The Geographical Review

T H E SPATIAL ALLOCATION O F MIGRANTS

I N ACCRA, GHANA*

MILTOX E. HARVFY Avn RICHARD R. BRAYD

I

N T H E past decade the pace of basic research on cities in developing countries has quickened markedly, but like the early stages of the development process itself, progress has tended to be regionally imbalanced. JVhereas fundamental questions regarding the nature of' residential growth and change are beginning to be ansrvered in Latin America and Asia, much less is known about similar processes in African cities.' Moreover, there is no certainty that the findings in Lima, Calcutta, or Manila are applicable in the African context. T h e spatial dynamics of uncontrolled settlements, for example, have scarcely been studied in African cities, although many norv rank among the rvorld's most rapidly growing and least well regulated urban areas. Where are the primary destinations of the nervcomers, in the center of the city o r on the urban fringe! Are neighborhood preferences different

' A grant-in-aid from the Institute of Social Science, T h e Hague, Netherlands, supported eleven students from the Universit) of Cape Coast during the period of ir~rer\ielr.ingin 1968 and ~ g t j y T . h e direction and encouragement of Professor E. V. U'. Vercruijsse is gratefully acknowledged by Dr. Brand. T h e authors \\.ish to thank Professors George Carey, Ernilio Cauetti, Akin Ilabogut!je, David Meyer, Margaret Peil, George Kengert, and J . Barry Kiddell for comr~ientsmade on an earlier draft. ' For a sarilple of the published research in these regions, see Stanley D. Brunn: Urbanization in Developing Countries: An International Bibliography, Latin Amer. Studies C ~ n t fResearch l Rept. 'Yo. 8, Latin Amer. Studies Center and T h e Center for Urban Affairs, Michigan State Uni\-.. East I.ansing, Mich., 1971; and hprodicio A. Laquian: Slums and Squatters in South and Southeast Asia, in Urbanization and National Development (edited b) Leo Jakobson and Ved Prakash; Sage Publications, Beverly Hills, Calif., 1 9 7 1 )pp. ~ I 89-203.

)DR. H . ~ R V EisYa professor of geography at Kent State University, Kent, Ohio. DR. BRAND is a research associate in geography at T h e Pennsylvania State University, University Park. Copyright O 1971 b y the Atnericntr Gt,02rat~ltical Society d'.Teru York

2

T H E GEOGRAPHICAL REVIEW

for skilled, semiskilled, and unskilled migrants? How have these locational tendencies changed over time? Jl'hat attributes structure the environments that are attractive to migrants with different lengths of urban experience and from different sociocultural backgrounds? These basic questions present im~nediatepolicy implications, yet they remain unanswered for most African cities. I n this paper we address ourselves to the foregoing questions in the context of migrant residential patterns in Accra, Ghana, and we seek to analyze the relationships among migrant status, residential characteristics, and observed locational patterns. O u r study is based on data drawn fro111 a sample survey and fi-om published census figures. Multivariate analysis of variance, stepwise discriminant analysis, and regression analysis are used to identify significant subsets of variables that structure the observed migrant

M I G R A N T S I N ACCRA

3

allocation surface, to extrapolate a potential surface for all of Accra from the sample data, and then to derive empirical generalizations about differences between two broad groups of migrants (foreign Africans and interregional migrants) in terms of neighborhood characteristics.

TRACT NUMBER AND STATISTICAL AREA

MIGRANTS

LOCAL BORN

TRACT TOTAL

yo M I G R A N T

Ja~nestown Jarnestown 3 Jamestown 4 Ussher Town 5 Ussher Town 6 Ussher Town 7 Ussher Town 8 Adabraka (Tudu) g Adabraka 1 0 Adabraka I I Mamprobi I 2 Sabon Zongo I 3 Abossey Okai 1 4 Kaneshie 15 Kaneshie 16 Kokornlernle 1 7 New Town 1 8 Nima 19 Labadi TOTAL MEAN I 2

* The respondents are male and female heads-of-household. Tracts I through I o are in central Accra, which is composed of the Jamestown, Ussher Town, and Adabraka statistical areas (see Fig. I ). SPATIAL DISTRIBUTION OF MIGRANTS It io hypothesized initially that variation in the length of urban experience is associated with significant differences in the socioeconomic composition and the areal pattern of migrant communities and that these variations are caused by the operation of complex processes of initial site selection and subsequent intraurban movement. Underlying these hypotheses is the belief that the longer the residence in Accra and the higher the socioeconomic status of migrants, the greater the range of choice for relocating to increase personal satisfaction. Data on length of residence in Accra were gathered in a survey of heads-of-household between December, 1968, and April, 1969. Eleven of the fifteen built-up statistical areas of the city (comprising

4

T H E GEOGRAPHICAL REVIELV

NEOPHYTES

TRACT NIJMBER AND STATISTICAL AREA

Ja~nestown Jarnestown 3 Jarnestown 4 Ussher Town 5 Ussher Town 6 Ussher Town 7 Ussher Town 8 Adabraka (Tudu) g Adabraka 1 0 Adabraka I I Mamprobi I 2 Sabon Zongo 1 3 Abosse Okai 1 4 Kaneshie 15 Kaneshie I 6 Kokomlernle 17 New Town I 8 Nirna 19 Labadi TOTAL I 2

DISTANCE FROM CBD

(fn

thousa~~Os of f e e t )

2.8 1.7 2.4 0.4 1.2

1.9 3.1 4.2 6.9 8.5 '3.5 I 0.8 I 3.8 14.0 13.5 12.0 '5.3 16.0 24.0

Nurnber of responses

yo of neophyte ~nierants

12

5.9 3.0 I .o

6 2

o 3 2

o o 6 3 16 I8 7 6 23 10

60 12

I5 200

YEAR) %,of total mlgrants and local Number of born in responses tracta

(LESS THAN I

0.0

1.5 I .o 0.0 0.0 3.0 1.5 7.9 8.9 3.5 3.0 11.5 5.0 29.9 6.0 7.4 100.0 u = 7.07 c . v . ~= 134.4

4.7 2.8 I .2 0.0 2.5 2.2

0.0 0.0 5.3 1 .4 6.5 3.7 3.3 5.4 21.5 2.1 8.3 5.0 4.3

TRANSITIONAL

% of transitional rnierants

I6

2.0

2o 26 12

2.4

3.2

1.5

9 I5 34

1.1

I .8

4.1

1.5

12

33 29 25 93 32 7 22

I55 222

44 I4 820

4.0 3.5 3.0 11.3 3.9 0.8 2.7 I 8.9 27.1 5.4 1 .7 100.0 o = 6.74 c.v.* = 128.1

* T h e total of 2,941 responses are by heads-of-household born outside Accra.

