Click for next page ( 109


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 108
4 Urban Population Dynamics: Models, Measures, and Forecasts This chapter brings analytic tools to bear on the urban levels and trends described in the previous chapter. Our treatment of the issues is narrowly demographic, particularly at the outset, focusing on what might be termed the proximate causes of urban growth. Rural-to-urban migration is one of these proximate causes; of equal importance are the rates of urban and rural natural increase and the relative sizes of the urban and rural populations. Territorial reclassification must also be considered. In placing emphasis on this small set of demographic variables, we are mindful of their uncertain causal status. Rates of migration and natural increase are at once the cause and the consequence of larger social and economic forces. Even reclassification touches on economic, fiscal, and political concerns. In the chapters to follow, the socioeconomic content of the demographic variables will be explored in depth. This chapter begins by describing the features of urban population dynamics that can be seen even with the simplest of analytic mechanisms. A model of urban and rural population growth is developed to show how an initial urban and rural population distribution, when subjected to fixed demographic rates, can produce a variety of demographic outcomes: annual increases in the total urban population, the share of those increases due to migration, rates of urban growth, levels of urbanization, and their rates of change. Using projections, we highlight several regularities that can help in understanding the empirical record. The chapter then draws upon the Demographic and Health Surveys (DHS) for evidence on the main contributors to urban population growth fertility, mortal- ity, and migration which act together to determine urban age composition. As will be seen, urban populations are much more concentrated in the productive and reproductive ages than are rural populations. Where age is concerned, ur- ban populations are configured for higher potential economic productivity. With 108

OCR for page 108
URBAN POPULATION DYNAMICS 109 respect to the reproductive ages, however, these populations are also configured in a way that enhances the potential for high fertility although later marriage and greater contraceptive use generally keep this potential from being realized and that raises the profile of reproductive health concerns and other diseases affecting young adults. To shed light on the contribution of migration, we analyze the DHS data on recent moves, linking these data to measures of city size. This analysis exposes several weaknesses in the basic infrastructure of urban population research. First, because the DHS surveys are generally restricted to women of reproductive age, they reveal little about the situations of men and other migrants. Second, the ge- ographic identifiers supplied with DHS datasets are so coarse that it is difficult to determine even the name of the city in which a survey respondent lives, unless that city happens to be the nation's capital. Third, the population of the city in ques- tion is available from United Nations sources if the city is a capital, but otherwise can be determined only for cities above 100,000 in population. Moreover, only "raw" estimates of city size, taken from the United Nations Demographic Year- books, are available for cities in the range of 100,000 to 750,000 population. The expertise of the United Nations Population Division, as expressed in its influential series World Urbanization Prospects, is focused only upon cities larger than this. In short, a rather mundane analysis task brings alarming research gaps into view. The next section turns a critical eye on the two United Nations databases used in urban population research the annual Demographic Yearbooks and the bien- nial World Urbanization Prospects. In this field, World Urbanization Prospects has assumed the role of a standard reference work; it is the authoritative, com- prehensive source of urban population estimates and projections upon which most researchers and institutions rely. Because it assembles data over time, World Ur- banization Prospects provides an especially rich set of materials on urban growth, with detailed time series for all of the world's large cities. But the occasional user of these data is apt to be misled by their attractive packaging, and may need to be reminded of the weaknesses and heterogeneities of the population series that are available to the United Nations and the difficulties it faces in adequately estimat- ing and projecting urban populations. The chapter then examines the record of urban population projections, an issue of fundamental importance that has attracted curiously little attention apart from the efforts of the United Nations and Brockerhoff (1999~. As will be seen, the his- tory of city size projections does not inspire confidence. The urban record of suc- cess is so thin that a recent authoritative assessment of national-level demographic projections (National Research Council, 2000) does not even consider the possi- bility of projecting the national populations of developing countries by conducting separate rural and urban projections. Evidently we are still some distance from being able to apply modern statistical techniques to urban population time series. Having reviewed the aggregate databases, we turn to new developments in the area of spatially disaggregated geographic information systems (GIS). As

OCR for page 108
110 CITIES TRANSFORMED a locus of several new technologies, GIS holds promise for restoring a spatial dimension to the (typically) aspatial data gathered through demographic surveys and related sources. There is at least the possibility that data collected according to GIS principles might strengthen the foundation for urban population projection. GIS technology is perhaps even more promising as a political device, that is, as a mechanism for fostering dialogue among the units of government that collect data, those that supply services, and the urban residents who wish to make use of such services. We document several of the encouraging efforts now under way. The chapter ends by emphasizing the infrastructure that will be needed to support urban population research. THE SIMPLE ANALYTICS Five demographic indicators are required to sketch the main developments in an urban transition: the absolute annual increase in the urban population, the share of that increase attributable to migration, the urban growth rate, the level of ur- banization, and the rate of urbanization. In the model presented below, all five indicators result from the repeated application of constant demographic rates to an initial population distribution. ~ Key Concepts and Notation To highlight the essentials, we abstract from the problem of reclassification and focus on a hypothetical country divided into rural and urban sectors that are fixed in geographical terms. Age dependence in fertility, mortality, and migration rates is initially ignored. The results thus obtained would generally continue to hold in an age-differentiated simulation (Rogers, 1995~. In this stylized representation, national population size in year t is denoted by Pi, the size of the urban population by Us, and that of the rural population by Rt. Given the sizes of the rural and urban populations in a base year Ro and TO, respectively the totals Us and Rt evolve in a manner determined by four demographic rates, each of which is expressed on a per annum basis: raft the rate of natural increase in the urban population, that is, the difference between urban birth and death rates Ilr the rate of- natural increase in the rural population mama the migration rate from rural to urban areas, expressed per rural resident mll,r the migration rate from urban to rural areas, expressed per urban resident iFor more detail on the equations and their derivation, see Appendix B and United Nations (1974, 1998b).

