Urban travel in Nairobi, Kenya: analysis, insights, and opportunities

Urban travel in Nairobi, Kenya: analysis, insights, and opportunities

Journal of Transport Geography 22 (2012) 65–76 Contents lists available at SciVerse ScienceDirect Journal of Transport Geography journal homepage: w...

815KB Sizes 10 Downloads 87 Views

Journal of Transport Geography 22 (2012) 65–76

Contents lists available at SciVerse ScienceDirect

Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo

Urban travel in Nairobi, Kenya: analysis, insights, and opportunities Deborah Salon a,⇑, Eric M. Aligula b a b

The Earth Institute at Columbia University, 405 Low Library, MC 4335, 535 West 116th Street, New York, NY 10027, USA Infrastructure and Economic Services Division, Kenya Institute for Public Policy Research and Analysis, P.O. Box 56445-00200, Nairobi, Kenya

a r t i c l e Keywords: Africa Public transport Travel survey Logit model

i n f o

a b s t r a c t We use a unique travel survey data set from Nairobi, Kenya to explain why, where and how people in Nairobi travel and the implications of this behavior pattern. We provide both an in-depth exploration and analysis of the travel patterns and preferences of Nairobi residents and a discussion of the implications of these results for transport policy in this city. The data show that the lack of suitable transport infrastructure exacerbates travel challenges for residents across all income groups. A substantial portion of the local population cannot regularly afford any form of motorized transportation. They thus are forced to locate in slums near sources of employment, and the widespread lack of pedestrian and bicycle infrastructure increases the risk that they face when traveling. The middle income group who cannot afford private cars is almost completely dependent on the informal public transport system, which provides good geographic service coverage at the expense of service quality. Approximately 15% of Nairobi’s households own cars. Our analysis shows that without policies that make non-motorized transport safer and public transport service better, car ownership and use will increase sharply as the city’s residents become wealthier, further congesting already-overloaded roadways. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction In transport planning, there are two basic goals: to provide sufficient mobility to a population that they have access to jobs, goods, and services that they need and want, and to provide this mobility in such a way that minimizes the negative environmental impact of travel. Transport planning can also be applied to shape the pattern of spatial development and therefore the direction and pace of transport investment. In many wealthier cities, residents already have basic mobility, and transport planners have the luxury to focus on preserving that mobility while reducing the externalities associated with the system. In poorer cities, providing basic mobility and access is the main planning focus. In Nairobi, Kenya, the challenge for planners is that they must strive toward both of these goals at the same time. Nairobi is a city of contrasts. On the one hand, the city is the capital of Kenya, and downtown Nairobi boasts paved streets, skyscrapers, and busy street life – at least during regular business hours. Nairobi also serves as a gateway to East and Central Africa for both tourists and corporations with business interests in the region, as well as for international aid workers. On the other hand, a large fraction of the residents of the city live in extreme poverty. Nairobi’s informal settlements (a.k.a. slums) are estimated to house approxi⇑ Corresponding author. Present address: Institute of Transportation Studies, University of California, Davis, Davis, CA 95616, USA. E-mail address: [email protected] (D. Salon). 0966-6923/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtrangeo.2011.11.019

mately 1 million of the city’s residents (CBS, 1999; World Bank, 2006), and some consider this estimate to be conservative (e.g. Amnesty International, 2009). One of the key challenges for the Nairobi metropolitan area is the provision of reliable and efficient transportation for all members of society, both today and in the future. Transport planners in Nairobi have the problem of substantial numbers of people needing basic mobility services at the same time that the roads are highly congested and the social and environmental externalities of their transport system are large. The fact that Nairobi is a rapidly growing city makes the transportation challenge significantly larger. Its current population is approximately 3.2 million people, and this number is projected to grow to 7.4 million by the year 2030.1 This article describes urban travel in Nairobi using household travel survey data from 2004. Highlighting the most salient details, we then discuss the implications of current travel patterns for transport policy in the city. 2. Existing literature Most of the limited literature on urban transport in African cities focuses on either the mobility challenge of the extreme poor (e.g. Bryceson et al., 2003; Venter et al., 2007; Salon and Gulyani, 2010) or the challenge posed by the largely-informal public transport sector 1 The current population estimates, including the growth rate projections, have been compiled by the authors from the Kenya National Census Data of 1999.

66

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

(e.g. Kamuhanda and Schmidt, 2009). Much of this literature and the policy recommendations contained therein are based on general observations on the plight of the urban poor rather than on specific data and analysis that quantifies the extent of that plight. Those articles and reports that do contain specific data usually report only summary statistics of indicator variables such as transport mode shares, expenditure on transport, and car ownership levels, rather than use multivariate statistical methods to gain an understanding of how these variables interact (e.g. Behrens, 2004, 2009; Kaltheier, 2002; Bryceson and Howe, 2000; Bryceson et al., 2003; Palmer et al., 1997). The larger literature on urban transport in the developing world expands this focus to include the infrastructure and environmental challenge posed by rapid motorization (e.g. Gakenheimer, 1999; Gwilliam, 2003). As discussed by Gwilliam (2003), successful policy responses to the joint challenges of rapid motorization, road safety, and declining transport options for the poor must be integrated rather than separate. This is especially true in cities like Nairobi that have a substantial portion of the population living in poverty at the same time that wealthier residents are rapidly increasing their levels of car ownership and use, causing severe air pollution and roadway congestion. Travel survey-based analyses for African cities are rare, due largely to a lack of data. Three examples from the literature are Behrens (2004) who focused on Capetown, South Africa, Gebeyehu and Takano (2007) who studied Addis Ababa, Ethiopia, and Abane (2011) who studied four cities in Ghana. Behrens’s (2004) focus was to demonstrate and test advanced travel survey designs, and Gebeyehu and Takano (2007) focus on using advanced data analysis methods. Neither of these authors makes transport policy recommendations. Abane (2011) does aim to inform policy, but focuses only on public transport. None of these studies is based on a representative sample of the population. In comparison with these existing studies, ours contributes to the literature in two key ways. First, this is the only travel behavior study that we are aware of that uses a large representative sample of the population of Nairobi – or of any African city, for that matter. This means that our results are generalizable to the city’s population, and policy recommendations are robust. Second, we use our results to illuminate the transport planning challenges in Nairobi in a holistic way – not just focusing on the motorization rate or the Nairobi residents who live in poverty – and we provide a discussion of potential policy solutions.

