This appendix lists the data sources used to estimate the population figures in the GIS database and other relevant information pertaining to the demographic estimates and GIS data sets. The data for the nineteen West African countries have been estimated by Benoit Ninnin for the West Africa Long Term Perspective Study (WALTPS) carried out by the Club du Sahel/OECD (see Ninnin 1994). All other figures have been estimated by the author at NCGIA. Unless otherwise stated the administrative boundaries are the same as those in the African Data Sampler (WRI 1995) or in the WALTPS database (see Brunner et al 1995).
Kenya
Division (3rd level) totals were available only for1989.
Areal interpolation of the 1979 figures using GIScoverages of locations for 1979 and sublocations for 1989 producedreasonable estimates only for some districts. For others, districtgrowth rates were used to produce division figures for 1979. For 1962 division estimates growth rates by district for the 1962-79period were used.
P60-90 were computed using the resulting intercensalgrowth rates 1962-79 and 1979-89. However, the figures are veryaccurate at the division level only for 1990; for the other estimatedyears, the division level data should be interpreted with caution.
Boundaries for 1989 were obtained from UNESCO/TSBFNairobi and rubber-sheeted to match the 1979 coverage which wasassumed to have higher accuracy. Accuracy reached in this operationis unknown. The sublocation boundaries were dissolved to yielda division level coverage.
Lesotho
1) contains total pop by district and sex for 1976and 1986 (86 = preliminary figures). 2) contains pop for 1966(in 000).
P60-90 based on intercensal growth rates.
GIS coverage obtained from the EDC/USAID/FEWS project.
Liberia
WALTPS pop estimates based on:
Libya
P60-90 based on growth rates between 1973 from 2)and 1984 from 1). Data for 1964 were also available, but namesin the list did not match the names in the 1984 list. Similarly,data for 1973 had to be aggregated to the 1984 units.
Boundaries were digitized from the CIA map "Libya"dated 5-93. See http://www.lib.utexas.edu/Libs/PCL/Map_collection/africa/Libya.GIF. Projection informationwas included on the map, but the scale was not. Due to the smallcartographic scale of the map, however, the accuracy of the boundariesis very low. This may not matter much in the uninhabited desertareas, but may lead to errors in the urbanized coastal regions- particularly around Tripolis.
A note on the map says: "Presently Libya has25 municipalities. An unconfirmed press article reports thatthe municipalities were replaced with 1.500 communes in 1992."(?!)
Madagascar
Statistique et de la Recherche Economique, Tananarive.
Pop figures and areas by administrative unit wereavailable at the sous-prefecture level for 1966, 1975 and 1993. However, boundaries changed, especially between the 1975 and1993 censuses. 1966 and 1975 figures thus had to be estimatedfor areas created in 1993 and figures for those units that wereaffected by these changes needed to be adjusted. In cases wherea major urban center was split from its surrounding area, the66 and 75 populations of the city from a secondary source wastaken for the urban part and the residual from its total populationin 66 and 75 were used for the rural areas. In cases, where areaswere split, one of two methods was used: either a simple arealinterpolation, where population is distributed in proportion toarea; or the 66 and 75 populations were used using average growthrates for the surrounding areas, and the resulting figures weresubtracted from those units from which the new units were created.
The resulting estimates were adjusted for two reasons:
- the 1966 figures appear to represent estimates(no census was taken in the 1960 census round), and
- the period between 1975 and 1993 is very long,such that average annual growth rates are likely to underestimatethe acceleration of population growth in the last few years.
All figures were thus adjusted uniformly to matchthe figures for Madagascar from the UN World Population Prospects(1994 revision).
Boundary data set produced by the National StatisticalOffice in collaboration with the UN Statistics Division's SoftwareDevelopment Project.
Malawi
Boundary changes between 1966 and 1977 censuses: 1966 population for Chitipa, Karonga and Rumphi districts adjustedbased on population and area figures from 1. above: P66adj =Pop66old/Land66 * Land77.
P60-90 calculated by using unadjusted intercensalgrowth rates for 66-77 and 77-87.