OCR for page 108
URBAN POPULATION DYNAMICS 111 As discussed in Appendix B. which presents more of the mathematical detail, from these simple ingredients the model can generate time paths for the five urban outcome measures mentioned above. To proceed, we must define some terms. The urban population increment is the net addition to the urban population from year t1 to year t, or /\Ut = UtUt-1 - We make use of an equivalent representation, /\Ut = Ut_1 (~nmarry + Rt-l mr'n' (4.1) in which the roles of the demographic rates and the population distribution are more clearly evident.2 The urban growth rate is another measure of change in the total urban population over a single year; it is expressed not in absolute terms, but as a fraction of the initial urban population: UGR' This, too, can be given a useful alternative form: UGRt = ratm~,r + /\Ut Ut_ - Rt-1 U mr'n (4.2) The migrant share of urban growth, denoted MSt, is the proportion of net urban growth that is due to migration from rural areas. The share can be written as MSt=(l+Ut-l nnmUrN 1 mr'~ (4.3) As is evident in equations (4.2) and (4.3), the urban growth rate and the share of growth due to migration are determined by several constants the rate of urban natural increase and rates of migration to and from urban areas as well as a time-varying factor, Ut- 1 /Rt - 1, the urban/rural population balance. Because urban population increments, growth rates, and migrant shares are closely related, they are often discussed as a group. So, too, are the following measures, which concisely summarize levels and trends. The level of urbanization is simply the urban proportion (or percentage), Aft= p, 2A counterpart expression for the rural increment would include the term formom. This term is generally positive. See, for instance, Oucho and Gould (1993) on net rural increase in sub-Saharan Africa.

OCR for page 108
112 CITIES TRANSFORMED and the rate of urbanization is the rate of growth in this proportion over time. It can be written as /\P7l,,t UtUt- = Pll,,t- 1 It- 1 PtPt- Pt1 which is the difference between the urban and national population growth rates. Even a model as simple as this can help clarify otherwise puzzling and coun- terintuitive aspects of urban transitions. We use it to address several questions. First, if urban growth results from both migration and natural increase, which of these accounts for the greater share of growth? Should the migration share be re- garded as a constant or as a time trend? How are urban growth rates linked to the rate of urbanization? Do national population growth rates translate directly into rates of urban growth (according to Preston [1979], a 1-point decline in national growth reduces urban growth by the same amount), or can we antici- pate a systematic change in the relationship between the rates as urbanization proceeds? City Growth: Migration or Natural Increase? Much of the concern surrounding urban growth has to do with the annual additions to their populations that cities must somehow absorb, and with the contribution to growth that is made by rural-to-urban migrants. As discussed earlier, demogra- phers often find themselves emphasizing the role of urban natural increase (see United Nations, 1980; Chen, Valente, and Zlotnik, 1998), if only to counter the impression that migration must be the dominant factor. To disentangle migration from natural increase is more difficult than might be supposed; an analytic model is helpful in showing just where the problems lie. As Rogers (1982) explains, to understand whether migration or natural in- crease is the dominant source of growth, one must first decide on terms. Much depends on whether one conceives of the problem in terms of JqOWS or stocks. Flows are, by definition, short-term measures. Their empirical counterparts are found in the decompositions of intercensal urban growth that separate urban nat- ural increase on the one hand from the sum of net migration and reclassification on the other. According to such flow estimates as discussed in Chapter 3 the share of migration in urban growth is in the neighborhood of 40 percent in most developing countries. Stocks, by contrast, are cumulative measures. If the mi- grant contribution is to be assessed in terms of stocks, the estimate should take into account not only the migrants themselves, but also their descendants. To understand the cumulative contribution of migration to urban growth, one would compare the size of an urban population with what it would have been in the ab- sence of rural-to-urban migration (i.e., with mr,~' = 0) or with lower rates of

OCR for page 108
URBAN POPULATION DYNAMICS 113 migration than occurred. Of course, if migration is permitted cities will grow larger than they would otherwise, but the size of the difference is of interest.3 Because flows depend on stocks for example, the flow of rural-to-urban mi- grants is mr,ll Rt_l, with Rt_1 being the stock of rural population one cannot cleanly separate them. For analytic purposes, it is preferable to distinguish direct elects from feedback elects and to focus attention on the implications of changes in the fundamental rates. Consider, then, the consequences of a change in rail, the urban rate of natural increase. Returning to the migration share equation (4.3), we see that at time t, the direct effect of an increase in rail is to reduce the share of migration in urban growth, as would be expected. Likewise, the direct effect of an increase in the rural-to-urban migration rate, m,, is to increase the migration share, again as would be expected. In either case, the amount of change produced in the migration share depends on several factors, one of which is the urban/rural balance, U~_~/R~_~. Once rates have changed, feedback effects come into play, and these effects exert further influence on the migrant share. Higher rates of urban natural in- crease, rail, tip the population balance toward urban areas, causing U~_~/R~_~ to rise with time. The more rapid population shift toward cities diminishes the rel- ative size of the rural sector, and this in turn diminishes the relative contribution of rural migrants to city growth. The direct and feedback effects of rail work in the same direction; through both routes, a higher rate of urban natural increase reduces the migrant share of urban growth. Applying the same kind of analysis to the rural-to-urban migration rate, we find that an increase in mr,~' generates feedbacks that work against the direct ef- fect. As explained above, the direct effect is to increase the share of urban growth due to migrants; but with faster rural outmigration, the population balance begins to shift toward urban areas. Over time, this feedback acts to reduce the migrant share. The opposition of forces can be seen in Figure 4-1, which shows the share of urban growth due to migration for values of mr,~' ranging from 0.5 percent per annum to 3.0 percent.4 Higher migration rates are associated initially, and through most of the projection period, with larger migrant shares. As feedback 3 The flows-stocks perspective has often proven helpful; for instance, it was used by the National Research Council (1997) to analyze the full contribution of international migration to the population of the United States. 4 All projections begin with an urban proportion of Pu,o = 0.15. For the other parameters of the projections, we have been guided by Livi Bacci (1997) for natural increase and by the United Nations (1980) and Chen, Valente, and Zlotnik (1998) for migration. Chen, Valente, and Zlotnik (1998) present regional estimates of mr,U that range from 0.5 percent (Africa in the 1980s) to 2.65 percent (Latin America in the 1970s). Earlier, the United Nations (1980) provided estimates ranging from 0.05 percent for Nepal in the 1960s to 3.7 percent for Venezuela in the 1950s. We restrict mr,U to lie between 0.5 and 3 percent. The reverse urban-to-rural migration rate, mu, r, is set to 0.25 percent throughout.