3. A Nairobi travel survey The household travel survey that is the focus of this paper was conducted in Nairobi in 2004 by a research team led by the Kenya Institute for Public Policy Research and Analysis (KIPPRA). The sample includes 2105 households and nearly 7500 individuals, and was derived from the National Sample Survey and Evaluation Program (NASSEP IV) to be representative of the population of Nairobi. The NASSEP IV master sampling frame is a subset of 1999 census enumeration areas,2 maintained by Kenya’s National Bureau of Statistics and used for many household surveys in Kenya. For more information on master sampling frames, see United Nations (2005).

in the city are relatively large, averaging 3.5 people. Less than 20% of survey respondents are over 35 years of age, and only 3% of respondents are over 55 years of age (see Fig. 1). Part of the explanation for this is that the official retirement age in Kenya is 55, and many Kenyans will move out of Nairobi when they retire. As Fig. 3 makes clear, approximately two-thirds of Nairobi residents live extremely modest lives, spending less than two times the amount that the Kenyan government has deemed to be the ‘‘urban poverty line’’ for expenditure.3 We use this established urban poverty line as a benchmark to describe the relative wealth of households in our sample throughout this analysis. To do this, we divide households into expenditure categories that are multiples of this poverty line, as in Fig. 3. For an in-depth discussion of the merits of using expenditure rather than income as an indicator of wealth, see Deaton and Zaidi (2002). A spatial representation of poverty in Nairobi is given in Fig. 4. In this map, each zone is shaded according to household expenditure per adult-equivalent as reported by our sample households (including rent). The labels indicate the percent of households in each zone whose expenditure level falls below the Kenyan urban poverty line. As in most cities, poverty and wealth are both spatially concentrated in Nairobi. A city’s wealth, transport system, and pattern of land uses are codetermined. In cities such as Nairobi, where much of the population cannot afford motorized forms of transportation, physical proximity – especially to employment centers – becomes an important determinant of urban form. Fig. 5 depicts the pattern of population density in Nairobi, as given by the 1999 Kenyan Census. There is high variation in Nairobi’s density, with certain slum areas reported to reach densities of more than 100,000 people per square kilometer4 (Marras, 2011) – all housed in single-story structures – while much of the city is populated at densities comparable to US suburbs. The density of development in the slum areas is striking, and the people who live in these areas are regularly subject to highly unsanitary and unsafe conditions. However, these people do not have better options, largely because Nairobi’s transport system (due to expense and/ or reliability and/or availability) does not allow them to live in better conditions and also have access to employment opportunities. 3.2. Transport mode choices Transport mode is an important indicator of both the level of mobility that people enjoy and the social and environmental externalities associated with transportation. The remainder of this paper will explore the factors that determine mode choices in Nairobi, and discuss the transport policy implications of our findings. In general, we expect that in choosing a travel mode for a particular trip, people consider both travel time and cost, together with factors such as the comfort, safety, and reliability of travel using that mode. Travelers choose the mode that is within their travel budget (for both time and money) and offers the best experience while traveling. We also expect that travelers make their mode choices within the context of longer-term related choices about vehicle ownership and home location relative to other often-visited locations such as work or school. A more complete model of mode choice, therefore, would explicitly model all of these related choices. In this research, we do explicitly model both

3.1. Profile of the respondents Figs. 1–5 provide a basic profile of the households in the survey sample. Survey respondents were relatively young and households 2 Enumeration areas are the smallest units of census geography in Kenya, each containing approximately 100 households. They are roughly equivalent to census ‘‘blocks’’ in the United States.

3 In 2004, the urban poverty line for Kenya was 3174 Kenyan Shillings (Ksh) of monthly household expenditure per adult equivalent, excluding rent. This represents the cost of buying the amount of calories sufficient to meet a person’s recommended daily nutritional requirement (2250 calories) and minimal non-food expenditures (e.g. clothes, primary school expenses such as uniforms). For a more detailed discussion, please see World Bank (2008). 4 For reference, the population density of New York’s Manhattan Island is estimated to be just over 27,000 people per square kilometer (US Census 2009).

67

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

55+

7+

36−54

1

6 Under 16 5 2

24−35 4 16−23

Fig. 1. Age distribution in survey sample (N = 7445).

car ownership and mode choices. However, we do not explicitly model home location or destination location choices because, as discussed below, our location data are imprecise. We begin by exploring the factors determining mode choice that survey respondents reported directly. The household travel survey conducted by KIPPRA included an unusual set of questions about travel preferences, asking respondents directly about the characteristics of transport modes that are most important to them. The key question was worded as follows: ‘‘Rank the following factors or qualities in the order of importance in influencing your preferred mode of travel’’. The first column of Table 1 lists the ranked factors. Of these factors, the most important attributes are clearly physical access to the mode of travel and whether the mode is affordable to the traveler. This is reflected in the survey data, as more than half of the respondents ranked each of these characteristics in their top three. Looking at these data, the most striking point is that preference ranking of these factors is similar for most respondent groups – except that there is a clear difference between the travel preferences of the wealthiest Nairobi residents and those of everybody else. Table 1 illustrates this by listing the percent of the poorest and richest subset of households that ranked each factor in the top 3, top 5, and bottom 3. Contrasting the importance of factors for the more and less wealthy segments of Nairobi’s population, the first point that emerges is that cost is much less important for wealthier travelers than for poorer travelers. There is also a large difference between the importance placed on comfort, reliability, safety, personal security, and speed by the two groups. These differences suggest that there may be demand for two types of public transport services in Nairobi – one that is more comfortable, safe, and reliable for a higher fare, and one that is closer to Nairobi’s status quo for a lower fare. The KIPPRA travel survey also included the typical questions about where respondents go and the mode of transport they use. One variable that we use extensively in our analysis that warrants further explanation is ‘‘traveler’’. A person is classified as a ‘‘traveler’’ in our data if they indicated in the survey that their work or school location is in a different transportation zone from their home. Likewise, ‘‘non-travelers’’ work or go to school in the same transportation zone where they live. It is remarkable that only 42% of those sampled are travelers – 49% of adults and 27% of children under 16. This means that the majority of Nairobi residents have a low mobility lifestyle. Although this lack of physical mobility is a choice for some, it is likely forced upon others because they cannot afford to pay for motorized transport modes.

