Urbanisation and digitalisation are two of the megatrends characterising contemporary human society. Digital broadband access is an essential enabler, and despite its large growth potential, it can differ across territories. Taking a comparative approach from a global perspective, this chapter studies the relationship between urbanisation and digitalisation by looking at the quality of broadband access in urban centres using geospatial data processing. It is based on a combination of open and free data sourced from earth observation (Copernicus and Landsat programmes) to map and classify human settlements, with social sensing data to assess broadband quality with open data released by Ookla® at the grid level. We analyse the database in a stratified way to identify whether urban centres in high-income countries are better in terms of broadband connectivity compared to those in developing economies; whether urban centre population size is an advantage in the regions of the world where connectivity is low; and whether urban centres that have experienced stronger population growth in recent years display an advantage in terms of digitalisation. This work sheds light on the nature and type of deprivation related to uneven access to infrastructure, especially digital ones. The results indicate significant geographical and income disparities in terms of internet download speeds across the world. The performance of mobile and fixed broadband connectivity is different, and mobile connectivity offers a higher performance alternative to fixed networks in less affluent countries.
2.1 Introduction
Inequality is a relational concept concerning the difference in opportunities and outcomes between and within groups and territories. The Agenda 2030 set several ambitious goals to reduce inequalities and disparities across the world. In particular, Sustainable Development Goal 10 (SDG10) calls for a reduction in inequality to achieve sustainable development from the economic, social, environmental, and institutional points of view (European Commission. Directorate General for International Partnerships 2021). Together with climate change and other major anthropogenic and natural conditions, digitalisation contributes to the creation of a new generation of inequalities not previously considered. In this respect, the SDG17 aims, among other aspirations, to enhance the use of enabling technologies and to contribute to the reduction of inequality of opportunity (UN Department of Economic and Social Affairs 2015).
A growing literature has found that digital technologies reduce information friction, trade costs, and access to services (Broadbent and Papadopoulos 2013; Freund and Weinhold 2004; Zhou et al. 2022) and contribute to increasing the number of workplaces (Stenfeldt and Andersson 2016; Katz et al. 2010). Instead, there is less consensus on the relationship between broadband and economic growth, as well as broadband and population growth, especially when the endogeneity of the broadband process is taken into consideration (Mahasuweerachai et al. 2009). The availability, quality, and accessibility of digital technologies and the opportunities they offer for access to services and to work in a knowledge economy are two intertwined elements (European Commission. Directorate General for International Partnerships 2021). Therefore, these key conditions should be met for digital technologies to have a positive impact: the supply of adequate infrastructure (i.e. the hardware of single households and firms) and the digital literacy that permits widespread use of digital technology at both basic and highly advanced levels (Steyaert 2002).
Anzeige
From the infrastructural point of view, the expansion of broadband technology is significantly driven by private-sector firms focused on developing infrastructure and selling subscription services where the number of potential customers is higher, potentially strengthening pre-existing inequalities across territories and population groups. On the other hand, according to the political priorities of the administrations, recent years have witnessed the deployment of international, national, and local programmes and funds which might either exacerbate inequalities (Soja 2009) or reduce the digital divide (Graham 2005; Valenzuela-Levi 2021; Graham and Mann 2013), and favour agglomeration economies or convergence (Picot and Wernick 2007; van Winden and Woets 2004).
Since the net effect is very heterogeneous across countries, this work centres on deepening evidence on the geographic distribution of broadband availability and quality, together with an improved understanding of the relationship between broadband quality and income, population, and population growth. This chapter, therefore, aims to answer the following three research questions: (1) Are the urban centres (UCs) located in higher-income countries also performing better both in mobile and fixed broadband quality? (2) Do UCs with a higher population today perform better in broadband quality? (3) What is the relationship between population dynamics and broadband speed in UCs?
Previous studies analysed broadband connectivity from an infrastructure standpoint using official data with a low disaggregation scale (e.g. at the country level), also focusing on the rural–urban divide in certain specific countries (Mahasuweerachai et al. 2009; Arribas-Bel 2014; Otioma et al. 2019) or regions of the world (Perpiña Castillo et al. 2021). In many cases, remote sensing technology was used to directly proxy inequality (McCallum et al. 2022).