Sarnple sizes for each tract appear in the "tract total" column in Table I .

a

268 of the 290 enumeration areas delineated by the 1960 Census of Population) lvere arbitrarily selected, and rvithin these a random sample of enunieratiori areas was drawn until approximately 10 percent of the total population of each statistical area were represented. In some instances, meeting this criterion required the selection of several tracts in a given statistical area because tract sizes varied greatly. In all, nineteen tracts were selected and 4,646 responses from male and female household heads were generated. Stratifying the sample in this manner insured a fairly representative cross section of the urban population in the inner city, in the intermediate distance, and in peripheral areas (Fig. 1 ) . A variety of socioeconomic neighborhood types were also encompassed, including inner-city slums (in parts of Jamestown and Ussher Town), workingmen's communities (northern Adabraka and Mamprobi), middle-class housing estates (Kaneshie), a privately developed middle-class area (Kokomlemle), and low-

5

MIGRANTS I N ACCRA

(1-5

YEARS)

LONG TERM

(6-19

yoo f total mlgrants and local born in tracta

YEARS)

PERMANENT

(20

YEARS OR MORE)

yoof

Number of responses

% of long-term migrants

total mlgrants and local born in tracta

yo of total Number of responses

% of permanent miarants

migrants and local born in tracta

"he coefficient of variation is derived by dividing the standard deviation by the mean and multiplying by loo. The larger the coefficient of variation, the greater the spatial concentration.

status settlements on the periphery (Sabon Zongo, New Toivn, Nima, and Labadi). I n Table I some preliminary facets of the sanlple are manifest. First, the overall importance of migration is apparent: 63 percent of the persons interviewed were born outside Accra. Second, the overall distribution of ~nigrantsand indigenes is not a simple one. In most tracts the passage of time has produced a complex mixture of migrants and indigenes that defies facile generalization. T h e exceptions are I,abacli, which is dominated by local-born Ga, and Nima, New Town, and Kokornlemle, which are inhabited overwhelmingly by migrants. Since we assert that the length of urban experience is an important dimension in the migrant allocation process, the target population of 2,94 1 migrants was partitioned into four subsets neophytes (less than one year's residence in Accra), transitional (one to five years), long-term (six to nineteen years), and

6

THE GEOGRAPHICAL REVIEW

permanent (twent? years or more) - in order to observe their distributional patterns (Table 11). T h e decision to adopt this fourfold typology ivas largely subjective, but it was guided by suggestions gleaned from various published sources and from field observations. In establishing the cutoff point for neophytes we wished to iclentifj initial residential preferences. T h e break at five years corresponds to the approximate length of an apprenticeship, with its semi-independent financial status, and also covers the period of maximum residential mobility for skilled migrants. In establishing the ttvent?-year figure rve hoped to discriminate between those who had long experience else~vhere(long-term migrants) and those rvho spent the majority of their adult years in Accra (permanent migrants). T h e first and most important inference to be dralvn from these data is summarized by the diminution of'the coefficient of variation over time, ~vhichshows that the degree of spatial concentration is greatest among rierv migrants and fBlls off regularly to a minimum among permanent ~nigrants.'It is also instructive to observe that the latter are only a quarter as numerous as the long-term migrants. This reflects, among other things, the sudden increase in net migration to Accra that followed World War 11 and the relative recency of' the demand for housing on the urban periphery. Table I1 also shows that most newcomers to Accra are found in peripheral rather than central-city locations; only 16.5 percent of the neoph?tes were enumerated in central Accra (Tracts 1 through 1 o), ~ v h e r e the average proportion of' neophytes per tract was only 2 percent. This peripheral orientation of neophytes can also be seen in the tract proportions in Table 11. T h e representation of both neophyte and transitional migrants tends to increase as the distance fi-om the indigenous core increases, the principal anomalies being Tracts 14 and 19, which are incligenous strongholds. In fact, five of the six tracts with the highest combined proportions of recent migrants lie more than two miles fi-om downtown. Of these, Tract 17 in Nerv Town, which alone accounts for more than a quarter of all migrants with five or fewer years of urban residence, is a prime example of the rate at which The size of the population tracts may hale some effect on the percentage concentration of the migrant groups. However, since the same nineteen tracts were used in all cases and since few tracts emerged as important concentrations for all four migrant types, the effect of tract population ma) not he as strong a determinant in migrant allocation as certain socioecono~nicvariables used later in this paper.

M I G R A N T S I N ACCRA

I I

7

ACCRA EXPANSION OF RESIDENTIAL AREAS ,1900-1969

low-income communities on the periphery are growing. Quite a different locational pattern emerges when the proportional representation of long-term and of permanent migrants is examined: they are less concentrated than the more recent migrants, and large proportions are found in central Accra and at intermediate locations as well as on the periphery. Without delving further into the detail contained in Tables I and 11, it is clear that the areal pattern of recent in-migration tends to be markedly concentrated and generally decentralized. Thus Accra seems to be growing more by what Mabogunje refers to as "spatial expansion" than by "fission" in the central city." In this process in-migration appears to be the dominant source of growth; secondary inputs come fi-om natural increase and intraurban relocation (although no estimates have been made o f their relative contributions).

:' ..\kin hfaboguri~e:T h e Growth of Residential Distric-ts in Ibadan, Geogr-.RPZI.,Vo1. 1962, pp. 36-77.

52.

8

THE GEOGRAPHICAL REVIEW

FIG. 3 - Popular responses to con~n~ercial invasion on the fringe of the rapidly expanding CBD. T h e shanties that line ever) available space in what were formerly interior court)ards house people a,ho are closely tied to the economy of the core.

An indication of the urban sprawl is given by Figure 2 , which depicts the rapidity of' areal expansion and infilling in many, if not all, outlying areas since MTorldMTar11. Labadi, Airport Residential Area, Nima-Maamobi, Kotobabi, New Town, Kokornlemle, Kaneshie North, and Bubiashi grew tremelldously during the 1960's and now constitute integral parts ofthe built-up portions of the city. Unfortunatel) , not all of' this growth has been controlled. In the postwar generation Accra grew by leaps and bounds, averaging more than l o percent per year. Nima, for example, was depicted on the 1949 Gold Coast Survey map of Accra as only a name on a virtually uninhabited exurban area.* On the 1957 edition of the same series5 the area was packed with an unordered mass of irregularly shaped buildings without access roads or signs of urban infrastructure, and in 1958 it was officially designated as a ".4ccra," 1 :6,250 series (Gold Coast Survey Dept., Accra, 1949)

".4ccra," 1 :6,250 series (Ghana Surve) Dept., Accra, 1957).

MIGRANTS I N ACCRA

9

FIG. 4 - Subdivision of the shanties shown in Figure 3. In the long run this highly dilersified, multiethnic component in the indigenous core Inay produce a degree of mutual undet-standing and tolerance that outweighs the enlironrnental deficiencies.

slum in urgent need of remedial treatment."The fact that the bulk of Nima (and New Town) lay beyond the municipal boundary at the onset of the post-World War 11 migration boom allowed illegal occupance, construction, and subdivision to proceed with impunity, much as they did in Lagos and in other African c i t i e ~ . ~ T h e phenomenon of urban sprawl is being studied and directed by several African governrrlents for the first time;%orvever, there is fi "Accra: A Plan for the Town" (Xlinistry of Housing, l-o\\ 11 and Chuntl-y Planning Division, Accra, 1958), p. j n . ' P. 0. Sada: Residential Land Use in Lagos: .4n 1riclui1-yInto the Relevance of Traditional 5lodels,~4fi.zcnnL'rban Sottr, Vol. 7, N o . 1, I '3 72, pp. 3-25, For f ~ ~ r t hdiscussion er of the effect of municipal boundaries on the location of peripheral sha~ltyto\vnssee Kichard Brand: 0 1 1 the Problems and Promise of Lo\\--Income Areas i l l African Cities, in Corlte~iiporary Africa: Geographical Interpretations of a Continent in Change (edited by C:. GI-egory Knight and James L. Ne\vrnan; Prentice-Hall, Inc., Englewood (:lift's, N.J., in press). See, tor example, "Urban Land Policies and Land I'se C:ontrol Xfeasures in Africa" (a paper presented to the Interregional Serllinar on Urban Land Policies and Land I'se Clontrol Measures, Xladrid, No\.1-13, 1971).