3

Fig. 2. Household size distribution in survey sample (N = 2104).

>5X 4−5X

Below Poverty

3−4X Poverty

2−3X Poverty

1−2X Poverty

Fig. 3. Household wealth distribution in survey sample relative to Kenyan urban poverty level (N = 2104).

As is evident from Fig. 10, many of the transportation zones used in this analysis are geographically large, making it likely that some of the respondents who cross zone boundaries – and are therefore in the traveler category – may actually be making shorter trips than some in the non-traveler category who might travel from one end of a single zone to the other. Unfortunately, this inaccuracy cannot be helped because we do not have geographic information about respondents’ origins and destinations that is more precise than these zones. The vast majority of Nairobi residents walk or use public transport as their ‘‘most frequent means of travel’’.5 The overall mode split that we observe in Nairobi is similar to that reported for many 5 Although the survey asked specifically about which particular form of public transport people used, here all of the public transport modes are aggregated into a single category (including matatus, Kenya bus service, metro shuttle, other city buses, Kenya railway, and both school and company transport). Likewise, private transport aggregates private car and taxi (12 out of 7215 sampled individuals chose taxi). Although some Nairobi residents do use bicycles for transport, the survey questionnaire did not include the option of bicycle for ‘‘most frequent means of travel’’, so bicyclists are not included in these mode choice charts.

68

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

28%

29%

29%

12%

12%

48%

66%

35%

7% 38%

35% 12% 25% 21% 16% 10% 12%

45%

28%

39%

44%

36%

17%

0%

3% Median Monthly Expenditure

17% Nairobi National Park

2950 - 4525 Ksh 4525 - 5900 Ksh

0 2.5 5

5900 - 8925 Ksh

10 Kilometers

8925 - 10350 Ksh 10350 - 34400 Ksh

Notes: Categories are data quantiles Labels indicate the percent of sample households below poverty line. Fig. 4. Spatial distribution of wealth in Nairobi (based on survey sample).

Central Business District

People Per Square Km 79 - 1000

Nairobi National Park

1100 - 3500 3600 - 8800 8900 - 20000 21000 - 55000

0 2.5 5

10 Kilometers

Note: Categories are data quantiles

Fig. 5. Spatial distribution of Nairobi’s population (based on 1999 Kenya Census data).

Table 1 Factors that influence preferred modes of travel in Nairobi. Influencing factor

Accessibility from home, work, etc. Fare/cost/expense to Destination Speed (short travel time) Frequency (short waiting time) Safety on board vehicle Reliability (dependable) Security Comfort (cleanliness, seating) Regularity (on time) Crew hospitality and Behavior Comprehensiveness of Network Flexibility/maneuverability Music in the vehicle

<3X poverty (N = 1300)

>5X poverty (N = 223)

% In top 3

% In top 5

% In bottom 3

% In top 3

% In top 5

% In bottom 3

59 55 30 27 24 22 19 19 14 11 9 8 6

76 66 48 50 42 45 34 36 37 21 18 21 10

6 18 16 11 15 13 25 21 15 39 42 33 73

63 33 22 20 35 34 28 34 15 7 5 9 3

72 41 37 43 53 60 52 55 41 13 14 18 5

4 33 17 12 13 8 14 7 15 45 40 35 80

69

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

Non−Traveler

Traveler

136

Below Poverty

NT T

1−2xPoverty

NT T

2−3xPoverty

NT T

3−4xPoverty

NT T

4−5xPoverty

NT T

>5xPoverty

NT T

307 594

1325

1086

1546

0

Walk

Public Transport

20

40

Walk

Private Transport

60

80

Public Transport

100

Car

NT=Non−Traveler, T=Traveler

Note: The numbers within each pie piece indicate the number of people in that portion of our sample.

Fig. 8. Wealth and mode choice percent – adults 16 and older.

Fig. 6. ‘‘Most frequent means of travel’’ adult mode split, age 16 and older.

Non−Traveler

Below Poverty

NT T

1−2xPoverty

NT T

2−3xPoverty

NT T

3−4xPoverty

NT T

4−5xPoverty

NT T

>5xPoverty

NT T

Traveler

34

55

246

208

325

1214

0

20

40

Walk

60

80

Public Transport

100

Car

NT=Non−Traveler, T=Traveler

Walk

Public Transport

Private Transport

Note: The numbers within each pie piece indicate the number of people in that portion of our sample. Fig. 7. ‘‘Most frequent means of travel’’ child mode split, under 16.

other African cities (see, e.g. Bryceson et al., 2003; Behrens, 2004; Palmer et al., 1997). Cities in other parts of the developing world also have high levels of walking and public transport use, but there is often also significant use of bicycling and motorized two- and threewheelers (Hook and Howe, 2005). We have separated this mode choice information by travelers and non-travelers for children and adults (see Figs. 6 and 7), and then further by household expenditure groups (see Figs. 8 and 9). Despite the above caveat regarding zone size, as expected, travelers are more likely to use public transport, while non-travelers are more likely to walk. This makes sense, as walking is a more reasonable choice of transport mode for those traveling shorter distances. Comparing Figs. 6 and 7, it is also clear that children are more likely to walk than are adults – in both the traveler and non-traveler categories. Figs. 8 and 9 illustrate mode choices by traveler status and expenditure group, for adults and children, respectively. Among adults, the mode split is as expected, with travelers walking less than non-travelers in every expenditure category, and wealthier people walking less than poorer people. Among children, these relationships are less clearly defined. The percentage of traveler children who walk is constant through the middle wealth levels and drops for the highest two expenditure groups. The percentage of non-traveler children who walk drops over the first three expenditure groups and then levels off at approximately half.