This study focuses on a spatial feature that is not directly observable by satellite sensors and presents several innovations compared with most of the previous work on the geographic distribution of digital infrastructure.
Anzeige
First, it combines three different data sets sourced from both traditional and emerging data: (a) remote sensing data combined with official population statistics to delineate the urban centres of the planet along with their population dynamics (Global Human Settlement Layer – GHSL – data); (b) traditional data on countries and territories borders (Global Administrative layer, GADM) and income by country (World Bank, WB); and (c) social sensing global data on broadband speed collected through the Speedtest® by Ookla® application.
Second, this work analyses broadband accessibility and performance in a specific geographical realm: the urban centres. Urban centres are often identified as cities in common terminology and are part of the degree of urbanisation global definition of cities as settlements (European Commission, Statistical Office of the European Union 2021). Contemporary human society is remarkably urban, and urban centres are the most densely populated places on Earth. Urban centres occupy just 1% of the world’s land area but account for more than 60% of its population (Dijkstra et al. 2020) and are responsible for 50% of anthropogenic air pollution (Crippa et al. 2021). Therefore, our investigation of broadband inequality targets the most innovative, economically vibrant, and service-endowed places on Earth, where the best conditions for broadband connectivity should be present. Previous work in the European Union showed how urban areas are generally well covered by spatial information on broadband quality (Proietti et al. 2022). This work also applies the GHSL approach of data integration to produce new comparable urban indicators with global coverage (Melchiorri 2022).
Third, this work analyses continuous values of average speed and also broadband classified in different speed classes, avoiding the common binary approach of the literature looking at the digital divide (‘available’, ‘not available’). It, thus, broadens the focus on digital differentiation, which includes more information on the quality of broadband provision (Riddlesden and Singleton 2014). This aspect is particularly informative given that, according to Alvin and Heidi Toffler, economies of speed are a very important dimension and can replace economies of scale (Riddlesden and Singleton 2014; Toffler and Toffler 1994).
2.2 Data and Methods
2.2.1 Global Human Settlement Layer Urban Centres
The United Nations Statistical Commission designated urban centres (UCs) as one of the classes of the definition of urban areas adopted for international statistical comparison (UN. Statistical Commission 2020). UCs are defined according to a set of population-based rules, as described in the degree of urbanisation methodology (European Commission, Statistical Office of the European Union 2021). This methodology identifies UCs as contiguous cells (i.e. sharing a pixel side) of 1 square kilometre having a population density of at least 1500 inhabitants per km2 of land or at least 50% of land surface covered by built-up areas, which altogether add up to 50,000 inhabitants or more (Dijkstra et al. 2020). At the basis of UCs’ delineation, there is an extensive use of optical and radar earth observation data from Copernicus Sentinel satellites to map built-up surfaces at a global scale. Extraction of built-up surfaces from satellite data leverages a global Sentinel-2 mosaic and a combination of multiple training datasets (mainly building footprints datasets). The resulting Global Human Settlement Layer Built-up surface grid (GHS-BUILT-S) layer is used as input to produce a population grid Global Human Settlement Layer Built-up population grid (GHS-POP) by means of a dasymetric disaggregation of population counts from census data targeting a built-up surface density distribution. The dataset built to conduct this study consists of circa 9000 UCs, extracted from the Global Human Settlement Layer (GHSL) classification in 2020 Global Human Settlement Layer Built-up settlement classification grid (GHS-SMOD R2022 (Schiavina et al. 2022a)), the implementation of the degree of urbanisation method (European Commission. Joint Research Centre 2022) is based on the GHS-POP R2022 (Schiavina et al. 2022b) for 2020. The UCs’ boundaries obtained are characterised by measuring their surface and population size in different years (2000–2020, in 5-year intervals), and by assigning the country or territory and the income group they belong to through zonal statistics processing in the Geospatial Information System (GIS) environment. Country data are assigned according to the Global Administrative layer (GADM 3.6 https://gadm.org/index.html), and population count is calculated using the GHS-POP R2022 layers for the respective year. According to figures and thresholds for gross national income (GNI, World Bank, NY.GNP.ATLS.CD), each country and territory is assigned to an income group, ranging from the low income (i.e. lower GNI) to the high income (i.e. higher GNI), with two intermediate groups: lower middle and upper middle.
