Where to draw the line: Data problems and other difficulties estimating urbanisation in Africa

ABSTRACT

The purpose of this article is to review the current state of knowledge on the pitfalls around definitions of the urban and the use of census data in estimating and tracking changes in urban populations. Differing criteria for the urban population and changing definitions make comparisons of levels of urbanisation very difficult across countries. Where censuses are held infrequently and administered less rigorously, accurate data on the urban population are a particular problem. This is especially the case in sub-Saharan Africa. Secondary African cities are increasingly important sites of urbanisation in sub-Saharan Africa, yet there is far less knowledge about these smaller cities. This article therefore looks at issues around estimates of populations in sub-Saharan cities and why this is a particular problem in smaller cities. Some of the implications of these problems are discussed, as well as measures to improve our understanding of the urbanisation process in smaller cities.

KEYWORDS:

• Urbanisation
• census
• sub-Saharan Africa
• secondary cities

1. Introduction

According to World Urbanisation Prospects: The 2014 Revision (United Nations, 2015), 54% of the world’s population lived in urban areas in 2014.¹ The vast majority of the world’s population growth is predicted to occur in the urban areas of less developed countries (projected growth from 2.9 billion people in 2014 to 5.2 billion by 2050). Africa and Asia are expected to continue to urbanise more rapidly than the rest of the world, with urban residents expected to account for 56% of Africa’s population by 2050. Approximately half of all urban residents lived in smaller cities of fewer than 500 000 people in 2014 (43% lived in cities of fewer than 300 000 inhabitants). These are amongst the fastest growing urban agglomerations, with populations of fewer than one million people, and are found in Asia and Africa predominantly. While the 2011 edition of World Urbanisation Prospects expected that the rural population of African countries would decline, albeit it after an increase in the first several years, between 2011 and 2050 (United Nations, 2012), the 2014 revision predicts that 79% of African countries will experience simultaneous urban and rural population growth between 2014 and 2050 (United Nations, 2015). These are substantial disjunctures in projections over a very short time period, but which estimate is used has profound implications for policy and planning. This article investigates the methodological issues and assumptions that have made projections of urbanisation difficult and how both these assumptions and their subsequent projections have informed policy-making. The article then concludes with suggestions for ways forward.

2. Age and reliability of census data

The afore-mentioned predictions are based on the observed changes in the proportion of a country’s population living in urban areas, gleaned mostly from national population censuses (United Nations, 2015). Most countries conduct national population censuses only once every 10 years, and in many no census data are available for even longer periods. The accuracy of these censuses also varies substantially from one country to the next (Satterthwaite, 2010). Censuses are also conducted at different times in different countries, making comparisons at specific points in time impossible. When there are no available census data, estimates and projections are used instead. This is a long-standing, ongoing problem. For example, only 18% of the database used in the 1999 revision of the World Urbanisation Prospects was based on data that were less than three years old (Cohen, 2004). Most of the data were much older than this, with 45% of cases based on data that were three to eight years old and 38% on data older than eight years (Cohen, 2004). In the latest revision of the United Nations estimates of urbanisation, 45% of the countries only had data from 2009 and earlier, and 20% of all countries’ data preceded 2005 (United Nations, 2014). Overall, higher-income countries have the most up-to-date and complete data, and poorer countries have the most outdated and inadequate data – African countries in particular (Cohen, 2004). The United Nations 1999 report was based on data over eight years old for more than half of the African countries. In the case of the Democratic Republic of Congo, the most recent data still refer to the 1980s. Given this fact, questions arise about the prediction in the 2011 world urbanisation estimates that the Democratic Republic of Congo is one of nine countries which will contribute 26% of the world’s urban population growth between 2011 and 2030 (United Nations, 2012). Nigeria is the other sub-Saharan African country counted among the nine biggest contributors. However, the results of all population censuses since 1963 have been roundly rejected (many would even argue that the 1963 census considerably overestimates the population, with the degree of overestimation varying widely between regions and cities) (Simon, 1997; Moriconi-Ebrard et al., 2008). The census of 1991 employed an overnight curfew to ameliorate this situation. The final population figure was more than 20 million people short of predictions, causing many to also view this census with scepticism. Discrepancies this large have wider consequences than just affecting per-capita statistics for Nigeria. They affect the group data for the whole of sub-Saharan Africa, because Nigeria represents approximately 20% of the region’s population (Tiffen, 2003).