Fig. 9. Wealth and mode choice percent – children under 16.

We have also created a map to display the spatial pattern of adult traveler status and mode choices in the city. Fig. 10 displays both pieces of information. Zone shading indicates the percent of adults who travel outside their home zone for work or school. Those zones in the lower two quintiles for travelers are either centrally located or in the poorer portions of the outskirts of the city (see Fig. 4). This makes sense, as those in the city center can access jobs within their home zones, and the poor who reside in the outskirts are likely to work locally as well, since they cannot afford to travel the long distance to jobs in the city center. The zones in the highest quintile for travelers are largely the wealthier zones. Pie charts in each zone of Fig. 10 represent the mode split for adults in that zone – including information from both travelers and non-travelers. Not surprisingly, these indicate higher car and public transport use in zones that are relatively wealthy and/or far from the city center. Zones where both the traveler and the walking percentages are high are cause for concern, as those who walk long distances are likely spending substantial time and physical energy traveling. There are a number of such zones throughout the city. A major factor that determines transport mode choice is trip distance. Fig. 11 illustrates trip distance distributions in the traveler portion of the KIPPRA sample for each mode.6 As is evident

6 These distances are calculated from the center of the reported origin KIPPRA transport zone to the center of the destination transport zone. The problem with this approach is that the transport zones are large and the centers of the zones are therefore poor approximations of actual origins and destinations. Nonetheless, these values provide useful ballpark estimates for trip distances.

70

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

Public Transport

Private Transport

2

4

30

Walk

100

Fig. 10. Spatial distribution of mode choice and traveler status in Nairobi.

13

80

32 40

Percent

60

20

59

98

96

40

87 68

20

10

60

0

41

0

Below Poverty 1−2xPoverty 2−3xPoverty 3−4xPoverty 4−5xPoverty >5xPoverty

0

5

10

15

0

5

10

15

0

5

10

15

No Cars

Car Owner

Distance Fig. 12. Household car ownership by wealth level in Nairobi.

5

11 18

23

26

77

74

5

6

80

19

91

40

95

89 82

20

81

0

3.2.1. Car ownership and use in Nairobi Approximately 14% of households in our sample own cars, and 9% of adult survey respondents reported that the car was their most frequent mode of transport. Car ownership is clearly associated with wealth all over the world, and households in Nairobi follow this pattern (see Fig. 12). It is generally also true that it makes more economic sense for a household to own a car if it is a larger household. Fig. 13 makes it clear that this is true of Nairobi households – the percentage of households that own cars is higher for larger households than for small households. The spatial distribution of car ownership in the city is depicted in Fig. 14. As expected, zones with higher levels of car ownership tend to be those with higher wealth levels (as shown in Fig. 4), and also those that are farther from the city center.

9

60

from this figure, even for travelers, most trips by all modes in Nairobi are quite short: less than 5 km. The next four subsections summarize what our data reveal about Nairobi resident transport choices for each major travel mode: car, public transport, bicycle and walking. We then present evidence on household-level expenditure on transport. Finally, we present an integrated model of mode choice in Nairobi, and conclude the paper with a discussion of the policy implications of our results.

100

Fig. 11. ‘‘Traveler’’ distribution of trip distances by mode on a percentage basis.

1

2

3

No Cars

4

7 or more

Car Owner

Fig. 13. Household car ownership by household size.

As is clear from Fig. 15, there is a reason that car ownership is associated with wealth – cars are expensive. Car owning households spend far more for transport in Nairobi than car free house-

71

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

Fig. 14. Spatial distribution of car ownership and transport expenditure in Nairobi.

Car−Owner

Total

0

20

Percent

40

60

Car−Free

0

5000

10000 0

5000

10000 0

5000

10000

Ksh Fig. 15. Distribution of monthly household transport expenditures per adultequivalent by car ownership status.

holds do. The median car owner expenditure on fuel per week is 1800 Ksh. The median car owner expenditure on maintenance per month is 3000 Ksh. About a third of car owning households do not pay to park at their workplace. Of those who pay for parking, however, almost all of them pay 70 Ksh per day.7 In contrast, those who ride matatus pay either 20 or 50 Ksh per ride (depending on distance travelled), and of course those who walk do not pay any money for their transportation. Interestingly, despite the high cost of car ownership and use relative to all other modes, these survey data indicate that median expenditures on transport as a percent of total expenditures are remarkably constant across expenditure categories in Nairobi. From the poorest to the second-wealthiest group (4–5 times the poverty level of expenditure), the median percent of total expenditure on transport is 15%. This value drops slightly for the highest expenditure group – to 13%. It is worth noting here that these are relatively high percentages of total expenditure to be devoting

7

This figure has subsequently been raised to 140 Ksh per day, with indications that it may go higher. The focus on the part of the local authorities though seems to be on revenue enhancement rather than traffic management.

to transport. Kaltheier (2002) reports comparable values ranging from 8% to 15% for seven West African cities. The logit model results reported in Table 2 capture what this data can tell us about the factors that contribute to car use at the household level. The dependent variable here is binary, taking the value of one if at least one member of the household reported that a private vehicle was their most frequent mode of transport, and zero otherwise. The column ‘‘p-value’’ indicates statistical significance; if the p-value is less than 0.10, the coefficient can be interpreted as being significantly different from zero, and therefore affecting the dependent variable. All of the included variables that are statistically significant have the expected signs. Car owners, larger households, and households with more travelers are more likely to use cars, as indicated by the positive estimated coefficients for each of these variables. Estimated coefficients on the expenditure category variables should be interpreted relative to the excluded category – ‘‘Below Poverty’’. As expected, households in the upper expenditure categories are more likely to use cars. The lack of statistical significance of the coefficients on ‘‘1–2X Poverty’’ and ‘‘2–3X Poverty’’ indicate

Table 2 Logit model of the likelihood of car use in a household.