10

T H E GEOGRAPHICAL REVIEW

as yet little geographical research on urban growth processes even though land use and development is an important urban policy issue. Iliith respect to the central city, a region more frequently studied by geographers, the finding that the ten tracts in central Accra accounted for only 16.; percent of the neophytes in the sample takes on more significance when this figure is compared to the region's share of transitional migrants (25.1 percent), longterm migrants (32.10 percent), and permanent migrants (54.4 percent). By inference from the survey data and from other available studies we may safely say that at a previous stage of urban development the central city was more attractive (or available) to newcomers than it is today. Moreover, the diminishing rate of' in-migration to central Accra suggested by our data coincides with the relocation of upwardly mobile persons, some of whom build houses for rent in low-income suburbsg Rather than creating a vacuum for newcomers to fill, however, the available living space in the core has actually diminished, particularly on the fringe of the CBD, where commercial invasion is proceeding most avariciously. Figures 3 and 4 depict living conditions in Burgess's area of' transition,1° where the unregulated competition for space has driven residents to upper stories of buildings and to interior courtyards. From the street only a commercial ficade is visible, but a jumble of'shacks, shanties, and other extralegal subdivisions can be seen from above. In the absence of' enforced zoning ordinances and building codes, mixed and conflicting land uses have come into being and residential crowding has become severe. The difficulties of' providing municipal services to the uncontrolled settlements that abound in central Accra are great, for most are without running water, adequate ventilation, sewerage, electricity, and a degree of' privacy satisfactory to the residents. Tudu (Tract 8), in southern Adabraka, is another example of' the diminishing availability of rental space in the core to new migrants. As the predecessor of' Sabon Zongo, New Town, Marion Kilson: Continuity and Change in the Ga Residential System, Ghana Jaunt. ojSociolog~,Vol. 3, 1967, pp. 81-97; and Fred T. Sai, "Preface," in Survey into Housing and Household Conditions in LaRone, Jamestotvn and Nima (by N. 0. Addo and G. R.I. K. Kpedekpo; Ghana Medical School, Community Health Survey Reports, Project No. I , Accra, 1967). '" Robert E. Park, Ernest W. Burgess, and Roderick D. McKenzie: The City (The University of Chicago Press, Chicago, 1gz5),p. 51.

MIGRANTS I N ACCRA

11

Nima-Maamobi, and New Fadama (the present-day strangers' quarters), T u d u has long been a popular destination for Nigerians and "northerners" from up-country and beyond. Our data suggest that Tudu's drawing power is diminishing, since the proportions of long-term and of permanent migrants are considerably larger than the proportion of recent migrants, both in absolute terms and in terms of intratract proportions (Table 11).The explanation for this phenomenon in Tudu is similar to that given for Tract 6: both have experienced the pressure of an expanding business center in need of Lebensraum and both have seen rents bid u p beyond the economic reach of many newcomers, particularly the poorly skilled northerners who lack urban experience and financial security. In 1969, when the mean monthly wage for unskilled and semi-skilled workers in Accra averaged the equivalent of $18 to $ 2 0 , room rents ran as high as $6 or $7 in central Accra, but comparable rooms in Sabon Zongo, Nima, and New Town, on the periphery, cost only from $2. j o to $4.00 a month. Although our data are inadequate for a rigorous analysis and exposition of the intraurban migration histories of the respondents, it was evident from the interviews that a number of migrants enumerated in outlying communities, particularly those in the transitional, long-term, and permanent categories, had moved from locations in central Accra. Our data appear to be consistent with this qualitative assessment, for they indicate a phase shift in the migrant allocation pattern, suggesting that the migration stream into central Accra, which remained strong for a generation after World War 11, has diminished." The economic forces unleashed by city growth have shifted the locus of in-migrant concentration to more distant areas. In broad, comparative terms these findings are basically similar to the generalized growth patterns observed by Sada in Lagos, Tiivari in Nairobi, and Harrison in Tripoli, but they contrast with the findings of Knoop in Kinshasa and those of Abu-1,ughod in Cairo," where the traditional migrant destinations in the central l ' This conclusion was recently corroborated by an analysis of unpublished data from the 1370 census, which revealed that only about 5 percent o f t h e 250,000 people added to the municipal population during the intercensal decade were found in tracts within the region w-e have defined as central Accra. A preliminary inspection of our 1973 survey data on intraurban migration indicates that the vast majority- of the population growth occurred in areas which \rere peripheral and exurban in 1960. l 2 Sada, Residential Land Cse (see footnote 7 above], p. 23; R. C. Tiwari: Some Aspects of the Social Geography of Nairobi, Kenya, .!/ricnr~ CTrbnn note^, Vol. 7, So. I , 1972, pp.

12

T H E GEOGRAPHICAL REVIEW

city continue to attract large proportions of the newcomers who swell the urban population. However, much additional empirical work needs to be done before more accurate and useful comparisons can be drawn.

T o identify the variables that structure the environmerital niches of our four classes of migrants, we shall test the proposition that associated with concentrations of one or another of these migrant classes are different sets of social indicators that are both causes and effects of the presence of such persons. Th7eexpect that neighborhoods in which Europeans, Asians, elite Africans, and skilled interregional migrants from southern Ghana are found will display heterogenetic social indicators, since traditional sociocultural values are giving way to more cosmopolitan behavioral patterns. O n the other hand, tracts dominated by long-distance migrants from rural areas are more likely to have a tradition-bound orthogenetic character, wherein long-established mores typical o f a "folk society" tend to prevail.13 At this stage we cannot estimate the variations in social indicators that correspond to length of urban experience. For each of the four migrant length-of-experience groups, the nineteen sample tracts were divided into three levels of migrant concentration: primary (high), secondary (medium), and tertiary (low). Primary tracts were defined as having a migrant concentration exceeding one positive standard deviation from the mean; secondary tracts ranged u p to one positive deviation; and 36-61; and Robert S. Harrison: Sligrants in the City of Tripoli, Libya, Geogr. Rer~.,Vol. 57, 1967, p p 397-425 T h e relatively well balanced age and sex structures of' the peripheral bidonr~ill~sof Kinshasa indicate that these communities have grown mainly through intraurban migration frorn the core and through natural increase (see Henri Knoop: The Sex Ratio of an African Squatter Settlement: A n Exercise in Hypothesis-Building, Ajricctn Lirbarl,Votes, Vol. 6, No. 1 , 1971, pp. 19-24). Thecontinuing flood of'poor in-migrants to the central-city slu~nsof' Cairo is documented in Janet Abu-Lughod: Xligrant Adjustment to City Life: The Egyptian (:ase,Amer. Jourrz. of Sociology. Vol. 67, 1961-1962, pp. 22-32. l 3 Our probabilistic phrasing stems from our awareness that generalizations about spatial units which contain an average of' about 1 ,000 persons cannot hope to be precise. Furthermore, f o l l o ~ i n gHollnsteiner's interpretations in her study of a low-status area in Manila, we believe that although many attributes of rural or folk society prevail in long-distance migrant subareas, the stereotype of the "folk in the city" leaves much to be desi'red as a framework for analysis (see Mary R. Hollnsteiner: Becoming an Urbanite: T h e Neighbourhood as a Learning Environment, in The City as a Centre of Change in Asia [edited by D. J. Dwyer; Hong Kong University Press, Hong Kong, 19721, pp. 29-40).

