Content uploaded by Jacqueline M. Klopp
Author content
All content in this area was uploaded by Jacqueline M. Klopp on Nov 08, 2023
Content may be subject to copyright.
Click here to enter text.
Cape Town, South Africa, 5 – 7 March 2024
Lessons in Traffic: Nairobi’s School Term Congestion and Equity
Challenges
Charles Hatfielda,b, Anna Kustarc, Marcel Reinmutha, Constant Capd, Agraw Ali Beshire,
Jacqueline M. Kloppf, Alexander Zipfa, James Risingg, Thet Hein Tunc
*
aHeidelberg Institute for Geoinformation Technology (HeiGIT), Heidelberg University, Berliner Str. 45 (Mathematikon)
D-69120 Heidelberg, Germany
bInterdisciplinary Center for Scientific Computing (IWR), Heidelberg University, Im Neuenheimer Feld 205
69120 Heidelberg, Germany
cRoss Center for Sustainable Cities, World Resources Institute (WRI), 10 G Street NE Suite 800. Washington DC 20002, USA
dUN Environment Programme, United Nations Avenue, Gigiri P.O. Box 30552, 00100, Nairobi, Kenya
eRoss Center for Sustainable Cities, World Resources Institute (WRI) Africa, Main Bole Road, Olympia roundabout, DRC Street, First Floor,
P.O. Box 60130, Addis Ababa, Ethiopia
fCenter for Sustainable Urban Development, Climate School, Columbia University, 475 Riverside Drive, Suite 520 New York, NY 10115, USA
gSchool of Marine Science & Policy, University of Delaware, 261 S. College Ave, 111 Robinson Hall, Newark, DE 19716-1304, USA
Abstract
On January 2nd, 2019 at 8am, Kenyatta Ave. between Waberra St. and Muindi Mbingu St. in downtown Nairobi ground to a halt,
with speeds crawling 14.8 kilometers per hour (kph) (Uber, 2023). The reason? It was the first day of school in this rapidly growing
and thriving city. The specific needs of children – and the impacts of road design, traffic, and congestion on them – tend to be
poorly addressed in transport planning, including in cities like Nairobi (Klopp, 2016). While a growing body of research on the
geography of education in African cities has delved into aspects of school travel, equity, and their effects on learning (De Kadt,
2014, 2019; Macharia et al., 2023), the particular influence of school sessions, which induces unique trip dynamics, remains largely
unexplored. This paper aims to address this gap through a data-driven analysis of traffic effects when schools are in session,
compared to holidays in Nairobi. We leverage real-time road speed information from the publicly available Uber Movement data
(Uber, 2023) to model congestion spatially and temporally. Our objective is to assess the location and intensity of congestion, as
well as its distribution across different road types. Through this analysis, we aim to gain deeper insights into the potential equity
and economic impacts of congestion, which may be closely connected to inadequate land use and planning regarding children’s
education and school travel.
©"2024"Elsevier"Inc."All"rights"reserved."
Keywords: Nairobi; congestion; Uber; speed; school; equity
* Corresponding author. Tel.: +1-202-459-7092.
E-mail address: thet.tun@wri.org
http://dx.doi.org/10.1016/j.ijepes.2021.00.000
0142-0615 /© 2024 Elsevier Inc. All rights reserved.
2 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
1. Introduction
On January 2nd, 2019 at 8am, Kenyatta Ave. between Waberra St. and Muindi Mbingu St. in downtown Nairobi
slowed to a veritable crawl of 14.8 kilometers per hour (kph) (Uber, 2023). Nearly 30% slower than Eliud Kipchoge’s
world record pace of 21.02 kph at the 2022 Berlin Marathon (McAlister, 2022). The occasion? The first day of school
in a rapidly growing, young, and thriving city. Now, Eliud Kipchoge is a world-class athlete and as such perhaps it is
an unfair comparison. But the point stands, how bad must the congestion be that a marathon runner, not even a sprinter,
can outclass a car on a downtown road by 7 kph.
Nairobi isn’t alone in these struggles; traffic congestion is the defining feature of the car-centric cities (Hrelja,
2023). The problem though is that the often-typical response to congestion tends to be to build ever bigger highways
and more roads – often at the expense of adequate facilities for buses, bikes and walking – even though evidence shows
this road building induces more car traffic and leads back to congestion in a positive feedback loop (Speck, 2018).
Moreover, building wider roads for personal vehicle travel is also linked to poor land-use patterns and access that
induce the need for longer trips to key locations (Ommeren et al., 2022). To make matters worse, these strategies are
often deeply inequitable, particularly in cities like Nairobi where most people walk or take shared means of travel,
particularly minibuses, to commute to work and carry out their daily activities (Klopp, 2012; Dixon et al., 2018).
Forced by poor land-use, traffic regulation and inadequate pedestrian and public transport infrastructure, people often
share part of the road with vehicles, increasing the risk of collisions and health issues related to air pollution (Ngo et
al., 2015; Das, 2023).
Rather than focus on road and infrastructure, it is thus critical to more carefully unpack the dynamics and specific
factors that feed into congestion and other related problems in a specific city context while taking into account travel
needs of different demographics including children (Porter and Abane, 2009; Theißen and Louen, 2018). The particular
needs of children – and the impacts of road design, traffic and congestion on them – tend to be poorly catered for in
transport planning, including in cities like Nairobi (Klopp, 2016). A growing body of important work on the geography
of education in African cities is looking at school travel, equity and impacts on learning (De Kadt, 2014, 2019;
Macharia et al., 2023). Further, a number of accessibility (Nakamura and Avner, 2021; Campbell et al., 2019; Fried et
al, 2020; Nyamai and Schramm, 2023) and congestion and travel time studies (Kasuku, 2022; KNBS, 2019) exist for
Nairobi. However, the impact on travel times of public school being in session, which induces different trip dynamics,
has not been well explored. Anecdotal evidence based on traffic complaints suggests a problem, but this is largely a
neglected area of study.
This paper thus aims to provide a more data-driven analysis on traffic effects when public schools are in session
compared to when children in these schools are on holiday in Nairobi. Specifically, we hypothesize that a combination
of poor school transport, maldistribution of high-quality schools across the city, and the reliance on personal cars for
school commutes by middle and upper class Nairobians fosters additional congestion which leads to less sleep and
more in vehicle time for all children including those in school buses and minibuses (matatus). While data on school
travel and choice is sparse, we leverage real-time road speeds from the openly available Uber Movement data (Uber,
2023) to spatially and temporally model congestion when school is in and out of session as a proxy. The goal is to
evaluate the location and intensity of congestion as well as the distribution of congestion across road types to gain a
better understanding of the potential equity and economic impacts of congestion during the school term. Specifically,
by analyzing changes in road speed on a segment-by-segment basis during the morning rush hour between the school
term and holiday periods, we start to unveil spatial and temporal patterns of congestion that we believe are linked to
poor land use and planning regarding children’s education and school travel.