Number of travelers in HH Car owner HH size Number of children in HH 1–2X Poverty 2–3X Poverty 3–4X Poverty 4–5X Poverty >5X Poverty Distance to the CBD Area of home zone Density of home zone Constant

Coefficient

P value

0.18** 4.05*** 0.16* 0.10 0.43 0.46 1.10** 1.22** 2.01*** 0.05 0.005 3.54E5*** 4.59***

0.021 0.000 0.081 0.378 0.295 0.288 0.011 0.015 0.000 0.392 0.265 0.000 0.000

Dependent variable: household car use (0, 1) N = 2091, standardized weights are used Pseudo R-squared = 0.62 *

Statistical significance of coefficient estimates for 0.10 level. Statistical significance of coefficient estimates for 0.05 level. *** Statistical significance of coefficient estimates for 0.01 level. **

72

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

that there may be a wealth threshold below which there is little car use. Households living in more densely populated zones are less likely to use cars. 3.2.2. Traveling by bicycle in Nairobi Bicycle was not included as a mode option for ‘‘most frequent mode of travel’’ on the survey, so we do not have information about actual bicycle use. However, the survey did include a set of questions about bicycle ownership and attitudes toward bicycling. About 10% of survey respondents own bicycles. For the 90% who do not own bicycles, the survey asked why, whether they would use a bicycle for commuting if they had one, and if not, why not. The message from the survey respondents is that Nairobi roads are not safe for cyclists. Almost half of respondents who do not own bicycles report that their reason is that cycling is not safe in Nairobi, and predictably most of these respondents would not commute by bicycle even if they had one. An additional quarter of non-bike owners report that their reasons are personal security concerns or that they do not know how to ride a bicycle. Less than 20% of these respondents would use a bicycle for commuting if they had one. Of all those who said they would not commute by bicycle even if they had one, 85% cited safety as their reason. Until roads are safer for cyclists – or cycling paths are constructed – bicycling is unlikely to become common in the city, despite the fact that Nairobi’s mild climate and relatively flat topography are both favorable to bicycling. It is worth mentioning that some of the poorest residents would be willing to bicycle commute in spite of the risk, but many of them cannot afford bicycles. Just over a quarter of those who do not own a bicycle report that their reason is affordability, and two-thirds of these respondents would cycle-commute if they had a bicycle. 3.2.3. Use and perceptions of matatus Matatus are privately-owned and operated public transport vehicles, usually 14- and 25-seater vans and small buses. They are the most common transport mode in Nairobi, used by approximately two-thirds of adult travelers and almost half of adult nontravelers. Here, we summarize survey findings regarding respondents’ suggestions for improving matatu transport. Later in this paper, we’ll present findings about matatu mode choice as part of a larger statistical model of car ownership and mode choice. Matatus constitute the paratransit system in the city; there is no formal transit system to speak of. One advantage of an informal transit system is that the drivers can choose the routes where they know there will be passengers, rather than providing service only on a given route network. It is perhaps for this reason that among surveyed households, an impressive 99% reported that matatus are available to them. Although matatus clearly provide a valuable service in Nairobi, there are also problems with this mode of transportation. Most of the literature on informal paratransit systems cites problems that appear to be solvable by increased governmental oversight. These include unsafe and highly-polluting vehicles, driving patterns that endanger others on the roadways, congesting roads in central areas, poor working conditions, and mafia-style business structures (see e.g. Kaltheier, 2002). Our data show that survey respondents in Nairobi echoed these concerns, as well as the suggestion that government oversight might solve these problems. The survey asked a set of questions asking respondents’ opinions about public transport in the city. Table 3 and Table 4 summarize responses to questions about desired changes to the public transport system, and conditions given for switching to public transport. There is substantial agreement among Nairobi residents. For all respondents, the most often-suggested future change is for the government to have a larger and more direct role in the public

transport system – though the transit riders and walkers feel more strongly about this than the car drivers. The second most often suggested change is to improve the overall road network. It is not surprising that car drivers are especially aware of the need for this improvement, since they are personally navigating their vehicles away from the potholes. Beyond these two main suggestions, there is support from all respondents for stricter enforcement of existing regulations, and varying support for other actions, many of them to be led by the government. Hook and Howe (2005) provide a nuanced discussion of the prospects for successful formal public transport in the African context. Their main point is that African cities – including Nairobi – are both lower income overall, less dense, and have less roadway infrastructure per capita than cities elsewhere in the developing world. Relatively low densities mean that it is expensive to provide service to all areas, low incomes put downward pressure on fares (both through the political process and through market demand), and the relative lack of roadway infrastructure means that public transport vehicles are operating on already (and increasingly) congested routes. This means that despite clear evidence that the general population is supportive, more formal (i.e. more regulated) forms of public transport are unlikely to be successful in Africa without complementary government actions that change these underlying conditions. As for the conditions under which respondents would switch to using public transport, this question produced substantial differences of opinion on public transport improvement priorities. Walkers and current transit users cited fare regulations as their highest priority, while car drivers were much more concerned with public transport vehicle quality, and to a lesser extent with improved safety and security. As expected, car drivers and walkers both desire more and differentiated public transport service more than existing transit users do. These differences could indicate differences in the public transport policies and investments that each group will support going forward. 3.2.4. Traveling on foot Walking is the oldest and one of the most environmentallyfriendly ways of getting around. In wealthier cities, transport planners expend substantial effort to encourage walking as a transport mode by creating pedestrian-friendly environments. In Nairobi, foot travel is a sizable share of all travel. Approximately 1=4 of traveler adults and over half of non-traveler adults in our sample reported that walking was their most frequent mode of transport. This is undoubtedly a good thing from an environmental perspective. However, to the extent that people in Nairobi are forced to walk long distances because they cannot afford motorized modes, the high fraction of walking also reflects a physical and time burden on the poor for basic travel needs. Travelers who walk in our sample are indeed disproportionately those who are extremely poor – those living below the Kenyan urban poverty line. Of adult travelers who live below the Kenyan urban poverty line, 41% ‘‘choose’’ walking as their most frequent means of travel (see Fig. 8). Walking as the primary transport mode for adult travelers drops off dramatically just above the poverty line – to 24% and then down to approximately 10% of wealthier travelers. In the child traveling population (under 16 years old), there is more walking overall at every wealth level, but the basic pattern remains (see Fig. 9). Table 5 presents the results of a simple logistic regression of the probability of walking in the adult traveler portion of our sample. Most of the statistically significant coefficients have the expected signs. Those traveling longer distances, those from wealthier households, and car owners are less likely to walk, all else equal. Those from larger households are also less likely to walk. Those who live in denser areas are more likely to walk, as are those

73

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76 Table 3 Desired future changes to public transport in Nairobi.