MIGRANTS IN ACCRA

V A R I A B L E NUMBER AND N A M E

Migration status I Local borna 2

Foreign African males

3 Non-Africans Population structure 4 Child-woman ratio

5 Adult sex ratio Education 6 Educated adult males

7 Unschooled children 8 Educated adult females Occupation g Working-age males Io

Unemployment

I I

Nonemployed females

Spatial characteristics I 2 Population density I 3 Distance from CBD

13

DESCRIPTION O F VARIABLE

Ratio of persons born in Accra Capital District to

total population

Ratio of males born in other African countries to all

males Number of persons born outside continental Africa Ratio of children less than 5 years old to women between 15 and 44 years old Number of males more than 15 years old per I O O females more than 15 years old Ratio of males more than 15 years old with some schooling to all males more than 1 5 years old Ratio of children between 6 and I 4 years old with no schooling to total population Ratio of females more than 15 years old presently in school to all females more than 1 5 years old Ratio of males between 1 5 and 44 years old to all males Ratio of unemployed males more than 1 5 years old to all males lnore than 15 years old Ratio of economically inactive females more than 15 years old to all females more than 15 years old Gross residential density per acre Straight-line distance from the General Post Office (in Ussher Town near the CBD)

Source: "1960 Population Census of Ghana, Special Report 'A': Statistics of Towns with ~ o , o o oPopulation or More" (Census Office, Accra, 1962), pp. 196-216. a Includes those born in the Accra Municipal Council (or, simply, Accra) as well as those born in what was then known as the Tema Development Council, Tema Rural and Tema Manhean. Only 27,000 (6.4 percent) of the 415,000 people in the Accra Capital District were enumerated outside the Accra Municipal Council at the time of the 1960 census.

tertiary tracts had below average numbers of respondents.14 Census data on migration status, population structure, education, occupation, and spatial characteristics were also tabulated for the nineteen tracts (Table 111). T h e first of the three migration variables assesses the relative importance of residents born in Accra in the population mix of a tract. We expect that this variable will be usef~ilin discriminating between potential and nonpotential areas of migrant attraction, under the assumption that tracts with a high proportion of migrants have strong information fields \vhich link origin nodes with potential reception sites. T h e relationship is not universal, however, since a number of migrants - particularly those with urban histories that predate the saturation of residential " The extreme skewness of the distributions of responses (Table 11) was reduced by a log transformation, X , = log,,(X, + 1).

T H E GEOGRAPHICAL REVIEW

14

space in central Accra - are found not in tracts dominated by migrants but in heavily indigenous Ga neighborhoods near the central markets. If this proposition is valid, the effect of the second variable, foreign-born Africans, should complement the first. ?'he third variable, non-African migrants, is used as a surrogate for TABLEIV- MULTIVARIATE ANALYSIS O F VARIANCE FOR

THE FOURSAMPLE XilTH DIFFERENT LF,NGTHS O F URBANEXPERIENCE

POPULATIONS c

LENGTH O F U R B A N EXPERIENCE

Less than I year I to 5 years 6 to 19 years 2 0 years or more a

TEST O F EO_UALITY O F DISPERSIONS

.\I

2086.8 1984.3 2644.9 1944.4

F

ASSOCIATED DEGREES O F FREEDOM

1 2 . 6 0 ~ I l o and 6.34" 91 and 8.64" 91 and 7.83" 91 and

182 881 881 881

TEST O F EQUALITY O F CENTROIDS q2 F

O5 .08 .12 .06

'95 .92 .88 .94

-

ASSOCIATED DEGREES O F FREEDOM

':

9.39" 6.84"

8 8 8 8

and and and and

26 26 26 26

Significant at the . o o ~level of confidence.

residential status and neighborhood environmental quality, since Europeans (and to a slightly lesser extent Levantines and Indians) usually pay a high price for accommodations in Accra. In terms of population structure, fertility should be highest in stable indigenous tracts and lowest in transient migrant ones; the opposite should pertain to the adult sex ratio. These two variables are expected to discriminate between tracts that house new migrants and those with a larger proportion of migrants with longer histories of residence in Accra. This distinction reflects the general aversion of tradition-bound Africans to the practice of fa~nilyplanning15 and the normally lower fertility ratio in migrant tracts where single males predominate. Education is crucial in discriminating bet\veen orthogenetic and heterogenetic value orientations. T h e close relationship between a Western form of education and income in developing countries implies that with education the individual not only has a greater potential for self-advancement but also experiences a considerable rise in his aspirations and an increase in his range of locational alternatives.lfi Variable 8, educated \voInen, is thought to be a ' j I n a conversation in Ibadan in August, 1973,Jo11n and Patricia Caldwell reported to us that in their detailed sample survey of Ibadan women with five or fewer children only a miniscule percentage tliti not want any more children and were taking measures to prevent conception. l 6 Lawrence A. Brown and Eric <;. Moore: The Intra-urban Migration Process: '4 Perspective, Geografirka Annaler, Ser. B, Human Geog., Vol. gnB, 1970, pp. 1-13.

MIGRANTS I N ACCRA

1

5

particularly- sensitive indicator of relative position on the traditional-technocratic continuum, one that will discriminate between orthogenetic htuslim communities and areas that are more responsive to the potential of educated women in modern society. If migrant tracts differ in terms of demographic structure and education, occupational contrasts are inevitable. Variables g through 11 are thus pertinent. Finally, population density and linear distance from the CBD are included as general measures of the intensity of residential space occupance and relative location vis-a-vis the downtown commercial hub. Before identifying significant variables that discriminate among tracts with high, medium, and low concentrations for each of the four migrant length-of-experience groups, the partitioned data sets were subjected to a ~nultivariate analysis of variance (MANO\.'A) to test the strength of the newly divided data set." Table IV shows that in all four cases the partitioning of the nineteen tracts was significant at more than the .oi level of confidence. T o determine the socioeconomic character of the