We defined six UC population size classes for UCs, ranging from the extra small class (XS, 50–100 K people) to megacities (more than 10 M people) with the following intermediate classes: small (S, 100–250 K people), medium (M, 250 K–1 M people), large (L, 1–5 M people), extra large (XL, 5–10 M people).
2.2.2 Ookla® Global Fixed and Mobile Network Performance Maps
Data on broadband access and performance are provided by Ookla®. The data contain spatial information regarding global fixed broadband and mobile (cellular) network performance metrics at the grid level (with a grid size of 18 arc-seconds). Information on network performances is collected through the Speedtest® by Ookla® application, as a voluntary social sensing procedure, and aggregated into each square of the grid. Download speed, upload speed, and latency are averaged for each tile. For this work, the data employed refer to the last quarter of the year 2021 (October–December) and include information on (1) the average download speed of all tests performed in the tile, in kilobits per second, and (2) the number of tests taken by users in the tile. For each tile, data are available on (1) the fixed network performance for tests taken from a non-cellular connection type (e.g., Wi-Fi and Ethernet) and (2) the mobile network performance for tests taken from a cellular connection type (e.g., 4G LTE and 5G NR). To support this study, we prepared a dataset combining the GHSL UC with Ookla® data (Fig. 2.1). Tiles are spatially linked to the geographical boundaries of each urban centre to extract the corresponding data associated with population and income. Following the spatial link, we observed that of the 9119 UCs in the GHS-SMOD 2020 dataset, 988 UCs (~10%) have no associated information on fixed networks, and 684 UCs (~8%) have no associated information on mobile networks.
A flowchart of urban centers and their network performances. It starts with rasterization on the OOKLA grid, followed by U C raster, and zonal statistics. It ends with statistics per group.
Fig. 2.1
Processing flowchart: the boundaries of urban centres (UCs) are rasterised on a grid co-registered to the Ookla® network performance raster. The two rasters are then combined in a zonal statistic geo-processing to obtain the average performance per urban centre. Results are finally aggregated at different scales (groups: countries, SDG regions, and income group), weighting UCs by the number of Ookla® data
×
A visual inspection of the geographic distribution of the data showed that missing data relate to small UCs in Southeast Asia and Africa. Furthermore, some UCs are associated with few tiles within their geographic boundary, only partially covering the UCs’ entire surface and concentrated in a specific area (no isolated tiles). The lack of data for some areas in the UCs may be related to different aspects, for example, the location of urban functions in the UC (i.e. residential or services areas), the difference in income and affluence across areas in the city, etc. For fixed broadband and mobile networks, the following attributes have been calculated for each UC: (1) the average download speed (Mbps); (2) the number of tests (sum); (3) the speed class. Average speed is weighted by the number of tests performed in each UC to consider the uneven geographic distribution of measurements available in some UCs (therefore balancing out high-speed measurements with a low number of tests). Speed has been classified into three groups according to the standard approaches to average download speed (Spadafora 2022; Feijóo et al. 2018), as follows: from 0 to 30 Mbps (web surfing, e-mail, social networking, and moderate video for one or two devices); from 30 to 100 Mbps (multiple devices and online multiplayer gaming and 4 K streaming); and over 100 Mbps (more devices and sharing large files and live streaming video). The same indicators have been computed by aggregating UCs by country, country-income group, and UC size class.
2.3 Results
This section presents the results of the broadband indicators compiled for urban centres and aggregated per GADM entity, by geographical region of the world, by income group, and by population. With these results, we propose a multi-scale analysis of broadband connectivity for fixed and mobile networks. The first part of this section presents the results aggregated per country, followed by a more disaggregated analysis of urban centres.
