Journal of New Frontiers in Spatial Concepts KIT Scientific Publishing

ISSN 1868-6648 | Volume 4(2012), 53-59 http://ejournal.uvka.de/spatialconcepts/archives/1536

The spatial specificities of SC and ST population in Kolkata city: approaches, techniques and analysis
Swasti Vardhan Mishra
Indic Knowledge Operations network, E-Mail: swastivardhanmishra@gmail.com

Abstract The paper aims at achieving two objectives, firstly, to acquaint the researchers in the field of spatial demography with the techniques of residential segregation; pattern of population distribution based on the concept of spatial statistics and secondly, the practical evaluation of the techniques through real life data. The study queries about any residential segregation and pattern of location of the scheduled caste (SC) and the scheduled tribe (ST) population in Kolkata city by concentrating on the measures of Entropy Index (EI), Atkinsons Index, Absolute Centralization Index (ACE) and the Global and Local Spatial Autocorrelation (SA) measures of Morans I and LISA. The outcome of the study clarifies the fact that caste and race based segregation is being diluted in the wake of Indian urbanization and the spatial clustering tendency of the Concerned Social Group (CSG) is as mild. Keywords: Residential Segregation, Spatial Autocorrelation, Scheduled Caste, Scheduled Tribe
Manuscript recieved 23 April 2012, revised 25 April 2012, accepted 04 June 2012. Copyright note: This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided that the original work is properly cited.

INTRODUCTION Spatial Demography1 has gained a recent prominence among the spatial scientists trained in the disciplines of Geography, Economics, Sociology and Demography. Since the last few decades availability of micro data based on individual behavior (for example worldwide Demographic and Health surveys)2 has somewhat aloof the discipline of Spatial Demography as the individual centric information had nothing to be evaluated,3 on average and based on spatial implication. The recent vigour for the discipline has again

1 John R. Weeks, Population: An Introduction to Concepts and Issues, California 2008. John R. Weeks, The Role of Spatial Analysis in Demographic Research, in: Michael F. Goodchild/Donald G. Janelle (Ed.), Spatially Integrated Social Science, New York 2004, 381-399.
3 Paul R. Voss/Katherine J. Curtis White/Roger B. Hammer, The (Re-) Emergence of Spatial Demography, in: Working Paper 4. 2004, 1-4. 2

delved life into the same, bringing in more sophisticated and error free spatial statistics. Anthropocentric data has always been spatially integrated, where space is not a variable to be excluded,4 intentionally, from description, analysis and evaluation. A second point of research-oriented discontentment is with the Genre of Research. Recent demographic research is intensively focused on Reproductive Health, Maternal and Child health and Ageing with a hairline slot for population distribution and their pattern. Population Distribution Techniques (PDT) are the most powerful tool, which help largely in generalization and policy framing for specific group of spatial units. Health data, in a broader class, is spatially integrated and do have spatial underpinnings. PDT helps not only in recognizing any specific pattern of
4

Guy Stuart, Research and Publications: Harvard Kennedy School, Harvard Kennedy School, 2004.

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

54

distribution but also in chalking c out much m needed d strategic plan n. Residenti ial segregation n is a common n term applied d for clustering g of a minority y social group in contiguous s neighborhoods; for instanc ce, if a minori ity population n of a city is c clustered in particular p mun nicipal wards,, which are sp patially contigu uous in natur re, Residential l segregation is s said to be ex xisting. Reside ential segregation and its r related concep pts like Ghetto o, hypersegregation, etc h have their root ts in Urban Sociology S and d Urban Ecolog gy domains. It t is a theme of much interest t among Socio ologist, Urban n Geographers s, Economists s and Spatial D Demographers. The follow wing literatur re zeroes in the distribution n and pattern of location of schedule ed caste and d scheduled tr ribe populatio on of Kolkata city.5 At the e initial level, two measur res of Evenn ness and one e measure of C Centralization n of Residentia al segregation n have been at ttempted,6 wh hich is followe ed by evaluation of any p pattern of dist tribution of ab bove or below w average (i.e., the city avera age) proportion n of the Concerned Socia al Group (CS SG). Massey and Denton n (1988), throu ugh extensive literature sur rvey and employing clust tering techniq que demarcated five aspects s of segregation n: (a) Measur re of Eveness: measures the e differential distribution of CSG; (b) Mea asure of Exposure: measur res the poten ntial contact of CSG with h other groups; ; (c) Measure of o Concentration: measures s the relative a amount of phy ysical space sh haring among g groups; (d) M Measure of Centralization: C accounts for r the centripeta al (towards CB BD) or centrif fugal (towards s fringes) distr ribution of CS SG; (e) Measure of Clustering: account ts for the int tensity of disp proportionate e living of CSG G in contiguou us areas. The pattern of distributio on of CSG is evaluated on n theoretical gr rounds of Spa atial Autocorr relation (SA),, employing G Global and Loc cal statistics, Morans M I and d 7 LISA, respe ectively. Spat tial autocorre elation exists s when an obs servation of a variable is in n spatial relation with ot ther observati ions of the sa ame variable.. The observa ations must have h spatial character c and d could be rep presented on space. s For exa ample, in the e
5