" Paul K. Lohnes (Test Space anti Discriminant Space Classification Models anti Kelateti Significance Tests, L'dztcntio~~nlnnd Ps?chol. .2lpa.\lr1erne11t. \'ol. 2 1, 1 96 I , pp. 559-574) advocates performing MANOVA to test the stt-ength of the pat-titioning of the subpopulations before attempting tliscritninant anal)sis. BrieRv, RIANOVA tests for equalit) of the tht-ee centroitis b, computing an F-ratio approximation of LVilk's lambda, A, and of the tlispersions b) calculating the F-ratio equivalent of Box's '21. For the three-subgt-oup problems, testing the latter involves computing three vat-iance-covariat~ce matrixes, L.,. L.?, and L., anti the pooled variance-covariat~cematt-ix, L.,. Theit- I-elateti tleterminantal values, / D l 1, D , ID,/, anti I D , are used in developing Box's statistic:

1,

w h e r e S j = san~plesize ofthe i"' subgroup, and

Depending on certain assumptions, an appropriateF-I-atiois computeti for a given.14-value. LVilks's lambda is defined as .1 = CI. I T . where 1 1 . anti T are the tleterminants of the matrix of deviation sums of squares and cross products of the subjects from their group means and grantl mean respectivel!. Rao has developed an F-ratio approximation of. LVilks's lambda in terms of '1,the number of variables and subgt-oups anti the sire of the sample population. For detailed tiiscussions of M,4NO\'A see S. S. LVilks: Certain Grnet-alizations in the Anal\sis of \'at-iance,Hlorrtc>trlkn, Vol. 24, 1932, pp. 47 1-494; G. Roy: A Genet-al Distt-ibution 'Theot-s fot- a Class of Likelihooti Ct-itct-ia,ibzd., \'ol. 36, 1949, pp. 3 17-346; anti C:. R. Rao: Atlvancetl Statistical Rlethotis in Biomett-ic Research (John LViley and Sons, Inc., New I'ork, 19521, pp. 261-2 71

1

16

THE GEOGRAPHICAL REVIEW

neighborhoods attractive to the various classes of migrants the data sets used in MANOVA were standardized and used in a stepwise discriminant analysis.18 This procedure isolated the significant variables that structure the neighborhoods where particular concentrations of migrants are found. NEOPHYTE RIIGRANT CONCENTRATIONS

T h e distances among the three group centroids and their associated members in discriminant space (Fig. j) gives a visual impression of the strong differences among the three concentration levels of neophytes. Statistically, the high canonical correlation (.94)on the first linear discriminant function, which also has a high proportion (.838) of discriminant power and a low R " - v a l ~ e , ' ~ indicates that intraclass variability is low and that interclass differences are relatively high. Of the thirteen variables used in the analysis only two, population density (XI,) and adult sex ratio (X,), had F-ratios that were significant at more than the .05 level of confidence. T h e computed first linear discriminant function2' is

This function indicates that high concentrations of new migrants are directly related to areas with large excesses of adult males over adult females but negatively associated with population density. These two variables indicate clearly the rapid growth of the lowand medium-density areas on the periphery. T h e relationship has also been influenced by the small number of new migrants who locate initially in central Accra, a recent development in no small --

For a general application of discriminant anal)sis in geograph) see Leslie J. King: Discriminatory Analysis of UI-ban GI-owth Patterns in Ontario and Quebec, 1951-1961, Annak A J J ~of . ,4mcr. Grogr~.,\'(>I. 57, 1967, pp. 568-578; and E~nilioCasetti: Classificatory anti Regional i\nalysis b) Discriminant Iterations, Tech. Rept. No. 12, Office of Naval Research Task No. 389-135, Northwestern University, Evanston, Ill., n.d. l 9 The Ro-value of intraclass variabilit) to total variability indicates the strength of the classification. As R" approaches zero the discrimination becomes stronger. This ratio is easily computed for a set of m eigenvalues, A,(i = 1, . . . , mi: Is

R"=I/

[I - E X] ~ ;

o
See LeslieJ. King: Statistical Analysis in Geography (Prentice-Hall. Inc., Englewood Cliffs, N.J., 1969), pp. 214-215. 2 0 Because of the large proportion of the dispersion accounted for by the first linear discriminant function, the loadings of the variables on the second function are not discussed. Unless otherwise stated, we have followed this practice in the subsequent discriminations.

'7

M I G R A N T S I N ACCRA

MIGRANT GROUPS: PARTITIONING OF SAMPLE TRACTS ON TWO DISCRIMINANT FUNCTIONS

4

NEOPHYTES

-0 ............

: Second

?...................

I

;*. .+ 4

' ; Z

'

2I First l~neard~sci~m~nant tuoct~on Second ,i 4,

2.

LONG-TERM

n K

F~rstlhnaar dlscr~m~nant functloo

,,

So,nd

94

838

7 7

1 000

99 95

Y

80 1 00

I

'

1

'

II I

I

!

A

+

A

I

?

Cumulat~vc Canonical proponloo of correiallon d~spers~oo

2

'

932 1 000

..

Canon~cal CumU'a11v8 correlat~on proponlon Of d~sperslon

5' 4

11 1

..

Zl

I

98 7 8

t

,

b

5

rr

.*= /

0

5

Cumulatlw Canonical piopomon of cofrelatlon dispersion

First llnear dlscrlmlnant functlon

I

, '

AA

-- -

-

TRANSITIONAL

I

I

p I.

PERMANENT

t

.....

i

.

.I

J.

i

*..L ?

.

........................

I

.o-

.....

*. .

.................................

First llnear d~scr~m~nant funct~on Second ,, J,

97 85

1 000

SECOND LINEAR DISCRIMINANT FUNCTION

ILGEOGR

REV

*Centro~ds JAN

.

Primary

Secondary

1974

A

Tert~ary

I

FIG.5

way explained by the encroachment of the CBD and the general scarcity of inexpensive rooms for rent in the core. T h e adult sex ratio varies considerably when it is partitioned into primary, secondary, and tertiary concentrations (Table V). It is extremel) high when migrant concentrations are greatest, but falls off by about two-thirds in tracts with only tertiary migrant levels. T h e opposite is true for population density. O n the periphery, where neophytes are most heavily concentrated, densities are relatively low (though crowding is considerable); densities increase in tracts with secondary concentrations of migrants and peak in tracts where migrants are fewest.

18

THE GEOGRAPHICAL R E V I E ~ I ' TRANSITIONAL MIGRANT CONCENTRATIONS

For transitional migrants, Figure 5 also demonstrates the strong partitioning of the nineteen sample tracts into high, medium, and low concentrations, concentration being defined as the proportion

CONCENTRATION TYPES

MIGRANT GROUPS AND VARIABLES

Primary

Secondary

Tertiary

Ro

~Veophytemigrants X g Adult sex ratio X12 Population density Transitional migrants X I , Nonemployed females X12 Population density Long-term migrants Xl Local born X2 Foreign African males X4 Child-woman ratio Xg Educated adult males Xg Working-age males Xu Nonemployed females X I Z Population density Permanent migrants Xz Foreign African males Xc Child-woman ratio X12 Population density

* T h e means are computed from normalized data sets; hence the original variability in units of measurement has been removed and individual mean values are comparable. of a migrant group's sample population in a tract. T h e high canonical correlation (.98), which explains 93.2 percent of the dispersion on the first linear function, implies that analysis can be focused on this function with minimum loss of information. Only two variables, nonemployed females (X,,) and population densitv (XI,) are significant discriminators at more than the .oc; level of confidence. T h e derived function is

This equation implies that the concentration of transitional migrants (one to five years in Accra) is directly related to the number of nonemployed females but, like the first linear discriminant function, bears some inverse relationship to population density. T h e distribution of means (Table V) indicates that the relationship between population density (XI,) and concentration of transitional migrants is an inverted J-shaped curve. T h e densities are lowest on the periphery, increase in the intermediate-distance zone, and then decline again in the core.