2.3.1 Cities in 125 Countries Fall Below the Global Median Download Speed
The global median download speed from fixed lines in urban centres for 2021 is 86.45 Mbps. This value has been calculated by aggregating the 9119 urban centres for which data are available by GADM entity. Figure 2.1 displays the geographic distribution of the 125 GADM entities that fall below the median speed (in shades of red). The map clearly shows that the entire African continent is below the global median speed, with countries like the Central African Republic and Eritrea accounting for a median speed below 3 Mbps and seven other countries not achieving a median of 7 Mbps (i.e. South Sudan, Sudan, Comoros, Niger, Equatorial Guinea, Guinea-Bissau, and Chad). Overall, 50 African countries have a median speed below 30 Mbps. On the contrary, the highest median speed is reached in Reunion (> 100 Mbps), South Africa (51 Mbps), Côte d’Ivoire (43 Mbps), and Ghana (38 Mbps). Another geographical zone where median fixed broadband connectivity falls below the global median value ranges across the Middle East, Central Asia, and South East Asia. Places such as Turkmenistan (2.8 Mbps) and Afghanistan (5 Mbps) have the lowest median speed in the region. Despite falling below the global median, UCs in other developing economies in this region are well positioned in terms of download speed, i.e. Philippines (67 Mbps), India (70 Mbps), and Vietnam (77 Mbps). In the remaining regions of the world, several countries fall below the global median speed (i.e. those in Latin America and the Caribbean, Eastern Europe), yet broadband download speed is mainly above 10 Mbps (excluding Cuba <5 Mbps) (Fig. 2.2).
A map of the world with internet download speed distribution. It is divided into below and above-median speed distributions. Countries such as U S A and Russia have median speeds of above 86.45 megabits per second, whereas countries such as Australia and India have below median speeds.
Fig. 2.2
Map of GADM entities above (shades of blue) and below (shades of red) the global median fixed line download speed of 86.45 Mbps in 2021. Dark colours highlight the tails of the distribution showing the lowest and highest quintiles
×
2.3.2 Download Speed Varies by Region of the World and Income Group
About 40% of urban centres have Internet connectivity exceeding 100 Mbps, whereas the remaining 60% are equally split between the slowest connectivity class (0–30 Mbps) and the intermediate (30–100 Mbps). However, the geographical distribution of these classes shows a strong regional diversification. More than 70% of the urban centres with connection speeds below 30 Mbps are between Africa (42%) and Central and Southern Asia. In Sub-Saharan Africa, 88% of the urban centres fall in the 0–30 Mbps class. This percentage reduces to 76% in Northern Africa and Western Asia and goes down further to 41% in Central and Southern Asia. In Latin America and the Caribbean, about 23% of the urban centres belong to the 0–30 Mbps class, whereas only 13% of the urban centres in Eastern and Southern Asia are in this class. None of the urban centres in Europe or North America has an average speed between 0 and 30 Mbps. The class of intermediate download speed includes 77% of the urban centres in Australia and New Zealand, 58% of the urban centres in Central and Southern Asia, 43% of the centres in Latin America and the Caribbean, and about one-third of the urban centres in Europe. Moreover, 11% of the urban centres in Eastern and Southern Asia, Oceania, and Sub-Saharan Africa have an average download speed between 30 and 100 Mbps. Most urban centres in Eastern and Southern Asia (76%), Europe (63%), and North America (100%) are in the highest speed class (> 100 Mbps). One-fifth of the urban centres in Australia and New Zealand and over one-third of the urban centres in Latin America and the Caribbean belong to this class. The three cities in Sub-Saharan Africa that achieve a speed above 100 Mbps are all located in Reunion (Table 2.1).