1.1. Background: Urban mobility, land use and congestion in Nairobi
Nairobi is a good example of a rapidly growing city that faces congestion–and a myriad of other problems emerging
from its land-use and transport system and infrastructure. Some studies suggest that residents lose 1-2 hours during
their daily commutes at peak time (Kasuku, 2022; KNBS, 2019) and one study suggests addressing this congestion
could result in an annual saving of over US$50 million (World Bank, 2016). In its pursuit of improving the urban
transportation system, the county of Nairobi has explored various approaches, hoping to tackle the issue of congestion
effectively–with varying degrees of success and setbacks (Nairobi County, 2017, 2023).
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 3
Nairobi is characterized by a clustering of numerous opportunities and transport connections in a few core areas
including a historic central business district, and a radial road system that was designed to bring people from distant
residential areas to this core. In addition, most motorized trips are provided by a minibus (matatu) system that is
privately driven and operates on fragmented routes with many connections, centered on the Nairobi core (Walker,
2014; Klopp, 2021). Due to this spatial form and lack of a well-planned multimodal public transport system, an average
Nairobi resident has access to just 12% of all opportunities within a travel duration of 45 minutes (Fried et al., 2020).
An accessibility analysis using grid points found that within one hour of travel in Nairobi, one can reach 2% of
locations on foot and 15% by matatu as one could by car (Campbell et al., 2019). The study also highlights the impact
of approximated traffic congestion, which further reduces the average number of accessible locations by a third. This
has profound equity and transport justice impacts with those few wealthier residents in cars having substantially more
access than the average and poorer Nairobian. However, all suffer from lost time and emissions from idling cars.
Looking at the modal split of trips in Nairobi, 39% are entirely by foot, while 46% by public transport (matatus,
buses, ride hailing, or boda bodas), and only 13% by private cars (Dixon et al., 2018). In practice, most trips are multi-
modal and almost always involve some walking. A survey of urban school children’s transport estimates that 58%
take a car or bus to school and 42% travel by foot (Onywera et al., 2012). In the poorer, underserved settlements,
walking is even more prevalent with over 65% of adults and 96% of school children traveling on foot (Salon &
Gulyani, 2019).
Despite the high number of pedestrians, less than 2% of the annual road infrastructure budget is typically dedicated
to non-motorized and public transport (Nairobi County, 2017). Despite the county non-motorized transport (NMT)
policy which requires at least 20 percent of the transport budget be allocated to these modes, the vast majority of the
national transport budget goes into new road construction including in Nairobi, and these roads, such as the expressway
opened in Nairobi 2022, fail to take pedestrian safety into consideration (Nyamai, 2022). Road construction without
public transport and NMT improvements and the resulting induced demand can make travel conditions for most
residents, especially the smallest, worse. Concern arose, for example, when video showed primary school children
crossing the eight-lane highway to get from a matatu stop to their school, rather than walking an additional 600 meters
to the nearest footbridge (Lutta, 2022) but poor road safety for children is a long-standing problem (Klopp, 2016).
1.2. Going to schools in Nairobi
In Kenya, pre-primary and primary education are mandatory, and since 2003 education at public schools has been
free, successfully raising primary school enrollment to 93% before COVID-19 (UNICEF, 2021). In Nairobi City in
2019, almost 1.2 million students are attending pre-primary (263,792), primary (644,279), and secondary (263,299)
educational institutions (KNBS, 2019). (See Fig. 1 for school locations.) However, without adequate resources to cater
for all children, the quality of public education remains poor, and many believe it has declined with an inadequate
number of schools available for all children. Hence, many families, even from poor households, enroll their children
in low-cost private schools that are not always conveniently located (Zuilkowski, 2018). The number of public and
private primary schools in Nairobi has grown to over 350 schools in 2022 (Education News, 2022), providing
fundamental education to over 1 million children (KNBS, 2019).
Parents who choose where to enroll their children are more likely to select private schools for their teaching quality,
helpful teachers, and affordability than for proximity, leading to a great deal of travel to and from schools. Data from
the underserved neighborhood of Kibera in 2007 found that private schools had an average student-teacher ratio of
28:1, compared to public schools with an average ratio of 88:1 (Zuilkowski, 2018). While this difference may not be
as notable in Nairobi as a whole, perceptions of school quality impact parents’ decisions. Moreover, although public
schools are technically free, additional expenses for required uniforms, books, and examinations can make them less
affordable for poor families than some low-cost private schools, which vary in cost. Some students, particularly those
in informal settlements, may also be forced to attend private schools because public institutions are not evenly
distributed geographically to be accessible by all school-age children (Zuilkowski, 2018).
Therefore, it is not only commuters who are burdened with traffic, but also primary school children who have to
take buses or matatus to class every day, sometimes waking up at 4:30am to make it to class on time (Mwangi, 2015).
In the evening, students are also caught up in traffic, and may not arrive home until 6:30 pm or later, with little time
to do their homework. This deeper societal problem around education and access is thus likely to contribute to
4 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
congestion and this should translate into less congestion when school is out of session: during two-week breaks in
April and August, and during the longer end of year break from late October to January (Education News Hub, 2019).
Fig. 1. School locations in Nairobi. Data derived from World Bank. (https://datacatalog.worldbank.org/search/dataset/0038039)
1.3. Scalability in congestion analysis
Numerous studies have evaluated congestion using “big data” from other sources like TomTom and Google Maps
API (Christodoulou and Christidis, 2021). Many of these studies, however, primarily focus on urban areas in advanced
economies (Moya-Gómez and Palomares, 2017; Moya-Gómez et al., 2017; Moyano et al., 2021). Moreover, several
of these data sources used are either not freely available (e.g., Inrix, Mapbox, TomTom) or lack scalability (e.g.,
Google Maps API’s limitations on the number of API calls, which renders it impractical for large scale applications
involving thousands of origin-destination pairs). The pilot data from Uber Movement offers a unique and valuable
opportunity, providing open and granular speed and travel time details for select cities, including Nairobi, spanning
from January 2018 and March 2020. While previous studies have used Uber Movement speed data for analysis (see:
Mahfouz et al., 2023) for micromobility accessibility in San Francisco Bay Area), to the best of our knowledge, this
paper is the first to utilize the dataset in the context of an emerging economy to explore travel speeds when public
schools are in session in comparison to when they are in holiday.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 5
2. Methodology
2.1. Overall approach
Using Uber Movement data, the study begins with an exploratory examination that compares hourly traffic
congestion patterns across Nairobi during the school terms and the school holiday periods in 2019. Subsequently, we
conduct a more in-depth analysis of morning rush hours, involving four main assessments aimed at characterizing the
city’s congestion:
1. A descriptive time series analysis of daily morning rush hour speeds during these two periods.
2. An analysis of speed differences among different road typologies, as determined by OpenStreetMap (OSM)
highway tags, to understand the distribution and degree of congestion across various road hierarchies.