I9

MIGRANTS I N ACCRA

Since the concentration of transitional migrants is highest in Tract 16 (Kokomlemle), follo~vedby Tract g in Adabraka, both of which attract skilled migrants from southern Ghana, the large number of nonemployed females may be indicative of rising socioeconomic status rather than of hidden unemployment, which ~vouldbe more likely in lolver-class neighborhoods. In general terms, transitional migrants seem to have enlarged their action space beyond the regions that tend to draw most newcomers. LONG-TERM M I G R A N T CONCENTRATIONS The discriminant analvsis showed that seven significant variables were involved in the discrimination of long-term migrant concentrations. The low R0-value(Table V) and the high canonical correlation (.99) on the first linear discriminant function (Fig. 5 ) are indicative of the strong partitioning among the three concentration groups. The function was computed as follows:

Interestingly, although these migrants have resided in Accra for a minimum of six years they still are not found in strongly local-born tracts. This also explains the negative relationship with the child-woman ratio (X,), which is highest in traditional Ga areas like Jarnestown, Ussher Town, and Labadi, and the positive association with foreign African male migrants (X2).The negative coefficient of educated adult males (Xs) is suggestive of the fact that many migrants do not increase their educational status with the passage of time. However, individuals who rise in status as a result of educational achievement or acquired wealth often tend to elevate their residential aspiration levels as well. Frequently, but not always, the older, more traditional environs are cast from their pool of' possible relocation sites. Subsequent evidence from low-status areas in Kumasi21 and preliminary glimpses at the raw data from our 1973 survey of conlparable areas in Accra22point to a recurrence of the pattern of selective relocation preferences. Although it is too early to say anything definitive about the actual pattern of intraurban migration and site selection among the -

-

Richard R. Brand: Migration and Residential Site Selection in Five Low Income Communities in Kumasi (Ghana),Afriran Crban1Votrs, Vol. 7, No. I . 1972,pp. 73-94. 2 2 Ttie authors spent July and August of 1973 directing a sample survey in Accra, aimed at understanding the causes and correlates of residential mobility into and out of low-status com~nunities. ?'

20

T H E G E O G R A P H I C A L REl'IETV

lowest status groups in Accra, our field experiences suggest that such a process of upward spatial mobility is indeed at work. If o u r impressions are substantiated by the data we collected in the summer of 1973, the implications would be profound. I n that event, the role of so-called slums and shantytowns would haxe to be reinterpreted in light of their demonstrated performance as agents of sociocultural transformation. By the same token, however, the out-migration of the upwardly mobile from source areas like Bukom Square, in the tradition-bound heart of Ussher Town, is not in itself a panacea, for it leaves behind a weakened residual population that remains veritably homeostatic. More will be said about these phenomena in subsequent reports. PERMANENT MIGRANT CONCENTRATIONS

Figure 5 again shows that the discrimination among the three concentration levels of permanent migrants is strong. T h e analysis revealed three significant variables:

This function gives evidence of a distinction between permanent migrants and all other classes in that it shows that many of the former are foreign Afi-ican males (X2)who reside in high-density areas (XI,) where family size is large (X4). Identified here are the ethnic enclaves in the central city that retain their character even though the city is changing around them. Put in another way, this suggests support for the proposition that group cohesiveness and spatial clustering among older migrants increase as the sociocultural distance bridged by foreign African migrants widens. Schildkrout has shown, for example, that aliens in Kumasi face major obstacles of political, cultural, and economic i n t e g r a t i ~ n . ~ ~ Similar conclusions were reached by Peil in her interpretation of the circumstances leading u p to the expulsion of ~ f i - i c a naliens from Ghana in November, 1969 (a little more than six months after even as early as 1950 Busia .~~ our survey was c o i ~ c l u d e d )Indeed, observed the poorly integrated status of strangers in Ghanaian cities.25 23 Enid Schildkrout: Strangers a n d Local Government in Kumasi, Journ. of Modrrn African Studirs, Vol. 8 , '970, p p 25 1-269. 2 4 Margaret Peil: T h e Expulsion of West Afric-an Aliens,ibid.. Vol. 9 , 197 I , p p 205-229. 2 5 K. A. Husia: Report o n a Social Survey of Sekondi-Takoradi (Crown Agents, London, 1 9 5 0 ) ~p p 67-68.

MIGRANTS IN ACCRA

21

Overall, inferences from the analysis of the socioeconomic characteristics of migrant destination nodes suggest that in Accra -and probably in other capital cities in Africa -the "melting pot" process, which implies the gradual evolution of a highly variegated and richer urban culture from the interaction of the various groups within it, has not completely superseded cultural pluralism. I n Accra both neophytes and permanent migrants tend to be spatially nucleated, whereas the two intermediate experience categories are more diffuse in the urban area.

If a set of a priori variables adequately discriminates among groups of tracts with varying degrees of migrant concentration, then any new tracts can be assigned to existing groups based on their score-profiles on the variables involved. Although it cannot be assumed that such newly allocated tracts will invariably have the same migrant concentrations as the sample tracts in the same category, the procedure enables us to isolate potential concentration areas for migrants with particular lengths of urban experience. T h e development of such an allocation procedure has obvious implications for planning in rapidly expanding cities like Accra. O u r methodology could be used, for example, to identify compatible neighborhoods for families required to relocate in the course of renewal project^.^" In assigning the remaining 249 tracts in Accraz7 our strategy again involves linear discriminant analysis. I n this case discriminant functions were computed for the nineteen sample tracts on each of the four length-of-experience categories at three '"I'here is a long history of urban renewal projects in Accra, beginning with the Improvement Scheme of GovernorJ. P. Roger in I 908. In almost every case the government made provision for resettlement, usually- in publicly developed housing estates. At the present time, however, a proposal is being dralvn up in the Ministry of Works and Housing for massive redevelopment of the low-status area of Nima-Xlaa~nobi(Daily Graphir [Accra], Sept. 4, 1973. p. 5), but without provision for government-assisted resettlement. T h e sixty thousand persons who will eventually- be affected by this scheme are presumably to find new residences elsewhere in Accra on their owrn. If this myopic scheme is inipleniented, our tnethodology could be used by community development organizations, like Operation Help Nima, to provide interested families with a list of compatible neighborhoods within which they might search for newr housing. '' Recall our initial elimination of 2 predominantly rural statistical areas and their 2 2 enumeration areas from the universe of 17 statistical areas and 290 enumeration areas that constitute the Accra Municipal Council. Thus 290 minus 22, minus our19 sample tracts, leaves 249 unallocated tracts.