One of the greatest challenges to long-term trend prediction and analysis in Africa is the recognised dearth of precise and reliable data, especially for the sub-Saharan region (Satterthwaite, 2010). The lack/unreliability of these data is often due to several, interrelated issues, revolving mainly around limited capacity and resources, and political instability. Worsening economic situations in many countries lead to fewer resources for research and data collection being available (Rakodi, 1997). In the late colonial/early post-colonial period, the statistical capability of African states was enhanced greatly (Jerven, 2013). However, during the structural adjustment of the 1980s and 1990s, much of this capacity was eroded due to neglect. Political insecurity has also reduced government capability to administer data collection in some countries, while civil unrest and war have prevented it entirely in other places (Rakodi, 1997). Censuses are also not purely administrative exercises, but can be highly politicised. This is particularly the case in Nigeria, where census data are used to determine political districts, the distribution of oil revenues and civil service hiring practices (federal appointments must reflect the relative ethnic and religious make-up of Nigeria) (Lalasz, 2006). The controversy and problems with the various population censuses of Nigeria have been well documented (Jerven, 2013). Thus, for example, the results of the 2006 population census were highly controversial, with many rejecting the figures for certain states as having been falsified (Bamgbose, 2009; Jerven, 2013).² These problems are not unique to Nigeria. More recently, Kenya’s 2009 census also produced some highly contested results with the figures from certain districts even being cancelled (Jerven, 2013). Censuses are also increasingly expensive, due to factors such as relatively high population growth in developing countries and costly labour-saving and time-saving equipment. Often the amounts requested from donors to fund censuses are a large proportion of their budget for a country and are viewed as too high. This leads to ‘donor fatigue’ in funding them (Leete, 2001:3). Thus, urban population statistics for many sub-Saharan African nations are often based on projections from urban trends in preceding decades.

3. Problems defining, measuring and comparing urbanisation within and between countries

Data access issues aside, in trying to assess how accurate or plausible urban population estimates and predictions of increase are, one of the most crucial factors is how the urban population is identified. Urban areas are defined differently in various countries’ censuses, and often vary from one census to another in any given country (Cohen, 2006). Urban areas experience both demographic growth and in-migration, and thus definitions of the urban population dependent on the size of the population need to be updated accordingly (United Nations, 2014). Urban areas also tend to expand physically, so definitions dependent on their boundaries need to be changed concomitantly. This can lead to reclassification of rural areas as urban and complicate the identification of trends over time. These are just some of the factors that can have a huge impact on the estimation of the size of the urban population and how it changes over time.

Currently, there are no internationally accepted, uniform criteria to define an urban area. One form of definition relies on population size and density criteria (McGranahan & Satterthwaite, 2014). Some definitions include a requirement for most of the employed in an urban area to be engaged in non-agricultural economic activity. In others, an urban area is one that performs an administrative function of the state (e.g. is an administrative centre, has a municipality, a town committee or a cantonment board) or has certain facilities and infrastructure (e.g. schools, street lighting, paved streets, medical centres, piped water and sewerage) (Cohen, 2004). Whatever criteria are used, there are a wide variety of definitions of urban and rural areas, making estimates of different countries’ and cities’ populations largely incomparable.

Cities can also have several different urban population estimates, depending on which boundary is used. Is urbanisation being gauged at the national level, over a wide regional administrative area, over an extended metropolitan area, at the individual city scale, over a central area only or even at the neighbourhood level (National Research Council, 2003; Satterthwaite, 2010)? Different scales of measuring urbanisation entail different data sources and techniques. In 2002, the United Nations estimated the population of Johannesburg, South Africa to be 2.3 million people (Cohen, 2004). Yet if one considered the Greater Johannesburg Metropolitan Region, including Pretoria, and cities in the East and West Rand and Vaal area, the population was already far greater than this by 1996, at 7.3 million (Crankshaw & Parnell, 2004). Changing city boundaries over time can also cause large disparities in numbers of urban residents between censuses. Smaller urban settlements in particular require increasingly precise data and methods to generate accurate predictions due to smaller margins of error than with larger populations in bigger urban areas.