3. An analysis of percentage change in travel time between the two periods, by routing from centroids of regular
grid cells from the Central Business District (CBD) to all other hex grids.
4. A statistical analysis that employs a combination of a t-test and two bootstrap sampling methods to identify
statistically significant differences in mean road speeds between the two periods.
2.2. Data
The Uber Movement dataset is created through their ride hailing service with measured speeds coming from Uber
drivers during their trips1. Road speeds are calculated based on the aggregated and anonymized mean travel speeds of
drivers on a road segment over the course of an hour. The data is compiled as monthly datasets with every observation
as a single road segment identified by a pair of to and from segment Ids. Additional variables include OSM Way Ids,
UTC datetime, hour, hourly mean speed in kilometer per hour (kph) and standard deviation of the speed measurements.
Using OSM Way Ids, Uber Movement data can be directly related to OSM road networks, facilitating open-source
routing and accessibility analyses with Uber Movement road speeds. In total, there are 55.7 million road segment
measurements or observations in the 2019 Uber Mobility dataset for Nairobi, Kenya. In the morning rush hour, which
is the focus of this study and is defined as 6am - 9am, there are 4.2 million observations with 17,544 unique road
segments when considering directionality. In terms of coverage, the data includes 98% of the motorways, primary,
secondary, and trunk roads for Nairobi, 88.7% of tertiary roads, and 9.5% of residential roads as measured by their
total length available in OSM in 2019. More simply, if there are 100 km of motorways in Nairobi, then 98 km is
included in our dataset.
For this study, all three school semesters (known locally as “terms”) are included as well as the three holiday periods
(see Table 1). Note that the third school term (August 26 to October 25) is shorter than the others as we wanted to
control for the national secondary exam period in November.
Table 1. Dates of school terms and holiday periods for the year 2019.
School terms
Holiday periods
January 2 – April 5
April 29 – August 2
August 26 – October 25
April 6 – April 28
August 3 – August 25
October 26 – December 31
2.3. Procedures
2.3.1. OSM road network matching
For all the four main morning-rush-hour analyses, the initial step involves matching Uber Movement data to OSM’s
road network. We rely on Uber’s NPM Toolkit (Schnurr, 2019), a command line toolkit, to join the two datasets. This
1 As of October 2023, public access to the data has been closed. The white paper is still available on the Internet Archives (Uber Movement,
2019). All data will be made available publicly.
6 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
allows us to associate road segment Ids and their respective speeds to OSM highway types, and geometries. The toolkit
extracts the OSM road network and slices the network by nodes to create shorter road segments as OSM road objects
can vary considerably in length. Nodes in this case refer to any given intersection or available turn on the road. This
ensures a degree of regularity to road segment lengths and allows for greater granularity of mean road speeds as road
speeds are captured between each intersection and turn.
2.3.2. Road network routing
For the travel time difference analysis, our main tool is Openrouteservice (Ludwig et al., 2023; Zia et al., 2022). It
is a free and open-source routing engine that allows users to calculate point-based (single origin) travel times and
routes, as well as time- and distance-based isochrones (Petricola et al., 2022; Klipper et al., 2021; Geldsetzer et al.,
2020). By using OSM data, the engine compiles road network data into network graph weights, considering different
mobility profiles (car, bicycle, pedestrian). Of relevance to our study is its capability to incorporate third-party data
sources, in this case, speeds recorded by Uber Movement.
Openrouteservice requires the following inputs: An area of interest (through which the routing engine extracts the
underlying road network via OpenStreetMap), a set of origin points, a set of destination points, and a mobility profile.
In our study, our area of interest is defined by the boundary of the city of Nairobi, and our mobility profiles include
motorized travel and the measured speeds for the two time periods. To simulate representative traffic patterns, we
mapped our routes from a regular hexagonal grid with 914 cells to the CBD. The CBD was selected as the destination
because it serves as the city’s central transportation hub (Walker, 2014; Klopp, 2021).
In our analysis, travel time loss is determined by computing the percentage change in travel duration between school
days compared to holidays, using the latter as reference. A positive value thus indicates that travel times on school
days are longer than on holidays, while a negative value suggests that travel times on school days are shorter.
2.3.3. Statistical difference
We use a t-test as well as two bootstrapping approaches to examine statistical difference between the mean road
segments speeds for morning rush hours between the two periods. For the t-test, our large sample sizes of 3.2 million
observations for the school term and 1.04 million observations for the holiday help to satisfy assumptions of normality
and a representative sample–particularly, when considering coverage for major road types in the dataset. Specifically,
we use Welch’s t-test in place of a classic t-test since prior f-tests reveal statistically significant different variances
between the school term and holiday morning rush hour samples.
However, when dealing with a sizable sample, such as “big data,” one can almost always expect a statistically
significant result (Lin et al., 2013). To address this, we perform two bootstrap sampling approaches that control both
spatial clusters and temporal patterns in the data. To identify spatial clusters, hierarchical clustering is employed to
capture spatial groups among road segments using rescaled and recentered mean speed values (McQuitty, 1966). Road
segments with over 50% missing values are excluded, and clustering is based on dissimilarity calculated from pairwise
correlations. Additionally, specific hours and days are filtered during the bootstrap sampling process to account for
temporal patterns such as the unique dynamics of a Monday versus a Tuesday or Friday.
In the first bootstrap, we sample once randomly from each spatial cluster (n = 277) and filter for a random hour
during the morning rush hour, e.g., 6am, for each iteration and then sample across 20 randomly selected days for each
period. Therefore, one iteration will have a maximum of 5,440 observations all at the same random hour. In some
instances, there may be fewer observations in a given sample as not all clusters have observations in all combinations
of days and hours. This sampling is done simultaneously for both the school term and holiday period as such each
paired iteration uses the sample randomly selected hour. We then take the mean of each sample for each period and
then take the difference between the two. This ensures we are comparing the two periods at the same hour, but
randomly choosing an hour, while controlling for spatial clustering, we repeat this for 10,000 samples. Finally, we
take the mean difference between the two groups across all 10,000 samples.