22

THE GEOGRAPHICAL REVIEW

levels of migrant concentration. T h e functions were then used to calculate the probabilities that an unassigned tract belongs to one of the subclasses within the sample. Each tract was then assigned to the subclass with the highest probability of correct ~ l a s s i f i c a t i o n . ~ ~ T h e resulting spatial distribution was mapped for each length-of-experience category (Figs. 6 , 7 , 8 , and 9). T h e pattern of simulated neophyte allocation shown in Figure 6 is generally consistent with our earlier observations on the attractiveness of peripheral locations. Although the pattern is not a simple one, tracts with primary concentrations of new migrants form distinct clusters in the sprawling, low-status settlements of T o reduce the probabilit). of misclassification, which is possible when the allocation of a large number of new observations is based on a small sample, new discriminant functions were computed usingall 268 tracts. For a discussion of such classification problems see Leslie J. King: Discriminant Analysis: A Review of Recent Theoretical Contributions and Applications, Econ. Geogr., Vol. 46, 1970, pp. 367-378; reference on pp. 373-375.

MIGRANTS IN ACCRA

z2 3

Kotobabi, New Town, and Nima-Maamobi to the north of central Accra, to the east in the area between Osu and Labadi, and even farther north along the Nsawam Road on the rapidly growing fringes of the traditional settlement at Achimota Junction. Also of interest is the secondary cluster of tracts in Sabon Zongo and Abossey Okai, two areas that have traditionally been attractive to strangers in the city. Similarly, the two tracts constituting the former Muslim squatter community called Fadama,29 on the floodplain of the Korle Lagoon (Fig. I ) , also register as primary concentrations. No such clustering is found for higher-class tracts with primary ' W e r e the drawbacks of using census data collected a decade before are manifest. The entire community and its economic base, a sizable stockpile of spare and junk automobile parts, was removed to a remote area called Kew Fadama, about five miles from the core. Unhappily, enumeration area data from the 1970 census are not scheduled for publication until late in I 974.

*4

T H E GEOGRAPHICAL RE\'IEW

concentrations of neophytes, except perhaps the t ~ v otracts found in Kaneshie and those in northern Adabraka and Kokomlemle. T h e former offer residential site amenities as part of the government housing-estate package, and the latter are privately developed areas of middle- and upper-middle-class housing. Interregional migrants from other areas in southern Ghana are disproportionately represented in these post-World War I1 communities. Four of the eight tracts with low proportions of neophytes that are found amid the primary concentration nodes in New Town and Nima-Maamobi are areas of primary concentration for transitional migrants (Fig. 7). I n general these tend to be the core areas of their respective settlements, and as such also appear as primary concentrations for long-term and permanent migrants. Indicative of the northerly direction of residential expansion in these areas is the finding that ferv ti-acts in Kotobabi or Maamobi, to the north of these core areas, have primary concentrations of

MIGRANTS I N ACCRA

25

migrants with long urban histories. In essence these are largely u~lcontrolledsettlements nithout most forms of basic infrastructure, such as roads and municipal transportation. Among the most ~ a l n a b l e~nunicipal senices pro\ided for these areas are the periodic sprajings of the rat-infested drainage ditches and the garbage-lined, dried-up stream beds xvith DDT. H o ~ z e \ e r , residence here minimires rental expenses and continues to offer bridgeheaders from all over TYest Africa, especially northerners nith rural life-styles, a ~ a l u a b l ecultural matrix that eases their initial adjustment difficulties in the metropolis. I11 comparative terms the peripheral, loxv-status comnlunities in Accra appear to be similar in form and function to those of Lagos (Mushin, Idjoro, Law anson, So~nolu,and Ajegunle), ~vhichwere studied by Sada.30 Environmental conditions are usually substandard, chiefly because "'P. 0. Satla: T h e Rural-Urban Fringe of Lagos: Population a n d Land Use, Xig~,rlnn

Joltrn.

of

Econ. nnci Socrnl S t t i d ~ ~V01. s,

12, 1970,

pp

225-241.

26

THE GEOGRAPHICAL REVIEW

the lumpen proletariat inhabiting such areas are unpoliticized and largely powerless masses of semiliterate long-distance migrants who have little or no ability (or inclination) to lobby for the extension of even the basic municipal infiastructure. A comparison of Figures 6 through g in terms of primary concentrations of migrants in the core areas of Jamestown and Ussher Town shows that the simulation procedure allocated moderate numbers of neophytes and relatively large proportions of permanent migrants, but only small numbers of long-term migrants and even fewer transitional ones. Much o f t h e attraction of the core for newcomers lies in the fact that they can stay there rent-free with kin, whereas they must fend for themselves fBrther out. However, the expanding CBD and the extreme scarcity of inexpensive rooms in the core represent recent changes. T h e 1960's saw a progressive saturation take place, and today the core houses primarily older migrants who own or rent property and conduct businesses there. These relatively settled groups function in turn as touchstones in the chain migration process. I n this way relatives and fellow townsmen arriving in Accra have preselected destination nodes, where their problems of initial adjustment are mitigated by reliance on older migrants. T h e small number of tracts with primary concentrations of transitional migrants suggests that as neophytes gain familiarity and financial flexibility they tend to relocate in less expensive quarters even if this means tradeoffs such as lower amenity levels and longer journeys to work.31Contributing to the lessened attractiveness o f t h e core have been the decline in the demand for laborers with the termination of surf-port operations in 1962 and the progressive decentralization of large-scale manufacturing firms as zoning regulations have come into force since independence. Although chain migration undoubtedly still operates in Africa, particularly in lower-class communities where initial migrant economic flexibility is severely limited and where orthogenetic value systems tend to perpetuate ethnic ties,32 our projections 3 1 A similar process was observed by Sada (Residential Land Use [see footnote 7 above], p , 236) among low-income migrants in Lagos. A recent survey of 180 renters in the four lowest-income communities in metropolitan Kumasi showed that 72 percent were renting from fellow tribesmen. .4lso, about one-half of the Asante and one-third of the non-asante reported kinsmen as the most important consideration influencing their decision to settle in their present locations, and

MIGRANTS I N ACCRA

27

TABLE VI-REGRESSIONANALYSIS O F FOREIGN AND INTERREGIONAL MIGRANTS* VARIABLE NUMBER

AND NAME

FOREIGN AFRICAN MIGRANTS

INTERREGIONAL MIGRANTS

Local born Child-woman

ratio

Adult sex ratio

Educated adult

males

Unschooled

children

Educated adult

females

Working-age

males

Unemployment

Nonemployed females

Population

density

Distance from

CBD

* The numbers in parentheses below each equation are t-values. All correlation coefficients are significant at the .005 level. imply that the hard facts of urban economics may be replacing ethnicitp as a major determinant o f t h e migrant allocation process. More refined analyses are now under way to test such an hypothesis, using more detailed survey data recently collected in Accra, but Peil's findings lend weight to our tentative conclusions: ,411 parts of the town [Accra] now show considerable ethnic heterogeneity, though the core area. . . is less mixed than areas farther from the centre of town. Ne\+~comers are forced to scatter throughout the city because of the difficult): of getting housing. The): are free to scatter because of the lo\+, level of ethnic prejudice.33

URBAN NICHES FOR FOREIGN AND INTERREGIONAL MIGRANTS T o determine the validity of the proposition that migrants from different areas of origin are attracted to certain types of urban niches and to isolate the variables associated with particular types about three-fifths of both groups ranked kinship as one of the most important factors. Other compelling evidence of chain migration was found in the migration itineraries of the in-migrants. (Brand, Migration and Residential Site Selection [see footnote 2 I above]). 3 3 Margaret Peil: The Ghanaian Factory Worker: Industrial Man in Africa (Cambridge University Press, Cambridge, England, 1 9 7 2 ) p~. I 59.