Relying on different countries’ definitions of urban areas brings its own problems, especially in the case of smaller urban settlements. The official definition of an urban area includes settlements of over 20 000 people in most countries (Satterthwaite, 2010). There is much more variety in terms of what is included and excluded from the definition of an urban area with smaller settlements. A large swathe of the population of most countries lives in settlements of less than 20 000 people, seriously compromising the validity of international comparisons of urbanisation levels. In the sub-Saharan context, for example, a decision by an especially populous nation such as Nigeria to change the definition of an urban area in its census could dramatically alter the estimated level of urbanisation in the region (Satterthwaite, 2010). Table 1 presents the variety of definitions of an urban area that exist in the sub-Saharan region alone and how some of them have changed over time.

Table 1. Criteria for designation as an urban area in censuses in sub-Saharan Africa.

Type of definition	Parameters	Countries
Number of people in locality/agglomeration/town	An area is defined as urban if the population size is above a certain threshold^a	Angola, Benin, Côte d’Ivoire^d, Comores, Equatorial Guinea, Eritrea, Ethiopia, Benin, Ghana, Guinea-Bissau, Madagascar, Malawi, Mali, Mauritania, Nigeria, Senegal, Sierra Leone, Somalia, South Sudan
Mixed definition: number of people and economic activity	Minimum number of people^b and 50 to 75% of them engaged in non-agricultural economic activity	Botswana, Côte d’Ivoire, Democratic Republic of Congo, Zambia, Zimbabwe
Mixed definition: number of people and function of area	Minimum number of people^c and area is principle centre, municipality, town council, other urban centre or administrative headquarter	Central African Republic, Kenya^e, Niger, Uganda
Mixed definition: number of people and infrastructure	Localities of minimum 5 000–10 000 people and sufficient socio-economic and administrative infrastructure	Burkina Faso, Cameroon
Function of area	Administrative centres of prefectures, principle or district centres, capitals, district capitals, town planning areas	Cameroon, Chad, Comores, Côte d’Ivoire, Equatorial Guinea, Guinea, Guinea-Bissau, Malawi, Niger, Rwanda, Somalia, South Sudan, Togo
Mixed definition: function of area and economic activity	District headquarters and other settlements with rapid population growth and facilities that encourage people to engage in non-agricultural economic activities	Lesotho, Namibia
Places/localities officially designated as urban: Swaziland, Zimbabwe
Specific towns Burundi: Commune of Bujumbura
Cape Verde: as of 1990, municipalities of Praia, Mindelo, Sao Filipe and Espargos/Sal
Congo: as of 1984, six communes (Brazzaville, Pointe-Noire, Dolisie/Loubomo, Nkayi, Ouesso and Mossendjo)
Djibouti: for the 1956 and 1960 censuses, Djibouti (capital); for the 1983 census, the urban population of Ali-Sabieh, Dikhil, Tadjourah and Obock added; for the 1991 census, Arta, Damerjog, Yoboki and Randa added; for the 2009 census, Djibouti ville, and urban and rural sedentary populations of the regions of Ali Sabieh, Dikhil, Tadjourah, Obock and Arta added
Mozambique: for the 1980 census, 12 cities (Maputo, nine provincial capitals and the cities of Nacala-Porto and Chokwe); in the 1997 and 2007 censuses, 23 cities and 68 towns/vilas
Saint Helena: James Town (capital) and Half Tree Hollow
São Tomé and Príncipe: the district of Água Grande (São Tomé and Pantufo) and six other small settlements (as of 1991)
Other definitions Gambia: no official definition pre 1993 census; 2003 census states that a settlement is urban if it satisfies most of the following criteria: some commercial and institutional importance, non-agricultural work for most of the population, minimum 5000 people, high density, some degree of infrastructure
Mauritius: towns with proclaimed legal limits
Réunion: communes of minimum 2000 people living in houses no more than 200 m apart; or communes in which the majority are part of a multi-communal agglomeration
South Africa: classification is based on dominant settlement type and land use; cities, towns, townships, suburbs and so forth are typical urban settlements; enumeration areas with informal settlements, hostels, institutions, industrial and recreational areas, and smallholdings within or neighbouring a formal urban settlement are classified as urban
United Republic of Tanzania: 1957 and 1967 censuses, 16 gazetted townships; 1978 census onwards, definition includes several criteria, including regional and district headquarters, and wards with urban characteristics (exceed a minimal level of size-density criteria and/or many inhabitants in non-agricultural occupations); no specific numerical values of size and density identified; District/Regional Census Committees decide which wards are urban
No official definition: Seychelles, Mayotte

Source: United Nations World Urbanization Prospects: The 2014 Review (United Nations, 2015); McGranahan & Satterthwaite (2014); Tiffen (2003).

