Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
CLinking DHS Surveys to United Nations City Data Until 2000, the city population data published in the United Nations Demographic Yearbook were not computerized in any publicly accessible form. Thanks to the efforts of Alice Clague of the United Nations, the panel was able to secure a new database (United Nations, 2000) containing a prepublication version of the city population data that appear in the 1998 Demographic Yearbook. Although our computer file lacked some of the detail of the published table, it supplied the framework we needed to construct the merged city population dataset used throughout this report. To lengthen the time span of these data, we supplemented the computer file by keying in city population data from three earlier years of the Demographic Yearbook (United Nations, 1987, 1992, 1998a). Drawing on these data and another computer file (United Nations, 2001) con- taining the statistical annexes to a recent volume of World Urbanization Prospects, we assembled a composite city database that includes the following: · Raw population counts and estimates for capital cities and agglomerations of 100,000 population and above, taken from four recent volumes of the United Nations Demographic Yearbook · Extrapolated data series from World Urbanization Prospects (United Na- tions,2001) for cities of 750,000 population and above, covering the period 1950 to 2000 at 5-year intervals · Recent population estimates for capital city urban agglomerations, also taken from United Nations (2001) and earlier years of this publication For cities of under 750,000 population (other than capital cities), only the Demographic Yearbook population counts are available. 487
488 CITIES TRANSFORMED The panel then faced a problem in linking the United Nations city-level data to survey data on individuals and households from the Demographic and Health Surveys (DHS). Until recently the DHS program did not release information on the geographic locations of its sampling clusters, evidently to protect the confi- dentiality of the survey respondents. In the surveys available to the panel, not even the city name is provided for urban residents. This restriction means that if the DHS data are to be linked to data on cities, the only option is to forge the link using whatever geographic indicators are reported in the surveys.) Most DHS surveys provide information on the region of the country in which a sampling cluster is located, although the surveys vary greatly in the extent of regional detail, with some countries specifying regions very broadly in geograph- ical terms. All DHS surveys contain information on whether the sampling cluster is urban or rural in character, presumably following the definitions used by the country's national statistics office. Unfortunately, the population size of the urban area in which a cluster is located is not given; rather, its size is loosely character- ized in broad, almost qualitative terms. In Round 1 of the DHS program, urban areas were characterized as being either large or small cities. Since Round 2, three categories have been used: capitals and large cities, small cities, and towns. Large cities are defined as those with populations of 1 million or more; the small city category contains all cities in the wide range from 50,000 to 1 million pop- ulation; and towns are the residual category for urban areas below 50,000 in size (Demographic and Health Surveys, 1994, 2001~. THE MATCHING PROCEDURE Faced with these difficulties, the panel undertook to match the DHS data on region and city size, as broadly and qualitatively defined, to the United Nations estimates of the population sizes of specific individual cities.2 Using the available DHS data and reports, together with geographic atlases (e.g., Cohen, 1998b), we found that we could link the data in two ways. Often enough information was available for us to match a sampling cluster to a specific city. Sometimes, however, we lacked sufficiently detailed geographic variables from the DHS to identify a specific city of residence, but were able to winnow the possibilities to a set of cities. The latter type of linkage occurred more often in the Round 1 DHS surveys (whose urban characterizations were cruder than those in Rounds 2 and 3) and in more populous countries and those with higher levels of urbanization (which usually have more . . . . . . caches In any given geographic regions. iWe are indebted to Narayan Sastry of RAND Corporation for providing us with sampling cluster identifiers for the 1986 Brazilian DHS, a survey undertaken in an era in which the DHS program supplied such identifiers. Of the 90 DHS surveys used in our analyses, this is the only one allowing a precise match between sampling clusters and residence in particular cities. 2Edward Hui (Brown University) made a major contribution to this effort.