following study the variable is the Scheduled caste and d the Scheduled tribe pop pulation (CSG G) and the observations are e the Average Deviate Valu ue (ADV)8 of CSG located on space i.e.,, in different municipal wa ards of the city y. Now, CSG b being the varia able, if the High ADV lies contiguously c t to the High ADV or the Low ADV lies in n contiguity t to Low ADV then t positive spatial auto ocorrelation e exists (fig. 1 (a) and 1 (b) )). On the othe er hand, if Low w ADV lies in n contiguity to the High ADV A negative e spatial autoc correlation exi ists (fig. 1 (c)).

Fig gure 1. Identifying Positive an nd Negative Sp patial Autocor rrelation

Veronique Dup pont, Socio-spati ial differentiation and residential segregation in D Delhi: a question of o scale?, in: Geof forum 35. 2004, 157-175.
6

Spa atial Autocor rrelation can be further classified, bas sed on the des sired outcome e, into Global and Local sta atistics when n a single value ue is produced d to denote the e level of spatial autocorr relation, it is s called a Glo obal measure (Morans I) a and when a set s of outcom mes are produ uced (generally ly, in the form m of cluster ma ap and significance map), w we call it a Lo ocal measure e. Local meas sures are the e decomposed d form of Glo obal measure e, which goe es beyond th he average val lue (Morans I) I to highlight t the individu ual clusters of Hot spots, i.e e., the individ dual units con ntribution tow wards the aver rage value. Local Indica ators of Spati tial Associatio on (LISA) bei ing a decomp posed form of f the Global measure m is eas sily represented on maps and helps in n drawing val lue differences s with respect t to the averag ge value of the e Global measure.

Douglas S. Mas ssey/Nancy A. De enton, The Dimensions of Residential Segr regation, in: Socia al Forces 67. 1988 8, 281-315. Luc Anselin, Local Indicators of f Spatial Associat tion- LISA, in: nalysis 27, no. 2. 1995, 93-115. Geographical An
7

Au uthor has coined the term to show w how much indiv vidual units specific value differs from that of the city average, whi ich in other and high ADV un nits. sense also helps in demarcating low a

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

55

Figure 2. Form mula and related concept of En ntropy Index.

Figure 3. Form mula and relate ed concept of Absolute Centralization Ind dex.

Following p paragraphs are divided in nto data and d methodology y part, descrip ption part an nd concluding g views.

Th he formula for three measur res of Residen ntial segregat tion is structured in figure 2 2, 3,4 & 5.9 10 11 In calculating g ACE, the diistance of the respective wa ards from CBD D is calculated as a distance of o centroid of different ward ds from the cen ntroid of the CBD C ward. Th his technique of o estimation iis not error fre ee but it is sur rely the best po ossible alterna ative for the sa ame. In the pre eviously menti ioned measure e, the wards are arranged acc cording to thei ir distance from m the CBD before calcurtion of CSG population lati ing the cumula ative of propor p and d area. The main motive behind inclu uding two me easures of Eve eness (Entropy y Index and Atkinsons A Ind dex) is the cur riosity to dig o out any possi ible differenc ces when both h the CSG are e combined to o measure (EI I) and when both b of them a are independe ently evaluat ted for segrega ation (Atkinso ons Index). Spatial Autocorrelation m measures have been perfor rmed through h open source e spatial stati istics softwa are GeoDa12 pu ublished for th he same purpo ose by Luc An nselin (Arizon na State Unive ersitys GeoDa a center).13 It is i not feasible to t illustrate bot oth the measures in detail r vast availabi for r the lack of sp pace and their ility across the e online domain. The valu ue of Morans s I ranges bet tween -1 to 1, where 0 sign nifies no association, -1 sig gnifies severe negative n spatia al autocorrelat tion and 1 sig gnifies very stro ong positive sp patial autocorrelation. In the following study, SA is fr framed for ana alyzing the pat ttern of distr ribution of t those municip pal wards wh here the proportion of SC C and ST pop pulation is bet tter or worse than the city y average. The Average De eviate Value wards w are ident tified with the e help of a sim mple formula:

ADV
Figure 4. Form mula and related concept of tk kinsons Index.