Table 2.1
Share of urban centres per region of the world and per speed class
Region\speed class
0–30
Mbps (%)
30–100
Mbps (%)
> 100
Mbps (%)
Central and Southern Asia
34
47
1
Eastern and Southern Asia
14
12
57
Europe
0
16
19
Latin America and the Caribbean
9
17
9
Northern Africa and Western Asia
22
5
1
Sub-Saharan Africa
20
2
0
Northern America
0
0
12
*Australia and New Zealand, and Oceania <1% per class
Analysis by income classes clearly shows a high correlation between connectivity and affluence (Table 2.2). Urban centres in high-income countries (HICs) have significantly better connectivity than all other classes. More than 88% of urban centres in HICs belong to the >100 Mbps class, and another 12% of the class 30–100 Mbps. Conversely, more than 92% of the UCs in low-income countries (LICs) belong to the 0–30 Mbps class. Eighty-four per cent of the UCs in upper-middle-income countries (UMICs) have a connection exceeding 30 Mbps. In particular, 61% of UCs in UMICs exceed 100 Mbps, and 23% are assigned to the 30–100 Mbps class. About 17% of the UCs in this income group are in the lowest connectivity class 0–30 Mbps. Most of the UCs (51%) in lower-middle-income countries (LMICs) are in the 0–30 Mbps connectivity class, 48% are assigned to the 30–100 Mbps class, and the remaining 1% achieve a speed exceeding 100 Mbps.
Table 2.2
Share of urban centres per income group and per speed class
Income class/speed
0–30
Mbps (%)
30–100
Mbps (%)
> 100
Mbps
High-income countries (HICs)
0.2
11.6
88.2
Upper-middle-income countries (UMICs)
17
23
61
Lower-middle-income countries (LMICs)
51
48
1
Low-income countries (LICs)
92.4
7.3
0.4
2.3.3 Mobile Connectivity Offers a Higher Performance Alternative to Fixed Network in Less Affluent Countries
In the second phase of analysis, we explored the different relationships between broadband quality, income, and population size, presenting results at the level of the urban centre.
The first relationship explored is between broadband quality and income and whether UCs located in higher-income countries perform better in both fixed and mobile broadband quality. The analysis is centred on the comparison between the average speed for fixed broadband and mobile networks in the UCs and the income group to which each UC belongs. Figure 2.3 (lower panel) shows how low-income and lower-middle-income countries also have low-performance networks for speed, whereas upper-middle and high-income countries show higher speeds in the networks. Figure 2.3 (upper panel) displays the relationship between fixed and mobile connectivity holds at the urban centre level. In low- and lower-middle-income countries, the relationship appears more restricted compared to more affluent economies with a greater spread at higher speeds, and the difference between mobile and fixed broadband speed appears less visible than across more affluent economies (point cloud).
Five scatterplots of average mobile network speed versus average fixed network speed in megabits per second. Four data types including L I C, L M C, U M C, and H I C are plotted on the first graph and four individual graphs.
Fig. 2.3
Comparison of fixed versus mobile network connectivity performance by country (upper panel) and urban centre in each income group (lower panel)
×
The same analysis performed at the UC level also shows interesting results (Figs. 2.4 and 2.5), with a higher variance in speed for UCs belonging to HICs and UMICs, whereas UCs of LICs and LMICs are more regularly distributed in the low-left side of the quadrant. In high- and upper-middle-income countries, most UCs belong to the highest speed class (over 100 Mbps), whereas in the low and lower-middle-income countries, many UCs are below 30 Mbps, with only 27 UCs over 100 Mbps for lower-middle-income countries and only one for low-income countries. A significant difference is the one between China and India, where UCs displayed in the zoomed area in Figs. 2.4 and 2.5 belong to different speed classes, 30–60 Mbps in India and > 100 Mbps in China for fixed broadband, and 15–30 Mbps in India for mobile broadband.
A map of the world with internet speed distribution among population centers. It is divided among high-income, upper-middle-income, lower-middle-income, and low-income populations plotted on a bar graph. India has a high number of lower middle-income population centers.
Fig. 2.4
Map of urban centres classified by fixed network connectivity speed and population size class
A map of the world with internet speed distribution among population centers. It is divided among high-income, upper-middle-income, lower-middle-income, and low-income populations plotted on a bar graph. India has a high number of lower-middle-income population centers.
Fig. 2.5
Map of urban centres classified by mobile network connectivity speed and population size class
×
×
Moreover, significant disparities affect the share of the population located in UCs with different broadband speeds (charts in Figs. 2.4 and 2.5). For example, almost 90% of the UC population in HIC and about 60% of the one in Upper-middle income countries (UMC) is in UC with a fixed broadband class >100 Mbps. In LIC, about 50% of the UC population is in UC with a speed in the range 15–30 Mbps.
