternand Process

If you accept the definition of segregationemployed hereisolation of one group from

Figure 2. Residential concentration profiles for South Asians in Bradford, 199119962001.



anotherthen the evidence from Bradfordover the period 19912001 seems incontro-vertible: South Asians became more isolated.The average South Asian living in Bradfordin 2001 was more likely to meet anotherSouth Asian in her/his local neighbourhoodthen than he/she was a decade earlier andmany more South Asians lived in areas thatwere predominantly South Asian in theirethnic composition at the later of the twodates.

That conclusion appears to fly in the face ofSimpsons claim that Increasing residentialsegregation of South Asian communities is amythat least for Bradford. But, of course,that depends on what you mean by segre-gation! For Simpson, it is clearly a processor a changing pattern; for many peopleincluding index and concentration profile con-structorsit is a pattern. Of course, it is both:you cannot have one without the otherprocess produces pattern, which constrainsprocess and so on. Segregation of SouthAsians was high in Bradford at the threedates analysed and changed very little accord-ing to Simpsons chosen index. But otherindexes, and the associated graphical pro-cedures deployed here, tell a different story.

Which story is relevant? It depends. Ifyou want to show that South Asians arenot conforming to the North Americanghetto model, but are moving away in somenumbers from their original concentrations,then Simpsons approach and his dataconfirm that. The dispersal of South Asiansinto areas of Bradford where members ofother ethnic groups predominate hasensured that segregation has not increasedas much as it might, given the levels of immi-gration and natural growth experienced byBradfords South Asians over the decade.Nevertheless, as the depictions here indicate,that dispersal has not equalled the amount ofin-movement plus growth in the traditionalareas of South Asian concentration inBradford, and that increased concentrationis reflected in the greater segregationdisplayed here.

Increased segregationaccording to thePhilpottPeachPoulsen et al. definitioncomes about when growth in the core areasof ethnic concentration increases at a greaterrate than the dispersal of households through-out the urban fabric and away from those areasof concentration. This can be represented bya simple model (Figure 4), which shows

Figure 3. The exposure of South Asians to other groups in Bradford, 199119962001: concentrationprofiles for South Asians according to thresholds for other groups.



different ideal-type trajectories in ethnic con-centration in the years after the establishmentof a settlement core. (For more details, seePoulsen et al., 2004.) In a Type 1 trajectory,concentration builds up to a particular leveland then remains constanteither becausethe inflow to the core (both inmigration andnatural growth) is equalled by the outflow toother areas or because there is no furthergrowth in that group. In Type 2, the outflowexceeds the inflow (with the latter perhapstailing off over time, as replenishment of theoriginal core declines) and so concentrationlevels fallafter some 40 years in the idea-lised case. In Type 3, on the other hand,inflow exceeds outflow and so concentrationlevels continue to rise. In the context of thismodel, Bradford is experiencing a Type-3 tra-jectory, but the amount of outflow identifiedby Simpson could lead to a shift to a Type-2trajectory in the near future.

Segregation and Policy

Why does all this matterother than topurists regarding definitions and measurementof concepts? For Simpson (2004, p. 662)Social policy . . . must address the dynamicsof residential location and relocation, ratherthan simply the existence of segregation atany one point in time. Of course; and his evi-dence of rates of dispersal provides an opti-mistic picture of the evolving situation inBradford: over time, as more South Asians

move away from the core areas, issues relatingto segregation will become less pressing.

But, the evidence presented here suggeststhat this optimism may be slightly premature.Segregationdefined as the concentration ofSouth Asians into areas where their co-ethnics predominateincreased quite con-siderably in Bradford between 1991 and2001, because growth in those core areasexceeded outflows from them. If segregationhas negative effects, on feelings of self-esteem and identity among the minority popu-lation, for example, on conflict between themand the others who live elsewhere, and oneducational and other opportunitiesallclaims made by commentators and exempli-fied by empirical research (although notnecessarily on South Asians in Bradford)then the pattern now is possibly more impor-tant to policy-makers than the emergentprocess. Segregation, as defined here, notonly exists now but is becoming moremarked. Thus, it presents a challenge topolicy-makers, one which may be relativelyshort-livedalthough if it is not tackled, thedispersal trends that Simpson identifies maybe reduced and segregation would thenbecome an even greater problem.