ATA AND ME ETHODOLO OGY DA The data has s been obtain ned from the Office of the e Census Com mmissioner of India pertain ning to census s 2001. The spatial data i.e., the area and distance d from m the CBD is calculated with h the help of a GIS application. The me ethodology applied in arrivi ing at conclusions is facil le and availab ble through any a length of f segregation literature.

Th he value over 1 connotes a siituation of CSG proportion in the resp pective ward above the cit ty average (H High ADV), val lue of 1 denote tes equal propo ortion and val lue below 1 signifies com mparatively lo ower CSG pro oportion (Low w ADV).
9 Jacob S. Seigel/Dav vid A. Swanson, T The Methods and Materials of mography, Londo on 2004. Dem

Se ean F. Reardon/David OSullivan, M Measures of Spatia al Seg gregation, in: Socio ological Methodollogy 34, no. 1. 200 04, 121-162.
10

11

Le eah Platt Bouston n, Residential Segr regation in Amer rican cities, in: Nancy N Brooks/Ki iran Donaghy/Ge errit Jan Knaap (E Ed.), The Oxf ford Handbook of o Urban Econom mics and Planning g, New York 201 11, 318-339.
12 13

http://geodacenter h r.asu.edu/softwar re/downloads.

Luc L Anselin, Revie ew of Cluster Ana alysis Software, North N Am merican Associatio on of Central Can ncer Registries 20 004.

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

56

urb banization.16 With W this idea a of caste, trib bes can be rou ughly merged d without am mbiguity. The following gen neralization has h not been sought after through a com mprehensive and exhaustiive time series model, nev vertheless, the present stu udy do stress upon the sam me. Entropy Index, calculat ted for CSG as a a whole, sho ows a slight deviation d from m the situation n of absolut te even distrib bution with a value of 0.195. SC and ST groups have roughly equa al distribution n across all the e municipal wards of Kolkat ta city, implyin ng a situation of no segreg gation in relati tive terms. The e evenness quotient for bot th the CSG, c calculated sep parately, is mo ore intriguing. Atkinsons In ndex is perfor rmed with 17 18 fou ur shape para ameters ,h highlighting th he normative component of the measur re and its judic cious use.
Figure 5. Keys s to the Residen ntial Segregation n formulae.

Each measur re of Morans I and Lisa is accompanied d with a scatter rplot, LISA cluster map and d significance e 4 map, where t the significanc ce is kept at = 0.05 level.14 Besides, a mu ultivariate ana alysis is attem mpted between n SC and ST population. Th he SA measure e between the e 15 two identifies any spatial conformity c i.e e., possibility y of high prop portion of SC C population lying at the e same spatial unit as high proportion of o ST populae versa. The cluster maps also display y tion or vice iers (SO), sym mbolizing a pattern p comSpatial Outli mon to Nega ative Autocorr relation. A spatial unit of a certain value e lying in the e vicinity of spatial units,, having value opposite to it t, is a spatial outlier. The inter rpretation will be presente ed through a well-structur red format, in n which individual measure e is explained first followed by a combine ed interpretation taking in nto view all th he measures. DISCU USSION Even distribu ution of a po opulation sub bgroup across s spatial units is a normativ ve considerat tion. The two o measures of evenness i.e., Entropy Inde ex and Atkinsons Index h have been att tempted to em mphasize that t the significan nce of caste has h decreased d with Indian n

Fig gure 6. LISA Cl luster Map, Sign nificance Map and Scatter Plo ot for Scheduled d Caste Populat tion

16

Chandra C Bhan Pra asad, New order, in: Himal Southa asian April. 201 10.
17 14

Peter A. Roger rson, Statistical Methods M for Geog graphy: A Student's Guide, New Delhi 2001 1.
15 Rick Grannis, Segregation Indi ices and their Fun nctional Inputs, Methodology 32, no. 1. 2002, 69-8 84. The author in: Sociological M talked of spatial proximity to just tify location of minority units majority units, bu ut in this paper th he two CSG are with respect to m two kind of min nority social and cultural c group.