The second relationship investigated was that between broadband quality and population size, to determine whether UCs with a larger population would perform better in broadband quality and to observe any specific behaviour in low-speed/low-income countries.
Analysis of average speed by the UC population size class and speed class (Tables 2.3 and 2.4) reveals patterns of increase in average speed for larger urban centres, especially in the lower-speed classes.
Table 2.3
Average speed of fixed network connectivity per urban centre population size and speed class
Speed class (Mbps)
Urban centre size group
XS
S
M
L
XL
Megacity
50–100 K
100–250 K
250 K-1 M
1–5 M
5–10 M
>10 M
0–30
18.4
17.3
17.7
20.8
21.7
23.1
30–100
61.7
60.8
64.9
64.2
67.9
72.3
> 100
166.1
166.4
161.9
164.1
185.4
173.4
Table 2.4
Average speed of mobile network connectivity per urban centre group and speed class
Speed class (Mbps)
Urban centre size group
XS
S
M
L
XL
Megacity
50–100 K
100–250 K
250 K-1 M
1–5 M
5–10 M
>10 M
0–30
19.5
20.4
21.5
21.5
21.5
21.4
30–100
55.5
56.8
49.5
59.6
51.6
56.6
> 100
232.8
263.7
222.4
193.2
225.4
147.1
These trends are more evident in the low-speed classes of fixed connectivity (+34% and + 19% for the 0–30 Mbps and 30–100 Mbps classes, respectively, between size group S and megacities). For mobile connectivity only, the 0–30 Mbps class shows an increase in speed with UC size but only about 2 Mbps (+10%). The other classes (i.e. 30–100 Mbps and > 100 Mbps) do not show a stable increase. On the contrary, the highest speed class for mobile connectivity shows a decrease in speed with urban centre size (−44%). Therefore, we placed a specific focus on the lower-speed classes to understand these patterns in relation to the income classes of the UCs. To analyse this, we established a threshold for broadband speed to select countries with the lowest speed. The optimal choice is the median global speed: 86.45 Mbps (for fixed networks). In total, 125 countries showed a national average speed below this threshold. The results (Fig. 2.6 left) show that, for HIC, the size of the UC population does not significantly influence the average speed quality accessible in the UCs. The situation is different for the other income groups, with average speed increasing with larger sizes of the UC, especially for UMIC. Finally, LIC shows a far more moderate increase in average speed in relation to the UC population size. Moreover, the results also show that no megacity (more than 10 M people) in HIC has a speed lower than the global median (no LIC has megacities).
A bar graph of average network speed in megabits per second for fixed and mobile networks versus urban center population size class. 4 types of bars of H I C, U M C, L M C, and L I C are plotted on the graph.
Fig. 2.6
Average speed of fixed (left panel) and mobile (right panel) network connectivity in urban centres by population size class and country income group
×
Regarding mobile networks (Fig. 2.6 right), the median global speed is 33.5 Mbps. In this case, 91 countries have a national average speed below the threshold. Results in the bar plot show a different overview compared to fixed broadband. In HIC, only extra small UCs show an average speed lower than the median. In general, an increase in the size of the urban centre does not show a significant increase in the speed across the classes, and a decrease can be observed in urban centres above 5 million people in LIC. No extra-large UC or megacity in HIC has a speed lower than the global median.
In terms of lowest broadband quality, our dataset identifies that about 1.4% of the UC population in LICs has access to mobile broadband connectivity that does not reach the 3G standard (3 Mbps). This share rises to 7.8% of the UC population in LICs for a fixed broadband connectivity that does not reach the 3G standard. While at the global level, 0.3% of the UC population is in UC with an average broadband speed below the 3G standard (about 9 million people), this share of the UC population in such conditions reaches 2.1% in Northern Africa and Western Asia.