For the second bootstrap, we use the same spatial clusters and also filter for a random hour but in this case also
specifically sample randomly from the school term and holiday period every workday, i.e., one random Monday, one
random Tuesday, etc. Therefore, each iteration will have observations from every cluster at a random hour for each
day of the week, for a maximum of 1385 observations. Again, we simultaneously perform the sampling and filtering
for both the school term and holiday.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 7
3. Findings
3.1. Exploratory data analysis: hourly traffic speeds
Aggregated hourly mean speeds (in kph) in Nairobi reveal differences in mean speeds between the school term and
the holiday period during the morning rush hour (between 6am and 9am), suggesting that when school is in session
congestion increases overall across the city (Fig. 2). This contrasts with mean speeds from 10pm to 5am, when school
term road speeds are higher than the holiday period, and during the daytime when speeds are equal between the two
periods. Specifically, there is a sharp dip in mean road speeds from 4am to 6am when semester mean road speeds drop
from a mean of 50 kph to 36 kph, compared to the holiday period’s 48 kph to 38 kph. This sudden reversal in mean
speeds suggests that not only is there a clear distinction between off-hours traffic and peak traffic, but that the peak
starts very early in the morning. Further, that speeds effectively bottom out across both periods at 7am suggests that
Nairobi’s road network is at capacity throughout daytime hours.
Fig. 2. Aggregated hourly variations in mean traffic speeds (kph) from Uber for all measured road segments in Nairobi, Kenya.
3.2. A closer look: Morning rush hours
3.2.1. Daily speeds
The difference in morning rush hour speeds between the school terms and holiday periods becomes more evident
when observed daily across the entire year (Fig. 3a). Horizontal dashed bars in Fig. 3a indicates the mean road speeds
for each school term and holiday period in 2019. During the school terms, weekday morning rush hour speeds are
visibly lower than the holiday periods, while weekends and public holidays are substantially higher. The start of each
school term, compared to the end of the holidays, stands out with sharp drops in speed between the respective periods.
School Terms 1 and 3 appear to follow more similar temporal patterns than Term 2; this may be due to Term 2
overlapping with the summer holiday period for the northern hemisphere and the varying schedules of private and
public school terms. During this time, international schools that follow the northern hemisphere school calendar are
also closed.
8 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
Motorways exhibit the highest mean speeds, compared to other road types (Fig. 3b). For all road types, the
magnitude of change in daily mean morning rush hour speeds between the school term and holiday periods were
dampened. This suggests that the differences in mean road segment speed between the school term and holiday periods
may be cumulative in nature, with small differences aggregating to a more substantial difference between the two
periods.
Fig 3. (a). Top: Daily mean morning rush hour speeds across 2019. (b) Bottom: Daily mean morning rush hour speeds by OSM highway types.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 9
3.2.2. Road typologies
Not all road types contributed equally to the differences measured between the school term and holiday period.
After selecting the top 10 and 20 percentiles of road segments by the greatest increases in mean speed from the school
term to holiday period. We found that secondary roads, which in this context refer to arterial roads within the street
hierarchy methodology, were statistically significantly overrepresented in the top 10 and 20 percentiles and
underrepresented in the bottom 20 percentile according to a two-sided binomial test (Table 2). Similarly, tertiary roads
were overrepresented in the top 10 and 20 percentiles. Motorways, on the other hand, were statistically significantly
underrepresented in the top 20 percentile and primary roads in the top 10 percentile. The implication of which is that
we are seeing a structural overburdening of specific road types during the school term morning rush hour in Nairobi,
likely because of prior land use and transportation planning in the city.
Table 2. Under- and over-representation of OSM highway types in top quantiles for difference in speed between school term and holiday.
Highway Type
Top 10 (N = 1531)
Top 20 (N = 3062)
Bottom 10 (N = 1531)
Bottom 20 (N = 3062)
n
nexp
n
nexp
n
nexp
n
nexp
Motorway
27
49
**
Primary
68
97
**
Secondary
544
389
**
1025
777
**
698
777
**
Residential
185
331
**
485
663
**
Tertiary
387
347
*
760
695
**
Trunk
93
62
**
171
124
**
41
62
**
82
124
**
n = actual number of road segments; nexp = expected number of road segments; orange = overrepresented; blue = underrepresented; * P-value
<= 0.05; ** P-value <= 0.01.
The top 10 and 20 quantiles refer to roads that were faster during the holiday period compared to the school term while the bottom quantiles
refer to roads that were faster during the school term. Secondary, tertiary, and trunk roads were all overrepresented in the top 10 and 20
quantiles meaning that these are the roads facing the greatest congestion during the school term period compared to the holiday, while
motorways and primary roads were underrepresented. Among the roads that increased in speed during the school term, only secondary roads
and tertiary roads were underrepresented. This provides further evidence of their disproportionate congestion burden during the school term.
Overall, 5,363 road segments decreased in speed during the school term, which is more than double the 2,638 which
increased in speed during the school term (Fig. 4). 7,035 road segments experienced only +/- 1.5 kph change during
the two periods. Slower roads during the school term are concentrated in northwestern Nairobi. 54% of road segments
in Kileleshwa Ward and 43.8% of road segments in Kilimani Ward (both in Dagoretti North Constituency) were slower
during the school term. And in neighboring Westlands Constituency, 46.3% of road segments in Parklands/Highridge
Ward and 41.3% of road segments in Karura Ward were slower during the school term. This compares to just 6.6% of
road segments in Kileleshwa, 8.8% of road segments in Kilimani, 7.6% of road segments in Parklands/Highridge, and
10.4% of road segments in Karura increasing in speed during the school term. Notably, all four districts are among the
most measured wards in Nairobi and therefore most trafficked by Uber drivers. Additionally, they are high income
areas where people are more likely to own cars.
10 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
Fig 4. Change in mean speeds: school terms vs. holiday periods in Nairobi, Kenya.
3.2.3. Travel time loss analysis
We found that nearly the entirety of northern Nairobi as well as much of southern Nairobi experience longer travel
times to the CBD during the school term (Fig. 5). The map depicts the percentage change in travel time to the CBD
for the school term using the travel time of the holiday period as the reference. Negative numbers mean that for this
hexagon it was faster to travel to the CBD during the school term, while positive numbers mean it was slower during
the school term. A total of 914 hexagon locations were analyzed, for 81 no routes could be calculated to/from due to
access restrictions or distance from to the next road segment. 458 hexagons have longer travel times during the school
term, whereas 375 hexagons have shorter travel times.
The northern side of Nairobi is severely impacted by school term congestion with many origins experiencing
upwards of 5% slower travel speeds during the school term compared to the holiday period. On the other hand, the
eastern and central portions of the city experience much shorter drive times during the school term, while the western
experiences a wide range of changes from some of the fastest speeds during the school term to drastic decreases in
speed during the school term. Intriguingly, percent change in mean speeds is highly spatially autocorrelated with a
resultant global Moran’s I of 0.69, implying that any given hexagon is very likely to share similar percent change in
mean speeds as neighboring hexagons.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 11
Fig 5. Percent change in travel time to the Central Business Unit (CBD).