Conclusions

In almost all areas of social science, the resultsof a piece of research depend on how the ques-tions were phrased and how the relevant con-cepts were defined and measured. This iscertainly the case with ethnic residential

Figure 4. A model of changing patterns of ethnic residential concentration.



segregation, a concept that has a number of (atleast implicit) definitions and a series of quan-titative measures linked to them. Segregationcan mean either a process, which is whatSimpson focused on, or a patternand withthe latter, there are several competing (andto some extent complementary) definitionsof what that pattern comprises. According toSimpsons definition of segregation as pattern,the situation of Bradfords South Asianswas largely unchanged over the period19912001; according to our definition, it hasbeen exacerbated.

Evidence-based policy relies on the resultsof empirical social science research. Thusthe nature of that research must be carefullyprescribed; using different definitions of keyconcepts can lead to different results andsuggest different policy outcomes. There are,of course, no single right answers: each waya question is asked, each way a concept ismeasured, produces a different rightanswer. It all depends on the question. Theway Simpson phrased the question anddefined the concept led him to the positivemessage . . . that segregation is not theproblem it is perceived to be (p. 679). Theway we phrased it and made our measure-ments lead to a less sanguine conclusion:although residential segregationaspatterndid not increase as much as itmight have done for Bradfords SouthAsians between 1991 and 2001, because ofthe dispersal identified by Simpson, neverthe-less concentration of South Asians into a smallnumber of residential areas increased over thatperiod. The problemif segregation is aproblemhas not gone away yet. A mixedmessage perhaps, but if policy is to dealwith issues of today, while preparing fortomorrow, then perhaps we should not bequite as optimistic as Simpson.

References

CUTLER, D. M., GLAESER, E. L. and VIGDOR, J. L.(1999) The rise and decline of the Americanghetto, Journal of Political Economy, 107, pp.456506.

LIEBERSON, S. (1981) An asymmetrical approachto segregation, in: C. PEACH, V. ROBINSONand S. SMITH (Eds) Ethnic Segregation inCities, pp. 6182. London: Croom Helm.

MASSEY, D. and DENTON, N. A. (1988) Thedimensions of residential segregation, SocialForces, 67, pp. 182315.

NODEN, P. (2000) Rediscovering the impact ofmarketisation: dimensions of social segregationin Englands secondary schools, 199499,British Journal of Sociology of Education, 21,pp. 372390.

PEACH, C. (1996) The meaning of segregation, Plan-ning Practice and Research, 11, pp. 137150.

PEACH, C. (1999) London and New York: contrastsin British and American models of segregation,International Journal of Population Geography,5, pp. 319351.

PHILLIPS, D. (1998) Black minority ethnic concen-tration, segregation and dispersal in Britain,Urban Studies, 35, pp. 16811702.

PHILLIPS, D. (2002) Movement to opportunity?South Asian relocation in northern cities.ESRC end-of-grant report R000238038. Schoolof Geography, University of Leeds.

PHILPOTT, T. L. (1978) The Slum and the Ghetto:Neighbourhood Deterioration and MiddleClass Reform, Chicago 18801930. NewYork: Oxford University Press.

POULSEN, M. F., JOHNSTON, R. J. and FORREST, J.(2002) Plural cities and ethnic enclaves: introdu-cing a measurement procedure for comparativestudy, International Journal of Urban andRegional Research, 26, 229243.

POULSEN, M. F., JOHNSTON, R. J. and FORREST, J.(2004) Is Sydney a divided city ethnically?,Australian Geographical Studies, 42, pp.356377.

SIMPSON, L. (2004) Statistics of racial segregation:measures, evidence and policy, Urban Studies,41, pp. 661681.

SIN, C. H. (2003) The interpretation of segregationindices in context: the case of Singapore, TheProfessional Geographer, 42, pp. 422437.



Urban Stud 2005 Visual Methods Johnston 1221 7

Documents

Transcript of Urban Stud 2005 Visual Methods Johnston 1221 7