LINKING DHS SURVEYS TO UNITED NATIONS CITY DATA 489 In the case of the unique matches, the panel assigned a city size value taken from the merged United Nations database described above. Our aim was to use a city population estimate for 1990 a year roughly midway in the course of the DHS program or the nearest available year. In deciding among the different United Nations population estimates, we gave first preference to the World Urban- ization Prospects estimate, but if this was unavailable turned to the Demographic Yearbook estimate for the urban agglomerations or, if necessary, to the Yearbook estimate for the city proper. For the cases in which a DHS sampling cluster could be linked only to a set of cities, we constructed the weighted average of the 1990 populations for the cities in the set, using the city populations themselves as weights, and assigned this weighted average to the DHS sampling cluster. The idea was to assign the city population value that would be expected if a person were to be drawn at random from the set of cities. THE RESULT: CITY SIZE RANGES Clearly there is much room for subjective errors and imprecision in the final re- sults of this matching procedure. It is unfortunate that until recently the DHS program has had no mechanism in place to allow its survey clusters to be linked directly to city identifiers; having such a mechanism would have eliminated one source of error (see Appendix F for the panel's recommendation for meeting this need). As we have discussed, the United Nations population estimates are them- selves subject to errors and substantial differences in definitions. In view of the many uncertainties involved, all city sizes are classified in ranges of population size in the panel's analyses using DHS data, rather than being expressed as point estimates. We chose ranges that appeared to be narrow enough to be informative, yet broad enough to mitigate the effects of matching and other errors. In our anal- yses of DHS data, we do not make use of population growth rates for cities and sets of cities, although that possibility deserves further research attention. DHS SURVEYS USED IN THIS REPORT The following tables provide information on the DHS surveys examined in this report, which are limited to those datasets that had been released and were avail- able to the panel as of the beginning of 2000. (Many more DHS surveys have been made available since then, but they could not be included.) Table C-1 lists the countries whose surveys were examined by the panel, with the year of the sur- vey indicated. Table C-2 gives each country's urban definition precisely as it is recorded by the United Nations. Table C-3 lists the cities of these countries whose populations fell into the 1-5 million range in 1990 or the closest year to it for which a calculation was possible, and also lists the cities with more than 5 million population.
490 CITIES TRANSFORMED TABLE C-1 Countries with a Demographic and Health Survey Included in the Panel's Dataset, by Region (survey year in parentheses) Region Countries Surveyed North Africa Egypt (1988, 1992, 1995), Morocco (1987, 1992), Tunisia (1988) Sub-Saharan Africa Benin (1996), Botswana (1988), Burkina Faso (1993), Burundi (1987), Cameroon (1991, 1998), Central African Republic (1994-95), Chad (1996-97), Comoros (1996), Cole d'Ivoire (1994), Ghana (1988, 1993, 1998-99), Kenya (1989, 1993, 1998), Liberia (1986), Madagascar (1992, 1997), Malawi (1992, Mali (1987, 1995-96), Mozambique (1997), Namibia (1992), Niger (1992, 1998), Nigeria (1990), Rwanda (1992), Senegal (1986, 1992-93, 1997), Sudan (1989-90), Tanzania (1991-2, 1996), Togo (1988, 1998), Uganda (1988-89, 1995), Zambia (1992, 1996), Zimbabwe (1988, 1994) South-East Asia Indonesia (1987, 1991, 1994, 1997), Philippines (1993, 1998), Thailand (1987) South and West Asia Bangladesh (1993-94, 1996-97), India (1992), Kazakhstan (1995), Republic of Kyrgyzstan (1997), Nepal (1996), Pakistan (1990-91), Sri Lanka (1987), Turkey (1993), Uzbekistan (1996), Yemen (1991) Latin America Bolivia (1989, 1994, 1998), Brazil (1986, 1996), Colombia (1986, 1990, 1995), Dominican Republic (1986, 1991, 1996), Ecuador (1987), E1 Salvador (1985), Guatemala (1987, 1995, 1999), Haiti (1994-95), Mexico (1987), Nicaragua (1998), Paraguay (1990), Peru (1986, 1991-92, 1996), Trinidad and Tobago (1987) TABLE C-2 Urban Definitions in the Countries with a DHS Survey in the Panel's Dataset, as Recorded by the United Nations Country Definition Bangladesh Benin Bolivia Botswana Brazil Burkina Faso Burundi Places having a municipality (pourashava), a town committee (shahar committee), or a cantonment board. Localities with 10,000 or more inhabitants. Localities with 2,000 or more inhabitants. Agglomerations of 5,000 or more inhabitants where 75 percent of economic activity is nonagricultural. Area internal to the urban perimeter of towns and cities as defined by municipal law. The sum of 14 towns. Commune of Bujumbura.