3.2.4. Statistical analyses
Upon performing Welch's t-test, we found that both the morning rush hour mean traffic speeds were statistically
significantly different between the school term and holiday period with an estimated p-value of 2.2e-16, with the mean
speeds lower during the school term in comparison to the holiday period. A problem to consider is that t-tests assume
that observations are independent and identically distributed. However, there is both spatial dependence between
observations on connected road segments and temporal dependence by day-of-week.
In the more rigorous analyses, both bootstrapping approaches also found a statistically significant difference
between the school term and holiday period at the 95% confidence level. Results can be found in Table 3. Overall,
both bootstraps found that road speeds were slower by roughly 1 kph during the school term according to both mean
and median values.
Table 3. Mean difference in road speeds between school terms and holiday periods using bootstrapping approach.
First bootstrap
Second bootstrap
Description
Spatial clusters, and hourly filter
Spatial clusters, and hour and days filter
Observations per iteration
5,440*
1,385*
Iterations
10,000
10,000
Total observations
~54,000,000
~1,300,000
Samples below zero
2.30%
4.10%
Median difference in speed (kph)
1.12
0.99
Mean difference in speed (kph)
1.109
0.96
12 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
4. Discussion
We found that whether or not school is in session is impactful for congestion during the morning rush hours. Our
research reveals that the morning rush hours (6am - 9am) during the school term see noticeably slower road speeds
than during holidays. Importantly, these findings are underpinned by statistically significant evidence, as revealed by
both t-tests and bootstrapping approaches.
As far as the location and degree of intensity of congestion during the school term, a deeper dive into spatial patterns
unearthed disparities in congestion levels across different wards. Certain wards such as Kileleshwa Ward and Karura
Ward exhibit markedly elevated congestion during the school term. In stark contrast, other wards (e.g., Nairobi Central
Ward and Embakasi Ward) witness faster road speeds during the school term. It is worth mentioning that many of the
wards experiencing the greatest school term congestion are amongst Nairobi’s wealthier wards where parents are likely
to put children in better schools and drive them there. While this might be anticipated – given the propensity of affluent
groups to choose residences that prioritize certain amenities over proximity to the CBD – it also flags a potential
concern. Car owning elite concerns with congestion might reinforce the push for more road construction as a solution
leading to increasing car and fossil fuel dependencies, reminiscent of patterns seen in the USA.
Moreover, the data shows a varying impact of congestion based on road types. Secondary roads, which may be
linked to access to schools which are mostly located off major roads, face the most significant change in mean speeds,
while motorways and primary roads are less affected. A cursory glance might suggest that road expansion is the
solution to Nairobi's traffic woes. However, we contend that the differential congestion effects on specific roads and
in certain areas point towards the need for a more nuanced understanding of congestion and the deeper-rooted issues
behind its specific patterns. The patterns possibly hint at the commuting habits of Nairobi’s residents around education
access, and more importantly, a skewed distribution of land use and amenities as significant. There are, of course,
limitations to our findings:
• This research relied on a single year of Uber Movement data for mean road speeds, which by definition, represents
only a subset of trips, i.e., those of paying customers. As such the inherent biases of Uber Movement data, despite
its extensive coverage of Nairobi, may skew our results and give an incomplete picture of traffic dynamics in
Nairobi. As we used 2019 dataset, changes in the built environment outside this temporal scope are not included.
• As for our travel time loss estimates, we simulated trips to the CBD because the city operates on a roughly spoke-
hub model with the CBD acting as the central node in the road and transportation system. But the fact is we lacked
data on actual trips and commutes. When calculating routes, our routing engine takes the “optimal” route, but
people’s choices may differ. Such preference data would be invaluable for simulating realistic commutes and
teasing out structural patterns and likely inequities. We also were unable to consider income data in this research
that would also enable more insight around equity considerations.
• We also make various assumptions and decisions in regards to scale that may present challenges in regards to the
Modifiable Areal Unit Problem (MAUP) (Openshaw, 1983). Namely, our choice of hexagonal grid cells with
widths of 1000m may limit our ability to represent microscale dynamics as hexagonal geometries may well impact
spatial relationships and analytic results. The CBD as a single hexagon may be an oversimplification and disallows
intra-CBD commuting patterns. Additionally, by focusing our study on the administrative boundaries of Nairobi
we may be introducing the “zoning effect” through which we exclude important commuting patterns from
neighboring areas outside of the city. Nairobi has become a wider metropolitan area with a flow of commuters
from neighboring counties coming into the city to work. This in turn ties into contentious questions regarding what
is urban or in this case what is Nairobi (Brenner and Schmid, 2015)? Clearly future analyses should factor suburbs
and peri-urban areas into analyses such as this one.
Considering these limitations, future analyses could make several changes such as multi-scale analyses that
consider how congestion patterns vary across scales of spatial aggregations from micro to macro grids of differing
shapes and arrangements. The CBD and other destinations could also be decomposed to allow for more granular
analyses. Moreover, including peripheral areas outside of official administrative boundaries could also extend the
analyses to help match it to more realistic circumstances.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 13
5. Conclusions and Recommendations
Despite the limitations of this study, it provides clear evidence that congestion worsens in Nairobi during the school
term. This is likely due to the surge in vehicular volume associated with school commutes, which in turn is likely
linked to spatial mismatch between school children and school locations and the lack of reliable and safe shared school
travel whether by walking, cycling, school bus or matatu - even for the middle classes. This suggests that these factors
merit further investigation as generally unrecognized planning and social factors contributing to congestion.
It is crucial to underscore that our analysis does not in any way lead to placing blame on students, their families, or
schools for congestion. Instead, it points to more fundamental conditions and planning problems surrounding school
term-induced congestion. Indeed, this analysis points to an inherent contradiction: the very children that parents aim
to provide with the best opportunities are inadvertently exposed to heightened congestion and all its ill-effects
including less sleep and more exposure to pollution, as well as long, and potentially dangerous, commutes precisely
because of aspirations to provide them with the very best opportunities under the conditions of poor and child-
unfriendly land-use and transport planning. This underscores the overall insight that we need to go beyond a focus on
road expansion and widening to fundamentally address congestion (Ommeren et al., 2022; Speck, 2018).
Overall, this work points to the need for a paradigm shift in the approach to land-use and transport planning in
Nairobi and many other African cities, one that emphasizes better allocation of schools and other amenities and
improved overall land-use, non-motorized as well as school and public transport. This would offer students and
residents more efficient, cheaper, low emissions and healthier options for travel and an opportunity to avoid long trips.