28

THE GEOGRAPHICAL REVIEW

of migrant concentration, simple regressions were calculated using census data from Table 111, with migrants from other regions in Ghana (Y,)and foreign African males (Y,)as dependent variables. T h e results are shown in Table VI. These sets of equations point u p several salient differences between the interregional and foreign African subsets of the overall migration stream. However, the variability involved in o u r analysis of all 268 tracts tends to make interpretations more difficult, as does the fact that northern Ghanaians are recorded by the census as interregional migrants even though, in terms o f t h e extent of their participation in the modern economy of Accra, they would be better classed as foreign Africans. Census data make such an allocation impossible, so a certain amount of distortion is inevitable. Nonetheless, the main concentrations of each group are found in tracts with relatively small proportions of local-born persons. As expected, the degree of exclusiveness suggested by the negative coefficients is stronger among foreign Africans who have bridged the widest cultural distance in coming to Accra. T h e child-woman ratio and the adult sex ratio discriminate nicely between the two migrant sets, foreign Africans tending to have a younger and predominantly single-male-dominated p o p ~ l a t i o n . ~ ~ T h e two groups are also differentiated along educational lines, particularly with respect to women. T h e lower level of educational attainment among foreign African migrants contrasts with that of the interregional migrants in spite o f t h e imprecisioil created by the inclusion of poorly educated northerners in the latter group. As remarked earlier, the average migrant from southern Ghana has probably had some prior urban exposure and tends to exhibit quite a different set of attributes than the northerner or foreigner, who is more likely to be of rural background. Proceeding from this one would expect to find many interregional migrants living in medium- and low-density communities and to find foreigners gravitating toward lower-cost rentals, where densities are higher; indeed, this is clearly shown by the regressions. What we did not expect was the finding that unemployment was as much a problem in interregional migrant tracts as it was among foreigners. In fact, our analysis produced almost identical relationships for each group. T h e explanation lies partly in the widespread incidence of 3 4 .4lthough not attempted here, it would be interesting to run the same analysis using just inner-city tracts and low-status tracts o n the periphery. In this way more could be learned about the role migrants play in differentiating between these two settlement types.

MIGRANTS IN ACCRA

29

and partly in the fact that large male unemployment in the percentages of the migrants from northern Ghana are unemployed and are defined in the census as interregional ~nigrants.~~

We hope that our analysis will be found useful on both substantive and methodological grounds. Discriminant analysis was used for the first time in African urban geographical research as an aid to classification and as a means of extending our findings to the total urban population. It allowed us to identify characteristics that differentiate among tracts with various concentrations of migrants with different levels of urban experience and to allocate the remaining urban tracts to established classes on the basis of their salient characteristics. Finally, regression analysis was used to point out differences among tracts occupied by migrants from different regions of origin. Substantively, we have shown that variations both in length of urban residence and in regional origin are associated with significant differences in the socioeconomic composition and the areal pattern of migrant communities. Central Accra continues to attract a small segment of the low-income migrant stream, but the great majority of newcomers are now settling in peripheral locations, where trade offs are made between amenity levels and rental costs. These locations are without doubt the most rapidly growing subareas in the city.3i Unfortunately, the official response to the problems posed by these settlements seems to be rooted in 35 T h e incidence of unemployment among adult males in 1960 is mapped by tract and discussed in Richard R. Brand: The Spatial Organization of Residential Areas in Accra, Ghana, with Particular Reference to .4spects of Modernization, Econ. Geogr., Vol. 48, 1972, pp. 284-298; reference on pp. 294-295. A spatial analysis of the I-esiduals from the regressions on unemployment would no doubt lead to a more refined understanding of the differential impact of unemployment on various types of communities, but this lies beyond the scope of our paper. " he unpublished 1970 census figures for et~umerationareas in Accra show that the overall population increased by 229,000 (from 388,000 in 1960 to 6 17.000 in 1970);and g i g new enumeration areas appeared in 1970, bringing the total to 569. Mapping the new tracts revealed that only fourteen (5 percent) of the new tracts appeared in central Accra (as defined in this article). In other words. approximately 95 percent of the intercensal population increase took place in what, in I 960 terms, would have been middle-distance and peripheral locations. A large but as yet undetermined proportion of this population growth has occurred in the lowest-status groups and is manifested in the massive spreading of uncontrolled settlements on and beyond the previous urban margins.

"

30

T H E GEOGRAPHICAL REVIEW

the conventional negativistic view of low-income areas,38 as evidenced in the plans for urban renewal in Nima-klaamobi. However, demolition in the midst of a severe housing shortage and unaccompanied by a popularly based rehousing scheme is patently ill advised. If the evidence from Nairobi's hlathare Valley"" is relevant in the context of areas like Ni~na-hlaamobiin Accra, administrators would d o well to investigate ineans of encouraging and channeling the progressive forces at work in these settlements, rather than squelching them only to witness a resurgence e l ~ e w h e r e .A ~ "spate of land-use control measures are available, but the selection depends on goals set by the municipal authorities. Whatever the specific methods chosen, it seems clear that an approach based on the philosophy of "progressive improvement," with the community sharing in the decision-making power and with the government providing the overall financial and administrative substructure, holds more promise than slum clearance or other largely physical-planning approaches to redevelopment do. In African cities this necessitates a radical change in urban development philosophies - one that is, in our view, inevitable but not likel) to happen overnight. T h e adoption of a more holistic planning paradigm also requires a great deal more detailed research designed to monitor the changes in low-income communities through in-migration and relocation. It is here that the static quality of the data available for this paper is most limiting. Hopefilly the detailed grain of the time-trend data we recently gathered in Accra will make it possible to remedy this deficiency and will allow us to reconstruct the process of internal relocation and neighborhood change. Until this is done we will not know houl population movements cause neighborhoods to change -we only know that they d o indeed change. 3n For an excellent article on this topic, see Richard Stren: llrban Policy in Africa: A Political Analysis, African Studiec Re7t., VOI. I 5, 1972, pp. 489-516. 3 9 T h e population of this squatter colony actually doubled in sixteen months between 1969 and 1970, despite repeated repressive measures on the part of municipal authorities. For a detailed account of the problems and prospects of these people, see David Etherton: blathare \.'alley: A Case Study of Uncontrolled Settlement in Nairobi (Housing Research and Development Unit, University of Nairobi, 197 I ) . .4 short summary of the difficulties of the Nairobi City Council is found in Africa Report, \'ol. 16, No. 2, 1971, pp. '7-8. 'O Abundant evidence from other African cities indicates that this tends to happen when slash-and-burn tactics are used to remove squatters without provision for rehousing. T h e most recent incident in .4ccra involved a Ewe fishing community, in which the people were forced away from the Labadi Beach area, only to appear with similar habitations several miles to the west along the beach.