OCR for page 108
114 o.s - s 0.8- 0.7- o s "' 0.6- cat i= CITIES TRANSFORMED Ha, ~ . 0 005 -~` m: 0.030 mr U = 0.010 -. :'` mrU= 0.015 .. `' ~ i. . ~ 0.4 - 0 10 20 30 40 50 Projection Period FIGURE 4-1 Declining migrant shares of urban growth. effects exert their influence, however, all these curves decline and show a tendency to converge. The point at which the migrant share of growth reaches one-half has been termed the cross-over point by Keyfitz and Ledent (Rogers, 1982, citing their papers). As can be seen in the curve with mr,~' = 0.03, depicted by a dashed line in Figure 4-1, very high rates of rural-to-urban migration can hasten the arrival of the cross-over point and produce lower migrant shares thereafter than would be seen in a regime of lower migration rates. The apparent paradox is an expression of a feedback effect. Urban Growth and the Rate of Urbanization The rate of urban growth obeys a similar logic. Urban natural increase, rail, has a positive direct effect on the rate of urban growth (see equation (4.2~. The direct effect of the rural-to-urban migration rate, m,, is also positive, but its strength varies with the urban/rural population balance. When man' rises, the rural sector begins to decline in relative importance, and the urban growth rate then falls. Over the long term, the urban growth rate will approach UGR = ranm~,r + b

OCR for page 108
URBAN POPULATION DYNAMICS 0.08 - O 0.06- . _ N ~ 0.04- s a 0.02- 0.00 - 115 ~ \ \ ~;0.02 Rate of Urbanization Rates of Urban Growth By/ ~ _ ma_ _, ~ _ 0 10 20 30 Projection Period FIGURE 4-2 Rates of urban growth and urbanization. 40 50 where b is the long-term urban/rural balance, that is, the value taken by U~/R~ in the limit.5 As was shown earlier, high rates of natural increase have been a defining fea- ture of the demographic regimes of many developing countries and distinguish their urban transitions from the Western historical experience. As just noted, a higher rail produces more rapid urban growth; working indirectly through migra- tion, a higher rural Ilr also produces more rapid urban growth. Although they have powerful effects on the urban growth rate, rail and Or need not have any particular implications for the rate of urbanization. As can be seen in Appendix B. equa- tion (B.6), if roll and Ilr happen to be equal, they can be scaled up or down with no direct effect on the rate of urbanization. The projections depicted in Figure 4-2 illustrate the point. In this figure, the rates of natural increase are set to 2 percent (the upper curve) and 1 percent (the middle curve). These two curves exhibit high initial urban growth rates that are followed by growth rate declines. Meanwhile, the rate of urbanization shown with the dashed line remains wholly unaffected, adhering to the same trajectory whether the natural rates of increase are high or low. If equality in urban and rural rates of natural increase appears to be a spe- cial case, consider data from the United Nations (1980: Tables 10-12~. These 50f course, the rates must be such that the limit exists; see Appendix B.

OCR for page 108
116 CITIES TRANSFORMED data show that in the 1950s and 1960s, a number of developing countries had values of rail and fir that were roughly equal. In the 27 developing countries ex- amined by the United Nations, fertility and mortality rates were lower in urban than in rural areas, but the differences between fertility and mortality were about the same.6 When rail is approximately equal to fir, as in these cases, the rate of urbanization is all but entirely attributable to migration. If raft 76 rtr or if these rates change by different amounts, their effects will be expressed in the rate of ur- banization, producing a different path of urbanization than that which appears in Figure 4-2. For example, countries with slower rates of urban than rural natural increase (rat' < rtr) will tend to have slower rates of urbanization, other things being equal. Urban and National Population Growth As noted in Appendix B. equation (B.5), national population growth rates can be written as a weighted average of the urban and rural rates of natural increase, with the weights being the proportions of urban and rural residents in the national total. In the fixed-rate analytic model compare equations (B.2) and (B.5) in Appendix B it is clear that national and urban growth rates depend on rail and fir in much the same way, and this implies that national and urban growth will tend to be positively correlated.7 Figure 4-3 charts the relationship between the two rates of growth over the course of one projection.8 The two rates are positively associated, as expected, although they are linked in a nonlinear fashion. At the outset of the projection (see the upper right portion of the figure), both the urban and national growth rates are high; as the country urbanizes, both rates fall. With increasing urbanization, the slope of the relationship between urban and national growth rates flattens. One might well expect to see the main features of Figure 4-3 reproduced in empirical urban growth regressions. Indeed, demographers have uncovered strong regularities in the association between urban and national rates of growth. In one analysis using a sample of cities in both developing and developed countries (Pre- ston, 1979; United Nations, 1980), a regression of urban growth rates on national growth rates yielded a coefficient for national growth that was very close to unity. The relationship was revisited by Brockerhoff (1999) with a sample limited to 6The United Nations tables show that even in this era, rural rates of natural increase were often slightly higher than urban rates. It is not clear whether the generalization ran ~ rtr still stands. An inspection of recent data (United Nations, 2000: Tables 9 and 18) shows that in the 10 countries with data available for the l990s (most of these 10 being in West Asia), the urban rate of natural increase falls well short of the rural rate. However, according to Visaria (1997), rates of natural increase in India are about the same in urban and rural areas. 7Of course, if rid = rtr, the national growth rate would be invariant to the distribution of population between urban and rural areas. sin this projection, ran = 0.01, rtr = 0.02, and mr,U = 0.01. Recall that mU,r = 0.0025 in this and all other projections.