Enhance security for passengers and crew, remove criminal gangs Government direct role and intervention in public transport A professional association for urban public transport Public service vehicle redesign Expansion of the railway system Strict enforcement of regulations The government to create enabling environment for the development Improvement of the road network The government to institute a grand metropolitan transport Comprehensive review/reversal of all the new regulations Customer care + company provision for their staff Traffic management + traffic lights + police

Transit users and walkers (N = 1804)

Car drivers (N = 214)

Most important (%)

Mentioned (%)

Most important (%)

Mentioned (%)

2 52 6 2 1 9 9 12 3 1 1 2

7 74 16 5 5 25 21 35 9 3 2 12

3 26 6 2 4 12 8 23 6 1 2 5

6 46 17 2 10 24 17 41 19 2 3 20

Table 4 Conditions for ‘‘switching’’ to public transport among Nairobi residents. Percent who listed the condition in their Top 3

Regulate fares and introduce a fare policy Improve security Improve public transport network Improve public transport vehicle quality Improve transport infrastructure Provide more and differentiated transit services Improve traffic flow and traffic management Reform the matatu industry and public transport system Improve safety Quick compensation and improved handling of accident situations Have no choice and other alternative Vehicle restrictions in the CBD Crew welfare

Walkers (N = 179) (%)

Car drivers (N = 203) (%)

Transit users (N = 495) (%)

69 26 5 32 28 36 20 2 12 1 1 0 1

31 32 9 62 21 43 17 6 22 0 1 0 0

69 23 2 35 32 28 23 3 12 0 0 0 0

Notes: (1) As indicated by the sample sizes above (N), many respondents did not answer this question. (2) Many people who responded to this question are already matatu users. We interpret their responses to be conditions for switching from using matatus to using a formal public transport system.

Table 5 Logit model of the likelihood of walking for adult travelers aged 16 and older.

Distance Male HH size Car owner Age 24–35 Age 36–54 Age 55 and over 1–2X Poverty 2–3X Poverty 3–4X Poverty 4–5X Poverty >5X Poverty Central home zone Middle ring home zone Area of home zone Density of home zone Constant

Coefficient

P value

0.08** 0.06 0.13*** 0.48** 0.37*** 0.18 0.19 0.83*** 1.08*** 1.08*** 1.75*** 1.33*** 0.26 0.42*** 0.01*** 1.09E5*** 0.06

0.011 0.570 0.000 0.018 0.006 0.242 0.548 0.000 0.000 0.000 0.000 0.000 0.217 0.008 0.000 0.002 0.843

Dependent variable: walk as most frequent means of transport (0, 1) N = 2309, standardized weights are used Pseudo R-squared = 0.10 

Statistical significance of coefficient estimates for 0.10 level. Statistical significance of coefficient estimates for 0.05 level. *** Statistical significance of coefficient estimates for 0.01 level. **

who live in larger zones. In general, zone size increases with distance from the CBD. One of the most interesting results here is the explanatory variables that are not statistically significant – the gender of the traveler and most of the age categories. Although both gender and age are correlated with adult traveler status (men of all ages are more likely to be travelers than women, and older men are more likely to be travelers than younger men), this model shows that among adult travelers, men and women are equally likely to walk. Similarly, adults in all age categories with the exception of ages 24–35 are equally likely to walk, all else equal.

4. Multinomial logit model of transport mode and car ownership choice So far in this article, we have focused on each transport mode separately and described our findings. Here, we estimate a multinomial logit model of the joint choice of transport mode and car ownership for our sample. The results provide clues about substitution patterns among transport modes in Nairobi. For a detailed review of discrete choice multinomial logit methodology and theory, see Train (2003) or Ben-Akiva and Lerman (1985).

74

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

Table 6 Multinomial logit model of car ownership and mode choice for adults (age 16 and older).

Distance Male HH Size Traveler Bike Own Age 16–23 Age 24–35 Age 55+ 1–2X Poverty 2–3X Poverty 3–4X Poverty 4–5X Poverty >5X Poverty Central Middle ring Zone area Densitya Distance to CBD Constant

Car-free matatu

Car-free car

Car-own walk

Coef.

p-val.

Coef.

p-val.

Coef.

0.08 0.20 0.03 0.69 0.30 0.52 0.06 0.06 0.64 0.80 0.76 1.56 1.20 0.13 0.41 0.00 0.14 0.03 0.21

0.006 0.006 0.112 0.000 0.022 0.000 0.585 0.785 0.000 0.000 0.000 0.000 0.000 0.534 0.003 0.073 0.000 0.368 0.490

0.11 0.23 0.00 0.57 0.66 0.93 0.27 1.59 1.48 1.92 2.49 4.03 4.17 1.68 1.92 0.01 0.33 0.32 2.78

0.132 0.514 0.964 0.236 0.163 0.078 0.520 0.003 0.013 0.006 0.001 0.000 0.000 0.007 0.000 0.079 0.028 0.001 0.004

0.03 0.61 0.27 0.07 0.74 0.09 0.23 0.01 1.21 3.16 4.19 4.96 5.67 0.27 0.47 0.01 0.30 0.16 4.53

Car-own matatu

Car-own car

p-val.

Coef.

p-val.

Coef.

p-val.