LINKING DHS SURVEYS TO UNITED NATIONS CITY DATA TABLE C-2 continued Country Definition Cameroon Central African Republic Chad Colombia Comoros 491 Cole d'Ivoire Dominican Republic Ecuador Egypt E1 Salvador Ghana Guatemala Haiti India Indonesia Kazakhstan Kenya Kyrgyzstan Urban centers. Twenty principal centers with a population of over 3,000. Administrative centres of prefectures, sous-pre'fectures and administrative posts. Population living in a nucleus of 1,500 or more inhabitants. Administrative centers of prefectures and localities with 5,000 or more inhabitants. Urban agglomerations containing more than 10,000 inhabitants; agglomerations of from 4,000 to 10,000 persons with more than 50 percent of households engaged in nonagricultural activities; and the administrative centers of Grand Lahoun and Dabakala. Excludes the milieux urbain of Bouna, which has a population of 11,000. Administrative centers of comunas and municipal districts. Capitals of provinces and cantons. Governorates of Cairo, Alexandria, Port Said, Ismailia, and Suez; frontier governorates; and capitals of other governorates and district capitals (markaz). Administrative centers of municipios. Localities with a population of 5,000 or more. Municipio of Guatemala Department; officially recognized centers of other departments and municipalities. Urban population for 1981 is officially adjusted to include the urbanized suburbs bordering the municipio of Guatemala, consistent with the previous census. Administrative centers of communes. Towns (places with municipal corporation, municipal area committee, town committee, notified area committee, or cantonment board); and all places having 5,000 or more inhabitants, a density of not fewer than 1,000 persons per square mile or 390 per square kilometer, pronounced urban characteristics, and at least three-fourths of the adult male population employed in pursuits other than agriculture. Municipalities (kotamadya), regency (kabupaten) capitals and other places with urban characteristics. Cities and urban-type localities, officially designated as such, usually according to the criteria of number of inhabitants and predominance of agricultural or nonagricultural workers and their families. Towns with 2,000 or more inhabitants. Cities and urban-type localities, officially designated as such, usually according to the criteria of number of inhabitants and predominance of agricultural or nonagricultural workers and their families. (continued)
492 TABLE C-2 continued Country Definition Liberia Madagascar Malawi Mali CITIES TRANSFORMED Mexico Morocco Mozambique Namibia Nepal Nicaragua Niger Nigeria Pakistan Paraguay Peru Philippines Rwanda Senegal Sri Lanka Localities with 2,000 or more inhabitants. Centers with more than 5,000 inhabitants. All townships, town planning areas, and district centers. Localities with 5,000 or more inhabitants and district centers. Localities with 2,500 or more inhabitants. Urban centers. Conselho of Maputo and Beira. The censuses are thought to be underenumerated, but the underenumeration is almost solely in the rural areas. Hence, the census population counts for urban areas (1951 and 1960) and for Windhoek (1951, 1960, 1981) are accepted, but the 1992 Revision for the total population is used to estimate the "true" census count for all three dates. The 1951 and 1960 percentage urban is estimated by dividing the census count of urban population by the above-estimated "true" census counts. To obtain the urban population at the time of the 1981 census, it was noted that in 1951 and 1960, 30 percent of the urban population lived in Windhoek. Therefore for 1981, the census-counted Windhoek population was accepted and was divided by 0.30 to obtain an estimated urban population. Localities with 9,000 or more inhabitants (panchayats). Administrative centers of departments and municipios. Urban centers (27 towns). Towns with 20,000 or more inhabitants whose occupations are not mainly agrarian. Places with municipal corporation, town committee, or cantonment. Administrative centers of the official districts of the Republic. Populated centers with 100 or more dwellings grouped contiguously and administrative centers of districts. All cities and municipalities with a density of at least 1,000 persons per square kilometer; administrative centers, barrios of at least 2,000 inhabitants, and those barrios of at least 1,000 inhabitants which are contiguous to the administrative center, in all cities and municipalities with a density of at least 500 persons per square kilometer; and all other administrative centers with at least 2,500 inhabitants. Kigali; administrative centers of prefectures and important agglomerations and their surroundings. Agglomerations of 10,000 or more inhabitants. Municipalities, urban councils, and towns.