OCR for page 108
URBAN POPULATION DYNAMICS 0.06 - 0.05- 53 0.04- 0.03 - 0.02 - 117 Start of Projection (proportion urban = 0.15) / - 0.0160 0.0165 0.0170 0.0175 0.0180 0.0185 Rate of National Population Growth FIGURE 4-3 Rates of urban and national population growth, given roll < for. developing countries; his regression yielded a coefficient estimate of about 0.8. Such differences in regression coefficients are to be expected when the analy- sis samples differ in the average level of urbanization.9 In this case, however, the countries in the Brockerhoff sample are less urbanized on average, leaving it doubtful that the lower coefficient estimate found by Brockerhoff is due to the changing curvature of Figure 4-3. Note that when roll < for, a higher rural-to-urban migration rate mr,ll reduces the national population growth rate because it speeds the transfer of population to the urban sector where the natural rate of growth is lower. At the same time, a higher mr,~' increases the rate of urban growth. Hence, changes in the migra- tion rate would cause urban and national population growth rates to be negatively correlated, other things being held equal. Migration and Urban Age Structure As is well known, rural-to-urban migration rates are strongly influenced by age, and this age dependence is reflected in urban and rural age distributions. As 9In reaching these conclusions, both Preston (1979) and Brockerhoff (1999) use multivariate re- gression and include covariates other than national growth rates in their specifications. The initial level of urbanization, used in both analyses, proves to have a negative, strongly significant influence on the rate of urban growth. The link to the level of urbanization is implicit in Figure 4-3.

OCR for page 108
118 CITIES TRANSFORMED will be documented shortly using DHS data, in urban populations proportion- ately more residents are found in the prime working and reproductive ages, and proportionately fewer residents are children. Such differences are attributable to lower urban fertility and the age selectivity of rural-to-urban migration. But the relationship between migration and urban age composition is more complex than it might at first appear. In supplying cities with more young adults, rural-to-urban migration also tends to increase the urban crude birth rate, and higher birth rates, in turn, partly offsets the direct effects of migration on age structure. To further clarify the role of migration, we constructed an age-differentiated demographic simulation that makes use of model schedules of fertility, mortality, marriage, and migration. The approach is inspired by Rogers (1986) and draws on demographic schedules developed by Rogers (1995), Coale and Trussell (1974), and Ewbank, de Leon, and Stoto (1983~. Consider two hypothetical populations, one rural and the other urban, between which there is no migration. Both popula- tions share the same (high) mortality rate (eO = 45) and total fertility rate (TFR = 6.0), and we assume that they have each attained stable age distributions. We then open the border between the two populations and allow for migration in both di- rections. We assume, however, that the rural-to-urban migration rates are higher. it Figure 4-4 depicts the consequences. The first wave of rural-to-urban mi- grants increases the proportion of urban residents of reproductive age (shown on the right scale of the figure). As these new urban residents marry and bear chil- dren, the urban crude birth rate increases from its initial stable population value. Meanwhile the rural crude birth rate declines from its stable value; although there is migration in both directions, the schedules we have adopted ensure a net trans- fer of young adults to urban areas. Once the urban crude birth rates have been driven higher, the share of those aged 15-49 in the urban population begins to fall as the proportion of children rises (the latter proportion is not shown). The inflation of urban crude birth rates is temporary; it subsides as the first cohorts of rural migrants work their way through the urban age distribution, and the urban and rural birth rates then approach each other. The urban crude birth rate remains higher than the rural rate, however, as a result of the continuing influence of rural-to-urban migration. All this occurs even though the rural and urban rates of fertility, marriage, and mortality are assumed to be equal at each age. The effects displayed in the figure are purely compositional. Even so, they serve as a reminder of one important role of rural-to-urban migration: it increases the share of the urban population in the reproductive ages. The analytic models we have been using are based on assumptions of fixed demographic rates whether these are aggregate rates of natural increase and mi- gration in the simpler projection model, or fixed underlying schedules in the model with age structure. If such models are to help in sorting out the empirical record, iInitially, the urban share of the combined population is 15 percent; by the end of the projection, this share has risen to about 56 percent.