Jo ohn Iceland/Dani iel H. Weinberg/E Erika Steinmetz, Racial and Eth hnic Residential Segregation in the e United States: 19 980-2000, Wa ashington 2002.
18

Holger H Schlor/Wo olfgang Fischer/Ju urgen-Friedrich Hake, H Social We elfare, Income, Co onsumption, Ener rgy, and the Ineq quality Ave ersion of society A case study fro om Germany, in: Nicholas Tso ounis (Ed.), Intern national Conferen nce on Applied Economics E , Per rugia 2011, 567-57 77.

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

57

The value of f b is kept be etween 0.4 an nd 0.1 (below w 0.5) as the n no. of wards having above e city average e CSG populat tion is less; 42 for SC and 48 8 for ST popuvalue for the in ndex increases s as the shape e lation. The v parameter va alue goes down n, the only exception being g the SC popu ulation as the shape param meter is tuned d from 0.2 to 0 0.1. ons Index for r ST population increases s The Atkinso from 0.025 t to 0.90 to 0.25 5 to 0.53, accentuating the e point that as s the influenc ce of wards having h Below w Average CSG G population is i inflated (by y reducing the e value of b), the index inc creases. The values v for SC C ned level of b population, at the previously mention 007, 0.008 and d 0.006. Follow wing from the e are 0.005, 0.0 values at han nd, it can be interpreted i th hat in spite of f maximum cr redibility given n to BA wards s (b=0. 1), the e Index value d does not show w a noticeable e mark. There e exists close t to zero segreg gation for bot th CSG, however ST popu ulation does show s a feeble natured segregation (une evenness). In Human Geography, concepts c of urban u fringe,, ance and demographic gravitation g is s spatial dista overwhelmin ngly followed, which are of f fundamental l importance f for the exposit tion of Reside ential segregation in gene eral and mea asure of cent tralization in n particular. A Absolute Centr ralization Index (ACE) is a measure of i intensity of CSG C to live away a from or r close to the e central core e of the city (CBD). The e measure con nsiders the distance from the t CBD and d the proportio on of total are ea for each wa ard. From the e study it is fou und that the SC S population n tends to live e evenly throu ughout the cityscape, c with h a value of f 0.028. This g group does no ot seem to be spatially s marginalized or spatially dear r, for its value e lies close to o zero. The ST T population n of the city illumines its s peripheral lo ocation with an n ACE value at a -0.597. The e value is ser riously consid derable for detecting d the e probable cau uses of such a spatially s marg ginal character r of the CSG. I Innumerable literature on the t condition n of tribes in In ndia along a long span of ti ime has highlighted their r marginal na ature; their location in the e peripheral a and hostile te errain. Drawi ing from the e same notion n, it can be deduced d that the marginal l character of ST population n in Kolkata is i attributable e to their failur re in merging themselves with w the mainstream popul lation. Contrary y to this well-g grounded view w, is the view w of counterur rbanisation, which w asserts relocation of f

population from m urban core to fringe; fo or straying fro om the grime and congestiion of the city y. Kolkata city y has certainly a ring of po osh localities around a its urb ban core like Jadavpur, Kan nkurgachi, Ph hoolbagan, Jod dhpur Park and others, which work as Relief Ma agnets;19 to st tay away from m the city ch haos. It is im mperative to tak ke a field visit t and get accu urate justifica ation for such h a level of se egregation. The lack of res source and time prevented d me from ta aking this eff fort, which prompts one of f the most sev vere drawbac cks of a resea arch; neverthe eless, the poss sibility for suc ch an initiative will be bindiing on me.

Fig gure 7. Showing g the LISA C Cluster Map, Significance S Ma ap and Scatter Plot P for Schedulled Tribe Popu ulation

To oblers First la aw of Geogra aphy is at th he core of Spa atial Autocorr relation statis stics.20 The law w divulges tha at everything g is related t to everything g else, but nea ar things are more related d than distant things.21

19

Author A has coined d the term, keepin ng in mind the loc cal perspective e.
20

Harvey H J. Miller, Tobler's T First Law w and Spatial Anal lysis, in: Ann nals of the Associ iation of America an Geographers 94, 9 no. 2. 200 04, 284289.
21

W. W Tobler, A computer movie simu ulating urban gro owth in the Det troit region, in: Economic Geograp phy 46, no. 2. 197 70, 234-240.