Finally, we investigated the relationship between broadband quality and population growth to understand if UCs that have been growing the most in the recent past perform better in broadband quality. To analyse the more complex relationship between population, population growth, and income across urban centres, an ordinary least squares regression was tested, taking into consideration country-fixed effects. The results (Table 2.5) show that broadband speeds, both mobile and fixed, are heavily related to the income group of the country the UC belongs to, with UCs belonging to HICs having a higher average speed (reference class) compared to UCs in LICs, LMICs, and UMICs. Moreover, the regression shows that the UC population in 2000 is positively and significantly (p < 0.01) related to the average speed in UC, even though contributing only in marginal terms. A population size difference of 1% in 2000 might be related to only a slight increase in both mobile and fixed broadband (+0.06 Mbps and 0.05 Mbps, respectively). Finally, the UC population growth between 2000 and 2020 is positively related to the current average speed but in a non-significant way (p > 0.1). The same relations and significance levels are obtained by repeating the analysis separately for each region of the world, emphasising the robustness of the results.
Table 2.5
Regression results
Variables
Average speed mobile
Average speed fixed
lnPOP_2000
5.6374***
(0.5976)
5.4066***
(0.3228)
Pop growth 2000–2020
0.0171
(0.0143)
0.0081
(0.0077)
Low income
−69.7820***
(23.4511)
−114.8959***
(12.6677)
Lower middle
−57.6524**
(23.8569)
−109.4958***
(12.8869)
Upper middle income
−69.4959***
(21.1720)
−101.5476***
(11.4366)
Observations
7.967
7.967
R-squared
0.6963
0.8726
Notes: Standard errors in parentheses
Constant controls and country-fixed effects are not reported
Romania was excluded from this analysis due to data constraints
***p < 0.01, **p < 0.05, * p < 0.1
2.4 Discussion
Based on the results presented above, we can highlight some general facts regarding the disparities in access and quality of broadband among different countries and urban centres. These considerations pertain to geographical/spatial inequality, technological/infrastructural disparities, and the effects of economic affluence and the intensity of urbanisation.
The first finding from the global analysis is that access to high-speed mobile networks is more geographically widespread than that of fixed broadband, with only approximately 8% of the UC presenting no information on broadband quality. In fact, Ookla® data on mobile connectivity were available for more UCs compared to fixed lines data. On the other hand, the quality of fixed broadband appears higher than that of mobile networks, with the global median speed for fixed networks considerably higher (85.45 Mbps) than that for mobile connections (33.5 Mbps). However, this infrastructural aspect is mitigated by a non-linearity between fixed and mobile performance and infrastructural rollout cost. For example, many countries in the Global South achieve better mobile connections than fixed ones. For example, the share of the UC population in LICs in a UC with a speed between 15 and 30 Mbps is 50% for mobile broadband compared to 30% for fixed broadband and 21% and 14% for the broadband speed class 30–60 Mbps.
Not only does mobile broadband connectivity have better performance in several urban centres in the Global South, but also it is, on average, more affordable. The 2022 Facts and Figures report of the International Telecommunication Union (2022) reports that the cost of mobile broadband connectivity as a share of per capita gross national income is about half the one for a fixed broadband connection. In Africa, a mobile broadband connection weighs 5% of the per capita gross national income, compared to more than 15% for a fixed broadband connection.
Our results also show that income and urbanisation appear as drivers for access to better connections and higher speed. This aspect is more evident for fixed broadband rather than for mobile connections. In particular, even in countries below the global median and with lower income, fixed broadband speed appears to be higher for larger UCs than for smaller ones, and we also found that for developing economies, UCs with larger population sizes tend to have faster broadband connectivity.
A third finding from the analysis is that UCs with higher population growth in the recent past would not necessarily perform better in broadband quality.
With regard to the uneven distribution of access to fixed and mobile networks, three main conditions across countries can be identified:
1.
Countries showing a balanced comparison between fixed and mobile average speed (i.e. along the bisector line in Fig. 2.3) for which we can still observe a speed gradient in terms of speed, such as balanced and fast connections, for example, Sweden (177 Mbps and 173 Mbps for fixed and mobile, respectively), or balanced but slow like Burundi (8 Mbps and 9 Mbps, respectively).