OCR for page 108
144 CITIES TRANSFORMED TABLE 4-9 Mean Percentage Error (MPE) and Mean Absolute Percentage Error (MAPE) in Urban Population Projections for the Year 2000, by Length of Forecast, Region, Level of Development, and Size of Country Category Region East Asia and Pacific (EAP) EAP excluding China Europe Latin America and Caribbean Middle East and North Africa South Asia Sub-Saharan Africa Organisation for Economic Cooperation and Development Other High-Income Level of Development Low Lower Middle Lower Middle excluding China Upper Middle High Size of Country 0-2 million 2-10 million 10-50 million 50+ million 50+ million excluding China MPEa MAPEb 20 yearsC 10 yearsd 5 yearse 20 years 10 years 5 years 0.039 0.184 0.140 0.198 0.133 0.272 0.218 0.267 -0.028 0.098 -0.004 0.130 0.088 0.054 -0.009 0.068 0.085 0.197 0.027 0.234 0.055 0.068 -0.024 -0.018 -0.183 -0.102 -0.056 0.231 0.069 0.256 0.099 0.037 0.128 0.089 0.008 0.060 -0.027 -0.019 0.183 0.032 0.261 -0.013 World Excluding China 0.113 0.295 0.140 0.226 0.245 0.291 0.382 0.289 0.043 0.166 0.053 0.130 0.088 0.075 0.021 0.123 0.105 0.197 0.070 0.274 0.097 0.110 0.048 0.020 0.334 0.199 0.072 0.312 0.115 0.279 0.199 0.117 0.074 0.063 0.030 0.120 0.098 0.019 0.216 0.108 0.027 0.124 0.192 0.001 0.189 0.126 0.018 0.141 0.171 0.007 0.190 0.120 0.020 0.199 0.080 0.283 0.049 0.161 0.066 0.115 0.026 0.053 0.022 0.528 0.268 0.169 0.282 0.199 0.082 0.329 0.163 0.070 0.168 0.208 0.049 0.227 0.149 0.054 0.206 0.199 0.055 0.257 0.156 0.060 NOTE: Based on 169 countries and territories whose boundaries have not changed substantially over the last 20 years. Excludes former Soviet Union. All figures are weighted by population size. MPE = mean percentage error. Positive error associated with projections being too high and negative error with projections being too low. b MAPE = mean absolute percentage error. c 20-year comparison based on comparing 1980 projections for the year 2000 in United Nations (1980) with actual data in United Nations (2002a,b). 10-year comparison based on comparing 1990 projections for the year 2000 in United Nations (1991) with actual data in United Nations (2002a,b). e 5-year comparison based on comparing 1996 projections for the year 2000 in United Nations (1998b) with actual data in United Nations (2002a,b).

OCR for page 108
URBAN POPULATION DYNAMICS 145 and negative if they are too low; as such, the MPE is a measure of bias. By contrast, the MAPE is always positive and is usually taken to be a measure of . . . mpreclslon. As is evident from the positive values for MPEs in the first three columns of Table 4-9, urban projections have been too high more often than too low. This is attributable in part to the fact that fertility has declined in many places more rapidly than was expected. At the global level, the forecasts of the urban popula- tion in 2000 that were made 20 years ago were approximately 14 percent too high; forecasts made 10 years ago were approximately 17 percent too high; and fore- casts made 5 years ago were nearly perfect as a result of the fortuitous canceling out of roughly equal numbers of high and low errors. The general pattern is much as expected, in that projections typically are more accurate over shorter projection periods. The inclusion of China in the calculations makes a considerable difference to the results. Urbanization trends in China, which is home to 30 percent of the ur- ban population of Asia, have fluctuated greatly over the years. These fluctuations stem both from historical events, such as the Cultural Revolution and its aftermath, that retarded or even reversed urbanization in China, and from several revisions (since 1983) in the official criteria defining cities and towns (recall Box 4.2~. If the Chinese data are set aside, a more consistent pattern is revealed in which projec- tions over shorter periods are more accurate. Table 4-9 also shows that there has been considerable diversity in the quality of urban projections by geographic region, level of economic development, and size of country. The United Nations urban projections have been most reliable for Organisation for Economic Cooperation and Development (OECD) countries, on average, and least reliable for countries in sub-Saharan Africa and other high- income countries, many of which are quite small. United Nations projections also tend to be better for larger than for smaller countries, perhaps because larger countries tend to receive more attention. Similarly, focusing on the largest countries in each developing region and the largest city in each, Brockerhoff (1999) compares two projections for 2000 one taken from the 1980 United Nations projections (United Nations, 1980) and the other from the 1996 projections (United Nations, 1998b). He finds the 1996 pro- jections of city size to be far lower, implying that they had to be revised substan- tially downward. Among all cities of 750,000 or more residents whose popula- tions were projected in 1980, the median reduction in projected city size was 15.1 percentage points. Upward revisions in the projections were far less common than downward revisions. Such forecast errors stem from several sources. The baseline data can be ab- sent or unreliable. Censuses may well undercount urban populations: crowded cities with their mobile populations present a challenge to census takers in every part of the world. Recent censuses in Indonesia and Pakistan are believed to have seriously undercounted the populations of Jakarta and Karachi (Jones, 2002~. As

OCR for page 108
146 CITIES TRANSFORMED demographers incorporate the results of recent census rounds, they often find that their estimates of urban populations at earlier points in time need to be revised. Also, as noted above, total population growth in developing countries has been slower than projected, apparently as a result of fertility declines that were more rapid than anticipated. Many economic and social changes have taken place that simply could not have been foreseen 20 years ago. We conclude that urban and city size projections must be treated with a good deal of caution. It appears that projections at higher levels of aggregation (such as total urban populations) have been slightly more reliable. Regionally aggregated data may benefit to a certain extent from the cross-cancellation of country-level errors. Nevertheless, the urban future is highly uncertain even for some regions. The United Nations projection for Africa, for example, is that by 2025, the conti- nent will have become predominantly urban. This is a reasonable extrapolation of current trends, but of course one wonders whether the decoupling of urbanization from economic growth in Africa and the economic crises plaguing the continent's cities will again cause the level of urbanization to fall short of the prediction. STATISTICAL SYSTEMS FOR DISAGGREGATED DATA Evidently, the aggregate databases on city size and growth are in need of substan- tial repair. But the United Nations demographers cannot be charged with this task: although they have great expertise and a store of critical knowledge, they must de- pend on the figures contributed by national statistical agencies. What factors are likely to motivate these agencies to rethink and reform their procedures and give them the means to do so? One thinks first of the role of national censuses. Censuses are large and of- ten politically charged undertakings, and although they are regularly fielded in some developing countries, in others they are held irregularly, while in still oth- ers the census taking enterprise appears to have ground to a halt. In any case, not all statistical agencies will process census data into the small spatial units that are needed for accurate counts of city populations and informative assessments of socioeconomic conditions within cities. At the local level, planning is further hindered by limited information about local land markets. Most large cities lack sufficient, accurate, and current data on patterns of land conversion and infras- tructure deployment. Urban maps are not infrequently 20 to 30 years out of date, lacking any information on newly emerging periurban areas. We cannot hope to understand the motivations and constraints of national and local statistical agencies, but we can point to two developments that may be encouraging. The first is the move toward governmental decentralization that is occurring in many countries of the developing world, which places new re- sponsibilities in the hands of municipal and regional government units. In the national debates that accompany decentralization, the appropriate role for infor- mation in the processes of governance is often discussed. For instance, as noted in