This lowers congestion which in turn leads to children in shorter school commutes. Finally, last but not least, lower
and less polluting and stressful commutes lead to better educational and hence life outcomes (De Kadt, 2014, 2019;
Macharia et al., 2023). In essence, through this study, we advocate for an integrative approach to urban planning—
one that's proactive, sustainable, and puts the welfare of its residents, especially its children, at the forefront.
Acknowledgement and Author Contributions
The authors affirm that there are no potential conflicts of interest regarding the research, authorship, and publication
of this article. The authors confirm contribution to the manuscript as follows. Study conception and design: Charles
Hatfield, Marcel Reinmuth, Constant Cap, Jaqueline M. Klopp, Thet Hein Tun. Literature review and bibliography:
Anna Kustar. Data collection, visualization, and statistical modelling: Charles Hatfield, Marcel Reinmuth, Agraw Ali
Beshir, Alexander Zipf, James Rising. Contextualization and analysis of results: Anna Kustar, Constant Cap,
Jacqueline M. Klopp. Draft manuscript preparation and review: all authors.
Code and Data Availability
All the code used in producing this work is openly available on GitHub at https://github.com/GIScience/nairobi-
uber-access. All software utilized is free and open source. The analysis was conducted using R programming language
version 4.3.1 (R Core Team 2023). For data processing, the following libraries were employed: dplyr (Wickham 2023),
purrr (Wickham 2023), furrr (Vaughan 2022), lubridate (Grolemund & Wickham 2011) and for spatial data processing,
the sf package was used (Pebesma 2018, Pebesma & Bivand 2023). Additionally, the Openrouteservice
(https://github.com/GIScience/openrouteservice) routing software was utilized to generate routes and derive travel
time. The software was run within a Docker (Merkel 2014) container on an Ubuntu Linux machine (Canonical Ltd.
2023). Visualizations were created within R using the ggplot2 package (Wickham 2016), and desktop GIS software
QGIS version 3.32.3-Lima (QGIS Development Team 2023).
14 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
Appendix A. Mean morning rush hour speeds: school terms vs. holiday periods in Nairobi, Kenya.
The density plot compares the mean speeds of road segments measured during 6am to 9am during the school term
versus holiday period. Notably, the holiday period has a greater proportion of higher speed roads as compared to the
school term. The difference between the two periods is similarly apparent with the mean and median road speeds.
Appendix B. Percent change in travel time from the Central Business District to every school location.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 15
Appendix C. Average travel time gain and loss for travels across Nairobi
This map depicts the median difference in travel time from each hexagon to every other for the two periods school
term versus holiday. A total of 914 hexagon locations were analyzed, for 80 no routes could be calculated to/from due
to access restrictions or distance from to the next road segment. 416 hexagons increase their travel time to other
locations in the city on average, whereas 918 hexagons decrease. Negative numbers mean that for this hexagon it was
faster to travel to other segments during the school term, while positive numbers mean it was slower during the school
term.
Appendix D. Coverage of Uber speed measurements for Nairobi by major road type.
Except for unclassified and residential, all major road types are covered by more than 85% in Uber dataset.
16 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
Appendix E. Road lengths mapped for Nairobi in OpenStreetMap from 2014 to 2023.
References (for Manuscript)
Brenner, Neil, and Christian Schmid. 2015. “Towards a New Epistemology of the Urban?” City 19 (2–3). Routledge:
151–182. doi:10.1080/13604813.2015.1014712.
Campbell, Kayleigh B., James A. Rising, Jacqueline M. Klopp, and Jacinta Mwikali Mbilo. 2019. “Accessibility
across Transport Modes and Residential Developments in Nairobi.” Journal of Transport Geography 74
(January): 77–90. doi:10.1016/j.jtrangeo.2018.08.002.
Christodoulou, A., and P. Christidis. 2021. “Evaluating Congestion in Urban Areas: The Case of Seville.” Research
in Transportation Business & Management, Urban Transport Planning and Policy in a changing world: bridging
the gap between theory and practice, 39 (June): 100577. doi:10.1016/j.rtbm.2020.100577.
Das, Dillip Kumar. 2023. “Exploring the Significance of Road and Traffic Factors on Traffic Crashes in a South
African City.” International Journal of Transportation Science and Technology 12 (2): 414–427.
doi:10.1016/j.ijtst.2022.03.007.
de Kadt, Julia, Shane A. Norris, Brahm Fleisch, Linda Richter, and Seraphim Alvanides. 2014. “Children’s Daily
Travel to School in Johannesburg-Soweto, South Africa: Geography and School Choice in the Birth to Twenty
Cohort Study.” Children’s Geographies 12 (2). Routledge: 170–188. doi:10.1080/14733285.2013.812304.
de Kadt, Julia, Alastair van Heerden, Linda Richter, and Seraphim Alvanides. 2019. “Correlates of Children’s Travel
to School in Johannesburg-Soweto—Evidence from the Birth to Twenty Plus (Bt20+) Study, South Africa.”
International Journal of Educational Development 68 (July): 56–67. doi:10.1016/j.ijedudev.2019.04.007.
Dixon, Simon, Haris Irshad, and J-P Labuschagne. 2018. “Deloitte City Mobility Index - Nairobi.” Deloitte Insights.
https://www2.deloitte.com/content/dam/insights/us/articles/4331_Deloitte-City-Mobility-
Index/Nairobi_GlobalCityMobility_WEB.pdf.
Education News. 2022. “Primary Schools in Nairobi County; School Name, Sub County Location, Number of
Learners.” Educationnewshub.Co.Ke. https://educationnewshub.co.ke/primary-schools-in-nairobi-county-
school-name-sub-county-location-number-of-learners/.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 17
Education News Hub. 2019. “2019 Schools’ and Colleges’ Term Dates from the Ministry of Education.”
Educationnewshub.Co.Ke. https://educationnewshub.co.ke/2019-schools-and-colleges-calendar-from-the-
ministry-of-education/.
Fried, Travis, Thet Hein, Jacqueline Klopp, and Benjamin Welle. 2020. “Measuring the Sustainable Development
Goal (SDG) Transport Target and Accessibility of Nairobi’s Matatus.” Transportation Research Record:
Journal of the Transportation Research Board 2674 (5). doi:https://doi.org/10.1177/0361198120914620.
Geldsetzer, Pascal, Marcel Reinmuth, Paul O Ouma, Sven Lautenbach, Emelda A Okiro, Till Bärnighausen, and
Alexander Zipf. 2020. “Mapping Physical Access to Health Care for Older Adults in Sub-Saharan Africa and
Implications for the COVID-19 Response: A Cross-Sectional Analysis.” The Lancet Healthy Longevity 1 (1):
e32–e42. doi:10.1016/S2666-7568(20)30010-6.