Notes: ^aDifferent minimum number of people for different countries, e.g. 1000, 1500, 2000, 2500, 3000, 5000, 10 000 or 20 000.

^bDifferent minimum number of people for different countries, e.g. 2000, 2500, 4000, 5000 or 10 000.

^cDifferent minimum number of people for different countries, e.g. 2000, 2500 or 3000.

^dSeveral countries have more than one set of criteria for defining their urban areas, so they appear under multiple categories in this table.

^eBecause of substantial changes in the 1999 census delineations of urban areas in Kenya, only the population for the ‘urban core’ should be considered to ensure consistency with previous censuses.

It could also be argued that a single definition of an urban agglomeration is not even feasible or desirable, given the variety of urban settings worldwide. For example, in Botswana, which has a relatively small population (just over two million people in 2014), the main criterion for defining an urban area is that it has at least 5000 inhabitants (United Nations, 2015). This definition would be meaningless in a country with a much larger urban population (e.g. China), where settlements considered rural often have populations in excess of this. Uganda defined an urban area as having a population of as few as 1000 people up until the 1991 census, which was scaled up to 2000 people for the 2002 census. The definition of an urban area in Mali changed from a minimum of 5000 people in the 1987 census to a minimum of 30 000 in the 1998 census and 40 000 in the 2009 census (McGranahan & Satterthwaite, 2014). ‘Urban’ was defined as having a municipal commune in 1988 in Senegal (Tiffen, 2003). Touba, Senegal’s second largest city, was considered rural from an administrative point of view but already had a population of 183 000 people in 1988. This had reportedly risen to 500 000 by 1999 – an estimated 5% of the entire country’s population.

Yet there is a long-standing call by many researchers for a consistent, internationally comparable definition of urban areas and populations. Rozenfeld et al. (2011) considered various methods for defining cities. ‘Legal’ definitions of cities are arguably less desirable due the arbitrariness of legal boundaries. Metropolitan Statistical Areas, as used in the United States, have the advantage of combining administratively defined areas using their social or economic relationships to each other. However, Metropolitan Statistical Areas are very time consuming to create, and require good qualitative skills and high levels of analysis – capacity that one would not expect to be in great supply in many African cities, and especially not outside the major primary capital cities. The City Clustering Algorithm also has the advantage of taking into account relationships, because it defines a city as a connected cluster of populated areas (defined at high resolution). However, this definition of an area is not based on administrative boundaries, and instead uses a more geographical or economic base.

Yet internationally standardised methods for defining urban areas that do not include their administrative boundaries may not be as big a priority for a national data-collecting agency as it is for researchers. Administrative definitions are widely used to differentiate between urban and rural areas. Multiple boundaries (e.g. city versus metropolitan versus city-region) cause problems when trying to understand urban growth, but have practical and political logics for different types of governance (Buettner, 2014). These miss much of the spatial–temporal complexity and dynamics of settlements; but delineation by remote sensing and satellite imagery is difficult to align with the administrative status quo. Internationally comparable measurement systems are unlikely to be any government’s main priority. In-country comparability of data over the long term is a much greater concern in terms of its utility in planning and resource allocation (Buettner, 2014). Other possible reasons for the lack of movement in the direction of a standardised definition of urban areas may include the following:

• Complexity of definition: in demography, births and deaths are relatively easy to define and measure. The population of a country is also relatively easy to define and count. Urban and rural populations are much harder to define due to multiple criteria used to identify them.

• Changing context: a further complication is that these multiple criteria are modified over time in response to changing economic, historical, administrative and transportation circumstances (Buettner, 2014).