OCR for page 108
URBAN POPULATION DYNAMICS 147 Chapter 2, national budget allocations to regions can be based on regional popu- lation sizes and indicators of poverty, and similar criteria can be applied to the transfers from regional to municipal levels of government. The need for informa- tion exchange and feedback among units of government is often recognized in the national debates, although what is likely to come of the insight is generally less than clear. The second development, not unrelated to the first, is the rise of civil soci- ety and the recognition that when they can be assembled, proper socioeconomic maps can be powerful political tools. A city map of differences in service delivery or health conditions can give residents a means of assessing their relative stand- ing and staking claims to resources. For example, maps of Accra and Sao Paulo showing differentials in health status, mortality rates, and environmental condi- tions among city districts produced considerable local debate and some policy change in both of these cities (Stephens, Akerman, Avle, Maia, Campanareio, Doe, and Tetteh, 1997~. Maps highlighting the city neighborhoods with far-above- average mortality rates or unusual concentrations of environmental health prob- lems both inform and help mobilize the inhabitants of these areas and the politi- cians who represent them. Assembling such maps is a daunting task, however. In most large cities, each municipal agency or department maintains its own database, often organizing the data in an idiosyncratic manner and rarely sharing them with other agencies. Computerization of data is still relatively uncommon. Many city agencies con- tinue to rely on paper files and paper maps, which may be stored in formats and at scales that all but prohibit comparison, collation, and revision (Bernhardsen, 1999~. There are almost no examples of fully integrated databases for the con- stituent parts of large metropolitan regions. At best one finds data of reasonable quality for the central areas of the city, with little comparable information for the outlying areas. Looking to the skies for help, some countries have seen remotely sensed and geocoded data as an alternative to data gathered on the ground. Here there are encouraging technical developments. Sutton (2000), for example, has made esti- mates of intraurban population density based on measures of light intensity; total city populations were estimated by measuring the areal extent of the city in the imagery. These remotely sensed data alone were found to be strongly correlated with census population counts, and the use of ancilliary socioeconomic informa- tion further strengthened the correlation. Methods such as these have good po- tential to improve estimates of the spatial extent of city populations. They can be used to inform "smart interpolation" programs that can improve on existing maps and other population data in areas where good census data are unavailable. The possibilities are attracting considerable research interest Weeks (2002) provides a guide to some of the very recent technical developments. The essential principle of GIS is that when information is systematically geocoded, it becomes possible to integrate data from highly diverse sources. Many

OCR for page 108
148 CITIES TRANSFORMED recent GIS applications draw information from maps, satellite images, videos taken from low-flying aircraft, statistical data from tables, photographs, and other sources. When such data are overlaid, they permit cities to to be described and their socioeconomic conditions monitored more quickly and accurately than was previously possible. Geocoded data can assist governance in many ways, such as planning for infrastructure and transportation, tracking crime and improving law enforcement planning, allowing comparisons of program effectiveness across jurisdictions, strengthening taxation bases and record keeping, facilitating site se- lection for services, and promoting better evacuation plans in the event of emer- gencies (O'Looney, 2000~. Under ideal circumstances, common databases can help instill habits of cooperation among the units of local, regional and even na- tional governments, or at least among their technical departments. GIS technology is still in its early stages of development in most poor coun- tries, and even where the enterprise is under way, it tends to be a single-office op- eration, usually located in a planning or engineering department. And of course, the usefulness of GIS technology is dependent on the availability of appropriate GIS-coded data. But encouraging initiatives can be seen in a number of develop- ing countries, as described below. Qatar At the forefront of geographic information technology in the developing world stands Qatar, whose GIS activities began in 1988. The country is now completely covered by a high-resolution, digital topographic database, which draws together images; estimates of land elevation; and information on streets, buildings, zoning, land use, soils, and urban utilities (Tosta, 1997~. These data are meant to be avail- able to all government agencies that need them. In one successful application, the availability of digital parcel and building records for the entire country allowed the Central Statistical Organization to conduct an extremely comprehensive housing and population census in a single day. To be sure, Qatar's situation must be very nearly unique. Factors favoring the country's advanced use of GIS technology include its small geographic size, high-level political support for GIS initiatives (including the authority to mandate and enforce uniform standards), outstanding technical leadership, and adequate funding (Tosta, 1997~. African Initiatives Many African governments record substantial amounts of data in the form of maps. The major sources of spatial data are the national mapping agencies; many municipal authorities also gather spatial data, with particular attention to cadastral records. Efforts are under way in a number of African countries to create dig- ital databases through the conversion of such maps. In Botswana, for example,