Mal
WALTPS pop estimates based on:
Boundaries produced by the USGS/EDC USAID/FEWS project.
Mauritania
WALTPS pop estimates based on:
Boundaries produced by the USGS/EDC USAID/FEWS project.
Morocco
Census figures and estimates for 1982 to 1990 wereavailable from 1. Data for 1971 from 2. Where units were split,the population 1971 population was distributed using proportionsfor 1982.
Sidi Bernoussi-Zenata (geog code 5 in Centre) seemsto be the same unit as Mohammedia-Znata. No data for Al Fida/DerbSul -> aggregated with Casablanca.
For those admin units located in Western Sahara,no data were available for 1971. Instead the total UN estimatefor Western Sahara for that year was split to those four units(Boujdour, Es Semara, Laayoune and Oued Ed Dahab) with proportionsorresponding to 1982 figures.
Mozambique
The UNICEF figures were attached to a boundary filewhich was used as the basis for the administrative units. Thisfile was converted from MapInfo to Arc/Info format and reprojected. In order to reduce distortions, rubbersheeting was applied usingthe first level boundaries in the original Africa DB for guidance.
2nd level population figures were available for 1960,70, 80, and 91. However, the district boundaries and naming ofunits changed significantly between 1970 and 1980 (and to a lesserextent between 1960 and 1970). Since no detailed maps of adminunits for 60 and 70 were available (only listings), the data neededto be reconciled using published maps (and a fair bit of intuition!). The main map used was the 1:2 Million map of Mozambique publishedby Cartographia (Budapest) in 1993/94.
In most cases, the names in the older census pubscould be matched to the new names which usually are based on townnames. In a few cases this was not the case, and average provincelevel growth rates had to be applied to several units. Problematicprovinces were Niassa (formerly Lago and Niassa) and Gaza.
The published national totals were found to be toolow
in 1990 when compared to the UN estimates (1.168mio vs. 1.418 mio) as well as compared to a 1987 estimate publishedin the Europa Yearbook (1.436), which also contains province levelestimates. To adjust the 1990 total, first an adjustment factorfor each province was derived by assuming that the share of eachprovinces pop of the total national pop is correctly representedin the 1987 estimates published in the Europa Yearbook. Theseshares were then applied to the UN national estimate to yieldprovince totals. The district pop figures in each province werethen uniformly adjusted so that they match these province totals.
The 1960 and 70 figures were also somewhat lowerthan the UN totals (~10%), and the district totals were thereforeadjusted uniformly. The 1980 figures were very close to the UNestimates and thus not adjusted.
Since Mozambique experienced significant turmoilduring the last two decades, the published figures are probablynot very reliable. Furthermore, unusual fluctuations in districtpopulation (large decreases between 1970 and 80 followed by largeincreases between 80 and 90) could either be due to data errors,sudden population movements due to civil strife, or signficantredesign of district boundaries. In
interpreting the estimates for Mozambique, more thanthe
usual amount of caution and skepticism should beused!!!
Namibia
Data were available for 1970, 81 and 91 censuses. 1980 and 90 data were derived using average annual 70-81 and81-91 growth rates. The Namaland district was created between1970 and 1981. Estimates for this district were derived for 1970assuming the same share of it contained the same share of populationof the three districts in which it was included previously - Mariental,Bethanien, and Keetmanshoop. A corresponding number of peoplewere subtracted from these for 1970.
Due to lack of data, 1960 estimates had to be derivedusing average 1970-80 growth rates applied to the 70 figures.
Boundary data set updated using maps in census publications.
Niger
WALTPS pop estimates based on:
Nigeria
WALTPS pop estimates based on:
Boundary data set produced by WALTPS.
[ African Population Distribution Database |
UNEP/GRID-Sioux Falls ]
<URL: http://grid2.cr.usgs.gov/globalpop/africa/country-specific/k_n.htm>
Last modified: 20 February 1997.
Please address any comments or suggestions to
uwe@ncgia.ucsb.edu.