Some organisations are attempting a standardised definition of urban areas. The Africapolis project has identified 2588 agglomerations across West Africa and classified only those with over 10 000 inhabitants as urban to facilitate regional comparisons (Moriconi-Ebrard et al., 2008). Several projects are attempting to generate uniform criteria for urban population estimates using satellite imagery of land cover or night-time lights (United Nations, 2014). To estimate the level of urbanisation, the World Bank’s agglomeration index gives a multi-dimensional measure of urbanisation including spatial data such as geo-referenced population densities and travel time (Buettner, 2014). However, these methods have not yet, and may not in the future, be able to produce a long enough time-series of data for sufficient historical analysis.³ Given these constraints in estimating global urban populations, the United Nations takes the view that ‘overall, national statistical offices are in the best position to establish the most appropriate criteria to characterise urban areas in their respective countries’ (United Nations, 2014:3).

As has been discussed here, because there is no single definition of the urban population, each country relies on its own definition. In any country, this definition changes over time, which can have major implications for what proportion of the population is considered urban. Furthermore, scaling up from tracking urbanisation trends in one country to a region is almost impossible without a standard definition, but this is not possible for a range of reasons. These factors lead to three major problems. There is not only little reliability of urbanisation trends within countries due to changing definitions of what is urban, but also little chance of comparison across countries because they use different definitions. In addition, there is no capacity to develop regional or global trend lines because of these definitional problems, and a standardised definition would have little meaning within countries.

4. Implications of methodologies and assumptions

These issues around census estimates of urban populations have significant implications for predictions of urbanisation trends. A prime example of this was the assumption by many scholars that Africa was urbanising rapidly due to large-scale in-migration (and in a context of no economic growth, unlike the rest of the developed world where urbanisation has occurred generally during times of economic growth) (Satterthwaite, 2007; Potts, 2015). Table 2 presents the predicted urban population in the year 2000, estimated in 1980 and 1996, for selected sub-Saharan African countries. The drop in numbers of urban dwellers in the year 2000 estimated from 1996 versus 1980 in most of the countries illustrates slower than expected urban growth. The discrepancy in the predictions between the two years indicates that the earlier predictions were possibly either based on too few data or perhaps even refuted by unexpected patterns of migration, especially considering civil conflict in several of these countries (Brockerhoff, 1999). Another contributing factor may have been slower than expected rates of natural increase. However, one of the most important causes was lack of economic growth, either nationally or in a specific city ‘losing out’ on investment to other urban areas in the same country (Satterthwaite, 2007). The lower part of Table 2 also presents the mean percentage and the mean absolute percentage errors in urban population projections for the year 2000 by year of prediction for the sub-Saharan African region (where the greater regional boundary has remained mostly unchanged). These positive prediction errors indicate that the urban population projections were consistently too high, although better in the shorter term than several decades ago (comparing the five-year and 20-year errors) (Cohen, 2004).

Table 2. Change in United Nations urban population projections to 2000 (thousands).

Country name	Projection for 2000 made in 1980		Projection for 2000 made in 1996		Percentage change in projections between 1980 and 1996
Democratic Republic of Congo	27 839		15 670		−45.4
Ghana	10 843		7 644		−29.5
Lesotho	217		641		195.4
Malawi	6 489		1 686		−74.0
Mozambique	3 199		7 869		146.0
Nigeria	45 041		55 561		25.8
South Africa	30 109		23 291		−22.4
Sudan	16 551		10 772		−34.9
Region	Mean percentage error^a			Mean absolute percentage error^a
	20 years	10 years	5 years	20 years	10 years	5 years
Sub-Saharan Africa	0.218	0.234	0.055	0.382	0.274	0.097

Source: Adapted from Brockerhoff (1999), Cohen (2004).

Notes: ^aTwenty-year comparison, United Nations projections for year 2000 in 1980 compared with United Nations data in 2001; 10-year comparison, United Nations projections for year 2000 in 1991 compared with United Nations data in 2001; five

-year comparison, United Nations projections for year 2000 in 1998 compared with United Nations data in 2001.