OCR for page 108
URBAN POPULATION DYNAMICS 149 the Department of Town and Regional Planning has developed a digital database to monitor land use compliance in Gaborone. Lesotho's Mapping Agency is en- gaged in a large-scale digital mapping exercise for its urban areas. In most coun- tries, however, metadata the sets of organized spatial data and information about those data (e.g., where the data are located, how and by whom they were collected and maintained, how they can be accessed, and what their major characteristics are) remain in a rudimentary form. In Lagos, GIS technology has been employed effectively to resolve conflicts over land use. The goverment owns large portions of land in certain sections of the city, and residents are supposed to pay "ground rents." But owing to the multiple claims on many parcels and a history of poor record keeping, the government has been collecting only 5 to 10 percent of the rents it is due. To improve collection rates, a land information system is being developed that will provide access to all documents for each parcel of government-owned land. The geographic bound- aries of each parcel have been derived from digital orthophotos in conjunction with various legal plot maps that have been digitized. India Although circumstances in India appear to strongly favor GIS advances, the coun- try's spatial data infrastructure remains curiously limited. The great Survey of India, which dates to the mid-eighteenth century, covered the entire country with rigorous cartographic surveys. India is also the birthplace of the IRS (Indian Re- mote Sensing) series of satellites, which provide high-resolution remote sensing data to global markets. And India is home to a remarkable software industry. Why, then, has its use of GIS technology not progressed further? Most Indian government agencies simply do not understand the value of their data for government functions or for the private sector. Much as in African coun- tries, enormous quantities of valuable material are stored in paper form and sel- dom computerized. Security concerns have led to restrictions on access to maps, as well as to aerial photographs. Despite these obstacles, a number of diverse GIS initiatives are under way in some of India's largest cities; examples are described in Box 4.3. Malaysia Since the mid-1980s, several federal and state land agencies have explored GIS technology and developed stand-alone systems with valuable information. But these systems have not been integrated across agencies. In an effort to draw the information together, the Malaysian government is developing a national land in- formation system, which will provide access to spatial data for all levels of gov- ernment, the commercial sector, the nonprofit sector, academia, and the general

OCR for page 108
150 CITIES TRANSFORMED BOX 4.3 GIS Initiatives Across India Greater Mumbai Remote sensing and GIS are much involved in land use planning, with indicators ranging from soil type to air pollution. Maps have been produced on decadal population growth, population distribution, employment, the distribution of socioeconomic facilities, agriculture and forest land uses, and traffic patterns. These maps have illuminated spatial and temporal trends in each settlement within the Mumbai Metropolitan Region. GIS technology has also been used to assess al- ternative locations for a proposed second international airport and for a solid waste disposal site, and is assisting in the preparation of a rehabilitation and resettlement program for encroachments at Bandra-Kurla Complex, site of a planned interna- tional finance and business center. The government is using GIS technology to map features related to fire hazards and risk assessment, as well as service delivery. Perhaps the most interesting development in Mumbai is that a proposed land use plan for the metropolitan region has been transferred to village maps. Citizens and other concerned groups are thereby able to understand the implications of the pro- posed plan and to file objections and suggestions. Hyderabad The Center for Resource Education undertook a project on spatial mapping of industrial estates and environmental hazardous sites in or near residential areas. The maps depict the contribution of each industry to pollution, show the likely environ- mental impact, and indicate monitoring points. Hyderabad has also been develop- ing a GIS-based integrated emergency response management system for Hyderabad City. The project incorporates maps depicting land use, road networks (including travel time estimates), and the locations and numbers of fire and police stations and water-filling points. Chennai City (Madras) A GIS database has been developed for road networks. With this database, priorities can be assigned to road improvements in the context of an integrated transportation information system. Bangalore The Bangalore Development Authority has used GIS applications for route planning and tracking of 200 (and eventually 2,000) private buses. The Global Po- sitioning System (GPS) is used to monitor the locations of the buses and to generate appropriate bills based on distance traveled. public. The private sector is being encouraged to contribute products and services (Mahid Bin Mohamed, 1998). Other Applications Elsewhere around the world, GIS technology is being applied in innovative ways to improve urban management. The Kuwaiti Ministry of Public Works has launched a large-scale computerized management system to assist in the main- tenance of roadways, bridges, sanitary and storm sewers, and street rights-of-way. The Water Authority of Jordan has employed GIS technology to restructure the water supply network in Amman. This project has involved a complete redesign

OCR for page 108
URBAN POPULATION DYNAMICS 151 of the water supply system in the city's congested, densely populated core. In Dhaka, GIS technology has been applied to the problem of drainage in Dhaka City. A digital elevation model was established for the catchment area; inunda- tion maps were then produced, and various flooding scenarios of the past were simulated. Use of this technology, found to be cost-effective, has enabled the government to develop sustainable flood alleviation schemes. The Future of GIS Although the above are all promising developments, each involves substantial costs, ranging from those of personnel and training to those of purchasing and converting maps. GIS entails much more than technology, and because it requires cooperation among units and levels of government that have little experience in this regard, it may be perceived as threatening. Effective use of the technology requires both new organizational structures and experienced staff. The few case studies available do not demonstrate that the novel techniques and ways of think- ing about information and interrelated services made possible by GIS will nec- essarily make successful transitions from the technical staffs of engineering and planning departments to the broader (and more powerful) realms of government. But clearly this is a development that bears watching, and one that may well bring new energies to bear on the collection of spatially disaggregated data. CONCLUSIONS AND RECOMMENDATIONS This chapter has addressed a wide range of issues, touching on both methods and substance. In concluding, we pass rather lightly over the empirical findings presented earlier in the chapter and emphasize implications for the infrastructure needed to support urban population research. Conclusions The analytic models examined in this chapter highlight a point that is often over- looked: urban growth rates and the migrant shares of growth will both tend to be high when a country is in its initial stages of urbanization; both will then tend to decline as the level of urbanization rises. The linkages of urban natural in- crease to the rate of urbanization can also be misperceived. If urban natural in- crease happens to equal rural natural increase, rural-to-urban migration will be the dominant factor in urbanization. This is the argument of the United Nations (19804. Although there was evidence of equal rates of natural increase in the 1950s and 1960s, it is unclear whether equality in the rates persists. If it does not, differences in the rates of natural increase will also exert an important influ- ence on the rate of urbanization. Migration has a distinctive role to play in af- fecting urban age structures together with lower urban fertility, it confers upon