Hrelja, Robert, and Tom Rye. 2023. “Decreasing the Share of Travel by Car. Strategies for Implementing ‘Push’ or
‘Pull’ Measures in a Traditionally Car-Centric Transport and Land Use Planning.” International Journal of
Sustainable Transportation 17 (5). Taylor & Francis: 446–458. doi:10.1080/15568318.2022.2051098.
KARA. 2017. “Non-Motorized Transport Policy for Nairobi.”
http://www.kara.or.ke/Nairobi%20NMT%20Policy%20Popular%20Version.pdf.
Kasuku, Silvester. 2022. “Urban Transport Policy and Land Use Planning Accessibility Nexus in Nairobi City.”
AFRICA HABITAT REVIEW 17 (1): 2535–2548.
Klipper, Isabell G., Alexander Zipf, and Sven Lautenbach. 2021. “Flood Impact Assessment on Road Network and
Healthcare Access at the Example of Jakarta, Indonesia.” AGILE: GIScience Series 2 (June). Copernicus GmbH:
1–11. doi:10.5194/agile-giss-2-4-2021.
Klopp, Jacqueline M. 2012. “Towards a Political Economy of Transportation Policy and Practice in Nairobi.” Urban
Forum 23 (1): 1–21. doi:10.1007/s12132-011-9116-y.
Klopp, Jacqueline M. 2020. “Children’s Needs Ignored in Nairobi’s Physical Planning.” Nation.
https://nation.africa/kenya/life-and-style/dn2/children-s-needs-ignored-in-nairobi-s-physical-planning-1186834.
Klopp, Jacqueline. 2021. “From ‘Para-Transit’ to Transit? Power, Politics and Popular Transport.” In Advances in
Transport Policy and Planning. Vol. 8. doi:10.1016/bs.atpp.2021.07.002.
KNBS. 2019a. “Nairobi - Population Statistics, Charts, Map and Location.” City Population Index.
https://www.citypopulation.de/en/kenya/admin/nairobi/47__nairobi/.
KNBS. 2019b. “2019 Kenya Population and Housing Census Volume I: Population by County and Sub-County.”
https://www.knbs.or.ke/?wpdmpro=2019-kenya-population-and-housing-census-volume-i-population-by-
county-and-sub-county.
Ledant, Martin. 2013. “Water in Nairobi: Unveiling Inequalities and Its Causes.” Les Cahiers d’Outre-Mer. Revue de
Géographie de Bordeaux 66 (263). Presses universitaires de Bordeaux: 335–348. doi:10.4000/com.6951.
Lin, Mingfeng, Henry C. Lucas, and Galit Shmueli. 2013. “Research Commentary: Too Big to Fail: Large Samples
and the p-Value Problem.” Information Systems Research 24 (4). INFORMS: 906–917.
Ludwig, Christina, Julian Psotta, Anna Buch, Nikolaos Kolaxidis, Sascha Fendrich, Mohammed Zia, Johannes Fürle,
Adam Rousell, and Alexander Zipf. 2023. “Traffic Speed Modelling to Improve Traffic Speed Estimation in
Openrouteservice: Data and Source Code. FOSS4G 2023.” Prizren: Zenodo. doi:10.5281/zenodo.8077872.
Lutta, Geoffrey. 2022. “Nairobi Expressway: Dangerous Crossing Used by School Kids Causes Uproar [VIDEO] -
Kenyans.Co.Ke.” https://www.kenyans.co.ke/news/75717-nairobi-expressway-dangerous-crossing-used-
school-kids-causes-uproar.
Mahfouz, Hussein, Adham Kalila, Bishoy Kelleny, and Abdelrahman Melegy. 2023. All Possible Commutes: How
Micromobility and Realistic Car Travel Times Impact Accessibility Analyses. NUMO.
https://www.numo.global/resources/all-possible-commutes-accessibility-analysis-micromobility-paper.
Macharia, Peter M., Angela K. Moturi, Eda Mumo, Emanuele Giorgi, Emelda A. Okiro, Robert W. Snow, and Nicolas
Ray. 2023. “Modelling Geographic Access and School Catchment Areas across Public Primary Schools to
Support Subnational Planning in Kenya.” Children’s Geographies 21 (5). Routledge: 832–848.
doi:10.1080/14733285.2022.2137388.
McAlister, Sean. 2022. “How Fast Was Eliud Kipchoge’s 2022 Berlin Marathon World Record?” Olympics.Com.
https://olympics.com/en/news/how-fast-was-eliud-kipchoge-world-record.
McQuitty, Louis L. 1966. “Similarity Analysis by Reciprocal Pairs for Discrete and Continuous Data.” Educational
and Psychological Measurement 26 (4). SAGE Publications Inc: 825–831. doi:10.1177/001316446602600402.
18 Hatfield et al./ African Transport Research Conference 2024, Cape Town, South Africa
Moya-Gómez, Borja, and Juan Carlos García-Palomares. 2017. “The Impacts of Congestion on Automobile
Accessibility. What Happens in Large European Cities?” Journal of Transport Geography 62 (June): 148–159.
doi:10.1016/j.jtrangeo.2017.05.014.
Moya-Gómez, Borja, María Henar Salas-Olmedo, Juan Carlos García-Palomares, and Javier Gutiérrez. 2018.
“Dynamic Accessibility Using Big Data: The Role of the Changing Conditions of Network Congestion and
Destination Attractiveness.” Networks and Spatial Economics 18 (2): 273–290. doi:10.1007/s11067-017-9348-
z.
Moyano, Amparo, Marcin Stępniak, Borja Moya-Gómez, and Juan Carlos García-Palomares. 2021. “Traffic
Congestion and Economic Context: Changes of Spatiotemporal Patterns of Traffic Travel Times during Crisis
and Post-Crisis Periods.” Transportation 48 (6): 3301–3324. doi:10.1007/s11116-021-10170-y.
Mwangi, James. 2015. “Nairobi Traffic Nightmare Causes Sleepless Nights to Pupils.” VOA.
https://learningenglish.voanews.com/a/nairobi-traffic-nightmare-sleepless-pupils/2866738.html.
Nairobi County. 2023. “County Integrated Development Plan.” Nairobi Government.
https://nairobiassembly.go.ke/ncca/wp-content/uploads/paperlaid/2023/NAIROBI-CITY-COUNTY-
INTEGRATED-DEVELOPMENT-PLAN-FOR-2023-2027-1.pdf.
Nairobi County. 2017. “Transport Committee Report on Consideration of Sessional Paper on Non-Motorized Policy.”
Nairobi Government. https://nairobiassembly.go.ke/ncca/wp-content/uploads/paperlaid/2018/Transport-
Committee-Report-On-Consideration-Of-Sessional-Paper-No.1-Of-2017-On-Non-Motorized-Policy.pdf.