0.641 0.002 0.000 0.850 0.028 0.737 0.356 0.990 0.015 0.000 0.000 0.000 0.000 0.545 0.107 0.000 0.000 0.025 0.000

0.09 0.39 0.46 0.66 0.69 0.23 0.03 0.37 2.03 4.12 5.08 5.96 5.87 0.31 0.61 0.01 0.46 0.15 5.62

0.028 0.007 0.000 0.005 0.003 0.279 0.868 0.366 0.000 0.000 0.000 0.000 0.000 0.388 0.011 0.000 0.000 0.007 0.000

0.01 0.19 0.40 1.16 0.28 2.43 0.90 0.57 0.93 3.09 4.37 5.03 5.84 1.45 0.96 0.02 0.68 0.30 2.92

0.851 0.228 0.000 0.000 0.333 0.000 0.000 0.093 0.013 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Dependent variable: car ownership and most frequent mode of transport Car-Free and Walk is the base alternative N = 4818, standardized weights are used Pseudo R-squared from constants = 0.251 Pseudo R-squared from zero = 0.468 Boldface indicates statistical significance of coefficient estimates at the 0.10 level or higher. Listed p-values provide individual estimates of statistical significance. a Density in 10,000 people per square kilometer.

Table 6 presents the results of a multinomial logit model of the joint choice of car ownership and mode choice for Nairobi adults.8 Joint choice models have compound choice sets, meaning that each alternative in the choice set is composed of more than one subchoice alternative. In the present application, the alternatives are combinations of car ownership and mode choice, and there are six of them. Each set of columns in Table 6 reports estimated coefficients and their associated p-values that indicate how each variable affects the probability of choosing that alternative compared to the base alternative: car-free and walking. The coefficient signs indicate the direction of the estimated relationship between the variable’s magnitude and the likelihood of choosing the alternative compared to the base alternative. Overall, the statistically significant coefficients have the expected signs. For instance, the effect of all expenditure levels above the poverty line is to increase the likelihood of all alternatives relative to the base alternative of car-free and walking, which is also the only alternative that has zero cost. Increased density of the home zone reduces the likelihood of all alternatives relative to the car-free and walk alternative. We leave further interpretation of individual coefficient estimates to the reader, and turn to a discussion of the non-marginal effects predicted by this model. Table 7 shows the non-marginal effects of changes in selected independent variables on the likelihood of choosing each alternative, expressed as percentage point changes from the shares originally predicted by the model. These results clearly show the predicted substitution patterns between car-mode alternatives, which can be used as an indication of how car ownership and transport mode shares may change as Nairobi changes over time.

8 We also estimated a nested logit specification of this model, with the nests being the car-free alternatives and the car-owning alternatives. We do not present this as our preferred model because one of the inclusive value estimates was substantially greater than one, indicating that this specification is not consistent with utility maximization.

To calculate these effects, we used our model to predict how the probabilities of choosing each alternative would change for each individual in our sample with changes in selected independent variables. The non-marginal effects (NME) reported in Table 7 are weighted averages of the estimated individual effects, calculated as follows:

P NME ¼

i wi ðp1i

P

 p0i Þ

i wi

where wi are probability weights for each individual i, based on home location zone, p0i is the predicted probability of choosing the alternative for each individual i, p1i is the predicted probability of choosing the alternative for each individual i after the non-marginal change in the independent variable(s). Because most of our modeled scenarios are changes in circumstances for a subset of our full sample, these effect estimates are based on different numbers of observations, as specified in Table 7. For example, the scenario of increasing travel distance is modeled for all individuals (N = 4818), whereas the scenario of nontravelers becoming travelers is modeled only for those households that are non-travelers in our sample (N = 2547). There are three main substitution patterns that emerge from Table 7. The first is that as people move from non-traveler to traveler status, they substitute matatu use for walking. Among those in our sample who are non-travelers, the predicted share of car-free walkers would be reduced by 13.7 percentage points if they became travelers, and the predicted share of car-free matatu riders would increase by 12.1 percentage points. The second substitution pattern that emerges is that for all expenditure categories, there is a strong movement from car-free living toward a car-owning lifestyle as households move upward in expenditure level. As for transport modes, the dominant substitution pattern depends on the original expenditure level. When the poorest households move into the middle expenditure group and as middle expenditure households move to the 4–5 times poverty expenditure level, they

75

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76 Table 7 Estimated non-marginal effects in percentage point changes in the likelihoods of choosing each alternative for adults (age 16 and older).

Distance traveled increases by 1 km Non-travelers become travelers HH in poorest two expenditure groups move to 2–3X poverty HH in middle two expenditure groups move to 4–5X poverty HH in 4–5X poverty move to wealthiest group HH in wealthiest group move to 4–5X poverty

N

Car-free walk

4818 2547 3064

1.3 13.7 17.5

1.4 12.1 2.8

1035 203 516

12.4 0.4 0.6

3.4 12.5 12.1

mainly substitute matatu use for walking. When relatively wealthy people become even wealthier (i.e. move to the highest expenditure category), the dominant transport mode substitution is car use for matatu use.

5. Discussion The city of Nairobi and its metropolitan region are facing an enormous planning challenge in transportation, land use, and economic development. Rapid population growth for the region is expected, and roads are already congested with vehicle traffic, causing increasingly dangerous levels of air pollution and transport delays that hurt public health and the regional economy. Meanwhile, a substantial portion of the local population cannot regularly afford any form of motorized transportation. Approximately two-thirds of Nairobi residents live near or below the Kenyan urban poverty line. Unlike many other cities in the developing world, intermediate modes such as the bicycle or motorized two- and threewheelers are largely absent in Nairobi. This leaves Nairobi’s poor substantially mobility-constrained, and many of these people simply do not travel far from home; trips are short, and approximately half of adults do not regularly leave their home neighborhoods. Informal, unplanned, and underserved settlements (a.k.a. slums) line up along the main transport corridors of the city and near employment centers. Because the city largely lacks pedestrian and bicycle infrastructure to serve this segment of the population, they risk their lives walking (and occasionally cycling) alongside the congested roadways. In addition to the public safety problem that this presents, the mix of motorized and non-motorized traffic causes further congestion and delay. This article has presented an in-depth exploration of unique travel survey data for Nairobi, Kenya. Our goal was to use these data to inform future transport policies and investments in this city, striving to improve the system for all users. Overall, our findings are not surprising, as they are largely consistent with our expectations based on theory of travel behavior. The fact that we do not come to any new conclusions upon analyzing the data is reassuring. It means that in Nairobi, people make travel decisions more-or-less the way we would expect them to. The super-poverty experienced by much of Nairobi’s population does not radically change this fact. Our results show that as people in Nairobi become wealthier, they increasingly choose safer, more comfortable, and more reliable modes of transport. In Nairobi as in many other cities, this means that those who can afford them will purchase and use private cars. Those who cannot afford cars, but can afford to pay for transportation, choose a combination of the local public transport service provided by matatus and walking, depending on the particular trip. Those who cannot afford to regularly pay for transportation simply walk. Recall that the current situation in Nairobi is characterized by severe traffic congestion and associated negative impacts on the local economy and air quality in spite of the fact that the vast major-