African countries, those in sub-Saharan Africa in particular, have been especially plagued by poor predictions of urbanisation. Most African countries did urbanise rapidly from the 1950s to the 1970s. One of the main reasons for this in Southern Africa countries in particular was that many cities were in effect under-urbanised through the imposition of various influx control-type measures on the native population by European colonial powers (Satterthwaite, 2007). There were severe restrictions on people’s ability to live and work in many of Africa’s cities, especially women and children joining husbands and fathers employed in urban areas. These control measures were relaxed and/or removed from the 1950s onwards, especially with increasing numbers of African countries achieving political independence. This coincided with strong global economic growth from the 1960s, leading to large investments in public infrastructure, health care and so forth in urban areas in Africa (Potts, 2012b). This helped generate increasing employment, thereby causing a period of increased migration to urban areas. After this, the problems of poorly conducted, infrequent censuses and the lack of data became acute. International organisations such as the United Nations and the World Bank continued to make population predictions for these countries based on the assumption that urbanisation was continuing at the pre-1980s pace (Potts, 2012a). Unfortunately, use of population estimates generated from increasingly incorrect assumptions about growth endured among governments, donor organisations and researchers. It often took several years for even correct census data produced later to be incorporated into projections. The idea that sub-Saharan Africa was urbanising rapidly in the face of little economic growth during the 1990s – contrary to historical urbanisation patterns in Europe and the rest of the world – was the accepted norm (Satterthwaite, 2010). By the late 1990s and 2000s, many new censuses had been conducted and their results made public in sub-Saharan Africa. These data revealed that very few countries had continued to urbanise rapidly. There was very slow or no increase in the level of urbanisation for recent inter-census periods in several sub-Saharan countries (Potts, 2009). Some had even de-urbanised over this period (Potts, 2012a).⁴ Moreover, those countries that have experienced the most urbanisation tend to have the best economic growth (Satterthwaite, 2010).

A major contributing factor to these over-estimates of urbanisation in Africa was the assumption of sustained large-scale rural–urban migration over several decades. This assumption has been refuted, and evidence shows that economic changes caused major shifts in circular migration patterns (Potts, 2008). Favourable global economic conditions in the 1960s and 1970s led to development of African towns and increased livelihood opportunities, fuelling rural–urban migration. The 1970s oil crises put paid to the global economic boom, however. Greatly reduced spending under structural adjustment programme conditions and trade liberalisation led to substantially lower incomes and standards of living in many African cities. Rural–urban migration did not continue unabated during this time, and was instead reduced in response to changing urban fortunes (Potts, 2008). An extreme example can be seen in Zambia, with net out-migration occurring in many of the towns of the Zambian Copperbelt in the 1990s onwards. This process has been facilitated by an abundance of land and continued customary rural tenure. Yet changes to major land policy being contemplated by the government could have far-reaching implications for current migration patterns. Zambia’s counter-urbanisation also highlights the need for effective policy to address urban poverty (Potts, 2005). The correct understanding of these urbanisation dynamics can lead to very different estimates of urban populations, and, therefore, policy and development implications for urban agglomerations. In addition, conducting censuses every 10 years also means a lot of migration, including shorter-term changes in net migration and circular migration, is not captured. Recognising the complex urban–rural and rural–urban migration patterns in sub-Saharan Africa is vital to gauging urban poverty and livelihood insecurity and how to ameliorate the serious situation in certain urban areas (Potts, 2009).

5. Questions around smaller cities

A lot of attention has been paid to mega-cities and their supposed explosion in population numbers. It seems instead that the largest cities have experienced slowing urban population growth, reflecting slowing national population growth (Cohen, 2006). According to the United Nations (2015), by 2030 most of the world’s urban population will be living in cities with populations of less than 500 000 people (Table 3). Data for smaller cities are less available (National Research Council, 2003). United Nations estimates exist only for cities of more than 100 000 inhabitants, and the estimates for cities with a population between 100 000 and 500 000 are mostly imprecise at best. Where there are no data for cities with under half a million inhabitants, this missing city size class is calculated as the residual between the world’s urban population and the sum of cities with half a million or more people (Buettner, 2014). Global theoretical modelling allows estimations that have a reasonably small margin of error. However, such estimates can mask substantial errors on a smaller scale. Over-estimating the population in several large cities can compensate for omitting the populations of lots of smaller cities and towns in terms of global or even regional comparisons of urbanisation. At the smaller scale, however, in terms of governance, the economy and infrastructure provision, a city of two million people poses a very different challenge to that of 200 towns of 10 000 people each (Moriconi-Ebrard et al., 2008).