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

58

Spa atial autocorr relation is mo ore numerou us for low AD DV units (low w value lying iin contiguity to low) in con ntrast to the units u with high h ADV. For SC population an extensiv ve cluster of h high value spa atial autocor rrelation is ob bserved in th he eastern frin nge of the city y, otherwise th he pattern of h high values, in n both the cas ses, is disperse ed. In contras st, units with low value pre esent a more e clustered as ssociation in both the cas ses. The spatia al outliers or n negative autoc correlation is also a found to be present fo or both the CS SG, where units of high values v residing g in proximi ity to low val lues is a comm mon feature b but low value es lying in con ntiguity to hig gher values is s a distinctive feature of the e ST populatio on. Th he bivariate association be etween the SC C and ST population is more m weaker w with a value of o Morans I= 0.0621, impl lying a weake er correlation n between the e location of SC S and ST po opulation and d here also hig gh value units s are meager iin contrast to low value units and units with w negative spatial autoco orrelation. Th he association in any of th he cases is not n strong eno ough to confo orm to the viiew of Tobler. Presence of outnumbered non-signif ficant units show the dis scontinuity of f any spatial conformity, a weaker dem mographic gr ravitation with hin a group and a possibility of any eco onomic, other r social and ps sychological l motives behind the locatio on of CSG. SION CONCLUS Residential segregation, a uniiversal phenomenon, is a massive m conc cern when it is in tune with w lowinc come social group. g The m milieu of suc ch groups inf fluences and get influence ed so immen nse, as to cre eate chaos in city governan nce. Handling of Residen ntial Segregat tion, embedd ded in social structure, mu ust be the pr rerogatives of f the Local go overnance and d should be maneuvered m g gently and thr rough efficie ent Urban Plan nning. Population of o Kolkata city y has decreased for the firs st time in census history fro om 4.57 million to 4.48 mi illion,22 which h creates an opportunity for more com mprehensive study. The s scope of the study will loo ok for any specification in th he explosion of o population outwards and a how stron ng it is reflec cted in its seg gregation scen nario.
22

Figure 8. Show wing the LISA A Cluster Map p, Significance e Map and Sca atter Plot for Bivariate Ana alysis between n Scheduled Cas ste and Schedul led Tribe Popul lation

Spatial Autocorrelation, Global G and Lo ocal, has been n performed fo or the proport tion of CSG in n every ward,, above or belo ow the city pro oportion.

Figure 9. Show wing the Legends for interpreting LISA A Cluster map an nd LISA Signifi icance map.

The LISA clu uster map, sign nificant map and a Morans I,, shows an eng grossing view w of spatial ass sociation. It is s more marked d for the SC po opulation; bein ng testified by y the value of M Morans I (0.4 4305 for SC po opulation and d 0.0965 for ST T population) ). The signific cance of these e associations, obviously, is s well pronou unced for SC C population w with 3 units showing s significance at = = 0.001 level, w which in the case of ST po opulation is a single unit.

Census C Commisio oner of India, Prov ovisional Populati ion Totals Pap per 1 of 2011: West Bengal, Office of the Registrar General G and Cen nsus Commission ner of India 2011..

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/

Journal of New w Frontiers in Spatial S Concepts | ISSN 1868-6 6648 | Vol. 4(20 012), 53-59

59

It is desired t that the presen nt study has in nfluenced the e thought proc cess of researchers and mo otivated them m to delve deep per into the domain d of po opulation distribution. Th he concept of local measure es of segregation in addit tion to comp plex measures s of other aspects of Residential segr regation, are the field of f much interes st and should be explored. The meas sure of Spatia al Auocorrelat tion, having a wider range o of application n, is a successfu fully trod path h in spatial scie ence. This tec chnique can be successfully y applied to co oncepts of Hea alth Demography in spatial l context. Rec cent research and develop pment in the e same has brought sophi istication and d chances of f immense pos ssibilities. Social res search must in nternalize the dimension of f space and sh hould adhere to t its rightful application a at t the time of in nvestigation an nd analysis.

http://creativecommo ons.org/licenses s/by-nc-nd/3.0/ /de/