2.
In some countries, the average speed for the fixed network is higher (belonging to another category) than the mobile one, as observed in India and Brazil (see Figs. 2.2 and 2.3). This aspect can be related to the gradual speed in urbanisation and UC expansion that allowed for progressive installation of infrastructural services in the UC or to the presence of previously existing infrastructural networks (e.g., phone landlines) that have been adapted to the new technology.
3.
Other countries and UCs instead show the opposite trend, with mobile network speed higher than for fixed ones (this is the case in the West and South Africa and for some areas in Southeast Asia; see Figs. 2.2 and 2.3). This result can potentially be related to the very high speed in urban growth affecting the areas where mobile network antennas (and mobile phones) can be deployed more quickly than ground cables in several parts of the country and cities (see Table 2.5).
A closer look at the individual UC also shows an uneven distribution of broadband information within certain UCs. Data appear to be more densely concentrated in some areas of UCs, providing only partial coverage of the total area of the UCs. This lack of information for some areas of UCs can be related to the location of specific urban functions (e.g. public offices and government districts), or it can be an indicator of affluent areas within the UC boundary.
In terms of policy relevance, it should be noted how literature intertwines digitalisation and access to the Internet with economic development. For example, Van Winden and Woets (2004) investigated urban broadband in Europe in terms of broadband speed and the implications for integrated policies (i.e. social cohesion, regional development, and market effects) and the role of cities as an actor in broadband policies (infrastructure and services). To date, however, there are no global studies on this topic. Our results complement recent institutional literature showing that Internet users in urban areas are significantly higher than in rural areas and that mobile connectivity and availability of mobile phones are essential to allow access to the Internet to many (International Telecommunication Union 2021) by quantifying the connectivity speed in UCs and comparing mobile and fixed performances.
While our research aspires to mark a step forward in studying broadband inequality by adopting quantitative descriptors rather than an approach of broadband availability/unavailability, further research and different methodologies are needed to perform a more in-depth and multi-scale study of differentiated use as proposed by Dimaggio et al. (2004). In particular, research on digital inequality could focus on aspects of technological apartheid as proposed by Steyaert (2002) and be downscaled with a localised approach. Despite these constraints, our research is useful for setting the scene at a global level about broadband quality in urban centres and can help to identify case studies for more specialised local research. One of the limitations of this work is that the broadband data are not available for all UCs or the entire area within the UC boundary. Considering how the data are collected through the Speedtest® social sensing application, this can be related to different possibilities, including the absence of an Internet connection or resident population in the area so that no tests can be performed. Another hypothesis might also be that while an Internet connection exists, the users may not be interested or aware of the possibility of a speed test. This aspect might be related to the difference between digital connectivity and digital skills and literacy. In this sense, the number of tests or lack of information can be used, among other indicators, as a proxy for evaluating the digital literacy of some areas.
2.5 Conclusions
This chapter focused on broadband internet connectivity in urban centres of the world. Societies of the twenty-first century rely on digital spaces and the Internet for several essential activities. At the same time, cities are the most densely populated places on Earth and host significant innovation and progress. Our study addressed the quality of broadband internet connectivity in urban centres around the globe, leveraging geospatial data integration techniques together with several novel aspects. Primarily, it addressed broadband connectivity from a digital differentiation standpoint, moving beyond binary studies about the availability/unavailability of internet connectivity to quantifying download speeds for mobile and landline networks. Moreover, the study has a global focus and a multi-scalar approach, exploring connectivity for more than 9000 urban centres across the globe and analysing results per urban centre, country, geographical region, and income group.
Our study indicates significant disparities in the quality of broadband speed connectivity in urban centres worldwide. Disparities do not manifest solely as speed differentials between urban centres in the Global North and South or different economies but also in terms of infrastructure. While the median download speed is higher for landlines, mobile connectivity takes a higher speed to many more urban centres.
This work has significance for understanding future societal transformations and informing global policy frameworks such as the Sustainable Development Goals. Future work aims to further define inequality within urban centres, classify it in more detail, and eventually establish time series of broadband quality.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