Table 3. Population distribution of the world by area and size class of urban settlement (millions).

Area of residence and size class	1970	1990	2014	2030
Minimum 10 million	55	153	453	730
5–10 million	106	157	301	434
1–5 million	245	459	827	1 128
500 000–1 million	129	202	363	509
<500 000	816	1 313	1 936	2 257
Total urban	1 351	2 284	3 880	5 058
Rural	2 341	3 036	3 364	3 367
Total population	3 692	5 320	7 244	8 425

Source: Adapted from United Nations (2015).

There are also serious definitional issues around small towns. Changes in a country’s definition of ‘urban’ may result in an area being reclassified, but in the absence of the kinds of economic activities, administrative functions and specialisation usually associated with an urban area (Potts, 2008). The population of a settlement may reach a certain threshold and then be reclassified as an urban area. Yet this growth in population can be entirely natural and not due to in-migration. Rural areas can also be reclassified due to increasing economic activity that is classed as more ‘urban’. In-migration and different economic activity are related to structural change in an area. These are completely different dynamics from those in an area that has undergone mainly natural population growth, and therefore necessitate different planning and policy approaches. Potts (2008) argues that there is also evidence that many smaller town inhabitants make most of their living out of agriculture. Are these growing settlements actually rural, or urban areas that do not provide enough employment and housing for their populations who then engage in farming to survive? The suggestion is that farmers are migrating to small urban settlements mainly to access services, markets and transport. Thus there is a mixture of the advantages of the nature of the urban economy and access to services attracting migrants – knowledge of which is crucial for understanding the nature of the urbanisation process and instituting appropriate urban planning (Potts, 2008).

Data limitations aside, given the sheer numbers of people they currently house and are expected to in decades to come, smaller cities should arguably be prioritised in terms of development. Small cities in developing countries in particular tend to have less infrastructure, falling short of the basic services needs of many residents (Cohen, 2006). A 2003 study of 90 countries found that small cities have much less provision of piped water, refuse removal, electricity and schools than medium or large cities (National Research Council, 2003). There is also evidence that those living in smaller cities suffer higher rates of poverty and worse child mortality rates. Government and institutional capacity in smaller cities also tends to be lacking, making them less able to address their problems (Cohen, 2006).

6. Other consequences of incorrect population estimates

Whether dealing with small or large cities, the consequences of incorrect estimates of urbanisation can be far reaching. Population projections for urban population growth made in the 1970s and 1980s led to a popular misconception of an approaching urban ‘population explosion’ in developing nations (Brockerhoff, 1999). This provided many governments with the justification to disproportionately allocate resources and institute means of population control, for example restricting migration into cities (although these attempts were largely unsuccessful in the long term and eventually abandoned).

Estimates of urban population numbers are not just used to predict levels of urbanisation. The data regarding key social and health issues of urban populations are also lacking (Satterthwaite, 2010). The sample sizes of the national surveys from which these kinds of data are drawn are arguably insufficient for individual city-level estimation and analysis. Also, the conditions and trends in large urban conurbations are vastly different from those in smaller cities and towns. Yet it is these smaller urban areas that are seldom studied. A looming issue is also the assumption that greenhouse gas levels are high in urban areas in Asia and Africa even though the emissions data are sorely lacking in many cases. Cities are being encouraged to reduce greenhouse gas levels, rather than focusing on the potentially far more pressing consequences of climate change (Satterthwaite, 2010).

7. Improving the data

More regular censuses are needed in many countries (National Research Council, 2003). Ideally, statistical agencies need to process these data into smaller spatial units for more accurate population numbers and to assess the socio-economic situation in cities. Up-to-date data on changing patterns of land use and infrastructure development are also often lacking, especially the mapping of peri-urban areas. Remote sensing possibly provides one of the most promising steps going forward. Researchers have found that some of the data generated thus far for parts of the world other than Africa (e.g. measuring the extent of cities from aerial photographs or estimating population density from measures of light intensity) have compared well with population numbers derived from censuses (Sutton, 2000). In the case of Africa, however, projects such as Africapolis have provided quite different estimates of urbanisation than those generated from certain very controversial, highly contested census results (Potts, 2012b). Given the rigour in the processes used in Africapolis, these estimates arguably provide a more accurate picture of urbanisation than the census. High-spatial-resolution images are becoming more widely available, and continued improvements in image processing are expected to lead to improved population estimate accuracies (Wu et al., 2005). In addition, these approaches have the added advantage of locating the urban population in the natural geography, facilitating the study of people–environment interactions (Buettner, 2014).