OCR for page 108
152 CITIES TRANSFORMED city age distributions a distinctive shape in which greater proportions of the pop- ulation are found in the productive and reproductive ages. Hence, other things being held equal, rural-to-urban migration will tend to inflate urban fertility rates. Moving back one link in the chain, we find that rural natural increase, working through migration, exerts an influence on urban fertility. The interlinkages of ur- ban and rural populations are as clearly evident in analytic models as in empirical studies. Using data from the DHS, we have found that among urban women of repro- ductive age, nearly one woman in four is a recent migrant, having moved to her current city or town less than half a decade earlier. In studies of urban change, the term "migrant" calls up the image of a rural-to-urban migrant. The DHS data show, however, that most urban migrants come from other towns and cities; only about one migrant in three arrives directly from a rural area. It appears that the common view of migrants needs to be tempered by empirical realities. Re- searchers need to consider more carefully the implications of migration within the urban sector. There is little evidence to support another common perception- that migrants are more prevalent in the populations of large than small cities. For women, at least, the DHS data do not confirm this supposition. Turning to the aggregate data sources on urban and city populations, we un- derscored a familiar point, one that is mentioned in many scientific reviews of urbanization: countries define urban areas in a great variety of ways. This defi- nitional heterogeneity is of concern mainly with respect to small settlements, but because these are so numerous, differences in definition can have a large impact on the urban totals reported at national levels. In an ideal world, it might be thought desirable for countries to adopt a common definition, but this is unlikely to occur. How damaging is the absence of consensus? As discussed in Chapter 2, the theories that animate contemporary urban research are increasingly dismissive of simple urban/rural dichotomies, and point toward richer conceptualizations in- volving centrality, communication, and relational networks. These theoretical de- velopments would appear to be leading away from simple prescriptions and def- initions of urban areas. As discussed earlier, however, the measurement of such concepts is still in the early stages, and much remains to be learned about their value for empirical research, planning, and policy making. Furthermore, as can be seen throughout this report, simple urban and rural classifications retain con- siderable explanatory power. It is probably unwise to set such useful measures aside while better ones are being developed. In any case, because much of the international heterogeneity in definition applies to smaller settlements, analyses based on cities above a certain size (e.g., 100,000 population) can escape many of the difficulties. Unfortunately, the United Nations estimates of city size, as presented in World Urbanization Prospects, are more heterogeneous and subject to measurement error than is commonly realized. We reviewed several cases and found that only the most attentive and dogged researcher would be likely to understand the

OCR for page 108
URBAN POPULATION DYNAMICS 153 idiosyncrasies of the city population data. As discussed below, if the United Na- tions were to make its data and methods publicly available, a wider community of researchers could assist in improving measures and methods. Much the same can be said of the United Nations projections, which have often proven to be so far off the mark that consideration of alternative projection methods is now badly needed. Recommendations for Urban Research Infrastructure As countries urbanize, the proliferation of cities and increases in average city sizes heighten the need for adequate urban population data. There is, first, a need for acceptable estimates of city population size. Second, and especially for the larger cities, there is a greater need for intracity data, which are required both for un- derstanding social and economic diversity and for extending services. Spatial information is essential in both of these areas, for not even city sizes can be de- termined without good information on city boundaries. The potential for use of spatially collected data is perhaps even greater within cities. The difficulty lies in determining where among the many local, national, and international statistical systems there exists a combination of motivation and re- sources sufficient to generate such spatially disaggregated data. Of course, the major burden of responsibility must rest with the national statistical agencies themselves, but international researchers and agencies can make a contribution through focused research and coordination. The panel is hopeful that GIS and related technology advances will bring new energy and ideas to the problem, but sustained efforts and international technical assistance will clearly be required. The panel's impression is that where city and urban population data for de- veloping countries are concerned, most demographers believe the United Nations Population Division will somehow take care of things. Yet the Population Di- vision is but one small group of expert professionals with many responsibilities extending beyond the maintenance of urban databases. If the panel understands correctly, the Population Division manages to assemble its urban estimates and projections with very few resources, evidently dedicating less than the equivalent of a single full-time staffer to the task. The United Nations Statistical Office like- wise has many responsibilities. It would be unrealistic to suppose that these units are about to receive major new infusions of funds and personnel. Yet the status quo is a precarious arrangement. It places responsibility for urban databases essential to the demographic field on the shoulders of a very few individuals. If city and urban population data series are to be adequately and critically reviewed on an ongoing basis and if alternative forecasting methods are to be explored in any depth, a way must be found to bring greater resources to bear. More researchers, especially from the countries that contribute the data series, need to be involved, and more methodological perspectives taken into account. In the panel's view, on which we elaborate in this volume's concluding chapter, the

OCR for page 108
154 CITIES TRANSFORMED best way to attract more resources to the problem is to place the United Nations data and methods in the public domain, giving the Population Division the task of coordinating full-scale critical reviews. There is an urgent need for review of the empirical basis for city and urban population projections. United Nations projections of the populations of large cities have displayed a tendency toward upward bias. Total urban populations have also been projected to grow at rates that, in retrospect, were much too rapid. Although the United Nations has taken special care to restrain projections of city and urban population growth, it appears that these efforts have been insufficient. If the United Nations were to place its sources and methods fully in the public do- main, the problems that produce such projection errors might be diagnosed more effectively. Where the DHS program is concerned, the problem on which this chapter has shed light is the lack until quite recently of adequate spatial identifiers in the datasets released for public use. The problem, as we understand it, is that disclo- sure of the spatial locations of sampling clusters might compromise the privacy of respondents and threaten the exposure of confidential information. These are important concerns. As the research for this volume drew to a close, the DHS moved to address such concerns with a permissions policy that gives researchers access to GIS spatial identifiers for its recently fielded surveys. This is a welcome change in procedures; over time it will much enhance the value of DHS data for urban demography. We hope that the DHS will now do what it can to provide spatial identifiers for the surveys it fielded before 1999. There are potential ben- efits for rural analyses as well as urban. It may be that indicators of distance to nearby cities would suffice to measure concepts of rural "remoteness." In a world that is increasingly urban and in which rural areas are coming under the influ- ence of city economies and societies, it is difficult to imagine a next generation of demographic research that does not attend more closely to the implications of space.