Nakamura, Shohei, and Paolo Avner. 2021. “Spatial Distributions of Job Accessibility, Housing Rents, and Poverty:
The Case of Nairobi.” Journal of Housing Economics 51 (March): 101743. doi:10.1016/j.jhe.2020.101743.
Ngo, Nicole S., Michael Gatari, Beizhan Yan, Steven N. Chillrud, Kheira Bouhamam, and Patrick L. Kinneym. 2015.
“Occupational Exposure to Roadway Emissions and inside Informal Settlements in Sub-Saharan Africa: A Pilot
Study in Nairobi, Kenya.” Atmospheric Environment (Oxford, England : 1994) 111 (June): 179–184.
doi:10.1016/j.atmosenv.2015.04.008.
Nyamai, Dorcas Nthoki. 2022. “A Historical Account of Walking in Nairobi Within the Context of Spatial Justice.”
Urban Forum, November. doi:10.1007/s12132-022-09476-6.
Nyamai, Dorcas Nthoki, and Sophie Schramm. 2023. “Accessibility, Mobility, and Spatial Justice in Nairobi, Kenya.”
Journal of Urban Affairs 45 (3): 367–389. doi:10.1080/07352166.2022.2071284.
Ommeren, Jos van, Victor Mayland Nielsen, Francis Ostermeijer, and Hans Koster. 2022. “Cars Make Cities Less
Compact.” CEPR. https://cepr.org/voxeu/columns/cars-make-cities-less-compact.
Onywera, Vincent, Kristi Adamo, Andrew Sheel, Judith Waudo, Michael Boit, and Mark Tremblay. 2012. “Emerging
Evidence of the Physical Activity Transition in Kenya.” Journal of Physical Activity & Health 9 (May): 554–
562. doi:10.1123/jpah.9.4.554.
Petricola, Sami, Marcel Reinmuth, Sven Lautenbach, Charles Hatfield, and Alexander Zipf. 2022. “Assessing Road
Criticality and Loss of Healthcare Accessibility during Floods: The Case of Cyclone Idai, Mozambique 2019.”
International Journal of Health Geographics 21 (1): 14. doi:10.1186/s12942-022-00315-2.
Porter, Gina, and Albert Abane. 2009. “Increasing Children’s Participation in African Transport Planning: Reflections
on Methodological Issues in a Child-Centred Research Project.” In Doing Children’s Geographies. Routledge.
Salon, Deborah, and Sumila Gulyani. 2019. “Commuting in Urban Kenya: Unpacking Travel Demand in Large and
Small Kenyan Cities.” Sustainability 11 (14). Multidisciplinary Digital Publishing Institute: 3823.
doi:10.3390/su11143823.
Schnurr, D. 2019. “Uber Movement Data Toolkit.” NPM. https://www.npmjs.com/package/movement-data-toolkit.
Speck, Jeff. 2018. “Understand Induced Demand.” In Walkable City Rules: 101 Steps to Making Better Places, edited
by Jeff Speck, 64–65. Washington, DC: Island Press/Center for Resource Economics. doi:10.5822/978-1-61091-
899-2_27.
Theißen, Alexandra, and Conny Louen. 2019. “Are the Needs of Different People in Transport Planning Taken into
Account Today? A Case Study on Transport Development Plans in Germany.” Transportation Research
Procedia 41 (January). Elsevier: 283–291. doi:10.1016/j.trpro.2019.09.048.
Uber. 2023. “Uber Movement | Community.” Uber. https://www.uber.com/us/en/community/supporting-cities/data/.
Uber Movement. 2019. “Speeds Calculation Methodology.” Wayback Machine.
https://web.archive.org/web/20191126172634/https://movement.uber.com/_static/56b3b1999eb80fadffbeb9be
be9888a7.pdf.
UNICEF Kenya. 2021. “Education Programme.” https://www.unicef.org/kenya/education.
Hatfield et al./ Lessons in Traffic: Nairobi’s School Term Congestion and Equity Challenges 000 (2024) 000–000 19
Walker, Jarrett. 2014. “The Evolution of Logic in Privately Planned Transit.” Human Transit.
https://humantransit.org/2014/02/the-evolution-of-logic-in-developing-world-transit.html.
World Bank. 2016. Kenya Urbanization Review. AUS8099. World Bank, Washington, DC. doi:10.1596/23753.
World Bank. 2023. “Kenya - Schools | Data Catalog.” https://datacatalog.worldbank.org/search/dataset/0038039.
Zia, Mohammed, Johannes Fürle, Christina Ludwig, Sven Lautenbach, Stefan Gumbrich, and Alexander Zipf. 2022.
“SocialMedia2Traffic: Derivation of Traffic Information from Social Media Data.” ISPRS International Journal
of Geo-Information 11 (9). Multidisciplinary Digital Publishing Institute: 482. doi:10.3390/ijgi11090482.
Zuilkowski, Stephanie Simmons, Benjamin Piper, Salome Ong’ele, and Onesmus Kiminza. 2018. “Parents, Quality,
and School Choice: Why Parents in Nairobi Choose Low-Cost Private Schools over Public Schools in Kenya’s
Free Primary Education Era.” Oxford Review of Education 44 (2). Routledge: 258–274.
doi:10.1080/03054985.2017.1391084.
References (for Open-source Tools and Libraries)
Canonical Ltd. (2023). Ubuntu. https://ubuntu.com/
Garrett Grolemund, Hadley Wickham (2011). Dates and Times Made Easy with lubridate. Journal of Statistical
Software, 40(3), 1-25.
Merkel, D. (2014). Docker: Lightweight Linux Containers for Consistent Development and Deployment. Linux
Journal, 239. https://www.linuxjournal.com/content/docker-lightweight-linux-containers-consistent-
development-and-deployment
Pebesma, E. (2018). Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal, 10(1), 439-
446. https://doi.org/10.32614/RJ-2018-009
Pebesma, E., & Bivand, R. (2023). Classes and methods for spatial data in R. R package version 1.4-6.
https://CRAN.R-project.org/package=sf
QGIS Development Team. (2023). QGIS Geographic Information System. Open Source Geospatial Foundation
Project. https://qgis.org/
Vaughan, D. (2022). furrr: Apply Mapping Functions in Parallel using Futures. R package version 0.2.4.
https://CRAN.R-project.org/package=furrr
Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. Springer. https://ggplot2.tidyverse.org/
Wickham, H. (2023a). dplyr: A Grammar of Data Manipulation. R package version 1.1.1. https://CRAN.R-
project.org/package=dplyr
Wickham, H. (2023b). purrr: Functional Programming Tools. R package version 0.3.4. https://CRAN.R-
project.org/package=purrr