There are also relatively cheaper, simpler methods to improve the quality of urbanisation data. Community mapping, where members of the community and civil society collect data on socio-economic differences, health status, mortality rates, infrastructure, service delivery and so forth for a city, can add valuable information to spur debate and act as a powerful political and planning tool (National Research Council, 2003; Brown, 2009). However, it is difficult to ensure that this is city wide, covers every city and is internationally comparable. Other metrics relevant to urbanisation can be added to the existing census population counts for cities (Buettner, 2014). Many national statistical departments publish city populations by size classes. Tabulations of cities by size class could be added to the data on levels of urbanisation as a comparison. A measure of error can also be added to population estimates and projections. This treatment of the data as more probabilistic rather than deterministic would arguably give a better picture of the reliability of data on urbanisation and urban population projections (Buettner, 2014).

8. Conclusion

The accuracy of estimates and predictions of city populations and urbanisation varies between countries and regions of the world. This is due in large part to the irregularity of census administration and, in many cases, their lack of precision. Population estimates for countries and cities in sub-Saharan Africa have been particularly inaccurate, given the kinds of problems discussed in this review with the poor standard and availability of their census data in the past. Differing definitions of urban areas and populations also make comparability among countries and cities very difficult. Yet ways to make these data more comparable (e.g. using remote sensing and complementary metrics) have been mentioned. While mega-cities and their growth have garnered much attention, many predict that most urban growth will take place in smaller cities. Macro-scale interrogation of weak demographic and economic data has limited utility in understanding urbanisation in sub-Saharan Africa’s smaller cities. It is increasingly important to relate these smaller-scale urban growth trends and their associated patterns of poverty and inequality to public policy. Thus, greater emphasis on sub-national scales of interrogation of the available census and spatial data is necessary. In addition, better knowledge of the linkages between smaller cities’ changing spatial, economic and social characteristics and their hinterlands is also required. Fully comprehending the complexities of the urbanisation process in these cities is vital to the quality of life of millions of urban poor in sub-Saharan Africa and other developing nations.

Acknowledgements

This University of Cape Town-led project, informally called Consuming Urban Poverty, aims to study the relationship between urbanisation in sub-Saharan Africa, urban poverty and governance through the lens of food security in three secondary sub-Saharan African cities: Kisumu, Kenya; Kitwe, Zambia; and Epworth, Zimbabwe. The support of the Economic and Social Research Council (UK) and the UK Department for International Development is gratefully acknowledged.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work forms part of the Governing Food Systems to Alleviate Poverty in Secondary Cities in Africa project, funded under the ESRC-DFID Joint Fund for Poverty Alleviation Research (Poverty in urban spaces theme). Supported by the Economic and Social Research Council (UK) and the UK Department for International Development [grant number ES/L008610/1].

Notes

1 ‘Urbanisation is an increasing proportion of a population living in settlements defined as urban centres. The immediate cause of most urbanisation is the net movement of people from rural to urban areas. There are usually extensive urban-to-rural migration flows too, but urbanisation occurs when there is more migration from rural to urban areas than vice versa. Care is needed to avoid confusing urbanisation with “urban growth” or “growth in urban population”, both of which are absolute terms rather than proportions. Natural increase has had a very important role in the growth of urban populations but not in the increase in levels of urbanisation’ (Satterthwaite, 2007:2).

2 For a discussion of some of the 2006 Nigeria census results, comparing them with Africapolis population estimates in particular, please see Potts’ ‘Challenging the Myths of Urban Dynamics in Sub-Saharan Africa: The Evidence from Nigeria’ (Potts, 2012b).

3 The Africapolis project does, however, provide a substantial database for urban settlements across West, Central and East Africa going back to early censuses, thereby giving some basis for historical analysis.

4 Note, however, that the de-urbanisation measured in some countries is specifically for particular intercensal periods in the recent past.