Car-free matatu

Car-free car

Car-own walk

Car-own matatu

Car-own car

0.1 0 0.2

0.1 1.9 4.1

2.7 0.1 10.5

3.0 2.9 5.5

1.4 0.1 0.2

3.1 5.4 4.9

7.3 4.8 5.8

4.0 12.4 13.8

ity of Nairobi residents do not own cars and a large portion cannot afford matatu fares. If residents of Nairobi in fact do become wealthier, our results predict that there will be strong growth in both car ownership and use and in matatu demand. For instance, Table 7 indicates that in the absence of changes to the transport system in the city, relatively small absolute increases in expenditure levels for the poorest in Nairobi are predicted to lead to a net 13 percentage point reduction in walking among this subpopulation. Given that approximately two-thirds of Nairobi residents are in this category, this would put an additional 8% of the population on the roadways in motorized vehicles, worsening traffic congestion and air pollution further. Our survey data suggest some policy options that could temper this undesirable future scenario. These include making non-motorized modes more attractive by making them safer, making public transport more attractive by creating differentiated services that cater to high- and low-income riders and by increasing government oversight of the system, and introducing policies that will reduce traffic congestion such as road pricing for private cars, especially near the CBD. While these are not new policy suggestions for Nairobi, our data and analysis provide a stark prediction of a future without such changes to the city’s transportation system. Acknowledgements This work was made possible by funds provided to KIPPRA from the UMBRELLA Project focusing on improving the Enabling Business Environment in Kenya. The funds were applied to collect the data and implement the preliminary analysis. Further analysis of the data was enabled through financial support from the Volvo Research and Educational Foundation and from The Earth Institute at Columbia University. References Abane, A.M., 2011. Travel behaviour in Ghana: empirical observations from four metropolitan areas. Journal of Transport Geography 19 (2), 313–322. Amnesty International, 2009. The Unseen Majority: Nairobi’s Two Million SlumDwellers. Report AFR 32/005/2009. . Behrens, R., 2004. Understanding travel needs of the poor: towards improved travel analysis practices in South Africa. Transport Reviews 24 (3), 317–336. Behrens, R., 2009. What the NHTS reveals about non-motorised transport in the RSA. In: Khadpekar, N. (Ed.), Non-motorised Transportation: Making it a Viable Option. ICFAI University Press, Ahmedabad, India. Ben-Akiva, M., Lerman, S.R., 1985. Discrete Choice Analysis: Theory and Application to Travel Demand. MIT Press, Cambridge, MA. Bryceson, D.F., Howe, J., 2000. Poverty and Urban Transport in East Africa: Review of Research and Dutch Donor Experience. Report prepared for the World Bank, 147 pp. Bryceson, D.F., Mbara, T.C., Maunder, D., 2003. Livelihoods, daily mobility and poverty in sub-saharan Africa. Transport Reviews 23 (2), 177–196. Central Bureau of Statistics (CBS), 1999. Kenyan Census Data, . Deaton, A., Zaidi, S., 2002. Section 2.4: Income versus consumption. In Guidelines for Constructing Consumption Aggregates for Welfare Analysis. LSMS Working Paper Number 135. The World Bank, Washington, DC, pp. 11–13. Gakenheimer, R., 1999. Urban mobility in the developing world. Transportation Research Part A 33, 671–689.

76

D. Salon, E.M. Aligula / Journal of Transport Geography 22 (2012) 65–76

Gebeyehu, M., Takano, S., 2007. Diagnostic evaluation of public transportation mode choice in Addis Ababa. Journal of Public Transportation 10 (4), 27–50. Gwilliam, K., 2003. Urban transport in developing countries. Transport Reviews 23 (2), 197–216. Hook, W., Howe, J., 2005. Transport and the Millennium Development Goals: A Background Paper to the Task Force on Slum Dwellers of the Millennium Project, 100 pp, . Kaltheier, Ralf M., 2002. Urban Transport and Poverty in Developing Countries: Analysis and Options for Transport Policy and Planning. Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ), Eschborn, Germany, 48 pp. Kamuhanda, R., Schmidt, O., 2009. Matatu: a case study of the core segment of the public transport market of Kampala, Uganda. Transport Reviews 29 (1), 129– 142. Marras, S. et al., 2011. , (accessed 18.04.11). Palmer, C.J., Astrop, A.J., Maunder, D.A.C., 1997. Constraints, Attitudes and Travel Behavior of Low Income Households in Two Developing Cities. Transport Research Laboratory (TRL) Report 263, 25 pp.

Salon, D., Gulyani, S., 2010. Mobility, poverty, and gender: Travel ‘choices’ of slum residents in Nairobi, Kenya. Transport Reviews 30 (5), 641–657. Train, K., 2003. Discrete Choice Methods with Simulation. Cambridge University Press. United Nations, 2005. Sampling frames and master samples. In: Designing Household Survey Samples: Practical Guidelines. United Nations, New York, pp. 83–108 (Chapter 4), . US Census, 2009. American Community Survey Data for New York City. . Venter, C., Vokolkova, V., Michalek, J., 2007. Gender, residential location, and household travel: empirical findings from low-income urban settlements in Durban, South Africa. Transport Reviews 27 (6), 653–677. World Bank, 2006. Kenya: Inside Informality: Poverty, Jobs, Housing and Services in Nairobi’s Slums. Report No. 36347-KE, Water and Urban Unit 1, Africa Region. The World Bank, Washington, DC. World Bank, 2008. Kenya Poverty and Inequality Assessment. Report No. 44190-KE, Poverty Reduction and Economic Management Unit, Africa Region. The World Bank, Washington, DC.