LINKING DHS SURVEYS TO UNITED NATIONS CITY DATA TABLE C-2 continued Country Definition 493 Sudan Localities of administrative or commercial importance or with a population of 5,000 or more inhabitants. Thailand Municipalities. Togo Seven urban communes. Trinidad and Tobago Port-of-Spain, Arima borough, and San Fernando town. Tunisia Population living in communes. Turkey Population of the localities within the municipality limits of Uganda administrative centers of provinces and districts. Population of all settlements as small as trading centers with as few as 100 inhabitants. United Republic of Tanzania Gazetted townships. Uzbekistan Cities and urban-type localities, officially designated as such, usually according to the criteria of number of inhabitants and predominance of agricultural or nonagricultural workers and their families. The entire former colony of Aden, excluding the oil refinery and villages of Al Burayqah and Bi'r Fuqum for the former Democratic Yemen, and six main towns for the former Yemen. Yemen SOURCE: United Nations (1998b: 37-55~.
494 CITIES TRANSFORMED TABLE C-3 Large Cities in the Countries Covered by DHS Surveys in the Panel's Dataset, by Region Population circa 1990 Region North Africa 1 to 5 Million Egypt: Giza, Alexandria; Morocco: Rabat, Casablanca; Tunisia: Tunis Cameroon: Douala; Cole d'Ivoire: Abidjan; Ghana: Accra; Kenya: Nairobi; Mozambique: Maputo; Nigeria: Ibadan; Senegal: Dakar; Sudan: Omdurman, Khartoum; Tanzania: Dar es Salaam; Zimbabwe: Harare Indonesia: Palembang, Tangerang, Medan, Surabaya, Bandung; Philippines: Quezon City Over 5 Million Egypt: Cairo Sub-Saharan Africa Southeast Asia South, Central, West Asia Bangladesh: Chittagong; India: Ludhiana, Varanasi, Kalyan, Visakhapatnam, Bhopal, Ulhasnagar, Madurai, Patna, Coimbatore, Indore, Vadodara, Kochi, Surat, Jaipur, Lucknow, Nagercoil, Napur, Kanpur, Pune, Ahmedabad, Bangalore, Hyderabad; Kazakhstan: Almaty; Pakistan: Multan, Rawalpindi, Peshawar, Gujranwala, Faisalabad, Lahore; Turkey: Konya, Izmir, Ankara; Uzbekistan: Tashkent Bolivia: La Paz; Brazil: Santos, Belem, Campinas, Nova Igua5cu, Brasilia, Curitiba, Fortaleza, Salvador, Recife, Porto Alegre, Belo Horizonte; Colombia: Barranquilla, Cali, Medell~n, Santa Fe de Bogota; Dominican Republic: Santo Domingo; Ecuador: Quito, Guayaquil; El Salvador: San Salvador; Guatemala: Guatemala City; Haiti: Port-au-Prince; Mexico: Ecatepec, Netzahualcoyotl, Puebla de Zaragoza, Monterrey, Guadalajara, Latin America Nigeria: Lagos Indonesia: Jakarta; Philippines: Manila; Thailand: Bangkok Bangladesh: Dhaka; India: Madras, Delhi, Calcutta, Mumbai; Pakistan: Karachi; Turkey: Istanbul Brazil: Rio de Janeiro, Sao Paulo; Mexico: Mexico City; Peru: Lima